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> 10ce9adaa5SChris Mason #include <linux/workqueue.h> 11a74a4b97SChris Mason #include <linux/kthread.h> 125a0e3ad6STejun Heo #include <linux/slab.h> 13784b4e29SChris Mason #include <linux/migrate.h> 147a36ddecSDavid Sterba #include <linux/ratelimit.h> 156463fe58SStefan Behrens #include <linux/uuid.h> 16803b2f54SStefan Behrens #include <linux/semaphore.h> 17540adea3SMasami Hiramatsu #include <linux/error-injection.h> 189678c543SNikolay Borisov #include <linux/crc32c.h> 19b89f6d1fSFilipe Manana #include <linux/sched/mm.h> 207e75bf3fSDavid Sterba #include <asm/unaligned.h> 216d97c6e3SJohannes Thumshirn #include <crypto/hash.h> 22eb60ceacSChris Mason #include "ctree.h" 23eb60ceacSChris Mason #include "disk-io.h" 24e089f05cSChris Mason #include "transaction.h" 250f7d52f4SChris Mason #include "btrfs_inode.h" 260b86a832SChris Mason #include "volumes.h" 27db94535dSChris Mason #include "print-tree.h" 28925baeddSChris Mason #include "locking.h" 29e02119d5SChris Mason #include "tree-log.h" 30fa9c0d79SChris Mason #include "free-space-cache.h" 3170f6d82eSOmar Sandoval #include "free-space-tree.h" 32581bb050SLi Zefan #include "inode-map.h" 3321adbd5cSStefan Behrens #include "check-integrity.h" 34606686eeSJosef Bacik #include "rcu-string.h" 358dabb742SStefan Behrens #include "dev-replace.h" 3653b381b3SDavid Woodhouse #include "raid56.h" 375ac1d209SJeff Mahoney #include "sysfs.h" 38fcebe456SJosef Bacik #include "qgroup.h" 39ebb8765bSAnand Jain #include "compression.h" 40557ea5ddSQu Wenruo #include "tree-checker.h" 41fd708b81SJosef Bacik #include "ref-verify.h" 42aac0023cSJosef Bacik #include "block-group.h" 43b0643e59SDennis Zhou #include "discard.h" 44f603bb94SNikolay Borisov #include "space-info.h" 45eb60ceacSChris Mason 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 101141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 102141386e1SJosef Bacik { 103141386e1SJosef Bacik if (fs_info->csum_shash) 104141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 105141386e1SJosef Bacik } 106141386e1SJosef Bacik 107d352ac68SChris Mason /* 108d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 109d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 110d352ac68SChris Mason * just before they are sent down the IO stack. 111d352ac68SChris Mason */ 11244b8bd7eSChris Mason struct async_submit_bio { 113c6100a4bSJosef Bacik void *private_data; 11444b8bd7eSChris Mason struct bio *bio; 115a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 11644b8bd7eSChris Mason int mirror_num; 117eaf25d93SChris Mason /* 118eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 119eaf25d93SChris Mason * can't tell us where in the file the bio should go 120eaf25d93SChris Mason */ 121eaf25d93SChris Mason u64 bio_offset; 1228b712842SChris Mason struct btrfs_work work; 1234e4cbee9SChristoph Hellwig blk_status_t status; 12444b8bd7eSChris Mason }; 12544b8bd7eSChris Mason 12685d4e461SChris Mason /* 12785d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 12885d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 12985d4e461SChris Mason * the level the eb occupies in the tree. 1304008c04aSChris Mason * 13185d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 13285d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 13385d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 1344fd786e6SMisono Tomohiro * by btrfs_root->root_key.objectid. This ensures that all special purpose 1354fd786e6SMisono Tomohiro * roots have separate keysets. 1364008c04aSChris Mason * 13785d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 13885d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 13985d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1404008c04aSChris Mason * 14185d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 14285d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 14385d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 14485d4e461SChris Mason * 14585d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 14685d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 14785d4e461SChris Mason * needs update as well. 1484008c04aSChris Mason */ 1494008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1504008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1514008c04aSChris Mason # error 1524008c04aSChris Mason # endif 15385d4e461SChris Mason 15485d4e461SChris Mason static struct btrfs_lockdep_keyset { 15585d4e461SChris Mason u64 id; /* root objectid */ 15685d4e461SChris Mason const char *name_stem; /* lock name stem */ 15785d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 15885d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 15985d4e461SChris Mason } btrfs_lockdep_keysets[] = { 16085d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 16185d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 16285d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 16385d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 16485d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 16585d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 16660b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 16785d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 16885d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 16985d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 17013fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 1716b20e0adSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, 17285d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1734008c04aSChris Mason }; 17485d4e461SChris Mason 17585d4e461SChris Mason void __init btrfs_init_lockdep(void) 17685d4e461SChris Mason { 17785d4e461SChris Mason int i, j; 17885d4e461SChris Mason 17985d4e461SChris Mason /* initialize lockdep class names */ 18085d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 18185d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 18285d4e461SChris Mason 18385d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 18485d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 18585d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 18685d4e461SChris Mason } 18785d4e461SChris Mason } 18885d4e461SChris Mason 18985d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 19085d4e461SChris Mason int level) 19185d4e461SChris Mason { 19285d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 19385d4e461SChris Mason 19485d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 19585d4e461SChris Mason 19685d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 19785d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 19885d4e461SChris Mason if (ks->id == objectid) 19985d4e461SChris Mason break; 20085d4e461SChris Mason 20185d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 20285d4e461SChris Mason &ks->keys[level], ks->names[level]); 20385d4e461SChris Mason } 20485d4e461SChris Mason 2054008c04aSChris Mason #endif 2064008c04aSChris Mason 207d352ac68SChris Mason /* 208d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 209d352ac68SChris Mason * that covers the entire device 210d352ac68SChris Mason */ 2116af49dbdSDavid Sterba struct extent_map *btree_get_extent(struct btrfs_inode *inode, 21239b07b5dSOmar Sandoval struct page *page, size_t pg_offset, 21339b07b5dSOmar Sandoval u64 start, u64 len) 2145f39d397SChris Mason { 215fc4f21b1SNikolay Borisov struct extent_map_tree *em_tree = &inode->extent_tree; 2165f39d397SChris Mason struct extent_map *em; 2175f39d397SChris Mason int ret; 2185f39d397SChris Mason 219890871beSChris Mason read_lock(&em_tree->lock); 220d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 221a061fc8dSChris Mason if (em) { 222890871beSChris Mason read_unlock(&em_tree->lock); 2235f39d397SChris Mason goto out; 224a061fc8dSChris Mason } 225890871beSChris Mason read_unlock(&em_tree->lock); 2267b13b7b1SChris Mason 227172ddd60SDavid Sterba em = alloc_extent_map(); 2285f39d397SChris Mason if (!em) { 2295f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2305f39d397SChris Mason goto out; 2315f39d397SChris Mason } 2325f39d397SChris Mason em->start = 0; 2330afbaf8cSChris Mason em->len = (u64)-1; 234c8b97818SChris Mason em->block_len = (u64)-1; 2355f39d397SChris Mason em->block_start = 0; 236d1310b2eSChris Mason 237890871beSChris Mason write_lock(&em_tree->lock); 23809a2a8f9SJosef Bacik ret = add_extent_mapping(em_tree, em, 0); 2395f39d397SChris Mason if (ret == -EEXIST) { 2405f39d397SChris Mason free_extent_map(em); 2417b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 242b4f359abSTsutomu Itoh if (!em) 2430433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2445f39d397SChris Mason } else if (ret) { 2457b13b7b1SChris Mason free_extent_map(em); 2460433f20dSTsutomu Itoh em = ERR_PTR(ret); 2475f39d397SChris Mason } 248890871beSChris Mason write_unlock(&em_tree->lock); 2497b13b7b1SChris Mason 2505f39d397SChris Mason out: 2515f39d397SChris Mason return em; 2525f39d397SChris Mason } 2535f39d397SChris Mason 254d352ac68SChris Mason /* 2552996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 256d352ac68SChris Mason */ 257c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 25819c00ddcSChris Mason { 259d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 260e9be5a30SDavid Sterba const int num_pages = fs_info->nodesize >> PAGE_SHIFT; 261d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 26219c00ddcSChris Mason char *kaddr; 263e9be5a30SDavid Sterba int i; 264d5178578SJohannes Thumshirn 265d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 266d5178578SJohannes Thumshirn crypto_shash_init(shash); 267e9be5a30SDavid Sterba kaddr = page_address(buf->pages[0]); 268e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 269e9be5a30SDavid Sterba PAGE_SIZE - BTRFS_CSUM_SIZE); 27019c00ddcSChris Mason 271e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 272e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 273e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 27419c00ddcSChris Mason } 27571a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 276d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 27719c00ddcSChris Mason } 27819c00ddcSChris Mason 279d352ac68SChris Mason /* 280d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 281d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 282d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 283d352ac68SChris Mason * in the wrong place. 284d352ac68SChris Mason */ 2851259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 286b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 287b9fab919SChris Mason int atomic) 2881259ab75SChris Mason { 2892ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 2901259ab75SChris Mason int ret; 2912755a0deSDavid Sterba bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); 2921259ab75SChris Mason 2931259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 2941259ab75SChris Mason return 0; 2951259ab75SChris Mason 296b9fab919SChris Mason if (atomic) 297b9fab919SChris Mason return -EAGAIN; 298b9fab919SChris Mason 299a26e8c9fSJosef Bacik if (need_lock) { 300a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 301300aa896SDavid Sterba btrfs_set_lock_blocking_read(eb); 302a26e8c9fSJosef Bacik } 303a26e8c9fSJosef Bacik 3042ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 305ff13db41SDavid Sterba &cached_state); 3060b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3071259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3081259ab75SChris Mason ret = 0; 3091259ab75SChris Mason goto out; 3101259ab75SChris Mason } 31194647322SDavid Sterba btrfs_err_rl(eb->fs_info, 31294647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 31394647322SDavid Sterba eb->start, 31429549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 3151259ab75SChris Mason ret = 1; 316a26e8c9fSJosef Bacik 317a26e8c9fSJosef Bacik /* 318a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 319a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 32001327610SNicholas D Steeves * block that has been freed and re-allocated. So don't clear uptodate 321a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 322a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 323a26e8c9fSJosef Bacik * making everybody very sad. 324a26e8c9fSJosef Bacik */ 325a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 3260b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 32733958dc6SChris Mason out: 3282ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 329e43bbe5eSDavid Sterba &cached_state); 330472b909fSJosef Bacik if (need_lock) 331a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 3321259ab75SChris Mason return ret; 3331259ab75SChris Mason } 3341259ab75SChris Mason 335e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 336e7e16f48SJohannes Thumshirn { 337e7e16f48SJohannes Thumshirn switch (csum_type) { 338e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 3393951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 3403831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 341352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 342e7e16f48SJohannes Thumshirn return true; 343e7e16f48SJohannes Thumshirn default: 344e7e16f48SJohannes Thumshirn return false; 345e7e16f48SJohannes Thumshirn } 346e7e16f48SJohannes Thumshirn } 347e7e16f48SJohannes Thumshirn 348d352ac68SChris Mason /* 3491104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 3501104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 3511104a885SDavid Sterba */ 352ab8d0fc4SJeff Mahoney static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 353ab8d0fc4SJeff Mahoney char *raw_disk_sb) 3541104a885SDavid Sterba { 3551104a885SDavid Sterba struct btrfs_super_block *disk_sb = 3561104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 35751bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 358d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 359d5178578SJohannes Thumshirn 360d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3611104a885SDavid Sterba 3621104a885SDavid Sterba /* 3631104a885SDavid Sterba * The super_block structure does not span the whole 36451bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 36551bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 3661104a885SDavid Sterba */ 367fd08001fSEric Biggers crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, 368fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 3691104a885SDavid Sterba 37051bce6c9SJohannes Thumshirn if (memcmp(disk_sb->csum, result, btrfs_super_csum_size(disk_sb))) 371e7e16f48SJohannes Thumshirn return 1; 3721104a885SDavid Sterba 373e7e16f48SJohannes Thumshirn return 0; 3741104a885SDavid Sterba } 3751104a885SDavid Sterba 376e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 377ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 378581c1760SQu Wenruo { 379e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 380581c1760SQu Wenruo int found_level; 381581c1760SQu Wenruo struct btrfs_key found_key; 382581c1760SQu Wenruo int ret; 383581c1760SQu Wenruo 384581c1760SQu Wenruo found_level = btrfs_header_level(eb); 385581c1760SQu Wenruo if (found_level != level) { 38663489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 38763489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 388581c1760SQu Wenruo btrfs_err(fs_info, 389581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 390581c1760SQu Wenruo eb->start, level, found_level); 391581c1760SQu Wenruo return -EIO; 392581c1760SQu Wenruo } 393581c1760SQu Wenruo 394581c1760SQu Wenruo if (!first_key) 395581c1760SQu Wenruo return 0; 396581c1760SQu Wenruo 3975d41be6fSQu Wenruo /* 3985d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 3995d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 4005d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 4015d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 4025d41be6fSQu Wenruo */ 4035d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 4045d41be6fSQu Wenruo return 0; 40562fdaa52SQu Wenruo 40662fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 40762fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 40862fdaa52SQu Wenruo btrfs_err(fs_info, 40962fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 41062fdaa52SQu Wenruo eb->start); 41162fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 41262fdaa52SQu Wenruo return -EUCLEAN; 41362fdaa52SQu Wenruo } 41462fdaa52SQu Wenruo 415581c1760SQu Wenruo if (found_level) 416581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 417581c1760SQu Wenruo else 418581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 419581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 420581c1760SQu Wenruo 421581c1760SQu Wenruo if (ret) { 42263489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 42363489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 424581c1760SQu Wenruo btrfs_err(fs_info, 425ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 426ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 427ff76a864SLiu Bo first_key->type, first_key->offset, 428ff76a864SLiu Bo found_key.objectid, found_key.type, 429ff76a864SLiu Bo found_key.offset); 430581c1760SQu Wenruo } 431581c1760SQu Wenruo return ret; 432581c1760SQu Wenruo } 433581c1760SQu Wenruo 4341104a885SDavid Sterba /* 435d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 436d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 437581c1760SQu Wenruo * 438581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 439581c1760SQu Wenruo * @level: expected level, mandatory check 440581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 441d352ac68SChris Mason */ 4425ab12d1fSDavid Sterba static int btree_read_extent_buffer_pages(struct extent_buffer *eb, 443581c1760SQu Wenruo u64 parent_transid, int level, 444581c1760SQu Wenruo struct btrfs_key *first_key) 445f188591eSChris Mason { 4465ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 447f188591eSChris Mason struct extent_io_tree *io_tree; 448ea466794SJosef Bacik int failed = 0; 449f188591eSChris Mason int ret; 450f188591eSChris Mason int num_copies = 0; 451f188591eSChris Mason int mirror_num = 0; 452ea466794SJosef Bacik int failed_mirror = 0; 453f188591eSChris Mason 4540b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 455f188591eSChris Mason while (1) { 456f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 457c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); 458256dd1bbSStefan Behrens if (!ret) { 459581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 460b9fab919SChris Mason parent_transid, 0)) 461256dd1bbSStefan Behrens ret = -EIO; 462e064d5e9SDavid Sterba else if (btrfs_verify_level_key(eb, level, 463ff76a864SLiu Bo first_key, parent_transid)) 464581c1760SQu Wenruo ret = -EUCLEAN; 465581c1760SQu Wenruo else 466581c1760SQu Wenruo break; 467256dd1bbSStefan Behrens } 468d397712bSChris Mason 4690b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 470f188591eSChris Mason eb->start, eb->len); 4714235298eSChris Mason if (num_copies == 1) 472ea466794SJosef Bacik break; 4734235298eSChris Mason 4745cf1ab56SJosef Bacik if (!failed_mirror) { 4755cf1ab56SJosef Bacik failed = 1; 4765cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4775cf1ab56SJosef Bacik } 4785cf1ab56SJosef Bacik 479f188591eSChris Mason mirror_num++; 480ea466794SJosef Bacik if (mirror_num == failed_mirror) 481ea466794SJosef Bacik mirror_num++; 482ea466794SJosef Bacik 4834235298eSChris Mason if (mirror_num > num_copies) 484ea466794SJosef Bacik break; 485f188591eSChris Mason } 486ea466794SJosef Bacik 487c0901581SStefan Behrens if (failed && !ret && failed_mirror) 48820a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 489ea466794SJosef Bacik 490ea466794SJosef Bacik return ret; 491f188591eSChris Mason } 49219c00ddcSChris Mason 493d352ac68SChris Mason /* 494d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 495d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 496d352ac68SChris Mason */ 497d397712bSChris Mason 49801d58472SDaniel Dressler static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) 49919c00ddcSChris Mason { 5004eee4fa4SMiao Xie u64 start = page_offset(page); 50119c00ddcSChris Mason u64 found_start; 5022996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 5032996e1f8SJohannes Thumshirn u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); 50419c00ddcSChris Mason struct extent_buffer *eb; 5058d47a0d8SQu Wenruo int ret; 506f188591eSChris Mason 5074f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5084f2de97aSJosef Bacik if (page != eb->pages[0]) 5094f2de97aSJosef Bacik return 0; 5100f805531SAlex Lyakas 51119c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 5120f805531SAlex Lyakas /* 5130f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5140f805531SAlex Lyakas * It is useful to know what went wrong. 5150f805531SAlex Lyakas */ 5160f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5170f805531SAlex Lyakas return -EUCLEAN; 5180f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5190f805531SAlex Lyakas return -EUCLEAN; 5200f805531SAlex Lyakas 521de37aa51SNikolay Borisov ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 5229a8658e3SDavid Sterba offsetof(struct btrfs_header, fsid), 5239a8658e3SDavid Sterba BTRFS_FSID_SIZE) == 0); 5240f805531SAlex Lyakas 525c67b3892SDavid Sterba csum_tree_block(eb, result); 5262996e1f8SJohannes Thumshirn 5278d47a0d8SQu Wenruo if (btrfs_header_level(eb)) 5288d47a0d8SQu Wenruo ret = btrfs_check_node(eb); 5298d47a0d8SQu Wenruo else 5308d47a0d8SQu Wenruo ret = btrfs_check_leaf_full(eb); 5318d47a0d8SQu Wenruo 5328d47a0d8SQu Wenruo if (ret < 0) { 533c06631b0SQu Wenruo btrfs_print_tree(eb, 0); 5348d47a0d8SQu Wenruo btrfs_err(fs_info, 5358d47a0d8SQu Wenruo "block=%llu write time tree block corruption detected", 5368d47a0d8SQu Wenruo eb->start); 537c06631b0SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 5388d47a0d8SQu Wenruo return ret; 5398d47a0d8SQu Wenruo } 5402996e1f8SJohannes Thumshirn write_extent_buffer(eb, result, 0, csum_size); 5418d47a0d8SQu Wenruo 5422996e1f8SJohannes Thumshirn return 0; 54319c00ddcSChris Mason } 54419c00ddcSChris Mason 545b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 5462b82032cSYan Zheng { 547b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 54801d58472SDaniel Dressler struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 54944880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 5502b82032cSYan Zheng int ret = 1; 5512b82032cSYan Zheng 5529a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 5539a8658e3SDavid Sterba BTRFS_FSID_SIZE); 5542b82032cSYan Zheng while (fs_devices) { 5557239ff4bSNikolay Borisov u8 *metadata_uuid; 5567239ff4bSNikolay Borisov 5577239ff4bSNikolay Borisov /* 5587239ff4bSNikolay Borisov * Checking the incompat flag is only valid for the current 5597239ff4bSNikolay Borisov * fs. For seed devices it's forbidden to have their uuid 5607239ff4bSNikolay Borisov * changed so reading ->fsid in this case is fine 5617239ff4bSNikolay Borisov */ 5627239ff4bSNikolay Borisov if (fs_devices == fs_info->fs_devices && 5637239ff4bSNikolay Borisov btrfs_fs_incompat(fs_info, METADATA_UUID)) 5647239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 5657239ff4bSNikolay Borisov else 5667239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 5677239ff4bSNikolay Borisov 5687239ff4bSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) { 5692b82032cSYan Zheng ret = 0; 5702b82032cSYan Zheng break; 5712b82032cSYan Zheng } 5722b82032cSYan Zheng fs_devices = fs_devices->seed; 5732b82032cSYan Zheng } 5742b82032cSYan Zheng return ret; 5752b82032cSYan Zheng } 5762b82032cSYan Zheng 577facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 578facc8a22SMiao Xie u64 phy_offset, struct page *page, 579facc8a22SMiao Xie u64 start, u64 end, int mirror) 580ce9adaa5SChris Mason { 581ce9adaa5SChris Mason u64 found_start; 582ce9adaa5SChris Mason int found_level; 583ce9adaa5SChris Mason struct extent_buffer *eb; 58415b6e8a8SDavid Sterba struct btrfs_fs_info *fs_info; 58515b6e8a8SDavid Sterba u16 csum_size; 586f188591eSChris Mason int ret = 0; 5872996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 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; 59415b6e8a8SDavid Sterba fs_info = eb->fs_info; 59515b6e8a8SDavid Sterba csum_size = btrfs_super_csum_size(fs_info->super_copy); 596d397712bSChris Mason 5970b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 5980b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 5990b32f4bbSJosef Bacik */ 60067439dadSDavid Sterba atomic_inc(&eb->refs); 6010b32f4bbSJosef Bacik 6020b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 603727011e0SChris Mason if (!reads_done) 604727011e0SChris Mason goto err; 605f188591eSChris Mason 6065cf1ab56SJosef Bacik eb->read_mirror = mirror; 607656f30dbSFilipe Manana if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 608ea466794SJosef Bacik ret = -EIO; 609ea466794SJosef Bacik goto err; 610ea466794SJosef Bacik } 611ea466794SJosef Bacik 612ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 613727011e0SChris Mason if (found_start != eb->start) { 614893bf4b1SSu Yue btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", 615893bf4b1SSu Yue eb->start, found_start); 616f188591eSChris Mason ret = -EIO; 617ce9adaa5SChris Mason goto err; 618ce9adaa5SChris Mason } 619b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 62002873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 62194647322SDavid Sterba eb->start); 6221259ab75SChris Mason ret = -EIO; 6231259ab75SChris Mason goto err; 6241259ab75SChris Mason } 625ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6261c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 627893bf4b1SSu Yue btrfs_err(fs_info, "bad tree block level %d on %llu", 628893bf4b1SSu Yue (int)btrfs_header_level(eb), eb->start); 6291c24c3ceSJosef Bacik ret = -EIO; 6301c24c3ceSJosef Bacik goto err; 6311c24c3ceSJosef Bacik } 632ce9adaa5SChris Mason 63385d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 63485d4e461SChris Mason eb, found_level); 6354008c04aSChris Mason 636c67b3892SDavid Sterba csum_tree_block(eb, result); 637a826d6dcSJosef Bacik 6382996e1f8SJohannes Thumshirn if (memcmp_extent_buffer(eb, result, 0, csum_size)) { 63935be8851SJohannes Thumshirn u8 val[BTRFS_CSUM_SIZE] = { 0 }; 6402996e1f8SJohannes Thumshirn 6412996e1f8SJohannes Thumshirn read_extent_buffer(eb, &val, 0, csum_size); 6422996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 64335be8851SJohannes Thumshirn "%s checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", 6442996e1f8SJohannes Thumshirn fs_info->sb->s_id, eb->start, 64535be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, val), 64635be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 64735be8851SJohannes Thumshirn btrfs_header_level(eb)); 6482996e1f8SJohannes Thumshirn ret = -EUCLEAN; 6492996e1f8SJohannes Thumshirn goto err; 6502996e1f8SJohannes Thumshirn } 6512996e1f8SJohannes Thumshirn 652a826d6dcSJosef Bacik /* 653a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 654a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 655a826d6dcSJosef Bacik * return -EIO. 656a826d6dcSJosef Bacik */ 6571c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 658a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 659a826d6dcSJosef Bacik ret = -EIO; 660a826d6dcSJosef Bacik } 661ce9adaa5SChris Mason 662813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 663053ab70fSLiu Bo ret = -EIO; 664053ab70fSLiu Bo 6650b32f4bbSJosef Bacik if (!ret) 6660b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 66775391f0dSQu Wenruo else 66875391f0dSQu Wenruo btrfs_err(fs_info, 66975391f0dSQu Wenruo "block=%llu read time tree block corruption detected", 67075391f0dSQu Wenruo eb->start); 671ce9adaa5SChris Mason err: 67279fb65a1SJosef Bacik if (reads_done && 67379fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 674d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 6754bb31e92SArne Jansen 67653b381b3SDavid Woodhouse if (ret) { 67753b381b3SDavid Woodhouse /* 67853b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 67953b381b3SDavid Woodhouse * again, we have to make sure it has something 68053b381b3SDavid Woodhouse * to decrement 68153b381b3SDavid Woodhouse */ 68253b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6830b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 68453b381b3SDavid Woodhouse } 6850b32f4bbSJosef Bacik free_extent_buffer(eb); 686ce9adaa5SChris Mason out: 687f188591eSChris Mason return ret; 688ce9adaa5SChris Mason } 689ce9adaa5SChris Mason 6904246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 691ce9adaa5SChris Mason { 69297eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 693ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 6949e0af237SLiu Bo struct btrfs_workqueue *wq; 695ce9adaa5SChris Mason 696ce9adaa5SChris Mason fs_info = end_io_wq->info; 6974e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 698d20f7043SChris Mason 69937226b21SMike Christie if (bio_op(bio) == REQ_OP_WRITE) { 700a0cac0ecSOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) 7019e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 702a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) 7039e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 704a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 7059e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 706a0cac0ecSOmar Sandoval else 7079e0af237SLiu Bo wq = fs_info->endio_write_workers; 7089e0af237SLiu Bo } else { 7095c047a69SOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 7109e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 711a0cac0ecSOmar Sandoval else if (end_io_wq->metadata) 7129e0af237SLiu Bo wq = fs_info->endio_meta_workers; 713a0cac0ecSOmar Sandoval else 7149e0af237SLiu Bo wq = fs_info->endio_workers; 7159e0af237SLiu Bo } 7169e0af237SLiu Bo 717a0cac0ecSOmar Sandoval btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); 7189e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 719ce9adaa5SChris Mason } 720ce9adaa5SChris Mason 7214e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 722bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7230b86a832SChris Mason { 72497eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7258b110e39SMiao Xie 72697eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 727ce9adaa5SChris Mason if (!end_io_wq) 7284e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 729ce9adaa5SChris Mason 730ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 731ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 73222c59948SChris Mason end_io_wq->info = info; 7334e4cbee9SChristoph Hellwig end_io_wq->status = 0; 734ce9adaa5SChris Mason end_io_wq->bio = bio; 73522c59948SChris Mason end_io_wq->metadata = metadata; 736ce9adaa5SChris Mason 737ce9adaa5SChris Mason bio->bi_private = end_io_wq; 738ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 73922c59948SChris Mason return 0; 74022c59948SChris Mason } 74122c59948SChris Mason 7424a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7434a69a410SChris Mason { 7444a69a410SChris Mason struct async_submit_bio *async; 7454e4cbee9SChristoph Hellwig blk_status_t ret; 7464a69a410SChris Mason 7474a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 748c6100a4bSJosef Bacik ret = async->submit_bio_start(async->private_data, async->bio, 749eaf25d93SChris Mason async->bio_offset); 75079787eaaSJeff Mahoney if (ret) 7514e4cbee9SChristoph Hellwig async->status = ret; 7524a69a410SChris Mason } 7534a69a410SChris Mason 75406ea01b1SDavid Sterba /* 75506ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 75606ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 75706ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 75806ea01b1SDavid Sterba * 75906ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 76006ea01b1SDavid Sterba * inserted into the tree. 76106ea01b1SDavid Sterba */ 7624a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7638b712842SChris Mason { 7648b712842SChris Mason struct async_submit_bio *async; 76506ea01b1SDavid Sterba struct inode *inode; 76606ea01b1SDavid Sterba blk_status_t ret; 7678b712842SChris Mason 7688b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 76906ea01b1SDavid Sterba inode = async->private_data; 7704854ddd0SChris Mason 771bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 7724e4cbee9SChristoph Hellwig if (async->status) { 7734e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 7744246a0b6SChristoph Hellwig bio_endio(async->bio); 77579787eaaSJeff Mahoney return; 77679787eaaSJeff Mahoney } 77779787eaaSJeff Mahoney 778ec39f769SChris Mason /* 779ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 780ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 781ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 782ec39f769SChris Mason */ 783ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 78408635baeSChris Mason ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 78506ea01b1SDavid Sterba if (ret) { 78606ea01b1SDavid Sterba async->bio->bi_status = ret; 78706ea01b1SDavid Sterba bio_endio(async->bio); 78806ea01b1SDavid Sterba } 7894a69a410SChris Mason } 7904a69a410SChris Mason 7914a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7924a69a410SChris Mason { 7934a69a410SChris Mason struct async_submit_bio *async; 7944a69a410SChris Mason 7954a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7968b712842SChris Mason kfree(async); 7978b712842SChris Mason } 7988b712842SChris Mason 7998c27cb35SLinus Torvalds blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 800c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 801c6100a4bSJosef Bacik u64 bio_offset, void *private_data, 802e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 80344b8bd7eSChris Mason { 80444b8bd7eSChris Mason struct async_submit_bio *async; 80544b8bd7eSChris Mason 80644b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 80744b8bd7eSChris Mason if (!async) 8084e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 80944b8bd7eSChris Mason 810c6100a4bSJosef Bacik async->private_data = private_data; 81144b8bd7eSChris Mason async->bio = bio; 81244b8bd7eSChris Mason async->mirror_num = mirror_num; 8134a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8144a69a410SChris Mason 815a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 816a0cac0ecSOmar Sandoval run_one_async_free); 8174a69a410SChris Mason 818eaf25d93SChris Mason async->bio_offset = bio_offset; 8198c8bee1dSChris Mason 8204e4cbee9SChristoph Hellwig async->status = 0; 82179787eaaSJeff Mahoney 82267f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 8235cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 824d313d7a3SChris Mason 8255cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 82644b8bd7eSChris Mason return 0; 82744b8bd7eSChris Mason } 82844b8bd7eSChris Mason 8294e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 830ce3ed71aSChris Mason { 8312c30c71bSKent Overstreet struct bio_vec *bvec; 832ce3ed71aSChris Mason struct btrfs_root *root; 8332b070cfeSChristoph Hellwig int ret = 0; 8346dc4f100SMing Lei struct bvec_iter_all iter_all; 835ce3ed71aSChris Mason 836c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 8372b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 838ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 83901d58472SDaniel Dressler ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); 84079787eaaSJeff Mahoney if (ret) 84179787eaaSJeff Mahoney break; 842ce3ed71aSChris Mason } 8432c30c71bSKent Overstreet 8444e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 845ce3ed71aSChris Mason } 846ce3ed71aSChris Mason 847d0ee3934SDavid Sterba static blk_status_t btree_submit_bio_start(void *private_data, struct bio *bio, 848eaf25d93SChris Mason u64 bio_offset) 84922c59948SChris Mason { 8508b712842SChris Mason /* 8518b712842SChris Mason * when we're called for a write, we're already in the async 8525443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8538b712842SChris Mason */ 85479787eaaSJeff Mahoney return btree_csum_one_bio(bio); 85522c59948SChris Mason } 85622c59948SChris Mason 8579b4e675aSDavid Sterba static int check_async_write(struct btrfs_fs_info *fs_info, 8589b4e675aSDavid Sterba struct btrfs_inode *bi) 859de0022b9SJosef Bacik { 8606300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 8616300463bSLiu Bo return 0; 8629b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 863de0022b9SJosef Bacik return 0; 864de0022b9SJosef Bacik return 1; 865de0022b9SJosef Bacik } 866de0022b9SJosef Bacik 867a56b1c7bSNikolay Borisov static blk_status_t btree_submit_bio_hook(struct inode *inode, struct bio *bio, 86850489a57SNikolay Borisov int mirror_num, 86950489a57SNikolay Borisov unsigned long bio_flags) 87044b8bd7eSChris Mason { 8710b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 8729b4e675aSDavid Sterba int async = check_async_write(fs_info, BTRFS_I(inode)); 8734e4cbee9SChristoph Hellwig blk_status_t ret; 874cad321adSChris Mason 87537226b21SMike Christie if (bio_op(bio) != REQ_OP_WRITE) { 876cad321adSChris Mason /* 877cad321adSChris Mason * called for a read, do the setup so that checksum validation 878cad321adSChris Mason * can happen in the async kernel threads 879cad321adSChris Mason */ 8800b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 8810b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 8821d4284bdSChris Mason if (ret) 88361891923SStefan Behrens goto out_w_error; 88408635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 885de0022b9SJosef Bacik } else if (!async) { 886de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 887de0022b9SJosef Bacik if (ret) 88861891923SStefan Behrens goto out_w_error; 88908635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 89061891923SStefan Behrens } else { 891cad321adSChris Mason /* 89261891923SStefan Behrens * kthread helpers are used to submit writes so that 89361891923SStefan Behrens * checksumming can happen in parallel across all CPUs 894cad321adSChris Mason */ 895c6100a4bSJosef Bacik ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, 896e68f2ee7SNikolay Borisov 0, inode, btree_submit_bio_start); 89744b8bd7eSChris Mason } 89844b8bd7eSChris Mason 8994246a0b6SChristoph Hellwig if (ret) 9004246a0b6SChristoph Hellwig goto out_w_error; 9014246a0b6SChristoph Hellwig return 0; 9024246a0b6SChristoph Hellwig 90361891923SStefan Behrens out_w_error: 9044e4cbee9SChristoph Hellwig bio->bi_status = ret; 9054246a0b6SChristoph Hellwig bio_endio(bio); 90661891923SStefan Behrens return ret; 90761891923SStefan Behrens } 90861891923SStefan Behrens 9093dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 910784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 911a6bc32b8SMel Gorman struct page *newpage, struct page *page, 912a6bc32b8SMel Gorman enum migrate_mode mode) 913784b4e29SChris Mason { 914784b4e29SChris Mason /* 915784b4e29SChris Mason * we can't safely write a btree page from here, 916784b4e29SChris Mason * we haven't done the locking hook 917784b4e29SChris Mason */ 918784b4e29SChris Mason if (PageDirty(page)) 919784b4e29SChris Mason return -EAGAIN; 920784b4e29SChris Mason /* 921784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 922784b4e29SChris Mason * We must have no buffers or drop them. 923784b4e29SChris Mason */ 924784b4e29SChris Mason if (page_has_private(page) && 925784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 926784b4e29SChris Mason return -EAGAIN; 927a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 928784b4e29SChris Mason } 9293dd1462eSJan Beulich #endif 930784b4e29SChris Mason 9310da5468fSChris Mason 9320da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9330da5468fSChris Mason struct writeback_control *wbc) 9340da5468fSChris Mason { 935e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 936e2d84521SMiao Xie int ret; 937e2d84521SMiao Xie 938d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 939448d640bSChris Mason 940448d640bSChris Mason if (wbc->for_kupdate) 941448d640bSChris Mason return 0; 942448d640bSChris Mason 943e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 944b9473439SChris Mason /* this is a bit racy, but that's ok */ 945d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 946d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 947d814a491SEthan Lien fs_info->dirty_metadata_batch); 948e2d84521SMiao Xie if (ret < 0) 949793955bcSChris Mason return 0; 950793955bcSChris Mason } 9510b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9520da5468fSChris Mason } 9530da5468fSChris Mason 954b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 9555f39d397SChris Mason { 95671ad38b4SDavid Sterba return extent_read_full_page(page, btree_get_extent, 0); 9575f39d397SChris Mason } 9585f39d397SChris Mason 95970dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 9605f39d397SChris Mason { 96198509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 96298509cfcSChris Mason return 0; 9630c4e538bSDavid Sterba 964f7a52a40SDavid Sterba return try_release_extent_buffer(page); 965d98237b3SChris Mason } 966d98237b3SChris Mason 967d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 968d47992f8SLukas Czerner unsigned int length) 969d98237b3SChris Mason { 970d1310b2eSChris Mason struct extent_io_tree *tree; 971d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9725f39d397SChris Mason extent_invalidatepage(tree, page, offset); 9735f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 9749ad6b7bcSChris Mason if (PagePrivate(page)) { 975efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 976efe120a0SFrank Holton "page private not zero on page %llu", 977efe120a0SFrank Holton (unsigned long long)page_offset(page)); 978d1b89bc0SGuoqing Jiang detach_page_private(page); 9799ad6b7bcSChris Mason } 980d98237b3SChris Mason } 981d98237b3SChris Mason 9820b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 9830b32f4bbSJosef Bacik { 984bb146eb2SJosef Bacik #ifdef DEBUG 9850b32f4bbSJosef Bacik struct extent_buffer *eb; 9860b32f4bbSJosef Bacik 9870b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 9880b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 9890b32f4bbSJosef Bacik BUG_ON(!eb); 9900b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 9910b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 9920b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 993bb146eb2SJosef Bacik #endif 9940b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 9950b32f4bbSJosef Bacik } 9960b32f4bbSJosef Bacik 9977f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 998d98237b3SChris Mason .readpage = btree_readpage, 9990da5468fSChris Mason .writepages = btree_writepages, 10005f39d397SChris Mason .releasepage = btree_releasepage, 10015f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10025a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1003784b4e29SChris Mason .migratepage = btree_migratepage, 10045a92bc88SChris Mason #endif 10050b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1006d98237b3SChris Mason }; 1007123abc88SChris Mason 10082ff7e61eSJeff Mahoney void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) 1009090d1875SChris Mason { 10105f39d397SChris Mason struct extent_buffer *buf = NULL; 1011537f38f0SNikolay Borisov int ret; 1012090d1875SChris Mason 10132ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1014c871b0f2SLiu Bo if (IS_ERR(buf)) 10156197d86eSDavid Sterba return; 1016537f38f0SNikolay Borisov 1017c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(buf, WAIT_NONE, 0); 1018537f38f0SNikolay Borisov if (ret < 0) 1019537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 1020537f38f0SNikolay Borisov else 10215f39d397SChris Mason free_extent_buffer(buf); 1022090d1875SChris Mason } 1023090d1875SChris Mason 10242ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 10252ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 10260308af44SDavid Sterba u64 bytenr) 10270999df54SChris Mason { 10280b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 10290b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 10300b246afaSJeff Mahoney return alloc_extent_buffer(fs_info, bytenr); 10310999df54SChris Mason } 10320999df54SChris Mason 1033581c1760SQu Wenruo /* 1034581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 1035581c1760SQu Wenruo * verification. 1036581c1760SQu Wenruo * 1037581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 1038581c1760SQu Wenruo * @level: expected level, mandatory check 1039581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 1040581c1760SQu Wenruo */ 10412ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 1042581c1760SQu Wenruo u64 parent_transid, int level, 1043581c1760SQu Wenruo struct btrfs_key *first_key) 1044e20d96d6SChris Mason { 10455f39d397SChris Mason struct extent_buffer *buf = NULL; 104619c00ddcSChris Mason int ret; 104719c00ddcSChris Mason 10482ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 1049c871b0f2SLiu Bo if (IS_ERR(buf)) 1050c871b0f2SLiu Bo return buf; 1051e4204dedSChris Mason 10525ab12d1fSDavid Sterba ret = btree_read_extent_buffer_pages(buf, parent_transid, 1053581c1760SQu Wenruo level, first_key); 10540f0fe8f7SFilipe David Borba Manana if (ret) { 1055537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 105664c043deSLiu Bo return ERR_PTR(ret); 10570f0fe8f7SFilipe David Borba Manana } 10585f39d397SChris Mason return buf; 1059ce9adaa5SChris Mason 1060eb60ceacSChris Mason } 1061eb60ceacSChris Mason 10626a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 1063ed2ff2cbSChris Mason { 10646a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 106555c69072SChris Mason if (btrfs_header_generation(buf) == 1066e2d84521SMiao Xie fs_info->running_transaction->transid) { 1067b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1068b4ce94deSChris Mason 1069b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1070104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1071e2d84521SMiao Xie -buf->len, 1072e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1073b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 10748bead258SDavid Sterba btrfs_set_lock_blocking_write(buf); 10750b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1076925baeddSChris Mason } 10775f39d397SChris Mason } 1078ed7b63ebSJosef Bacik } 10795f39d397SChris Mason 1080da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1081e20d96d6SChris Mason u64 objectid) 1082d97e63b6SChris Mason { 10837c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 108496dfcb46SJosef Bacik root->fs_info = fs_info; 1085cfaa7295SChris Mason root->node = NULL; 1086a28ec197SChris Mason root->commit_root = NULL; 108727cdeb70SMiao Xie root->state = 0; 1088d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 10890b86a832SChris Mason 10900f7d52f4SChris Mason root->last_trans = 0; 109113a8a7c8SYan, Zheng root->highest_objectid = 0; 1092eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1093199c2a9cSMiao Xie root->nr_ordered_extents = 0; 10946bef4d31SEric Paris root->inode_tree = RB_ROOT; 109516cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1096f0486c68SYan, Zheng root->block_rsv = NULL; 10970b86a832SChris Mason 10980b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 10995d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1100eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1101eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1102199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1103199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1104d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 11052ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 11062ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 11075d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1108eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1109199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1110f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 11112ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 11122ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 11138287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1114a2135011SChris Mason mutex_init(&root->objectid_mutex); 1115e02119d5SChris Mason mutex_init(&root->log_mutex); 111631f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1117573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1118c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 11197237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 11207237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 11217237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 11228b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 11238b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 11247237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 11257237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 11267237f183SYan Zheng atomic_set(&root->log_writers, 0); 11272ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11280700cea7SElena Reshetova refcount_set(&root->refs, 1); 11298ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1130eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 11317237f183SYan Zheng root->log_transid = 0; 1132d1433debSMiao Xie root->log_transid_committed = -1; 1133257c62e1SChris Mason root->last_log_commit = 0; 1134e289f03eSFilipe Manana if (!dummy) { 113543eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 113643eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1137e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1138e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1139e289f03eSFilipe Manana } 1140017e5369SChris Mason 11413768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 11423768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 11436702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 11444d775673SChris Mason root->root_key.objectid = objectid; 11450ee5dc67SAl Viro root->anon_dev = 0; 11468ea05e3aSAlexander Block 11475f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1148370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1149bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1150bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1151bd647ce3SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 1152bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1153bd647ce3SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 1154bd647ce3SJosef Bacik #endif 11553768f368SChris Mason } 11563768f368SChris Mason 115774e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 115896dfcb46SJosef Bacik u64 objectid, gfp_t flags) 11596f07e42eSAl Viro { 116074e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 11616f07e42eSAl Viro if (root) 116296dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 11636f07e42eSAl Viro return root; 11646f07e42eSAl Viro } 11656f07e42eSAl Viro 116606ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 116706ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1168da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 116906ea65a3SJosef Bacik { 117006ea65a3SJosef Bacik struct btrfs_root *root; 117106ea65a3SJosef Bacik 11727c0260eeSJeff Mahoney if (!fs_info) 11737c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 11747c0260eeSJeff Mahoney 117596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 117606ea65a3SJosef Bacik if (!root) 117706ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1178da17066cSJeff Mahoney 1179b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1180faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 118106ea65a3SJosef Bacik 118206ea65a3SJosef Bacik return root; 118306ea65a3SJosef Bacik } 118406ea65a3SJosef Bacik #endif 118506ea65a3SJosef Bacik 118620897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 118720897f5cSArne Jansen u64 objectid) 118820897f5cSArne Jansen { 11899b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 119020897f5cSArne Jansen struct extent_buffer *leaf; 119120897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 119220897f5cSArne Jansen struct btrfs_root *root; 119320897f5cSArne Jansen struct btrfs_key key; 1194b89f6d1fSFilipe Manana unsigned int nofs_flag; 119520897f5cSArne Jansen int ret = 0; 119620897f5cSArne Jansen 1197b89f6d1fSFilipe Manana /* 1198b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1199b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1200b89f6d1fSFilipe Manana */ 1201b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 120296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1203b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 120420897f5cSArne Jansen if (!root) 120520897f5cSArne Jansen return ERR_PTR(-ENOMEM); 120620897f5cSArne Jansen 120720897f5cSArne Jansen root->root_key.objectid = objectid; 120820897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 120920897f5cSArne Jansen root->root_key.offset = 0; 121020897f5cSArne Jansen 12114d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0); 121220897f5cSArne Jansen if (IS_ERR(leaf)) { 121320897f5cSArne Jansen ret = PTR_ERR(leaf); 12141dd05682STsutomu Itoh leaf = NULL; 121520897f5cSArne Jansen goto fail; 121620897f5cSArne Jansen } 121720897f5cSArne Jansen 121820897f5cSArne Jansen root->node = leaf; 121920897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 122020897f5cSArne Jansen 122120897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 122227cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 122320897f5cSArne Jansen 122420897f5cSArne Jansen root->root_item.flags = 0; 122520897f5cSArne Jansen root->root_item.byte_limit = 0; 122620897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 122720897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 122820897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 122920897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 123020897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 123120897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 123220897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 123333d85fdaSQu Wenruo if (is_fstree(objectid)) 1234807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1235807fc790SAndy Shevchenko else 1236807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 123720897f5cSArne Jansen root->root_item.drop_level = 0; 123820897f5cSArne Jansen 123920897f5cSArne Jansen key.objectid = objectid; 124020897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 124120897f5cSArne Jansen key.offset = 0; 124220897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 124320897f5cSArne Jansen if (ret) 124420897f5cSArne Jansen goto fail; 124520897f5cSArne Jansen 124620897f5cSArne Jansen btrfs_tree_unlock(leaf); 124720897f5cSArne Jansen 124820897f5cSArne Jansen return root; 12491dd05682STsutomu Itoh 12501dd05682STsutomu Itoh fail: 12518c38938cSJosef Bacik if (leaf) 12521dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 125300246528SJosef Bacik btrfs_put_root(root); 12541dd05682STsutomu Itoh 12551dd05682STsutomu Itoh return ERR_PTR(ret); 125620897f5cSArne Jansen } 125720897f5cSArne Jansen 12587237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1259e02119d5SChris Mason struct btrfs_fs_info *fs_info) 12600f7d52f4SChris Mason { 12610f7d52f4SChris Mason struct btrfs_root *root; 12627237f183SYan Zheng struct extent_buffer *leaf; 1263e02119d5SChris Mason 126496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1265e02119d5SChris Mason if (!root) 12667237f183SYan Zheng return ERR_PTR(-ENOMEM); 1267e02119d5SChris Mason 1268e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1269e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1270e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 127127cdeb70SMiao Xie 12727237f183SYan Zheng /* 127392a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 127427cdeb70SMiao Xie * 127592a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 127692a7cc42SQu Wenruo * and they go away before a real commit is actually done. 127792a7cc42SQu Wenruo * 127892a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 127992a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 12807237f183SYan Zheng */ 1281e02119d5SChris Mason 12824d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 12834d75f8a9SDavid Sterba NULL, 0, 0, 0); 12847237f183SYan Zheng if (IS_ERR(leaf)) { 128500246528SJosef Bacik btrfs_put_root(root); 12867237f183SYan Zheng return ERR_CAST(leaf); 12877237f183SYan Zheng } 1288e02119d5SChris Mason 12897237f183SYan Zheng root->node = leaf; 1290e02119d5SChris Mason 1291e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1292e02119d5SChris Mason btrfs_tree_unlock(root->node); 12937237f183SYan Zheng return root; 12947237f183SYan Zheng } 12957237f183SYan Zheng 12967237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 12977237f183SYan Zheng struct btrfs_fs_info *fs_info) 12987237f183SYan Zheng { 12997237f183SYan Zheng struct btrfs_root *log_root; 13007237f183SYan Zheng 13017237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13027237f183SYan Zheng if (IS_ERR(log_root)) 13037237f183SYan Zheng return PTR_ERR(log_root); 13047237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13057237f183SYan Zheng fs_info->log_root_tree = log_root; 13067237f183SYan Zheng return 0; 13077237f183SYan Zheng } 13087237f183SYan Zheng 13097237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13107237f183SYan Zheng struct btrfs_root *root) 13117237f183SYan Zheng { 13120b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13137237f183SYan Zheng struct btrfs_root *log_root; 13147237f183SYan Zheng struct btrfs_inode_item *inode_item; 13157237f183SYan Zheng 13160b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13177237f183SYan Zheng if (IS_ERR(log_root)) 13187237f183SYan Zheng return PTR_ERR(log_root); 13197237f183SYan Zheng 13207237f183SYan Zheng log_root->last_trans = trans->transid; 13217237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 13227237f183SYan Zheng 13237237f183SYan Zheng inode_item = &log_root->root_item.inode; 13243cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 13253cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 13263cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1327da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 13280b246afaSJeff Mahoney fs_info->nodesize); 13293cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 13307237f183SYan Zheng 13315d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 13327237f183SYan Zheng 13337237f183SYan Zheng WARN_ON(root->log_root); 13347237f183SYan Zheng root->log_root = log_root; 13357237f183SYan Zheng root->log_transid = 0; 1336d1433debSMiao Xie root->log_transid_committed = -1; 1337257c62e1SChris Mason root->last_log_commit = 0; 1338e02119d5SChris Mason return 0; 1339e02119d5SChris Mason } 1340e02119d5SChris Mason 134162a2c73eSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1342cb517eabSMiao Xie struct btrfs_key *key) 1343e02119d5SChris Mason { 1344e02119d5SChris Mason struct btrfs_root *root; 1345e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 13460f7d52f4SChris Mason struct btrfs_path *path; 134784234f3aSYan Zheng u64 generation; 1348cb517eabSMiao Xie int ret; 1349581c1760SQu Wenruo int level; 1350cb517eabSMiao Xie 1351cb517eabSMiao Xie path = btrfs_alloc_path(); 1352cb517eabSMiao Xie if (!path) 1353cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 13540f7d52f4SChris Mason 135596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 1356cb517eabSMiao Xie if (!root) { 1357cb517eabSMiao Xie ret = -ENOMEM; 1358cb517eabSMiao Xie goto alloc_fail; 13590f7d52f4SChris Mason } 13600f7d52f4SChris Mason 1361cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1362cb517eabSMiao Xie &root->root_item, &root->root_key); 13630f7d52f4SChris Mason if (ret) { 136413a8a7c8SYan, Zheng if (ret > 0) 136513a8a7c8SYan, Zheng ret = -ENOENT; 1366cb517eabSMiao Xie goto find_fail; 13670f7d52f4SChris Mason } 136813a8a7c8SYan, Zheng 136984234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1370581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 13712ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 13722ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 1373581c1760SQu Wenruo generation, level, NULL); 137464c043deSLiu Bo if (IS_ERR(root->node)) { 137564c043deSLiu Bo ret = PTR_ERR(root->node); 13768c38938cSJosef Bacik root->node = NULL; 1377cb517eabSMiao Xie goto find_fail; 1378cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1379cb517eabSMiao Xie ret = -EIO; 138064c043deSLiu Bo goto find_fail; 1381416bc658SJosef Bacik } 13825d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 138313a8a7c8SYan, Zheng out: 1384cb517eabSMiao Xie btrfs_free_path(path); 1385cb517eabSMiao Xie return root; 1386cb517eabSMiao Xie 1387cb517eabSMiao Xie find_fail: 138800246528SJosef Bacik btrfs_put_root(root); 1389cb517eabSMiao Xie alloc_fail: 1390cb517eabSMiao Xie root = ERR_PTR(ret); 1391cb517eabSMiao Xie goto out; 1392cb517eabSMiao Xie } 1393cb517eabSMiao Xie 13942dfb1e43SQu Wenruo /* 13952dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 13962dfb1e43SQu Wenruo * 13972dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 13982dfb1e43SQu Wenruo */ 13992dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1400cb517eabSMiao Xie { 1401cb517eabSMiao Xie int ret; 1402dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1403cb517eabSMiao Xie 1404cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1405cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1406cb517eabSMiao Xie GFP_NOFS); 1407cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1408cb517eabSMiao Xie ret = -ENOMEM; 1409cb517eabSMiao Xie goto fail; 1410cb517eabSMiao Xie } 1411cb517eabSMiao Xie 1412dcc3eb96SNikolay Borisov /* 1413dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1414dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1415dcc3eb96SNikolay Borisov */ 1416dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1417dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1418dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1419dcc3eb96SNikolay Borisov if (ret) 14208257b2dcSMiao Xie goto fail; 14218257b2dcSMiao Xie 1422aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 1423aeb935a4SQu Wenruo root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) { 142492a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1425f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1426f39e4571SJosef Bacik } 1427f39e4571SJosef Bacik 1428cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 142957cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 143057cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1431cb517eabSMiao Xie 1432851fd730SQu Wenruo /* 1433851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1434851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1435851fd730SQu Wenruo */ 1436851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1437851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 14382dfb1e43SQu Wenruo if (!anon_dev) { 1439cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1440cb517eabSMiao Xie if (ret) 1441876d2cf1SLiu Bo goto fail; 14422dfb1e43SQu Wenruo } else { 14432dfb1e43SQu Wenruo root->anon_dev = anon_dev; 14442dfb1e43SQu Wenruo } 1445851fd730SQu Wenruo } 1446f32e48e9SChandan Rajendra 1447f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1448f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1449f32e48e9SChandan Rajendra &root->highest_objectid); 1450f32e48e9SChandan Rajendra if (ret) { 1451f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1452876d2cf1SLiu Bo goto fail; 1453f32e48e9SChandan Rajendra } 1454f32e48e9SChandan Rajendra 1455f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1456f32e48e9SChandan Rajendra 1457f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1458f32e48e9SChandan Rajendra 1459cb517eabSMiao Xie return 0; 1460cb517eabSMiao Xie fail: 146184db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1462cb517eabSMiao Xie return ret; 1463cb517eabSMiao Xie } 1464cb517eabSMiao Xie 1465a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1466cb517eabSMiao Xie u64 root_id) 1467cb517eabSMiao Xie { 1468cb517eabSMiao Xie struct btrfs_root *root; 1469cb517eabSMiao Xie 1470cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1471cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1472cb517eabSMiao Xie (unsigned long)root_id); 1473bc44d7c4SJosef Bacik if (root) 147400246528SJosef Bacik root = btrfs_grab_root(root); 1475cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1476cb517eabSMiao Xie return root; 1477cb517eabSMiao Xie } 1478cb517eabSMiao Xie 1479cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1480cb517eabSMiao Xie struct btrfs_root *root) 1481cb517eabSMiao Xie { 1482cb517eabSMiao Xie int ret; 1483cb517eabSMiao Xie 1484e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1485cb517eabSMiao Xie if (ret) 1486cb517eabSMiao Xie return ret; 1487cb517eabSMiao Xie 1488cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1489cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1490cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1491cb517eabSMiao Xie root); 1492af01d2e5SJosef Bacik if (ret == 0) { 149300246528SJosef Bacik btrfs_grab_root(root); 149427cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1495af01d2e5SJosef Bacik } 1496cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1497cb517eabSMiao Xie radix_tree_preload_end(); 1498cb517eabSMiao Xie 1499cb517eabSMiao Xie return ret; 1500cb517eabSMiao Xie } 1501cb517eabSMiao Xie 1502bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1503bd647ce3SJosef Bacik { 1504bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1505bd647ce3SJosef Bacik struct btrfs_root *root; 1506bd647ce3SJosef Bacik 1507bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1508bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1509bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1510bd647ce3SJosef Bacik btrfs_err(fs_info, "leaked root %llu-%llu refcount %d", 1511bd647ce3SJosef Bacik root->root_key.objectid, root->root_key.offset, 1512bd647ce3SJosef Bacik refcount_read(&root->refs)); 1513bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 151400246528SJosef Bacik btrfs_put_root(root); 151500246528SJosef Bacik btrfs_put_root(root); 1516bd647ce3SJosef Bacik } 1517bd647ce3SJosef Bacik #endif 1518bd647ce3SJosef Bacik } 1519bd647ce3SJosef Bacik 15200d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 15210d4b0463SJosef Bacik { 1522141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1523141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 1524141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dio_bytes); 1525141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1526141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1527141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1528141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 15290d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 15300d4b0463SJosef Bacik kfree(fs_info->delayed_root); 153100246528SJosef Bacik btrfs_put_root(fs_info->extent_root); 153200246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 153300246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 153400246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 153500246528SJosef Bacik btrfs_put_root(fs_info->csum_root); 153600246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 153700246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 153800246528SJosef Bacik btrfs_put_root(fs_info->free_space_root); 153900246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1540aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 1541bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 15423fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 15430d4b0463SJosef Bacik kfree(fs_info->super_copy); 15440d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 15450d4b0463SJosef Bacik kvfree(fs_info); 15460d4b0463SJosef Bacik } 15470d4b0463SJosef Bacik 15480d4b0463SJosef Bacik 15492dfb1e43SQu Wenruo /* 15502dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 15512dfb1e43SQu Wenruo * 15522dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 15532dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 15542dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 15552dfb1e43SQu Wenruo * 15562dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 15572dfb1e43SQu Wenruo * 15582dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 15592dfb1e43SQu Wenruo * same root objectid. 15602dfb1e43SQu Wenruo * 15612dfb1e43SQu Wenruo * @objectid: root id 15622dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 15632dfb1e43SQu Wenruo * pass 0 for new allocation. 15642dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 15652dfb1e43SQu Wenruo * for orphan roots 15662dfb1e43SQu Wenruo */ 15672dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 15682dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 15692dfb1e43SQu Wenruo bool check_ref) 15705eda7b5eSChris Mason { 15715eda7b5eSChris Mason struct btrfs_root *root; 1572381cf658SDavid Sterba struct btrfs_path *path; 15731d4c08e0SDavid Sterba struct btrfs_key key; 15745eda7b5eSChris Mason int ret; 15755eda7b5eSChris Mason 157656e9357aSDavid Sterba if (objectid == BTRFS_ROOT_TREE_OBJECTID) 157700246528SJosef Bacik return btrfs_grab_root(fs_info->tree_root); 157856e9357aSDavid Sterba if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 157900246528SJosef Bacik return btrfs_grab_root(fs_info->extent_root); 158056e9357aSDavid Sterba if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 158100246528SJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 158256e9357aSDavid Sterba if (objectid == BTRFS_DEV_TREE_OBJECTID) 158300246528SJosef Bacik return btrfs_grab_root(fs_info->dev_root); 158456e9357aSDavid Sterba if (objectid == BTRFS_CSUM_TREE_OBJECTID) 158500246528SJosef Bacik return btrfs_grab_root(fs_info->csum_root); 158656e9357aSDavid Sterba if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 158700246528SJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 1588bc44d7c4SJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 158956e9357aSDavid Sterba if (objectid == BTRFS_UUID_TREE_OBJECTID) 159000246528SJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 1591bc44d7c4SJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 159256e9357aSDavid Sterba if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 159300246528SJosef Bacik return btrfs_grab_root(fs_info->free_space_root) ? 1594bc44d7c4SJosef Bacik fs_info->free_space_root : ERR_PTR(-ENOENT); 15954df27c4dSYan, Zheng again: 159656e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 159748475471SStefan Behrens if (root) { 15982dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 15992dfb1e43SQu Wenruo ASSERT(!anon_dev); 1600bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 160100246528SJosef Bacik btrfs_put_root(root); 160248475471SStefan Behrens return ERR_PTR(-ENOENT); 1603bc44d7c4SJosef Bacik } 16045eda7b5eSChris Mason return root; 160548475471SStefan Behrens } 16065eda7b5eSChris Mason 160756e9357aSDavid Sterba key.objectid = objectid; 160856e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 160956e9357aSDavid Sterba key.offset = (u64)-1; 161056e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 16115eda7b5eSChris Mason if (IS_ERR(root)) 16125eda7b5eSChris Mason return root; 16133394e160SChris Mason 1614c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1615d68fc57bSYan, Zheng ret = -ENOENT; 1616d68fc57bSYan, Zheng goto fail; 1617d68fc57bSYan, Zheng } 1618d68fc57bSYan, Zheng 16192dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1620cb517eabSMiao Xie if (ret) 1621cb517eabSMiao Xie goto fail; 1622cb517eabSMiao Xie 1623381cf658SDavid Sterba path = btrfs_alloc_path(); 1624381cf658SDavid Sterba if (!path) { 1625381cf658SDavid Sterba ret = -ENOMEM; 1626381cf658SDavid Sterba goto fail; 1627381cf658SDavid Sterba } 16281d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 16291d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 163056e9357aSDavid Sterba key.offset = objectid; 16311d4c08e0SDavid Sterba 16321d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1633381cf658SDavid Sterba btrfs_free_path(path); 1634d68fc57bSYan, Zheng if (ret < 0) 1635d68fc57bSYan, Zheng goto fail; 1636d68fc57bSYan, Zheng if (ret == 0) 163727cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1638d68fc57bSYan, Zheng 1639cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 16400f7d52f4SChris Mason if (ret) { 164100246528SJosef Bacik btrfs_put_root(root); 16424785e24fSJosef Bacik if (ret == -EEXIST) 16434df27c4dSYan, Zheng goto again; 16444df27c4dSYan, Zheng goto fail; 16454df27c4dSYan, Zheng } 1646edbd8d4eSChris Mason return root; 16474df27c4dSYan, Zheng fail: 16488c38938cSJosef Bacik btrfs_put_root(root); 16494df27c4dSYan, Zheng return ERR_PTR(ret); 1650edbd8d4eSChris Mason } 1651edbd8d4eSChris Mason 16522dfb1e43SQu Wenruo /* 16532dfb1e43SQu Wenruo * Get in-memory reference of a root structure 16542dfb1e43SQu Wenruo * 16552dfb1e43SQu Wenruo * @objectid: tree objectid 16562dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 16572dfb1e43SQu Wenruo * one reference 16582dfb1e43SQu Wenruo */ 16592dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 16602dfb1e43SQu Wenruo u64 objectid, bool check_ref) 16612dfb1e43SQu Wenruo { 16622dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 16632dfb1e43SQu Wenruo } 16642dfb1e43SQu Wenruo 16652dfb1e43SQu Wenruo /* 16662dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 16672dfb1e43SQu Wenruo * the anonymous block device id 16682dfb1e43SQu Wenruo * 16692dfb1e43SQu Wenruo * @objectid: tree objectid 16702dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 16712dfb1e43SQu Wenruo * parameter value 16722dfb1e43SQu Wenruo */ 16732dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 16742dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 16752dfb1e43SQu Wenruo { 16762dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 16772dfb1e43SQu Wenruo } 16782dfb1e43SQu Wenruo 167904160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 168004160088SChris Mason { 168104160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 168204160088SChris Mason int ret = 0; 168304160088SChris Mason struct btrfs_device *device; 168404160088SChris Mason struct backing_dev_info *bdi; 1685b7967db7SChris Mason 16861f78160cSXiao Guangrong rcu_read_lock(); 16871f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1688dfe25020SChris Mason if (!device->bdev) 1689dfe25020SChris Mason continue; 1690efa7c9f9SJan Kara bdi = device->bdev->bd_bdi; 1691ff9ea323STejun Heo if (bdi_congested(bdi, bdi_bits)) { 169204160088SChris Mason ret = 1; 169304160088SChris Mason break; 169404160088SChris Mason } 169504160088SChris Mason } 16961f78160cSXiao Guangrong rcu_read_unlock(); 169704160088SChris Mason return ret; 169804160088SChris Mason } 169904160088SChris Mason 17008b712842SChris Mason /* 17018b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 17028b712842SChris Mason * functions. This is where read checksum verification actually happens 17038b712842SChris Mason */ 17048b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1705ce9adaa5SChris Mason { 1706ce9adaa5SChris Mason struct bio *bio; 170797eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1708ce9adaa5SChris Mason 170997eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1710ce9adaa5SChris Mason bio = end_io_wq->bio; 17118b712842SChris Mason 17124e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1713ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1714ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 17154246a0b6SChristoph Hellwig bio_endio(bio); 17169be490f1SOmar Sandoval kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 1717ce9adaa5SChris Mason } 171844b8bd7eSChris Mason 1719a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1720a74a4b97SChris Mason { 1721a74a4b97SChris Mason struct btrfs_root *root = arg; 17220b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1723d0278245SMiao Xie int again; 1724a74a4b97SChris Mason 1725d6fd0ae2SOmar Sandoval while (1) { 1726d0278245SMiao Xie again = 0; 17279d1a2a3aSDavid Sterba 1728fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1729fd340d0fSJosef Bacik 1730d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 17312ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1732d0278245SMiao Xie goto sleep; 1733d0278245SMiao Xie 173490c711abSZygo Blaxell /* 173590c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 173690c711abSZygo Blaxell * before we have finished mounting the filesystem. 173790c711abSZygo Blaxell */ 17380b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 173990c711abSZygo Blaxell goto sleep; 174090c711abSZygo Blaxell 17410b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1742d0278245SMiao Xie goto sleep; 1743d0278245SMiao Xie 1744dc7f370cSMiao Xie /* 1745dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1746dc7f370cSMiao Xie * during the above check and trylock. 1747dc7f370cSMiao Xie */ 17482ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 17490b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1750dc7f370cSMiao Xie goto sleep; 1751dc7f370cSMiao Xie } 1752dc7f370cSMiao Xie 17532ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1754c2d6cb16SFilipe Manana 17559d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 17560b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1757a74a4b97SChris Mason 1758d0278245SMiao Xie /* 175905323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 176005323cd1SMiao Xie * needn't do anything special here. 1761d0278245SMiao Xie */ 17620b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 176367c5e7d4SFilipe Manana 176467c5e7d4SFilipe Manana /* 176567c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 176667c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 176767c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 176867c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 176967c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 177067c5e7d4SFilipe Manana * unused block groups. 177167c5e7d4SFilipe Manana */ 17720b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 1773d0278245SMiao Xie sleep: 1774fd340d0fSJosef Bacik clear_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1775d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1776d6fd0ae2SOmar Sandoval kthread_parkme(); 1777d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1778d6fd0ae2SOmar Sandoval return 0; 1779838fe188SJiri Kosina if (!again) { 1780a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1781a74a4b97SChris Mason schedule(); 1782a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1783a74a4b97SChris Mason } 1784da288d28SFilipe Manana } 1785a74a4b97SChris Mason } 1786a74a4b97SChris Mason 1787a74a4b97SChris Mason static int transaction_kthread(void *arg) 1788a74a4b97SChris Mason { 1789a74a4b97SChris Mason struct btrfs_root *root = arg; 17900b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1791a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1792a74a4b97SChris Mason struct btrfs_transaction *cur; 17938929ecfaSYan, Zheng u64 transid; 1794a944442cSAllen Pais time64_t now; 1795a74a4b97SChris Mason unsigned long delay; 1796914b2007SJan Kara bool cannot_commit; 1797a74a4b97SChris Mason 1798a74a4b97SChris Mason do { 1799914b2007SJan Kara cannot_commit = false; 18000b246afaSJeff Mahoney delay = HZ * fs_info->commit_interval; 18010b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1802a74a4b97SChris Mason 18030b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 18040b246afaSJeff Mahoney cur = fs_info->running_transaction; 1805a74a4b97SChris Mason if (!cur) { 18060b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1807a74a4b97SChris Mason goto sleep; 1808a74a4b97SChris Mason } 180931153d81SYan Zheng 1810afd48513SArnd Bergmann now = ktime_get_seconds(); 18113296bf56SQu Wenruo if (cur->state < TRANS_STATE_COMMIT_START && 18128b87dc17SDavid Sterba (now < cur->start_time || 18130b246afaSJeff Mahoney now - cur->start_time < fs_info->commit_interval)) { 18140b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1815a74a4b97SChris Mason delay = HZ * 5; 1816a74a4b97SChris Mason goto sleep; 1817a74a4b97SChris Mason } 18188929ecfaSYan, Zheng transid = cur->transid; 18190b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 182056bec294SChris Mason 182179787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1822354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1823914b2007SJan Kara if (IS_ERR(trans)) { 1824354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1825914b2007SJan Kara cannot_commit = true; 182679787eaaSJeff Mahoney goto sleep; 1827914b2007SJan Kara } 18288929ecfaSYan, Zheng if (transid == trans->transid) { 18293a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 18308929ecfaSYan, Zheng } else { 18313a45bb20SJeff Mahoney btrfs_end_transaction(trans); 18328929ecfaSYan, Zheng } 1833a74a4b97SChris Mason sleep: 18340b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 18350b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1836a74a4b97SChris Mason 18374e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 18380b246afaSJeff Mahoney &fs_info->fs_state))) 18392ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 18408929ecfaSYan, Zheng if (!kthread_should_stop() && 18410b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1842914b2007SJan Kara cannot_commit)) 1843bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1844a74a4b97SChris Mason } while (!kthread_should_stop()); 1845a74a4b97SChris Mason return 0; 1846a74a4b97SChris Mason } 1847a74a4b97SChris Mason 1848af31f5e5SChris Mason /* 184901f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 185001f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 185101f0f9daSNikolay Borisov * anything. 1852af31f5e5SChris Mason * 1853af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1854af31f5e5SChris Mason * generation of the latest root in the array with the generation 1855af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1856af31f5e5SChris Mason */ 185701f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1858af31f5e5SChris Mason { 185901f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1860af31f5e5SChris Mason u64 cur; 1861af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1862af31f5e5SChris Mason int i; 1863af31f5e5SChris Mason 1864af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1865af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1866af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1867af31f5e5SChris Mason if (cur == newest_gen) 186801f0f9daSNikolay Borisov return i; 1869af31f5e5SChris Mason } 1870af31f5e5SChris Mason 187101f0f9daSNikolay Borisov return -EINVAL; 1872af31f5e5SChris Mason } 1873af31f5e5SChris Mason 1874af31f5e5SChris Mason /* 1875af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1876af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1877af31f5e5SChris Mason * done 1878af31f5e5SChris Mason */ 1879af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1880af31f5e5SChris Mason { 18816ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1882af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1883af31f5e5SChris Mason 1884af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1885af31f5e5SChris Mason 1886af31f5e5SChris Mason /* 1887af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1888af31f5e5SChris Mason * regardless of which ones we use today 1889af31f5e5SChris Mason */ 1890af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1891af31f5e5SChris Mason 1892af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1893af31f5e5SChris Mason 1894af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1895af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1896af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1897af31f5e5SChris Mason 1898af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1899af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1900af31f5e5SChris Mason 1901af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1902af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1903af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1904af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1905af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1906af31f5e5SChris Mason 1907af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1908af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1909af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1910af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1911af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1912af31f5e5SChris Mason 19137c7e82a7SChris Mason /* 19147c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 19157c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 19167c7e82a7SChris Mason */ 19177c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 19187c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 19197c7e82a7SChris Mason info->fs_root->node->start); 1920af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1921af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1922af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1923af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19247c7e82a7SChris Mason } 1925af31f5e5SChris Mason 1926af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1927af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1928af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1929af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1930af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1931af31f5e5SChris Mason 1932af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1933af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1934af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1935af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1936af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1937af31f5e5SChris Mason 1938af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1939af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1940af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1941af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1942af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1943af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1944af31f5e5SChris Mason 1945af31f5e5SChris Mason /* 1946af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1947af31f5e5SChris Mason * for the next commit 1948af31f5e5SChris Mason */ 1949af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1950af31f5e5SChris Mason &info->super_for_commit->super_roots, 1951af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1952af31f5e5SChris Mason } 1953af31f5e5SChris Mason 1954af31f5e5SChris Mason /* 1955bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1956bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1957bd2336b2SNikolay Borisov * 1958bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1959bd2336b2SNikolay Borisov * priority - priority of backup root required 1960bd2336b2SNikolay Borisov * 1961bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1962bd2336b2SNikolay Borisov */ 1963bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1964bd2336b2SNikolay Borisov { 1965bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1966bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1967bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1968bd2336b2SNikolay Borisov 1969bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1970bd2336b2SNikolay Borisov if (priority == 0) 1971bd2336b2SNikolay Borisov return backup_index; 1972bd2336b2SNikolay Borisov 1973bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1974bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1975bd2336b2SNikolay Borisov } else { 1976bd2336b2SNikolay Borisov return -EINVAL; 1977bd2336b2SNikolay Borisov } 1978bd2336b2SNikolay Borisov 1979bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1980bd2336b2SNikolay Borisov 1981bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1982bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1983bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1984bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1985bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1986bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1987bd2336b2SNikolay Borisov 1988bd2336b2SNikolay Borisov /* 1989bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1990bd2336b2SNikolay Borisov * need a fsck 1991bd2336b2SNikolay Borisov */ 1992bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1993bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1994bd2336b2SNikolay Borisov 1995bd2336b2SNikolay Borisov return backup_index; 1996bd2336b2SNikolay Borisov } 1997bd2336b2SNikolay Borisov 19987abadb64SLiu Bo /* helper to cleanup workers */ 19997abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20007abadb64SLiu Bo { 2001dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2002afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 20035cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2004fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2005fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 2006d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2007fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2008fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 20095b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2010e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2011736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2012a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2013fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2014b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 2015b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 2016a9b9477dSFilipe Manana /* 2017a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2018a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2019a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2020a9b9477dSFilipe Manana */ 2021a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2022a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 20237abadb64SLiu Bo } 20247abadb64SLiu Bo 20252e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20262e9f5954SRashika { 20272e9f5954SRashika if (root) { 20282e9f5954SRashika free_extent_buffer(root->node); 20292e9f5954SRashika free_extent_buffer(root->commit_root); 20302e9f5954SRashika root->node = NULL; 20312e9f5954SRashika root->commit_root = NULL; 20322e9f5954SRashika } 20332e9f5954SRashika } 20342e9f5954SRashika 2035af31f5e5SChris Mason /* helper to cleanup tree roots */ 20364273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2037af31f5e5SChris Mason { 20382e9f5954SRashika free_root_extent_buffers(info->tree_root); 2039655b09feSJosef Bacik 20402e9f5954SRashika free_root_extent_buffers(info->dev_root); 20412e9f5954SRashika free_root_extent_buffers(info->extent_root); 20422e9f5954SRashika free_root_extent_buffers(info->csum_root); 20432e9f5954SRashika free_root_extent_buffers(info->quota_root); 20442e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20458c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2046aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 20474273eaffSAnand Jain if (free_chunk_root) 20482e9f5954SRashika free_root_extent_buffers(info->chunk_root); 204970f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2050af31f5e5SChris Mason } 2051af31f5e5SChris Mason 20528c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 20538c38938cSJosef Bacik { 20548c38938cSJosef Bacik if (!root) 20558c38938cSJosef Bacik return; 20568c38938cSJosef Bacik 20578c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 20588c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 20591dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 20608c38938cSJosef Bacik if (root->anon_dev) 20618c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 20628c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2063923eb523SJohannes Thumshirn free_root_extent_buffers(root); 20648c38938cSJosef Bacik kfree(root->free_ino_ctl); 20658c38938cSJosef Bacik kfree(root->free_ino_pinned); 20668c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 20678c38938cSJosef Bacik spin_lock(&root->fs_info->fs_roots_radix_lock); 20688c38938cSJosef Bacik list_del_init(&root->leak_list); 20698c38938cSJosef Bacik spin_unlock(&root->fs_info->fs_roots_radix_lock); 20708c38938cSJosef Bacik #endif 20718c38938cSJosef Bacik kfree(root); 20728c38938cSJosef Bacik } 20738c38938cSJosef Bacik } 20748c38938cSJosef Bacik 2075faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2076171f6537SJosef Bacik { 2077171f6537SJosef Bacik int ret; 2078171f6537SJosef Bacik struct btrfs_root *gang[8]; 2079171f6537SJosef Bacik int i; 2080171f6537SJosef Bacik 2081171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2082171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2083171f6537SJosef Bacik struct btrfs_root, root_list); 2084171f6537SJosef Bacik list_del(&gang[0]->root_list); 2085171f6537SJosef Bacik 20868c38938cSJosef Bacik if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2087cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 208800246528SJosef Bacik btrfs_put_root(gang[0]); 2089171f6537SJosef Bacik } 2090171f6537SJosef Bacik 2091171f6537SJosef Bacik while (1) { 2092171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2093171f6537SJosef Bacik (void **)gang, 0, 2094171f6537SJosef Bacik ARRAY_SIZE(gang)); 2095171f6537SJosef Bacik if (!ret) 2096171f6537SJosef Bacik break; 2097171f6537SJosef Bacik for (i = 0; i < ret; i++) 2098cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2099171f6537SJosef Bacik } 2100171f6537SJosef Bacik } 2101af31f5e5SChris Mason 2102638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2103638aa7edSEric Sandeen { 2104638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2105638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2106638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2107638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2108638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2109638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2110ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2111638aa7edSEric Sandeen } 2112638aa7edSEric Sandeen 2113779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2114779a65a4SEric Sandeen { 2115779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2116779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2117779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2118779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2119779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2120779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2121779a65a4SEric Sandeen } 2122779a65a4SEric Sandeen 21236bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2124f37938e0SEric Sandeen { 21252ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 21262ff7e61eSJeff Mahoney 21272ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 21282ff7e61eSJeff Mahoney set_nlink(inode, 1); 2129f37938e0SEric Sandeen /* 2130f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2131f37938e0SEric Sandeen * the real end of the address space is determined by all of 2132f37938e0SEric Sandeen * the devices in the system 2133f37938e0SEric Sandeen */ 21342ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 21352ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2136f37938e0SEric Sandeen 21372ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 213843eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 213943eb5f29SQu Wenruo IO_TREE_INODE_IO, inode); 21407b439738SDavid Sterba BTRFS_I(inode)->io_tree.track_uptodate = false; 21412ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2142f37938e0SEric Sandeen 21432ff7e61eSJeff Mahoney BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops; 2144f37938e0SEric Sandeen 21455c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 21462ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 21472ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 21482ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2149f37938e0SEric Sandeen } 2150f37938e0SEric Sandeen 2151ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2152ad618368SEric Sandeen { 2153ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2154129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 21557f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2156ad618368SEric Sandeen } 2157ad618368SEric Sandeen 2158f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2159f9e92e40SEric Sandeen { 2160f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2161f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2162f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2163f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2164f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2165f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2166d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2167f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2168f9e92e40SEric Sandeen } 2169f9e92e40SEric Sandeen 21702a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 21712a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 21722a458198SEric Sandeen { 2173f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 21746f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 21752a458198SEric Sandeen 21762a458198SEric Sandeen fs_info->workers = 2177cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2178cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 21792a458198SEric Sandeen 21802a458198SEric Sandeen fs_info->delalloc_workers = 2181cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2182cb001095SJeff Mahoney flags, max_active, 2); 21832a458198SEric Sandeen 21842a458198SEric Sandeen fs_info->flush_workers = 2185cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2186cb001095SJeff Mahoney flags, max_active, 0); 21872a458198SEric Sandeen 21882a458198SEric Sandeen fs_info->caching_workers = 2189cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 21902a458198SEric Sandeen 21912a458198SEric Sandeen fs_info->fixup_workers = 2192cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 21932a458198SEric Sandeen 21942a458198SEric Sandeen /* 21952a458198SEric Sandeen * endios are largely parallel and should have a very 21962a458198SEric Sandeen * low idle thresh 21972a458198SEric Sandeen */ 21982a458198SEric Sandeen fs_info->endio_workers = 2199cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 22002a458198SEric Sandeen fs_info->endio_meta_workers = 2201cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2202cb001095SJeff Mahoney max_active, 4); 22032a458198SEric Sandeen fs_info->endio_meta_write_workers = 2204cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2205cb001095SJeff Mahoney max_active, 2); 22062a458198SEric Sandeen fs_info->endio_raid56_workers = 2207cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2208cb001095SJeff Mahoney max_active, 4); 22092a458198SEric Sandeen fs_info->rmw_workers = 2210cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 22112a458198SEric Sandeen fs_info->endio_write_workers = 2212cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2213cb001095SJeff Mahoney max_active, 2); 22142a458198SEric Sandeen fs_info->endio_freespace_worker = 2215cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2216cb001095SJeff Mahoney max_active, 0); 22172a458198SEric Sandeen fs_info->delayed_workers = 2218cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2219cb001095SJeff Mahoney max_active, 0); 22202a458198SEric Sandeen fs_info->readahead_workers = 2221cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2222cb001095SJeff Mahoney max_active, 2); 22232a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2224cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2225b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2226b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 22272a458198SEric Sandeen 22282a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 2229ba8a9d07SChris Mason fs_info->flush_workers && 22302a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 22312a458198SEric Sandeen fs_info->endio_meta_write_workers && 22322a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 22332a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 22342a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 22352a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 2236b0643e59SDennis Zhou fs_info->qgroup_rescan_workers && 2237b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 22382a458198SEric Sandeen return -ENOMEM; 22392a458198SEric Sandeen } 22402a458198SEric Sandeen 22412a458198SEric Sandeen return 0; 22422a458198SEric Sandeen } 22432a458198SEric Sandeen 22446d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 22456d97c6e3SJohannes Thumshirn { 22466d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2247b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 22486d97c6e3SJohannes Thumshirn 2249b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 22506d97c6e3SJohannes Thumshirn 22516d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 22526d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2253b4e967beSDavid Sterba csum_driver); 22546d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 22556d97c6e3SJohannes Thumshirn } 22566d97c6e3SJohannes Thumshirn 22576d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 22586d97c6e3SJohannes Thumshirn 22596d97c6e3SJohannes Thumshirn return 0; 22606d97c6e3SJohannes Thumshirn } 22616d97c6e3SJohannes Thumshirn 226263443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 226363443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 226463443bf5SEric Sandeen { 226563443bf5SEric Sandeen int ret; 226663443bf5SEric Sandeen struct btrfs_root *log_tree_root; 226763443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 226863443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2269581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 227063443bf5SEric Sandeen 227163443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2272f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 227363443bf5SEric Sandeen return -EIO; 227463443bf5SEric Sandeen } 227563443bf5SEric Sandeen 227696dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 227796dfcb46SJosef Bacik GFP_KERNEL); 227863443bf5SEric Sandeen if (!log_tree_root) 227963443bf5SEric Sandeen return -ENOMEM; 228063443bf5SEric Sandeen 22812ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 2282581c1760SQu Wenruo fs_info->generation + 1, 2283581c1760SQu Wenruo level, NULL); 228464c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2285f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 22860eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 22878c38938cSJosef Bacik log_tree_root->node = NULL; 228800246528SJosef Bacik btrfs_put_root(log_tree_root); 22890eeff236SLiu Bo return ret; 229064c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2291f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 229200246528SJosef Bacik btrfs_put_root(log_tree_root); 229363443bf5SEric Sandeen return -EIO; 229463443bf5SEric Sandeen } 229563443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 229663443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 229763443bf5SEric Sandeen if (ret) { 22980b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 229963443bf5SEric Sandeen "Failed to recover log tree"); 230000246528SJosef Bacik btrfs_put_root(log_tree_root); 230163443bf5SEric Sandeen return ret; 230263443bf5SEric Sandeen } 230363443bf5SEric Sandeen 2304bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 23056bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 230663443bf5SEric Sandeen if (ret) 230763443bf5SEric Sandeen return ret; 230863443bf5SEric Sandeen } 230963443bf5SEric Sandeen 231063443bf5SEric Sandeen return 0; 231163443bf5SEric Sandeen } 231263443bf5SEric Sandeen 23136bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 23144bbcaa64SEric Sandeen { 23156bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2316a4f3d2c4SDavid Sterba struct btrfs_root *root; 23174bbcaa64SEric Sandeen struct btrfs_key location; 23184bbcaa64SEric Sandeen int ret; 23194bbcaa64SEric Sandeen 23206bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 23216bccf3abSJeff Mahoney 23224bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 23234bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 23244bbcaa64SEric Sandeen location.offset = 0; 23254bbcaa64SEric Sandeen 2326a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2327f50f4353SLiu Bo if (IS_ERR(root)) { 2328f50f4353SLiu Bo ret = PTR_ERR(root); 2329f50f4353SLiu Bo goto out; 2330f50f4353SLiu Bo } 2331a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2332a4f3d2c4SDavid Sterba fs_info->extent_root = root; 23334bbcaa64SEric Sandeen 23344bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2335a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2336f50f4353SLiu Bo if (IS_ERR(root)) { 2337f50f4353SLiu Bo ret = PTR_ERR(root); 2338f50f4353SLiu Bo goto out; 2339f50f4353SLiu Bo } 2340a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2341a4f3d2c4SDavid Sterba fs_info->dev_root = root; 23424bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 23434bbcaa64SEric Sandeen 23444bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2345a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2346f50f4353SLiu Bo if (IS_ERR(root)) { 2347f50f4353SLiu Bo ret = PTR_ERR(root); 2348f50f4353SLiu Bo goto out; 2349f50f4353SLiu Bo } 2350a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2351a4f3d2c4SDavid Sterba fs_info->csum_root = root; 23524bbcaa64SEric Sandeen 2353aeb935a4SQu Wenruo /* 2354aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2355aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2356aeb935a4SQu Wenruo */ 235756e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 235856e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2359aeb935a4SQu Wenruo if (IS_ERR(root)) { 2360aeb935a4SQu Wenruo ret = PTR_ERR(root); 2361aeb935a4SQu Wenruo goto out; 2362aeb935a4SQu Wenruo } 2363aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2364aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 2365aeb935a4SQu Wenruo 23664bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2367a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2368a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2369a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2370afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2371a4f3d2c4SDavid Sterba fs_info->quota_root = root; 23724bbcaa64SEric Sandeen } 23734bbcaa64SEric Sandeen 23744bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2375a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2376a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 2377a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 23784bbcaa64SEric Sandeen if (ret != -ENOENT) 2379f50f4353SLiu Bo goto out; 23804bbcaa64SEric Sandeen } else { 2381a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2382a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 23834bbcaa64SEric Sandeen } 23844bbcaa64SEric Sandeen 238570f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 238670f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 238770f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 2388f50f4353SLiu Bo if (IS_ERR(root)) { 2389f50f4353SLiu Bo ret = PTR_ERR(root); 2390f50f4353SLiu Bo goto out; 2391f50f4353SLiu Bo } 239270f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 239370f6d82eSOmar Sandoval fs_info->free_space_root = root; 239470f6d82eSOmar Sandoval } 239570f6d82eSOmar Sandoval 23964bbcaa64SEric Sandeen return 0; 2397f50f4353SLiu Bo out: 2398f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2399f50f4353SLiu Bo location.objectid, ret); 2400f50f4353SLiu Bo return ret; 24014bbcaa64SEric Sandeen } 24024bbcaa64SEric Sandeen 2403069ec957SQu Wenruo /* 2404069ec957SQu Wenruo * Real super block validation 2405069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2406069ec957SQu Wenruo * 2407069ec957SQu Wenruo * @sb: super block to check 2408069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2409069ec957SQu Wenruo * 0 the primary (1st) sb 2410069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2411069ec957SQu Wenruo * -1 skip bytenr check 2412069ec957SQu Wenruo */ 2413069ec957SQu Wenruo static int validate_super(struct btrfs_fs_info *fs_info, 2414069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 241521a852b0SQu Wenruo { 241621a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 241721a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 241821a852b0SQu Wenruo int ret = 0; 241921a852b0SQu Wenruo 242021a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 242121a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 242221a852b0SQu Wenruo ret = -EINVAL; 242321a852b0SQu Wenruo } 242421a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 242521a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 242621a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 242721a852b0SQu Wenruo ret = -EINVAL; 242821a852b0SQu Wenruo } 242921a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 243021a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 243121a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 243221a852b0SQu Wenruo ret = -EINVAL; 243321a852b0SQu Wenruo } 243421a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 243521a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 243621a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 243721a852b0SQu Wenruo ret = -EINVAL; 243821a852b0SQu Wenruo } 243921a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 244021a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 244121a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 244221a852b0SQu Wenruo ret = -EINVAL; 244321a852b0SQu Wenruo } 244421a852b0SQu Wenruo 244521a852b0SQu Wenruo /* 244621a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 244721a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 244821a852b0SQu Wenruo */ 244921a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 245021a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 245121a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 245221a852b0SQu Wenruo ret = -EINVAL; 245321a852b0SQu Wenruo } 245421a852b0SQu Wenruo /* Only PAGE SIZE is supported yet */ 245521a852b0SQu Wenruo if (sectorsize != PAGE_SIZE) { 245621a852b0SQu Wenruo btrfs_err(fs_info, 245721a852b0SQu Wenruo "sectorsize %llu not supported yet, only support %lu", 245821a852b0SQu Wenruo sectorsize, PAGE_SIZE); 245921a852b0SQu Wenruo ret = -EINVAL; 246021a852b0SQu Wenruo } 246121a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 246221a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 246321a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 246421a852b0SQu Wenruo ret = -EINVAL; 246521a852b0SQu Wenruo } 246621a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 246721a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 246821a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 246921a852b0SQu Wenruo ret = -EINVAL; 247021a852b0SQu Wenruo } 247121a852b0SQu Wenruo 247221a852b0SQu Wenruo /* Root alignment check */ 247321a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 247421a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 247521a852b0SQu Wenruo btrfs_super_root(sb)); 247621a852b0SQu Wenruo ret = -EINVAL; 247721a852b0SQu Wenruo } 247821a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 247921a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 248021a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 248121a852b0SQu Wenruo ret = -EINVAL; 248221a852b0SQu Wenruo } 248321a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 248421a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 248521a852b0SQu Wenruo btrfs_super_log_root(sb)); 248621a852b0SQu Wenruo ret = -EINVAL; 248721a852b0SQu Wenruo } 248821a852b0SQu Wenruo 2489de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 24907239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 249121a852b0SQu Wenruo btrfs_err(fs_info, 24927239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2493de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 249421a852b0SQu Wenruo ret = -EINVAL; 249521a852b0SQu Wenruo } 249621a852b0SQu Wenruo 249721a852b0SQu Wenruo /* 249821a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 249921a852b0SQu Wenruo * done later 250021a852b0SQu Wenruo */ 250121a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 250221a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 250321a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 250421a852b0SQu Wenruo ret = -EINVAL; 250521a852b0SQu Wenruo } 250621a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 250721a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 250821a852b0SQu Wenruo btrfs_super_stripesize(sb)); 250921a852b0SQu Wenruo ret = -EINVAL; 251021a852b0SQu Wenruo } 251121a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 251221a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 251321a852b0SQu Wenruo btrfs_super_num_devices(sb)); 251421a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 251521a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 251621a852b0SQu Wenruo ret = -EINVAL; 251721a852b0SQu Wenruo } 251821a852b0SQu Wenruo 2519069ec957SQu Wenruo if (mirror_num >= 0 && 2520069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 252121a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 252221a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 252321a852b0SQu Wenruo ret = -EINVAL; 252421a852b0SQu Wenruo } 252521a852b0SQu Wenruo 252621a852b0SQu Wenruo /* 252721a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 252821a852b0SQu Wenruo * and one chunk 252921a852b0SQu Wenruo */ 253021a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 253121a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 253221a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 253321a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 253421a852b0SQu Wenruo ret = -EINVAL; 253521a852b0SQu Wenruo } 253621a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 253721a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 253821a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 253921a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 254021a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 254121a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 254221a852b0SQu Wenruo ret = -EINVAL; 254321a852b0SQu Wenruo } 254421a852b0SQu Wenruo 254521a852b0SQu Wenruo /* 254621a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 254721a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 254821a852b0SQu Wenruo */ 254921a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 255021a852b0SQu Wenruo btrfs_warn(fs_info, 255121a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 255221a852b0SQu Wenruo btrfs_super_generation(sb), 255321a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 255421a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 255521a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 255621a852b0SQu Wenruo btrfs_warn(fs_info, 255721a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 255821a852b0SQu Wenruo btrfs_super_generation(sb), 255921a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 256021a852b0SQu Wenruo 256121a852b0SQu Wenruo return ret; 256221a852b0SQu Wenruo } 256321a852b0SQu Wenruo 2564069ec957SQu Wenruo /* 2565069ec957SQu Wenruo * Validation of super block at mount time. 2566069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2567069ec957SQu Wenruo * flags will be skipped. 2568069ec957SQu Wenruo */ 2569069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2570069ec957SQu Wenruo { 2571069ec957SQu Wenruo return validate_super(fs_info, fs_info->super_copy, 0); 2572069ec957SQu Wenruo } 2573069ec957SQu Wenruo 257475cb857dSQu Wenruo /* 257575cb857dSQu Wenruo * Validation of super block at write time. 257675cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 257775cb857dSQu Wenruo * overwritten soon. 257875cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 257975cb857dSQu Wenruo */ 258075cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 258175cb857dSQu Wenruo struct btrfs_super_block *sb) 258275cb857dSQu Wenruo { 258375cb857dSQu Wenruo int ret; 258475cb857dSQu Wenruo 258575cb857dSQu Wenruo ret = validate_super(fs_info, sb, -1); 258675cb857dSQu Wenruo if (ret < 0) 258775cb857dSQu Wenruo goto out; 2588e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 258975cb857dSQu Wenruo ret = -EUCLEAN; 259075cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 259175cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 259275cb857dSQu Wenruo goto out; 259375cb857dSQu Wenruo } 259475cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 259575cb857dSQu Wenruo ret = -EUCLEAN; 259675cb857dSQu Wenruo btrfs_err(fs_info, 259775cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 259875cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 259975cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 260075cb857dSQu Wenruo goto out; 260175cb857dSQu Wenruo } 260275cb857dSQu Wenruo out: 260375cb857dSQu Wenruo if (ret < 0) 260475cb857dSQu Wenruo btrfs_err(fs_info, 260575cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 260675cb857dSQu Wenruo return ret; 260775cb857dSQu Wenruo } 260875cb857dSQu Wenruo 26096ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2610b8522a1eSNikolay Borisov { 26116ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2612b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2613b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2614b8522a1eSNikolay Borisov bool handle_error = false; 2615b8522a1eSNikolay Borisov int ret = 0; 2616b8522a1eSNikolay Borisov int i; 2617b8522a1eSNikolay Borisov 2618b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2619b8522a1eSNikolay Borisov u64 generation; 2620b8522a1eSNikolay Borisov int level; 2621b8522a1eSNikolay Borisov 2622b8522a1eSNikolay Borisov if (handle_error) { 2623b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2624b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2625b8522a1eSNikolay Borisov tree_root->node = NULL; 2626b8522a1eSNikolay Borisov 2627b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2628b8522a1eSNikolay Borisov break; 2629b8522a1eSNikolay Borisov 2630b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2631b8522a1eSNikolay Borisov 2632b8522a1eSNikolay Borisov /* 2633b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2634b8522a1eSNikolay Borisov * valid 2635b8522a1eSNikolay Borisov */ 2636b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2637b8522a1eSNikolay Borisov 2638b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2639b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2640b8522a1eSNikolay Borisov 2641b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26426ef108ddSNikolay Borisov backup_index = ret; 2643b8522a1eSNikolay Borisov if (ret < 0) 2644b8522a1eSNikolay Borisov return ret; 2645b8522a1eSNikolay Borisov } 2646b8522a1eSNikolay Borisov generation = btrfs_super_generation(sb); 2647b8522a1eSNikolay Borisov level = btrfs_super_root_level(sb); 2648b8522a1eSNikolay Borisov tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), 2649b8522a1eSNikolay Borisov generation, level, NULL); 2650b8522a1eSNikolay Borisov if (IS_ERR(tree_root->node) || 2651b8522a1eSNikolay Borisov !extent_buffer_uptodate(tree_root->node)) { 2652b8522a1eSNikolay Borisov handle_error = true; 2653b8522a1eSNikolay Borisov 26540465337cSJosef Bacik if (IS_ERR(tree_root->node)) { 2655b8522a1eSNikolay Borisov ret = PTR_ERR(tree_root->node); 26560465337cSJosef Bacik tree_root->node = NULL; 26570465337cSJosef Bacik } else if (!extent_buffer_uptodate(tree_root->node)) { 2658b8522a1eSNikolay Borisov ret = -EUCLEAN; 26590465337cSJosef Bacik } 2660b8522a1eSNikolay Borisov 2661b8522a1eSNikolay Borisov btrfs_warn(fs_info, "failed to read tree root"); 2662b8522a1eSNikolay Borisov continue; 2663b8522a1eSNikolay Borisov } 2664b8522a1eSNikolay Borisov 2665b8522a1eSNikolay Borisov btrfs_set_root_node(&tree_root->root_item, tree_root->node); 2666b8522a1eSNikolay Borisov tree_root->commit_root = btrfs_root_node(tree_root); 2667b8522a1eSNikolay Borisov btrfs_set_root_refs(&tree_root->root_item, 1); 2668b8522a1eSNikolay Borisov 2669336a0d8dSNikolay Borisov /* 2670336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2671336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2672336a0d8dSNikolay Borisov */ 2673b8522a1eSNikolay Borisov ret = btrfs_find_highest_objectid(tree_root, 2674b8522a1eSNikolay Borisov &tree_root->highest_objectid); 2675b8522a1eSNikolay Borisov if (ret < 0) { 2676b8522a1eSNikolay Borisov handle_error = true; 2677b8522a1eSNikolay Borisov continue; 2678b8522a1eSNikolay Borisov } 2679b8522a1eSNikolay Borisov 2680b8522a1eSNikolay Borisov ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 2681b8522a1eSNikolay Borisov 2682b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2683b8522a1eSNikolay Borisov if (ret < 0) { 2684b8522a1eSNikolay Borisov handle_error = true; 2685b8522a1eSNikolay Borisov continue; 2686b8522a1eSNikolay Borisov } 2687b8522a1eSNikolay Borisov 2688b8522a1eSNikolay Borisov /* All successful */ 2689b8522a1eSNikolay Borisov fs_info->generation = generation; 2690b8522a1eSNikolay Borisov fs_info->last_trans_committed = generation; 26916ef108ddSNikolay Borisov 26926ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26936ef108ddSNikolay Borisov if (backup_index < 0) { 26946ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26956ef108ddSNikolay Borisov } else { 26966ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26976ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26986ef108ddSNikolay Borisov } 2699b8522a1eSNikolay Borisov break; 2700b8522a1eSNikolay Borisov } 2701b8522a1eSNikolay Borisov 2702b8522a1eSNikolay Borisov return ret; 2703b8522a1eSNikolay Borisov } 2704b8522a1eSNikolay Borisov 27058260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2706eb60ceacSChris Mason { 270776dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2708f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 27098fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2710facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 271124bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2712eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 271311833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2714eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2715a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 271676dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 271724bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 27184cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2719ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2720f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 272147ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2722f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2723d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 272467c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 27257585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2726573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2727de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 272819c00ddcSChris Mason 27290b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 27306324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2731f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 273247ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 2733bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2734bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 27353fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27363fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2737bd647ce3SJosef Bacik #endif 2738c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 273966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 274066d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 274166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 274266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 274366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 274466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 274566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2746ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2747ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2748ba2c4d4eSJosef Bacik 2749771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27504cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 27512fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2752034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2753fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 275495ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27559ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27564cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2757a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2758f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27598b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2760f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 276190519d66SArne Jansen /* readahead state */ 2762d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 276390519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2764fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2765c8b97818SChris Mason 2766b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2767b34b086cSChris Mason num_online_cpus() + 2, 8); 27680afbaf8cSChris Mason 2769199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2770199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 277169fe2d75SJosef Bacik 2772638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 277321adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 277421adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 277521adbd5cSStefan Behrens #endif 2776779a65a4SEric Sandeen btrfs_init_balance(fs_info); 277721c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2778a2de733cSArne Jansen 27790f9dd46cSJosef Bacik spin_lock_init(&fs_info->block_group_cache_lock); 27806bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2781a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 27820f9dd46cSJosef Bacik 2783fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 2784fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 2785afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 278639279cc3SChris Mason 27875a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2788e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2789925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2790a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2791a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27921bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27939e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2794c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 279576dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2796803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2797fa9c0d79SChris Mason 2798ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2799f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2800b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2801416ac51dSArne Jansen 2802fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2803fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2804fa9c0d79SChris Mason 2805e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2806f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2807bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 28084854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2809034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 28103768f368SChris Mason 2811da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2812da17066cSJeff Mahoney fs_info->nodesize = 4096; 2813da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2814da17066cSJeff Mahoney fs_info->stripesize = 4096; 2815da17066cSJeff Mahoney 2816eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2817eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2818eede2bf3SOmar Sandoval 28199e967495SFilipe Manana fs_info->send_in_progress = 0; 28208260edbaSJosef Bacik } 28218260edbaSJosef Bacik 28228260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 28238260edbaSJosef Bacik { 28248260edbaSJosef Bacik int ret; 28258260edbaSJosef Bacik 28268260edbaSJosef Bacik fs_info->sb = sb; 28278260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 28288260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 28299e967495SFilipe Manana 2830c75e8394SJosef Bacik ret = percpu_counter_init(&fs_info->dio_bytes, 0, GFP_KERNEL); 2831ae18c37aSJosef Bacik if (ret) 2832ae18c37aSJosef Bacik return ret; 2833ae18c37aSJosef Bacik 2834ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2835ae18c37aSJosef Bacik if (ret) 2836c75e8394SJosef Bacik return ret; 2837ae18c37aSJosef Bacik 2838ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2839ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2840ae18c37aSJosef Bacik 2841ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2842ae18c37aSJosef Bacik if (ret) 2843c75e8394SJosef Bacik return ret; 2844ae18c37aSJosef Bacik 2845ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2846ae18c37aSJosef Bacik GFP_KERNEL); 2847ae18c37aSJosef Bacik if (ret) 2848c75e8394SJosef Bacik return ret; 2849ae18c37aSJosef Bacik 2850ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2851ae18c37aSJosef Bacik GFP_KERNEL); 2852c75e8394SJosef Bacik if (!fs_info->delayed_root) 2853c75e8394SJosef Bacik return -ENOMEM; 2854ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2855ae18c37aSJosef Bacik 2856c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2857ae18c37aSJosef Bacik } 2858ae18c37aSJosef Bacik 285997f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 286097f4dd09SNikolay Borisov { 286197f4dd09SNikolay Borisov struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; 286297f4dd09SNikolay Borisov int ret; 286397f4dd09SNikolay Borisov 286497f4dd09SNikolay Borisov /* 286597f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 286697f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 286797f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 286897f4dd09SNikolay Borisov */ 286997f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 287097f4dd09SNikolay Borisov if (ret < 0) { 2871c94bec2cSJosef Bacik if (ret != -EINTR) 2872c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2873c94bec2cSJosef Bacik ret); 287497f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 287597f4dd09SNikolay Borisov return ret; 287697f4dd09SNikolay Borisov } 287797f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 287897f4dd09SNikolay Borisov } 287997f4dd09SNikolay Borisov 288097f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 288197f4dd09SNikolay Borisov { 288297f4dd09SNikolay Borisov struct task_struct *task; 288397f4dd09SNikolay Borisov 288497f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 288597f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 288697f4dd09SNikolay Borisov if (IS_ERR(task)) { 288797f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 288897f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 288997f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 289097f4dd09SNikolay Borisov return PTR_ERR(task); 289197f4dd09SNikolay Borisov } 289297f4dd09SNikolay Borisov 289397f4dd09SNikolay Borisov return 0; 289497f4dd09SNikolay Borisov } 289597f4dd09SNikolay Borisov 2896ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 2897ae18c37aSJosef Bacik char *options) 2898ae18c37aSJosef Bacik { 2899ae18c37aSJosef Bacik u32 sectorsize; 2900ae18c37aSJosef Bacik u32 nodesize; 2901ae18c37aSJosef Bacik u32 stripesize; 2902ae18c37aSJosef Bacik u64 generation; 2903ae18c37aSJosef Bacik u64 features; 2904ae18c37aSJosef Bacik u16 csum_type; 2905ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 2906ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2907ae18c37aSJosef Bacik struct btrfs_root *tree_root; 2908ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 2909ae18c37aSJosef Bacik int ret; 2910ae18c37aSJosef Bacik int err = -EINVAL; 2911ae18c37aSJosef Bacik int clear_free_space_tree = 0; 2912ae18c37aSJosef Bacik int level; 2913ae18c37aSJosef Bacik 29148260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 291553b381b3SDavid Woodhouse if (ret) { 291683c8266aSDavid Sterba err = ret; 2917ae18c37aSJosef Bacik goto fail; 291853b381b3SDavid Woodhouse } 291953b381b3SDavid Woodhouse 2920ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 2921ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 2922ae18c37aSJosef Bacik GFP_KERNEL); 2923ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 2924ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 2925ae18c37aSJosef Bacik GFP_KERNEL); 2926ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 2927ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 2928ae18c37aSJosef Bacik err = -ENOMEM; 2929c75e8394SJosef Bacik goto fail; 2930ae18c37aSJosef Bacik } 2931ae18c37aSJosef Bacik 2932ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 2933ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 2934ae18c37aSJosef Bacik err = -ENOMEM; 2935c75e8394SJosef Bacik goto fail; 2936ae18c37aSJosef Bacik } 2937ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 2938ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 2939ae18c37aSJosef Bacik 29403c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 29411104a885SDavid Sterba 29421104a885SDavid Sterba /* 29431104a885SDavid Sterba * Read super block and check the signature bytes only 29441104a885SDavid Sterba */ 29458f32380dSJohannes Thumshirn disk_super = btrfs_read_dev_super(fs_devices->latest_bdev); 29468f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 29478f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 294816cdcec7SMiao Xie goto fail_alloc; 294920b45077SDave Young } 295039279cc3SChris Mason 29511104a885SDavid Sterba /* 29528dc3f22cSJohannes Thumshirn * Verify the type first, if that or the the checksum value are 29538dc3f22cSJohannes Thumshirn * corrupted, we'll find out 29548dc3f22cSJohannes Thumshirn */ 29558f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 295651bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 29578dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 295851bce6c9SJohannes Thumshirn csum_type); 29598dc3f22cSJohannes Thumshirn err = -EINVAL; 29608f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 29618dc3f22cSJohannes Thumshirn goto fail_alloc; 29628dc3f22cSJohannes Thumshirn } 29638dc3f22cSJohannes Thumshirn 29646d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 29656d97c6e3SJohannes Thumshirn if (ret) { 29666d97c6e3SJohannes Thumshirn err = ret; 29678f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 29686d97c6e3SJohannes Thumshirn goto fail_alloc; 29696d97c6e3SJohannes Thumshirn } 29706d97c6e3SJohannes Thumshirn 29718dc3f22cSJohannes Thumshirn /* 29721104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 29731104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 29741104a885SDavid Sterba */ 29758f32380dSJohannes Thumshirn if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { 297605135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 29771104a885SDavid Sterba err = -EINVAL; 29788f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 2979141386e1SJosef Bacik goto fail_alloc; 29801104a885SDavid Sterba } 29811104a885SDavid Sterba 29821104a885SDavid Sterba /* 29831104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 29841104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 29851104a885SDavid Sterba * the whole block of INFO_SIZE 29861104a885SDavid Sterba */ 29878f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 29888f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 29895f39d397SChris Mason 2990fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 2991fbc6feaeSNikolay Borisov 2992de37aa51SNikolay Borisov ASSERT(!memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2993de37aa51SNikolay Borisov BTRFS_FSID_SIZE)); 2994de37aa51SNikolay Borisov 29957239ff4bSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) { 2996de37aa51SNikolay Borisov ASSERT(!memcmp(fs_info->fs_devices->metadata_uuid, 2997de37aa51SNikolay Borisov fs_info->super_copy->metadata_uuid, 2998de37aa51SNikolay Borisov BTRFS_FSID_SIZE)); 29997239ff4bSNikolay Borisov } 30000b86a832SChris Mason 3001fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3002fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3003fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3004fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3005fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3006fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3007fbc6feaeSNikolay Borisov } 3008fbc6feaeSNikolay Borisov 3009fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3010fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3011de37aa51SNikolay Borisov 3012069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 30131104a885SDavid Sterba if (ret) { 301405135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 30151104a885SDavid Sterba err = -EINVAL; 3016141386e1SJosef Bacik goto fail_alloc; 30171104a885SDavid Sterba } 30181104a885SDavid Sterba 30190f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3020141386e1SJosef Bacik goto fail_alloc; 30210f7d52f4SChris Mason 3022acce952bSliubo /* check FS state, whether FS is broken. */ 302387533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 302487533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3025acce952bSliubo 302675e7cb7fSLiu Bo /* 302775e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 302875e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 302975e7cb7fSLiu Bo */ 303075e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 303175e7cb7fSLiu Bo 30322ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 30332b82032cSYan Zheng if (ret) { 30342b82032cSYan Zheng err = ret; 3035141386e1SJosef Bacik goto fail_alloc; 30362b82032cSYan Zheng } 3037dfe25020SChris Mason 3038f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 3039f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 3040f2b636e8SJosef Bacik if (features) { 304105135f59SDavid Sterba btrfs_err(fs_info, 304205135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 3043c1c9ff7cSGeert Uytterhoeven features); 3044f2b636e8SJosef Bacik err = -EINVAL; 3045141386e1SJosef Bacik goto fail_alloc; 3046f2b636e8SJosef Bacik } 3047f2b636e8SJosef Bacik 30485d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 30495d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 30500b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3051a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 30525c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 30535c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3054727011e0SChris Mason 30553173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 305605135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 30573173a18fSJosef Bacik 3058727011e0SChris Mason /* 3059727011e0SChris Mason * flag our filesystem as having big metadata blocks if 3060727011e0SChris Mason * they are bigger than the page size 3061727011e0SChris Mason */ 306209cbfeafSKirill A. Shutemov if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 3063727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 306405135f59SDavid Sterba btrfs_info(fs_info, 306505135f59SDavid Sterba "flagging fs with big metadata feature"); 3066727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3067727011e0SChris Mason } 3068727011e0SChris Mason 3069bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 3070bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 3071b7f67055SChandan Rajendra stripesize = sectorsize; 3072707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 3073963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 3074bc3f116fSChris Mason 3075da17066cSJeff Mahoney /* Cache block sizes */ 3076da17066cSJeff Mahoney fs_info->nodesize = nodesize; 3077da17066cSJeff Mahoney fs_info->sectorsize = sectorsize; 3078da17066cSJeff Mahoney fs_info->stripesize = stripesize; 3079da17066cSJeff Mahoney 3080bc3f116fSChris Mason /* 3081bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 3082bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 3083bc3f116fSChris Mason */ 3084bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3085707e8a07SDavid Sterba (sectorsize != nodesize)) { 308605135f59SDavid Sterba btrfs_err(fs_info, 308705135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 308805135f59SDavid Sterba nodesize, sectorsize); 3089141386e1SJosef Bacik goto fail_alloc; 3090bc3f116fSChris Mason } 3091bc3f116fSChris Mason 3092ceda0864SMiao Xie /* 3093ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 3094ceda0864SMiao Xie * update the flag. 3095ceda0864SMiao Xie */ 30965d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 30975d4f98a2SYan Zheng 3098f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 3099f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 3100bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 310105135f59SDavid Sterba btrfs_err(fs_info, 310205135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 3103c1c9ff7cSGeert Uytterhoeven features); 3104f2b636e8SJosef Bacik err = -EINVAL; 3105141386e1SJosef Bacik goto fail_alloc; 3106f2b636e8SJosef Bacik } 310761d92c32SChris Mason 31082a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 31092a458198SEric Sandeen if (ret) { 31102a458198SEric Sandeen err = ret; 31110dc3b84aSJosef Bacik goto fail_sb_buffer; 31120dc3b84aSJosef Bacik } 31134543df7eSChris Mason 31149e11ceeeSJan Kara sb->s_bdi->congested_fn = btrfs_congested_fn; 31159e11ceeeSJan Kara sb->s_bdi->congested_data = fs_info; 31169e11ceeeSJan Kara sb->s_bdi->capabilities |= BDI_CAP_CGROUP_WRITEBACK; 3117b5420237SNikolay Borisov sb->s_bdi->ra_pages = VM_READAHEAD_PAGES; 31189e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 31199e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 31204575c9ccSChris Mason 3121a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3122a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3123de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3124db94535dSChris Mason 3125925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 31266bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3127925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 312884eed90fSChris Mason if (ret) { 312905135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 31305d4f98a2SYan Zheng goto fail_sb_buffer; 313184eed90fSChris Mason } 31320b86a832SChris Mason 313384234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3134581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 31350b86a832SChris Mason 31362ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 31370b86a832SChris Mason btrfs_super_chunk_root(disk_super), 3138581c1760SQu Wenruo generation, level, NULL); 313964c043deSLiu Bo if (IS_ERR(chunk_root->node) || 314064c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 314105135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3142e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 3143e5fffbacSchandan free_extent_buffer(chunk_root->node); 314495ab1f64SZhao Lei chunk_root->node = NULL; 3145af31f5e5SChris Mason goto fail_tree_roots; 314683121942SDavid Woodhouse } 31475d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 31485d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 31490b86a832SChris Mason 3150e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3151c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3152c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3153e17cade2SChris Mason 31545b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 31552b82032cSYan Zheng if (ret) { 315605135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3157af31f5e5SChris Mason goto fail_tree_roots; 31582b82032cSYan Zheng } 31590b86a832SChris Mason 31608dabb742SStefan Behrens /* 31619b99b115SAnand Jain * Keep the devid that is marked to be the target device for the 31629b99b115SAnand Jain * device replace procedure 31638dabb742SStefan Behrens */ 31649b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 0); 3165dfe25020SChris Mason 3166a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 316705135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3168a6b0d5c8SChris Mason goto fail_tree_roots; 3169a6b0d5c8SChris Mason } 3170a6b0d5c8SChris Mason 3171b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 31724bbcaa64SEric Sandeen if (ret) 3173b8522a1eSNikolay Borisov goto fail_tree_roots; 31748929ecfaSYan, Zheng 317575ec1db8SJosef Bacik /* 317675ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 317775ec1db8SJosef Bacik * check the generation here so we can set the 317875ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 317975ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 318075ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 318175ec1db8SJosef Bacik * even though it was perfectly fine. 318275ec1db8SJosef Bacik */ 318375ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 318475ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 318575ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 318675ec1db8SJosef Bacik 3187cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3188cf90d884SQu Wenruo if (ret) { 3189cf90d884SQu Wenruo btrfs_err(fs_info, 3190cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3191cf90d884SQu Wenruo ret); 3192cf90d884SQu Wenruo goto fail_block_groups; 3193cf90d884SQu Wenruo } 319468310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 319568310a5eSIlya Dryomov if (ret) { 319605135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 319768310a5eSIlya Dryomov goto fail_block_groups; 319868310a5eSIlya Dryomov } 319968310a5eSIlya Dryomov 3200733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3201733f4fbbSStefan Behrens if (ret) { 320205135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3203733f4fbbSStefan Behrens goto fail_block_groups; 3204733f4fbbSStefan Behrens } 3205733f4fbbSStefan Behrens 32068dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 32078dabb742SStefan Behrens if (ret) { 320805135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 32098dabb742SStefan Behrens goto fail_block_groups; 32108dabb742SStefan Behrens } 32118dabb742SStefan Behrens 32129b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 1); 32138dabb742SStefan Behrens 3214c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3215b7c35e81SAnand Jain if (ret) { 321605135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 321705135f59SDavid Sterba ret); 3218b7c35e81SAnand Jain goto fail_block_groups; 3219b7c35e81SAnand Jain } 3220b7c35e81SAnand Jain 322196f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 32225ac1d209SJeff Mahoney if (ret) { 322305135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3224b7c35e81SAnand Jain goto fail_fsdev_sysfs; 32255ac1d209SJeff Mahoney } 32265ac1d209SJeff Mahoney 3227c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3228c59021f8Sliubo if (ret) { 322905135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 32302365dd3cSAnand Jain goto fail_sysfs; 3231c59021f8Sliubo } 3232c59021f8Sliubo 32335b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 32341b1d1f66SJosef Bacik if (ret) { 323505135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 32362365dd3cSAnand Jain goto fail_sysfs; 32371b1d1f66SJosef Bacik } 32384330e183SQu Wenruo 32396528b99dSAnand Jain if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { 324005135f59SDavid Sterba btrfs_warn(fs_info, 324152042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 32422365dd3cSAnand Jain goto fail_sysfs; 3243292fd7fcSStefan Behrens } 32449078a3e1SChris Mason 3245a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 3246a74a4b97SChris Mason "btrfs-cleaner"); 324757506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 32482365dd3cSAnand Jain goto fail_sysfs; 3249a74a4b97SChris Mason 3250a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3251a74a4b97SChris Mason tree_root, 3252a74a4b97SChris Mason "btrfs-transaction"); 325357506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 32543f157a2fSChris Mason goto fail_cleaner; 3255a74a4b97SChris Mason 3256583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3257c289811cSChris Mason !fs_info->fs_devices->rotating) { 3258583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3259c289811cSChris Mason } 3260c289811cSChris Mason 3261572d9ab7SDavid Sterba /* 326201327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 3263572d9ab7SDavid Sterba * not have to wait for transaction commit 3264572d9ab7SDavid Sterba */ 3265572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 32663818aea2SQu Wenruo 326721adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 32680b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 32692ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 32700b246afaSJeff Mahoney btrfs_test_opt(fs_info, 327121adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 327221adbd5cSStefan Behrens 1 : 0, 327321adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 327421adbd5cSStefan Behrens if (ret) 327505135f59SDavid Sterba btrfs_warn(fs_info, 327605135f59SDavid Sterba "failed to initialize integrity check module: %d", 327705135f59SDavid Sterba ret); 327821adbd5cSStefan Behrens } 327921adbd5cSStefan Behrens #endif 3280bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3281bcef60f2SArne Jansen if (ret) 3282bcef60f2SArne Jansen goto fail_trans_kthread; 328321adbd5cSStefan Behrens 3284fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3285fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3286fd708b81SJosef Bacik 328796da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 328896da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 32890b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3290e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 329163443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 329279787eaaSJeff Mahoney if (ret) { 329363443bf5SEric Sandeen err = ret; 329428c16cbbSWang Shilong goto fail_qgroup; 3295e556ce2cSYan Zheng } 3296e02119d5SChris Mason } 32971a40e23bSZheng Yan 32986bccf3abSJeff Mahoney ret = btrfs_find_orphan_roots(fs_info); 329979787eaaSJeff Mahoney if (ret) 330028c16cbbSWang Shilong goto fail_qgroup; 330176dda93cSYan, Zheng 3302bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 3303d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 330444c44af2SIlya Dryomov if (ret) 330528c16cbbSWang Shilong goto fail_qgroup; 330690c711abSZygo Blaxell 330790c711abSZygo Blaxell mutex_lock(&fs_info->cleaner_mutex); 33085d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 330990c711abSZygo Blaxell mutex_unlock(&fs_info->cleaner_mutex); 3310d7ce5843SMiao Xie if (ret < 0) { 331105135f59SDavid Sterba btrfs_warn(fs_info, "failed to recover relocation: %d", 331205135f59SDavid Sterba ret); 3313d7ce5843SMiao Xie err = -EINVAL; 3314bcef60f2SArne Jansen goto fail_qgroup; 3315d7ce5843SMiao Xie } 33167c2ca468SChris Mason } 33171a40e23bSZheng Yan 331856e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 33193140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 33203140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3321f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3322315bf8efSJosef Bacik fs_info->fs_root = NULL; 3323bcef60f2SArne Jansen goto fail_qgroup; 33243140c9a3SDan Carpenter } 3325c289811cSChris Mason 3326bc98a42cSDavid Howells if (sb_rdonly(sb)) 33272b6ba629SIlya Dryomov return 0; 33282b6ba629SIlya Dryomov 3329f8d468a1SOmar Sandoval if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 3330f8d468a1SOmar Sandoval btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 33316675df31SOmar Sandoval clear_free_space_tree = 1; 33326675df31SOmar Sandoval } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 33336675df31SOmar Sandoval !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 33346675df31SOmar Sandoval btrfs_warn(fs_info, "free space tree is invalid"); 33356675df31SOmar Sandoval clear_free_space_tree = 1; 33366675df31SOmar Sandoval } 33376675df31SOmar Sandoval 33386675df31SOmar Sandoval if (clear_free_space_tree) { 3339f8d468a1SOmar Sandoval btrfs_info(fs_info, "clearing free space tree"); 3340f8d468a1SOmar Sandoval ret = btrfs_clear_free_space_tree(fs_info); 3341f8d468a1SOmar Sandoval if (ret) { 3342f8d468a1SOmar Sandoval btrfs_warn(fs_info, 3343f8d468a1SOmar Sandoval "failed to clear free space tree: %d", ret); 33446bccf3abSJeff Mahoney close_ctree(fs_info); 3345f8d468a1SOmar Sandoval return ret; 3346f8d468a1SOmar Sandoval } 3347f8d468a1SOmar Sandoval } 3348f8d468a1SOmar Sandoval 33490b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 3350511711afSChris Mason !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 335105135f59SDavid Sterba btrfs_info(fs_info, "creating free space tree"); 3352511711afSChris Mason ret = btrfs_create_free_space_tree(fs_info); 3353511711afSChris Mason if (ret) { 335405135f59SDavid Sterba btrfs_warn(fs_info, 335505135f59SDavid Sterba "failed to create free space tree: %d", ret); 33566bccf3abSJeff Mahoney close_ctree(fs_info); 3357511711afSChris Mason return ret; 3358511711afSChris Mason } 3359511711afSChris Mason } 3360511711afSChris Mason 3361e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 33622b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 33632b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 33642b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 33656bccf3abSJeff Mahoney close_ctree(fs_info); 33662b6ba629SIlya Dryomov return ret; 33672b6ba629SIlya Dryomov } 3368e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 336959641015SIlya Dryomov 33702b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 33712b6ba629SIlya Dryomov if (ret) { 337205135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume balance: %d", ret); 33736bccf3abSJeff Mahoney close_ctree(fs_info); 33742b6ba629SIlya Dryomov return ret; 3375e3acc2a6SJosef Bacik } 3376e3acc2a6SJosef Bacik 33778dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 33788dabb742SStefan Behrens if (ret) { 337905135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume device replace: %d", ret); 33806bccf3abSJeff Mahoney close_ctree(fs_info); 33818dabb742SStefan Behrens return ret; 33828dabb742SStefan Behrens } 33838dabb742SStefan Behrens 3384b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 3385b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3386b382a324SJan Schmidt 33874bbcaa64SEric Sandeen if (!fs_info->uuid_root) { 338805135f59SDavid Sterba btrfs_info(fs_info, "creating UUID tree"); 3389f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 3390f7a81ea4SStefan Behrens if (ret) { 339105135f59SDavid Sterba btrfs_warn(fs_info, 339205135f59SDavid Sterba "failed to create the UUID tree: %d", ret); 33936bccf3abSJeff Mahoney close_ctree(fs_info); 3394f7a81ea4SStefan Behrens return ret; 3395f7a81ea4SStefan Behrens } 33960b246afaSJeff Mahoney } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 33974bbcaa64SEric Sandeen fs_info->generation != 33984bbcaa64SEric Sandeen btrfs_super_uuid_tree_generation(disk_super)) { 339905135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 340070f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 340170f80175SStefan Behrens if (ret) { 340205135f59SDavid Sterba btrfs_warn(fs_info, 340305135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 34046bccf3abSJeff Mahoney close_ctree(fs_info); 340570f80175SStefan Behrens return ret; 340670f80175SStefan Behrens } 3407f7a81ea4SStefan Behrens } 3408afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 340947ab2a6cSJosef Bacik 34108dcddfa0SQu Wenruo /* 34118dcddfa0SQu Wenruo * backuproot only affect mount behavior, and if open_ctree succeeded, 34128dcddfa0SQu Wenruo * no need to keep the flag 34138dcddfa0SQu Wenruo */ 34148dcddfa0SQu Wenruo btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 34158dcddfa0SQu Wenruo 3416ad2b2c80SAl Viro return 0; 341739279cc3SChris Mason 3418bcef60f2SArne Jansen fail_qgroup: 3419bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 34207c2ca468SChris Mason fail_trans_kthread: 34217c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 34222ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3423faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 34243f157a2fSChris Mason fail_cleaner: 3425a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 34267c2ca468SChris Mason 34277c2ca468SChris Mason /* 34287c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 34297c2ca468SChris Mason * kthreads 34307c2ca468SChris Mason */ 34317c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 34327c2ca468SChris Mason 34332365dd3cSAnand Jain fail_sysfs: 34346618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 34352365dd3cSAnand Jain 3436b7c35e81SAnand Jain fail_fsdev_sysfs: 3437b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3438b7c35e81SAnand Jain 34391b1d1f66SJosef Bacik fail_block_groups: 344054067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3441af31f5e5SChris Mason 3442af31f5e5SChris Mason fail_tree_roots: 34439e3aa805SJosef Bacik if (fs_info->data_reloc_root) 34449e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 34454273eaffSAnand Jain free_root_pointers(fs_info, true); 34462b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3447af31f5e5SChris Mason 344839279cc3SChris Mason fail_sb_buffer: 34497abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 34505cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 345116cdcec7SMiao Xie fail_alloc: 3452586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3453586e46e2SIlya Dryomov 34544543df7eSChris Mason iput(fs_info->btree_inode); 34557e662854SQinghuang Feng fail: 3456586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3457ad2b2c80SAl Viro return err; 3458eb60ceacSChris Mason } 3459663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3460eb60ceacSChris Mason 3461314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3462f2984462SChris Mason { 3463314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3464314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3465314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3466314b6dd0SJohannes Thumshirn struct page *page; 3467442a4f63SStefan Behrens 3468314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3469314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3470314b6dd0SJohannes Thumshirn 3471314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3472fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3473314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3474314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3475314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3476314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3477314b6dd0SJohannes Thumshirn SetPageError(page); 3478314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3479314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3480314b6dd0SJohannes Thumshirn } else { 3481314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3482f2984462SChris Mason } 3483314b6dd0SJohannes Thumshirn 3484314b6dd0SJohannes Thumshirn put_page(page); 3485314b6dd0SJohannes Thumshirn unlock_page(page); 3486314b6dd0SJohannes Thumshirn } 3487314b6dd0SJohannes Thumshirn 3488314b6dd0SJohannes Thumshirn bio_put(bio); 3489f2984462SChris Mason } 3490f2984462SChris Mason 34918f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 34928f32380dSJohannes Thumshirn int copy_num) 349329c36d72SAnand Jain { 349429c36d72SAnand Jain struct btrfs_super_block *super; 34958f32380dSJohannes Thumshirn struct page *page; 349629c36d72SAnand Jain u64 bytenr; 34978f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 349829c36d72SAnand Jain 349929c36d72SAnand Jain bytenr = btrfs_sb_offset(copy_num); 350029c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 35018f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 350229c36d72SAnand Jain 35038f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 35048f32380dSJohannes Thumshirn if (IS_ERR(page)) 35058f32380dSJohannes Thumshirn return ERR_CAST(page); 350629c36d72SAnand Jain 35078f32380dSJohannes Thumshirn super = page_address(page); 350829c36d72SAnand Jain if (btrfs_super_bytenr(super) != bytenr || 350929c36d72SAnand Jain btrfs_super_magic(super) != BTRFS_MAGIC) { 35108f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 35118f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 351229c36d72SAnand Jain } 351329c36d72SAnand Jain 35148f32380dSJohannes Thumshirn return super; 351529c36d72SAnand Jain } 351629c36d72SAnand Jain 351729c36d72SAnand Jain 35188f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3519a512bbf8SYan Zheng { 35208f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3521a512bbf8SYan Zheng int i; 3522a512bbf8SYan Zheng u64 transid = 0; 3523a512bbf8SYan Zheng 3524a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3525a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3526a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3527a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3528a512bbf8SYan Zheng */ 3529a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 35308f32380dSJohannes Thumshirn super = btrfs_read_dev_one_super(bdev, i); 35318f32380dSJohannes Thumshirn if (IS_ERR(super)) 3532a512bbf8SYan Zheng continue; 3533a512bbf8SYan Zheng 3534a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 35358f32380dSJohannes Thumshirn if (latest) 35368f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 35378f32380dSJohannes Thumshirn 35388f32380dSJohannes Thumshirn latest = super; 3539a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3540a512bbf8SYan Zheng } 3541a512bbf8SYan Zheng } 354292fc03fbSAnand Jain 35438f32380dSJohannes Thumshirn return super; 3544a512bbf8SYan Zheng } 3545a512bbf8SYan Zheng 35464eedeb75SHisashi Hifumi /* 3547abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3548314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 35494eedeb75SHisashi Hifumi * 3550abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3551abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3552abbb3b8eSDavid Sterba * same for write and wait phases. 35534eedeb75SHisashi Hifumi * 3554314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 35554eedeb75SHisashi Hifumi */ 3556a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3557abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3558a512bbf8SYan Zheng { 3559d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3560314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3561d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3562a512bbf8SYan Zheng int i; 3563a512bbf8SYan Zheng int errors = 0; 3564a512bbf8SYan Zheng u64 bytenr; 3565a512bbf8SYan Zheng 3566a512bbf8SYan Zheng if (max_mirrors == 0) 3567a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3568a512bbf8SYan Zheng 3569d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3570d5178578SJohannes Thumshirn 3571a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3572314b6dd0SJohannes Thumshirn struct page *page; 3573314b6dd0SJohannes Thumshirn struct bio *bio; 3574314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3575314b6dd0SJohannes Thumshirn 3576a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3577935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3578935e5cc9SMiao Xie device->commit_total_bytes) 3579a512bbf8SYan Zheng break; 3580a512bbf8SYan Zheng 3581a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3582a512bbf8SYan Zheng 3583fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3584fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3585fd08001fSEric Biggers sb->csum); 3586a512bbf8SYan Zheng 3587314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3588314b6dd0SJohannes Thumshirn GFP_NOFS); 3589314b6dd0SJohannes Thumshirn if (!page) { 3590fb456252SJeff Mahoney btrfs_err(device->fs_info, 3591314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3592f14d104dSDavid Sterba bytenr); 3593634554dcSJosef Bacik errors++; 3594634554dcSJosef Bacik continue; 3595634554dcSJosef Bacik } 3596634554dcSJosef Bacik 3597314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3598314b6dd0SJohannes Thumshirn get_page(page); 3599a512bbf8SYan Zheng 3600314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3601314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3602a512bbf8SYan Zheng 3603387125fcSChris Mason /* 3604314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3605314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3606314b6dd0SJohannes Thumshirn * checking. 3607387125fcSChris Mason */ 3608314b6dd0SJohannes Thumshirn bio = bio_alloc(GFP_NOFS, 1); 3609314b6dd0SJohannes Thumshirn bio_set_dev(bio, device->bdev); 3610314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3611314b6dd0SJohannes Thumshirn bio->bi_private = device; 3612314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3613314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3614314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3615314b6dd0SJohannes Thumshirn 3616314b6dd0SJohannes Thumshirn /* 3617314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3618314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3619314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3620314b6dd0SJohannes Thumshirn */ 3621314b6dd0SJohannes Thumshirn bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; 36221b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3623314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3624314b6dd0SJohannes Thumshirn 3625314b6dd0SJohannes Thumshirn btrfsic_submit_bio(bio); 3626a512bbf8SYan Zheng } 3627a512bbf8SYan Zheng return errors < i ? 0 : -1; 3628a512bbf8SYan Zheng } 3629a512bbf8SYan Zheng 3630387125fcSChris Mason /* 3631abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3632abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3633abbb3b8eSDavid Sterba * 3634314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3635abbb3b8eSDavid Sterba * date. 3636abbb3b8eSDavid Sterba */ 3637abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3638abbb3b8eSDavid Sterba { 3639abbb3b8eSDavid Sterba int i; 3640abbb3b8eSDavid Sterba int errors = 0; 3641b6a535faSHoward McLauchlan bool primary_failed = false; 3642abbb3b8eSDavid Sterba u64 bytenr; 3643abbb3b8eSDavid Sterba 3644abbb3b8eSDavid Sterba if (max_mirrors == 0) 3645abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3646abbb3b8eSDavid Sterba 3647abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3648314b6dd0SJohannes Thumshirn struct page *page; 3649314b6dd0SJohannes Thumshirn 3650abbb3b8eSDavid Sterba bytenr = btrfs_sb_offset(i); 3651abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3652abbb3b8eSDavid Sterba device->commit_total_bytes) 3653abbb3b8eSDavid Sterba break; 3654abbb3b8eSDavid Sterba 3655314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3656314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3657314b6dd0SJohannes Thumshirn if (!page) { 3658abbb3b8eSDavid Sterba errors++; 3659b6a535faSHoward McLauchlan if (i == 0) 3660b6a535faSHoward McLauchlan primary_failed = true; 3661abbb3b8eSDavid Sterba continue; 3662abbb3b8eSDavid Sterba } 3663314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3664314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3665314b6dd0SJohannes Thumshirn if (PageError(page)) { 3666abbb3b8eSDavid Sterba errors++; 3667b6a535faSHoward McLauchlan if (i == 0) 3668b6a535faSHoward McLauchlan primary_failed = true; 3669b6a535faSHoward McLauchlan } 3670abbb3b8eSDavid Sterba 3671314b6dd0SJohannes Thumshirn /* Drop our reference */ 3672314b6dd0SJohannes Thumshirn put_page(page); 3673abbb3b8eSDavid Sterba 3674314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3675314b6dd0SJohannes Thumshirn put_page(page); 3676abbb3b8eSDavid Sterba } 3677abbb3b8eSDavid Sterba 3678b6a535faSHoward McLauchlan /* log error, force error return */ 3679b6a535faSHoward McLauchlan if (primary_failed) { 3680b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3681b6a535faSHoward McLauchlan device->devid); 3682b6a535faSHoward McLauchlan return -1; 3683b6a535faSHoward McLauchlan } 3684b6a535faSHoward McLauchlan 3685abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3686abbb3b8eSDavid Sterba } 3687abbb3b8eSDavid Sterba 3688abbb3b8eSDavid Sterba /* 3689387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3690387125fcSChris Mason * for the barrier when it is done 3691387125fcSChris Mason */ 36924246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3693387125fcSChris Mason { 3694387125fcSChris Mason complete(bio->bi_private); 3695387125fcSChris Mason } 3696387125fcSChris Mason 3697387125fcSChris Mason /* 36984fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 36994fc6441aSAnand Jain * done in the waiting counterpart. 3700387125fcSChris Mason */ 37014fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3702387125fcSChris Mason { 3703c2a9c7abSAnand Jain struct request_queue *q = bdev_get_queue(device->bdev); 3704e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3705387125fcSChris Mason 3706c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 37074fc6441aSAnand Jain return; 3708387125fcSChris Mason 3709e0ae9994SDavid Sterba bio_reset(bio); 3710387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 371174d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 37128d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 3713387125fcSChris Mason init_completion(&device->flush_wait); 3714387125fcSChris Mason bio->bi_private = &device->flush_wait; 3715387125fcSChris Mason 371643a01111SLu Fengqi btrfsic_submit_bio(bio); 37171c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 37184fc6441aSAnand Jain } 3719387125fcSChris Mason 37204fc6441aSAnand Jain /* 37214fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 37224fc6441aSAnand Jain */ 37238c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 37244fc6441aSAnand Jain { 37254fc6441aSAnand Jain struct bio *bio = device->flush_bio; 37264fc6441aSAnand Jain 37271c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 372858efbc9fSOmar Sandoval return BLK_STS_OK; 37294fc6441aSAnand Jain 37301c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 37312980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 37324fc6441aSAnand Jain 37338c27cb35SLinus Torvalds return bio->bi_status; 3734387125fcSChris Mason } 3735387125fcSChris Mason 3736d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 3737401b41e5SAnand Jain { 37386528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 3739401b41e5SAnand Jain return -EIO; 3740387125fcSChris Mason return 0; 3741387125fcSChris Mason } 3742387125fcSChris Mason 3743387125fcSChris Mason /* 3744387125fcSChris Mason * send an empty flush down to each device in parallel, 3745387125fcSChris Mason * then wait for them 3746387125fcSChris Mason */ 3747387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3748387125fcSChris Mason { 3749387125fcSChris Mason struct list_head *head; 3750387125fcSChris Mason struct btrfs_device *dev; 37515af3e8ccSStefan Behrens int errors_wait = 0; 37524e4cbee9SChristoph Hellwig blk_status_t ret; 3753387125fcSChris Mason 37541538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3755387125fcSChris Mason /* send down all the barriers */ 3756387125fcSChris Mason head = &info->fs_devices->devices; 37571538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3758e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3759f88ba6a2SHidetoshi Seto continue; 3760cea7c8bfSAnand Jain if (!dev->bdev) 3761387125fcSChris Mason continue; 3762e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3763ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3764387125fcSChris Mason continue; 3765387125fcSChris Mason 37664fc6441aSAnand Jain write_dev_flush(dev); 376758efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 3768387125fcSChris Mason } 3769387125fcSChris Mason 3770387125fcSChris Mason /* wait for all the barriers */ 37711538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3772e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3773f88ba6a2SHidetoshi Seto continue; 3774387125fcSChris Mason if (!dev->bdev) { 37755af3e8ccSStefan Behrens errors_wait++; 3776387125fcSChris Mason continue; 3777387125fcSChris Mason } 3778e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3779ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3780387125fcSChris Mason continue; 3781387125fcSChris Mason 37824fc6441aSAnand Jain ret = wait_dev_flush(dev); 3783401b41e5SAnand Jain if (ret) { 3784401b41e5SAnand Jain dev->last_flush_error = ret; 378566b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 378666b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 37875af3e8ccSStefan Behrens errors_wait++; 3788387125fcSChris Mason } 3789401b41e5SAnand Jain } 3790401b41e5SAnand Jain 3791cea7c8bfSAnand Jain if (errors_wait) { 3792401b41e5SAnand Jain /* 3793401b41e5SAnand Jain * At some point we need the status of all disks 3794401b41e5SAnand Jain * to arrive at the volume status. So error checking 3795401b41e5SAnand Jain * is being pushed to a separate loop. 3796401b41e5SAnand Jain */ 3797d10b82feSQu Wenruo return check_barrier_error(info); 3798401b41e5SAnand Jain } 3799387125fcSChris Mason return 0; 3800387125fcSChris Mason } 3801387125fcSChris Mason 3802943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3803943c6e99SZhao Lei { 38048789f4feSZhao Lei int raid_type; 38058789f4feSZhao Lei int min_tolerated = INT_MAX; 3806943c6e99SZhao Lei 38078789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 38088789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 38098c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 38108789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 38118789f4feSZhao Lei tolerated_failures); 3812943c6e99SZhao Lei 38138789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 38148789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 38158789f4feSZhao Lei continue; 381641a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 38178789f4feSZhao Lei continue; 38188c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 38198789f4feSZhao Lei btrfs_raid_array[raid_type]. 38208789f4feSZhao Lei tolerated_failures); 38218789f4feSZhao Lei } 3822943c6e99SZhao Lei 38238789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3824ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 38258789f4feSZhao Lei min_tolerated = 0; 38268789f4feSZhao Lei } 38278789f4feSZhao Lei 38288789f4feSZhao Lei return min_tolerated; 3829943c6e99SZhao Lei } 3830943c6e99SZhao Lei 3831eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3832f2984462SChris Mason { 3833e5e9a520SChris Mason struct list_head *head; 3834f2984462SChris Mason struct btrfs_device *dev; 3835a061fc8dSChris Mason struct btrfs_super_block *sb; 3836f2984462SChris Mason struct btrfs_dev_item *dev_item; 3837f2984462SChris Mason int ret; 3838f2984462SChris Mason int do_barriers; 3839a236aed1SChris Mason int max_errors; 3840a236aed1SChris Mason int total_errors = 0; 3841a061fc8dSChris Mason u64 flags; 3842f2984462SChris Mason 38430b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 3844fed3b381SLiu Bo 3845fed3b381SLiu Bo /* 3846fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 3847fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 3848fed3b381SLiu Bo * been consistent on disk. 3849fed3b381SLiu Bo */ 3850fed3b381SLiu Bo if (max_mirrors == 0) 38510b246afaSJeff Mahoney backup_super_roots(fs_info); 3852f2984462SChris Mason 38530b246afaSJeff Mahoney sb = fs_info->super_for_commit; 3854a061fc8dSChris Mason dev_item = &sb->dev_item; 3855e5e9a520SChris Mason 38560b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 38570b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 38580b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 3859387125fcSChris Mason 38605af3e8ccSStefan Behrens if (do_barriers) { 38610b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 38625af3e8ccSStefan Behrens if (ret) { 38635af3e8ccSStefan Behrens mutex_unlock( 38640b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 38650b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 38665af3e8ccSStefan Behrens "errors while submitting device barriers."); 38675af3e8ccSStefan Behrens return ret; 38685af3e8ccSStefan Behrens } 38695af3e8ccSStefan Behrens } 3870387125fcSChris Mason 38711538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3872dfe25020SChris Mason if (!dev->bdev) { 3873dfe25020SChris Mason total_errors++; 3874dfe25020SChris Mason continue; 3875dfe25020SChris Mason } 3876e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3877ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3878dfe25020SChris Mason continue; 3879dfe25020SChris Mason 38802b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3881a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3882a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 38837df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3884935e5cc9SMiao Xie dev->commit_total_bytes); 3885ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3886ce7213c7SMiao Xie dev->commit_bytes_used); 3887a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3888a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3889a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3890a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 38917239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 38927239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 3893a512bbf8SYan Zheng 3894a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3895a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3896f2984462SChris Mason 389775cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 389875cb857dSQu Wenruo if (ret < 0) { 389975cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 390075cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 390175cb857dSQu Wenruo "unexpected superblock corruption detected"); 390275cb857dSQu Wenruo return -EUCLEAN; 390375cb857dSQu Wenruo } 390475cb857dSQu Wenruo 3905abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 3906a236aed1SChris Mason if (ret) 3907a236aed1SChris Mason total_errors++; 3908f2984462SChris Mason } 3909a236aed1SChris Mason if (total_errors > max_errors) { 39100b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 3911d397712bSChris Mason total_errors); 39120b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 391379787eaaSJeff Mahoney 39149d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 39150b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 39160b246afaSJeff Mahoney "%d errors while writing supers", 39170b246afaSJeff Mahoney total_errors); 39189d565ba4SStefan Behrens return -EIO; 3919a236aed1SChris Mason } 3920f2984462SChris Mason 3921a512bbf8SYan Zheng total_errors = 0; 39221538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3923dfe25020SChris Mason if (!dev->bdev) 3924dfe25020SChris Mason continue; 3925e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3926ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3927dfe25020SChris Mason continue; 3928dfe25020SChris Mason 3929abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 3930a512bbf8SYan Zheng if (ret) 39311259ab75SChris Mason total_errors++; 3932f2984462SChris Mason } 39330b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 3934a236aed1SChris Mason if (total_errors > max_errors) { 39350b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 39360b246afaSJeff Mahoney "%d errors while writing supers", 39370b246afaSJeff Mahoney total_errors); 393879787eaaSJeff Mahoney return -EIO; 3939a236aed1SChris Mason } 3940f2984462SChris Mason return 0; 3941f2984462SChris Mason } 3942f2984462SChris Mason 3943cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3944cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3945cb517eabSMiao Xie struct btrfs_root *root) 39462619ba1fSChris Mason { 39474785e24fSJosef Bacik bool drop_ref = false; 39484785e24fSJosef Bacik 39494df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 39502619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 39512619ba1fSChris Mason (unsigned long)root->root_key.objectid); 3952af01d2e5SJosef Bacik if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 39534785e24fSJosef Bacik drop_ref = true; 39544df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 395576dda93cSYan, Zheng 39561c1ea4f7SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 3957ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 39581c1ea4f7SLiu Bo if (root->reloc_root) { 395900246528SJosef Bacik btrfs_put_root(root->reloc_root); 39601c1ea4f7SLiu Bo root->reloc_root = NULL; 39611c1ea4f7SLiu Bo } 39621c1ea4f7SLiu Bo } 39633321719eSLiu Bo 3964faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3965581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3966faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3967581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 39680e996e7fSJosef Bacik if (root->ino_cache_inode) { 39690e996e7fSJosef Bacik iput(root->ino_cache_inode); 39700e996e7fSJosef Bacik root->ino_cache_inode = NULL; 39710e996e7fSJosef Bacik } 39724785e24fSJosef Bacik if (drop_ref) 397300246528SJosef Bacik btrfs_put_root(root); 39742619ba1fSChris Mason } 39752619ba1fSChris Mason 3976c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3977c146afadSYan Zheng { 3978c146afadSYan Zheng u64 root_objectid = 0; 3979c146afadSYan Zheng struct btrfs_root *gang[8]; 398065d33fd7SQu Wenruo int i = 0; 398165d33fd7SQu Wenruo int err = 0; 398265d33fd7SQu Wenruo unsigned int ret = 0; 3983c146afadSYan Zheng 3984c146afadSYan Zheng while (1) { 3985efc34534SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 3986c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3987c146afadSYan Zheng (void **)gang, root_objectid, 3988c146afadSYan Zheng ARRAY_SIZE(gang)); 398965d33fd7SQu Wenruo if (!ret) { 3990efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 3991c146afadSYan Zheng break; 399265d33fd7SQu Wenruo } 39935d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 399466b4ffd1SJosef Bacik 399565d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 399665d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 399765d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 399865d33fd7SQu Wenruo gang[i] = NULL; 399965d33fd7SQu Wenruo continue; 400065d33fd7SQu Wenruo } 400165d33fd7SQu Wenruo /* grab all the search result for later use */ 400200246528SJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 400365d33fd7SQu Wenruo } 4004efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 400565d33fd7SQu Wenruo 400665d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 400765d33fd7SQu Wenruo if (!gang[i]) 400865d33fd7SQu Wenruo continue; 4009c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 401066b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 401166b4ffd1SJosef Bacik if (err) 401265d33fd7SQu Wenruo break; 401300246528SJosef Bacik btrfs_put_root(gang[i]); 4014c146afadSYan Zheng } 4015c146afadSYan Zheng root_objectid++; 4016c146afadSYan Zheng } 401765d33fd7SQu Wenruo 401865d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 401965d33fd7SQu Wenruo for (; i < ret; i++) { 402065d33fd7SQu Wenruo if (gang[i]) 402100246528SJosef Bacik btrfs_put_root(gang[i]); 402265d33fd7SQu Wenruo } 402365d33fd7SQu Wenruo return err; 4024c146afadSYan Zheng } 4025c146afadSYan Zheng 40266bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4027c146afadSYan Zheng { 40286bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4029c146afadSYan Zheng struct btrfs_trans_handle *trans; 4030c146afadSYan Zheng 40310b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 40322ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 40330b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 40340b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4035c71bf099SYan, Zheng 4036c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 40370b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 40380b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4039c71bf099SYan, Zheng 40407a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 40413612b495STsutomu Itoh if (IS_ERR(trans)) 40423612b495STsutomu Itoh return PTR_ERR(trans); 40433a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4044c146afadSYan Zheng } 4045c146afadSYan Zheng 4046b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4047eb60ceacSChris Mason { 4048c146afadSYan Zheng int ret; 4049e089f05cSChris Mason 4050afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 4051d6fd0ae2SOmar Sandoval /* 4052d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4053d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4054d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4055d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4056d6fd0ae2SOmar Sandoval */ 4057d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4058a2135011SChris Mason 40597343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4060d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 40617343dd61SJustin Maggard 4062803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4063803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4064803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4065803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4066803b2f54SStefan Behrens 4067837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4068aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4069837d5b6eSIlya Dryomov 40708dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 40718dabb742SStefan Behrens 4072aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 40734cb5300bSChris Mason 40744cb5300bSChris Mason /* wait for any defraggers to finish */ 40754cb5300bSChris Mason wait_event(fs_info->transaction_wait, 40764cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 40774cb5300bSChris Mason 40784cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 407926176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 40804cb5300bSChris Mason 408121c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 408221c7e756SMiao Xie 4083b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4084b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4085b0643e59SDennis Zhou 4086bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4087e44163e1SJeff Mahoney /* 4088d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4089d6fd0ae2SOmar Sandoval * unused block groups. 4090e44163e1SJeff Mahoney */ 40910b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4092e44163e1SJeff Mahoney 4093f0cc2cd7SFilipe Manana /* 4094f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4095f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4096f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4097f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4098f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4099f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4100f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4101f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4102f0cc2cd7SFilipe Manana * This is a very rare case. 4103f0cc2cd7SFilipe Manana */ 4104f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4105f0cc2cd7SFilipe Manana 41066bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4107d397712bSChris Mason if (ret) 410804892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4109c146afadSYan Zheng } 4110ed2ff2cbSChris Mason 4111af722733SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || 4112af722733SLiu Bo test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) 41132ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4114acce952bSliubo 4115e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4116e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 41178929ecfaSYan, Zheng 4118e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4119afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4120f25784b3SYan Zheng 41215958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 41225958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 41235958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 41245958253cSQu Wenruo } 41255958253cSQu Wenruo 412604892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4127fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4128bcef60f2SArne Jansen 4129963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 413004892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4131963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4132b0c68f8bSChris Mason } 413331153d81SYan Zheng 41344297ff84SJosef Bacik if (percpu_counter_sum(&fs_info->dio_bytes)) 41354297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 41364297ff84SJosef Bacik percpu_counter_sum(&fs_info->dio_bytes)); 41374297ff84SJosef Bacik 41386618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4139b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 41405ac1d209SJeff Mahoney 41411a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 41421a4319ccSLiu Bo 4143de348ee0SWang Shilong /* 4144de348ee0SWang Shilong * we must make sure there is not any read request to 4145de348ee0SWang Shilong * submit after we stopping all workers. 4146de348ee0SWang Shilong */ 4147de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 414896192499SJosef Bacik btrfs_stop_all_workers(fs_info); 414996192499SJosef Bacik 4150afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 41514273eaffSAnand Jain free_root_pointers(fs_info, true); 41528c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 41539ad6b7bcSChris Mason 41544e19443dSJosef Bacik /* 41554e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 41564e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 41574e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 41584e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 41594e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 41604e19443dSJosef Bacik */ 41614e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 41624e19443dSJosef Bacik 416313e6c37bSJosef Bacik iput(fs_info->btree_inode); 4164d6bfde87SChris Mason 416521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 41660b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 41672ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 416821adbd5cSStefan Behrens #endif 416921adbd5cSStefan Behrens 41700b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 417168c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4172eb60ceacSChris Mason } 4173eb60ceacSChris Mason 4174b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4175b9fab919SChris Mason int atomic) 4176ccd467d6SChris Mason { 41771259ab75SChris Mason int ret; 4178727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 41791259ab75SChris Mason 41800b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 41811259ab75SChris Mason if (!ret) 41821259ab75SChris Mason return ret; 41831259ab75SChris Mason 41841259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4185b9fab919SChris Mason parent_transid, atomic); 4186b9fab919SChris Mason if (ret == -EAGAIN) 4187b9fab919SChris Mason return ret; 41881259ab75SChris Mason return !ret; 41895f39d397SChris Mason } 41906702ed49SChris Mason 41915f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 41925f39d397SChris Mason { 41930b246afaSJeff Mahoney struct btrfs_fs_info *fs_info; 419406ea65a3SJosef Bacik struct btrfs_root *root; 41955f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4196b9473439SChris Mason int was_dirty; 4197b4ce94deSChris Mason 419806ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 419906ea65a3SJosef Bacik /* 420006ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 420152042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 420206ea65a3SJosef Bacik * outside of the sanity tests. 420306ea65a3SJosef Bacik */ 4204b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 420506ea65a3SJosef Bacik return; 420606ea65a3SJosef Bacik #endif 420706ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 42080b246afaSJeff Mahoney fs_info = root->fs_info; 4209b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 42100b246afaSJeff Mahoney if (transid != fs_info->generation) 42115d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 42120b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 42130b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4214e2d84521SMiao Xie if (!was_dirty) 4215104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4216e2d84521SMiao Xie buf->len, 42170b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 42181f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 421969fc6cbbSQu Wenruo /* 422069fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 422169fc6cbbSQu Wenruo * but item data not updated. 422269fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 422369fc6cbbSQu Wenruo */ 422469fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4225cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4226a4f78750SDavid Sterba btrfs_print_leaf(buf); 42271f21ef0aSFilipe Manana ASSERT(0); 42281f21ef0aSFilipe Manana } 42291f21ef0aSFilipe Manana #endif 4230eb60ceacSChris Mason } 4231eb60ceacSChris Mason 42322ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4233b53d3f5dSLiu Bo int flush_delayed) 423435b7e476SChris Mason { 4235188de649SChris Mason /* 4236188de649SChris Mason * looks as though older kernels can get into trouble with 4237188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4238188de649SChris Mason */ 4239e2d84521SMiao Xie int ret; 4240d6bfde87SChris Mason 42416933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4242d6bfde87SChris Mason return; 4243d6bfde87SChris Mason 4244b53d3f5dSLiu Bo if (flush_delayed) 42452ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 424616cdcec7SMiao Xie 4247d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4248d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4249d814a491SEthan Lien fs_info->dirty_metadata_batch); 4250e2d84521SMiao Xie if (ret > 0) { 42510b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 425216cdcec7SMiao Xie } 425316cdcec7SMiao Xie } 425416cdcec7SMiao Xie 42552ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 425616cdcec7SMiao Xie { 42572ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 425835b7e476SChris Mason } 4259b53d3f5dSLiu Bo 42602ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4261b53d3f5dSLiu Bo { 42622ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4263d6bfde87SChris Mason } 42646b80053dSChris Mason 4265581c1760SQu Wenruo int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, 4266581c1760SQu Wenruo struct btrfs_key *first_key) 42676b80053dSChris Mason { 42685ab12d1fSDavid Sterba return btree_read_extent_buffer_pages(buf, parent_transid, 4269581c1760SQu Wenruo level, first_key); 42706b80053dSChris Mason } 42710da5468fSChris Mason 42722ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4273acce952bSliubo { 4274fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4275fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4276fe816d0fSNikolay Borisov 42770b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 42782ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 42790b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4280acce952bSliubo 42810b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 42820b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4283acce952bSliubo } 4284acce952bSliubo 4285ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4286ef67963dSJosef Bacik { 4287ef67963dSJosef Bacik struct btrfs_root *gang[8]; 4288ef67963dSJosef Bacik u64 root_objectid = 0; 4289ef67963dSJosef Bacik int ret; 4290ef67963dSJosef Bacik 4291ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4292ef67963dSJosef Bacik while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4293ef67963dSJosef Bacik (void **)gang, root_objectid, 4294ef67963dSJosef Bacik ARRAY_SIZE(gang))) != 0) { 4295ef67963dSJosef Bacik int i; 4296ef67963dSJosef Bacik 4297ef67963dSJosef Bacik for (i = 0; i < ret; i++) 4298ef67963dSJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 4299ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4300ef67963dSJosef Bacik 4301ef67963dSJosef Bacik for (i = 0; i < ret; i++) { 4302ef67963dSJosef Bacik if (!gang[i]) 4303ef67963dSJosef Bacik continue; 4304ef67963dSJosef Bacik root_objectid = gang[i]->root_key.objectid; 4305ef67963dSJosef Bacik btrfs_free_log(NULL, gang[i]); 4306ef67963dSJosef Bacik btrfs_put_root(gang[i]); 4307ef67963dSJosef Bacik } 4308ef67963dSJosef Bacik root_objectid++; 4309ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4310ef67963dSJosef Bacik } 4311ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4312ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4313ef67963dSJosef Bacik } 4314ef67963dSJosef Bacik 4315143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4316acce952bSliubo { 4317acce952bSliubo struct btrfs_ordered_extent *ordered; 4318acce952bSliubo 4319199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4320779880efSJosef Bacik /* 4321779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4322779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4323779880efSJosef Bacik */ 4324199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4325779880efSJosef Bacik root_extent_list) 4326779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4327199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4328199c2a9cSMiao Xie } 4329199c2a9cSMiao Xie 4330199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4331199c2a9cSMiao Xie { 4332199c2a9cSMiao Xie struct btrfs_root *root; 4333199c2a9cSMiao Xie struct list_head splice; 4334199c2a9cSMiao Xie 4335199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4336199c2a9cSMiao Xie 4337199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4338199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4339199c2a9cSMiao Xie while (!list_empty(&splice)) { 4340199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4341199c2a9cSMiao Xie ordered_root); 43421de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 43431de2cfdeSJosef Bacik &fs_info->ordered_roots); 4344199c2a9cSMiao Xie 43452a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4346199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4347199c2a9cSMiao Xie 43482a85d9caSLiu Bo cond_resched(); 43492a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4350199c2a9cSMiao Xie } 4351199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 435274d5d229SJosef Bacik 435374d5d229SJosef Bacik /* 435474d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 435574d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 435674d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 435774d5d229SJosef Bacik * actually get run and error out properly. 435874d5d229SJosef Bacik */ 435974d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4360acce952bSliubo } 4361acce952bSliubo 436235a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 43632ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4364acce952bSliubo { 4365acce952bSliubo struct rb_node *node; 4366acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4367acce952bSliubo struct btrfs_delayed_ref_node *ref; 4368acce952bSliubo int ret = 0; 4369acce952bSliubo 4370acce952bSliubo delayed_refs = &trans->delayed_refs; 4371acce952bSliubo 4372acce952bSliubo spin_lock(&delayed_refs->lock); 4373d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4374cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4375b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4376acce952bSliubo return ret; 4377acce952bSliubo } 4378acce952bSliubo 43795c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4380d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 43810e0adbcfSJosef Bacik struct rb_node *n; 4382e78417d1SJosef Bacik bool pin_bytes = false; 4383acce952bSliubo 4384d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4385d7df2c79SJosef Bacik href_node); 43863069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4387b939d1abSJosef Bacik continue; 43883069bd26SJosef Bacik 4389d7df2c79SJosef Bacik spin_lock(&head->lock); 4390e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 43910e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 43920e0adbcfSJosef Bacik ref_node); 4393d7df2c79SJosef Bacik ref->in_tree = 0; 4394e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 43950e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 43961d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 43971d57ee94SWang Xiaoguang list_del(&ref->add_list); 4398d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4399d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4400d7df2c79SJosef Bacik } 440154067ae9SJosef Bacik if (head->must_insert_reserved) 4402e78417d1SJosef Bacik pin_bytes = true; 440378a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4404fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4405d7df2c79SJosef Bacik spin_unlock(&head->lock); 4406acce952bSliubo spin_unlock(&delayed_refs->lock); 4407e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4408acce952bSliubo 4409f603bb94SNikolay Borisov if (pin_bytes) { 4410f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4411f603bb94SNikolay Borisov 4412f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4413f603bb94SNikolay Borisov BUG_ON(!cache); 4414f603bb94SNikolay Borisov 4415f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4416f603bb94SNikolay Borisov spin_lock(&cache->lock); 4417f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4418f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4419f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4420f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4421f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4422f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4423f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4424f603bb94SNikolay Borisov percpu_counter_add_batch( 4425f603bb94SNikolay Borisov &cache->space_info->total_bytes_pinned, 4426f603bb94SNikolay Borisov head->num_bytes, BTRFS_TOTAL_BYTES_PINNED_BATCH); 4427f603bb94SNikolay Borisov 4428f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4429f603bb94SNikolay Borisov 4430f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4431f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4432f603bb94SNikolay Borisov } 443331890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4434d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4435acce952bSliubo cond_resched(); 4436acce952bSliubo spin_lock(&delayed_refs->lock); 4437acce952bSliubo } 443881f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4439acce952bSliubo 4440acce952bSliubo spin_unlock(&delayed_refs->lock); 4441acce952bSliubo 4442acce952bSliubo return ret; 4443acce952bSliubo } 4444acce952bSliubo 4445143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4446acce952bSliubo { 4447acce952bSliubo struct btrfs_inode *btrfs_inode; 4448acce952bSliubo struct list_head splice; 4449acce952bSliubo 4450acce952bSliubo INIT_LIST_HEAD(&splice); 4451acce952bSliubo 4452eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4453eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4454acce952bSliubo 4455acce952bSliubo while (!list_empty(&splice)) { 4456fe816d0fSNikolay Borisov struct inode *inode = NULL; 4457eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4458acce952bSliubo delalloc_inodes); 4459fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4460eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4461acce952bSliubo 4462fe816d0fSNikolay Borisov /* 4463fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4464fe816d0fSNikolay Borisov * meanwhile. 4465fe816d0fSNikolay Borisov */ 4466fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4467fe816d0fSNikolay Borisov if (inode) { 4468fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4469fe816d0fSNikolay Borisov iput(inode); 4470fe816d0fSNikolay Borisov } 4471eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4472acce952bSliubo } 4473eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4474eb73c1b7SMiao Xie } 4475eb73c1b7SMiao Xie 4476eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4477eb73c1b7SMiao Xie { 4478eb73c1b7SMiao Xie struct btrfs_root *root; 4479eb73c1b7SMiao Xie struct list_head splice; 4480eb73c1b7SMiao Xie 4481eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4482eb73c1b7SMiao Xie 4483eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4484eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4485eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4486eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4487eb73c1b7SMiao Xie delalloc_root); 448800246528SJosef Bacik root = btrfs_grab_root(root); 4489eb73c1b7SMiao Xie BUG_ON(!root); 4490eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4491eb73c1b7SMiao Xie 4492eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 449300246528SJosef Bacik btrfs_put_root(root); 4494eb73c1b7SMiao Xie 4495eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4496eb73c1b7SMiao Xie } 4497eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4498acce952bSliubo } 4499acce952bSliubo 45002ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4501acce952bSliubo struct extent_io_tree *dirty_pages, 4502acce952bSliubo int mark) 4503acce952bSliubo { 4504acce952bSliubo int ret; 4505acce952bSliubo struct extent_buffer *eb; 4506acce952bSliubo u64 start = 0; 4507acce952bSliubo u64 end; 4508acce952bSliubo 4509acce952bSliubo while (1) { 4510acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4511e6138876SJosef Bacik mark, NULL); 4512acce952bSliubo if (ret) 4513acce952bSliubo break; 4514acce952bSliubo 451591166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4516acce952bSliubo while (start <= end) { 45170b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 45180b246afaSJeff Mahoney start += fs_info->nodesize; 4519fd8b2b61SJosef Bacik if (!eb) 4520acce952bSliubo continue; 4521fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4522acce952bSliubo 4523fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4524fd8b2b61SJosef Bacik &eb->bflags)) 4525fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4526fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4527acce952bSliubo } 4528acce952bSliubo } 4529acce952bSliubo 4530acce952bSliubo return ret; 4531acce952bSliubo } 4532acce952bSliubo 45332ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4534fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4535acce952bSliubo { 4536acce952bSliubo u64 start; 4537acce952bSliubo u64 end; 4538acce952bSliubo int ret; 4539acce952bSliubo 4540acce952bSliubo while (1) { 45410e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 45420e6ec385SFilipe Manana 4543fcd5e742SLu Fengqi /* 4544fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4545fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4546fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4547fcd5e742SLu Fengqi * the same extent range. 4548fcd5e742SLu Fengqi */ 4549fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4550acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 45510e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4552fcd5e742SLu Fengqi if (ret) { 4553fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4554acce952bSliubo break; 4555fcd5e742SLu Fengqi } 4556acce952bSliubo 45570e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 45580e6ec385SFilipe Manana free_extent_state(cached_state); 45592ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4560fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4561acce952bSliubo cond_resched(); 4562acce952bSliubo } 4563acce952bSliubo 4564acce952bSliubo return 0; 4565acce952bSliubo } 4566acce952bSliubo 456732da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4568c79a1751SLiu Bo { 4569c79a1751SLiu Bo struct inode *inode; 4570c79a1751SLiu Bo 4571c79a1751SLiu Bo inode = cache->io_ctl.inode; 4572c79a1751SLiu Bo if (inode) { 4573c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4574c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4575c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4576c79a1751SLiu Bo iput(inode); 4577c79a1751SLiu Bo } 4578bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4579c79a1751SLiu Bo btrfs_put_block_group(cache); 4580c79a1751SLiu Bo } 4581c79a1751SLiu Bo 4582c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 45832ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4584c79a1751SLiu Bo { 458532da5386SDavid Sterba struct btrfs_block_group *cache; 4586c79a1751SLiu Bo 4587c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4588c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4589c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 459032da5386SDavid Sterba struct btrfs_block_group, 4591c79a1751SLiu Bo dirty_list); 4592c79a1751SLiu Bo 4593c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4594c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4595c79a1751SLiu Bo list_del_init(&cache->io_list); 4596c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4597c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4598c79a1751SLiu Bo } 4599c79a1751SLiu Bo 4600c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4601c79a1751SLiu Bo spin_lock(&cache->lock); 4602c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4603c79a1751SLiu Bo spin_unlock(&cache->lock); 4604c79a1751SLiu Bo 4605c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4606c79a1751SLiu Bo btrfs_put_block_group(cache); 4607ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4608c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4609c79a1751SLiu Bo } 4610c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4611c79a1751SLiu Bo 461245ae2c18SNikolay Borisov /* 461345ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 461445ae2c18SNikolay Borisov * to use it without any locking 461545ae2c18SNikolay Borisov */ 4616c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4617c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 461832da5386SDavid Sterba struct btrfs_block_group, 4619c79a1751SLiu Bo io_list); 4620c79a1751SLiu Bo 4621c79a1751SLiu Bo list_del_init(&cache->io_list); 4622c79a1751SLiu Bo spin_lock(&cache->lock); 4623c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4624c79a1751SLiu Bo spin_unlock(&cache->lock); 4625c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4626c79a1751SLiu Bo } 4627c79a1751SLiu Bo } 4628c79a1751SLiu Bo 462949b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 46302ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 463149b25e05SJeff Mahoney { 4632bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4633bbbf7243SNikolay Borisov 46342ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4635c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4636c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4637c79a1751SLiu Bo 4638bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4639bbbf7243SNikolay Borisov post_commit_list) { 4640bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4641bbbf7243SNikolay Borisov } 4642bbbf7243SNikolay Borisov 46432ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 464449b25e05SJeff Mahoney 46454a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 46460b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 464749b25e05SJeff Mahoney 46484a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 46490b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 465049b25e05SJeff Mahoney 4651ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 465267cde344SMiao Xie 46532ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 465449b25e05SJeff Mahoney EXTENT_DIRTY); 4655fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 465649b25e05SJeff Mahoney 46574a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 46584a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 465949b25e05SJeff Mahoney } 466049b25e05SJeff Mahoney 46612ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4662acce952bSliubo { 4663acce952bSliubo struct btrfs_transaction *t; 4664acce952bSliubo 46650b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4666acce952bSliubo 46670b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 46680b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 46690b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4670724e2315SJosef Bacik struct btrfs_transaction, list); 4671724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 46729b64f57dSElena Reshetova refcount_inc(&t->use_count); 46730b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 46742ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4675724e2315SJosef Bacik btrfs_put_transaction(t); 46760b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4677724e2315SJosef Bacik continue; 4678724e2315SJosef Bacik } 46790b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4680724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 46810b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4682724e2315SJosef Bacik /* 4683724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4684724e2315SJosef Bacik * handle open currently for this transaction. 4685724e2315SJosef Bacik */ 4686724e2315SJosef Bacik wait_event(t->writer_wait, 4687724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4688724e2315SJosef Bacik } else { 46890b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4690724e2315SJosef Bacik } 46912ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4692724e2315SJosef Bacik 46930b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 46940b246afaSJeff Mahoney if (t == fs_info->running_transaction) 46950b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4696724e2315SJosef Bacik list_del_init(&t->list); 46970b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4698a4abeea4SJosef Bacik 4699724e2315SJosef Bacik btrfs_put_transaction(t); 47002ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 47010b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4702724e2315SJosef Bacik } 47030b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 47040b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4705ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4706ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 47070b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4708ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 47090b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4710acce952bSliubo 4711acce952bSliubo return 0; 4712acce952bSliubo } 4713acce952bSliubo 4714e8c9f186SDavid Sterba static const struct extent_io_ops btree_extent_io_ops = { 47154d53dddbSDavid Sterba /* mandatory callbacks */ 47160b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 47174d53dddbSDavid Sterba .readpage_end_io_hook = btree_readpage_end_io_hook, 47180da5468fSChris Mason }; 4719