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 538b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 54143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 55acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 562ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 57143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 582ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 59acce952bSliubo struct extent_io_tree *dirty_pages, 60acce952bSliubo int mark); 612ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 62acce952bSliubo struct extent_io_tree *pinned_extents); 632ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 642ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 65ce9adaa5SChris Mason 66d352ac68SChris Mason /* 6797eb6b69SDavid Sterba * btrfs_end_io_wq structs are used to do processing in task context when an IO 6897eb6b69SDavid Sterba * is complete. This is used during reads to verify checksums, and it is used 69d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 70d352ac68SChris Mason */ 7197eb6b69SDavid Sterba struct btrfs_end_io_wq { 72ce9adaa5SChris Mason struct bio *bio; 73ce9adaa5SChris Mason bio_end_io_t *end_io; 74ce9adaa5SChris Mason void *private; 75ce9adaa5SChris Mason struct btrfs_fs_info *info; 764e4cbee9SChristoph Hellwig blk_status_t status; 77bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata; 788b712842SChris Mason struct btrfs_work work; 79ce9adaa5SChris Mason }; 800da5468fSChris Mason 8197eb6b69SDavid Sterba static struct kmem_cache *btrfs_end_io_wq_cache; 8297eb6b69SDavid Sterba 8397eb6b69SDavid Sterba int __init btrfs_end_io_wq_init(void) 8497eb6b69SDavid Sterba { 8597eb6b69SDavid Sterba btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", 8697eb6b69SDavid Sterba sizeof(struct btrfs_end_io_wq), 8797eb6b69SDavid Sterba 0, 88fba4b697SNikolay Borisov SLAB_MEM_SPREAD, 8997eb6b69SDavid Sterba NULL); 9097eb6b69SDavid Sterba if (!btrfs_end_io_wq_cache) 9197eb6b69SDavid Sterba return -ENOMEM; 9297eb6b69SDavid Sterba return 0; 9397eb6b69SDavid Sterba } 9497eb6b69SDavid Sterba 95e67c718bSDavid Sterba void __cold btrfs_end_io_wq_exit(void) 9697eb6b69SDavid Sterba { 9797eb6b69SDavid Sterba kmem_cache_destroy(btrfs_end_io_wq_cache); 9897eb6b69SDavid Sterba } 9997eb6b69SDavid Sterba 100141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 101141386e1SJosef Bacik { 102141386e1SJosef Bacik if (fs_info->csum_shash) 103141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 104141386e1SJosef Bacik } 105141386e1SJosef Bacik 106d352ac68SChris Mason /* 107d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 108d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 109d352ac68SChris Mason * just before they are sent down the IO stack. 110d352ac68SChris Mason */ 11144b8bd7eSChris Mason struct async_submit_bio { 112c6100a4bSJosef Bacik void *private_data; 11344b8bd7eSChris Mason struct bio *bio; 114a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 11544b8bd7eSChris Mason int mirror_num; 116eaf25d93SChris Mason /* 117eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 118eaf25d93SChris Mason * can't tell us where in the file the bio should go 119eaf25d93SChris Mason */ 120eaf25d93SChris Mason u64 bio_offset; 1218b712842SChris Mason struct btrfs_work work; 1224e4cbee9SChristoph Hellwig blk_status_t status; 12344b8bd7eSChris Mason }; 12444b8bd7eSChris Mason 12585d4e461SChris Mason /* 12685d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 12785d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 12885d4e461SChris Mason * the level the eb occupies in the tree. 1294008c04aSChris Mason * 13085d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 13185d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 13285d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 1334fd786e6SMisono Tomohiro * by btrfs_root->root_key.objectid. This ensures that all special purpose 1344fd786e6SMisono Tomohiro * roots have separate keysets. 1354008c04aSChris Mason * 13685d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 13785d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 13885d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1394008c04aSChris Mason * 14085d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 14185d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 14285d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 14385d4e461SChris Mason * 14485d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 14585d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 14685d4e461SChris Mason * needs update as well. 1474008c04aSChris Mason */ 1484008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1494008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1504008c04aSChris Mason # error 1514008c04aSChris Mason # endif 15285d4e461SChris Mason 15385d4e461SChris Mason static struct btrfs_lockdep_keyset { 15485d4e461SChris Mason u64 id; /* root objectid */ 15585d4e461SChris Mason const char *name_stem; /* lock name stem */ 15685d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 15785d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 15885d4e461SChris Mason } btrfs_lockdep_keysets[] = { 15985d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 16085d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 16185d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 16285d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 16385d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 16485d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 16560b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 16685d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 16785d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 16885d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 16913fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 1706b20e0adSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, .name_stem = "free-space" }, 17185d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1724008c04aSChris Mason }; 17385d4e461SChris Mason 17485d4e461SChris Mason void __init btrfs_init_lockdep(void) 17585d4e461SChris Mason { 17685d4e461SChris Mason int i, j; 17785d4e461SChris Mason 17885d4e461SChris Mason /* initialize lockdep class names */ 17985d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 18085d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 18185d4e461SChris Mason 18285d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 18385d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 18485d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 18585d4e461SChris Mason } 18685d4e461SChris Mason } 18785d4e461SChris Mason 18885d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 18985d4e461SChris Mason int level) 19085d4e461SChris Mason { 19185d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 19285d4e461SChris Mason 19385d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 19485d4e461SChris Mason 19585d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 19685d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 19785d4e461SChris Mason if (ks->id == objectid) 19885d4e461SChris Mason break; 19985d4e461SChris Mason 20085d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 20185d4e461SChris Mason &ks->keys[level], ks->names[level]); 20285d4e461SChris Mason } 20385d4e461SChris Mason 2044008c04aSChris Mason #endif 2054008c04aSChris Mason 206d352ac68SChris Mason /* 2072996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 208d352ac68SChris Mason */ 209c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 21019c00ddcSChris Mason { 211d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 212e9be5a30SDavid Sterba const int num_pages = fs_info->nodesize >> PAGE_SHIFT; 213d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 21419c00ddcSChris Mason char *kaddr; 215e9be5a30SDavid Sterba int i; 216d5178578SJohannes Thumshirn 217d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 218d5178578SJohannes Thumshirn crypto_shash_init(shash); 219e9be5a30SDavid Sterba kaddr = page_address(buf->pages[0]); 220e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 221e9be5a30SDavid Sterba PAGE_SIZE - BTRFS_CSUM_SIZE); 22219c00ddcSChris Mason 223e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 224e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 225e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 22619c00ddcSChris Mason } 22771a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 228d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 22919c00ddcSChris Mason } 23019c00ddcSChris Mason 231d352ac68SChris Mason /* 232d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 233d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 234d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 235d352ac68SChris Mason * in the wrong place. 236d352ac68SChris Mason */ 2371259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 238b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 239b9fab919SChris Mason int atomic) 2401259ab75SChris Mason { 2412ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 2421259ab75SChris Mason int ret; 2432755a0deSDavid Sterba bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); 2441259ab75SChris Mason 2451259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 2461259ab75SChris Mason return 0; 2471259ab75SChris Mason 248b9fab919SChris Mason if (atomic) 249b9fab919SChris Mason return -EAGAIN; 250b9fab919SChris Mason 251a26e8c9fSJosef Bacik if (need_lock) { 252a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 253300aa896SDavid Sterba btrfs_set_lock_blocking_read(eb); 254a26e8c9fSJosef Bacik } 255a26e8c9fSJosef Bacik 2562ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 257ff13db41SDavid Sterba &cached_state); 2580b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 2591259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 2601259ab75SChris Mason ret = 0; 2611259ab75SChris Mason goto out; 2621259ab75SChris Mason } 26394647322SDavid Sterba btrfs_err_rl(eb->fs_info, 26494647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 26594647322SDavid Sterba eb->start, 26629549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 2671259ab75SChris Mason ret = 1; 268a26e8c9fSJosef Bacik 269a26e8c9fSJosef Bacik /* 270a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 271a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 27201327610SNicholas D Steeves * block that has been freed and re-allocated. So don't clear uptodate 273a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 274a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 275a26e8c9fSJosef Bacik * making everybody very sad. 276a26e8c9fSJosef Bacik */ 277a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 2780b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 27933958dc6SChris Mason out: 2802ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 281e43bbe5eSDavid Sterba &cached_state); 282472b909fSJosef Bacik if (need_lock) 283a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 2841259ab75SChris Mason return ret; 2851259ab75SChris Mason } 2861259ab75SChris Mason 287e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 288e7e16f48SJohannes Thumshirn { 289e7e16f48SJohannes Thumshirn switch (csum_type) { 290e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 2913951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 2923831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 293352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 294e7e16f48SJohannes Thumshirn return true; 295e7e16f48SJohannes Thumshirn default: 296e7e16f48SJohannes Thumshirn return false; 297e7e16f48SJohannes Thumshirn } 298e7e16f48SJohannes Thumshirn } 299e7e16f48SJohannes Thumshirn 300d352ac68SChris Mason /* 3011104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 3021104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 3031104a885SDavid Sterba */ 304ab8d0fc4SJeff Mahoney static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 305ab8d0fc4SJeff Mahoney char *raw_disk_sb) 3061104a885SDavid Sterba { 3071104a885SDavid Sterba struct btrfs_super_block *disk_sb = 3081104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 30951bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 310d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 311d5178578SJohannes Thumshirn 312d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3131104a885SDavid Sterba 3141104a885SDavid Sterba /* 3151104a885SDavid Sterba * The super_block structure does not span the whole 31651bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 31751bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 3181104a885SDavid Sterba */ 319fd08001fSEric Biggers crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, 320fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 3211104a885SDavid Sterba 32251bce6c9SJohannes Thumshirn if (memcmp(disk_sb->csum, result, btrfs_super_csum_size(disk_sb))) 323e7e16f48SJohannes Thumshirn return 1; 3241104a885SDavid Sterba 325e7e16f48SJohannes Thumshirn return 0; 3261104a885SDavid Sterba } 3271104a885SDavid Sterba 328e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 329ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 330581c1760SQu Wenruo { 331e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 332581c1760SQu Wenruo int found_level; 333581c1760SQu Wenruo struct btrfs_key found_key; 334581c1760SQu Wenruo int ret; 335581c1760SQu Wenruo 336581c1760SQu Wenruo found_level = btrfs_header_level(eb); 337581c1760SQu Wenruo if (found_level != level) { 33863489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 33963489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 340581c1760SQu Wenruo btrfs_err(fs_info, 341581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 342581c1760SQu Wenruo eb->start, level, found_level); 343581c1760SQu Wenruo return -EIO; 344581c1760SQu Wenruo } 345581c1760SQu Wenruo 346581c1760SQu Wenruo if (!first_key) 347581c1760SQu Wenruo return 0; 348581c1760SQu Wenruo 3495d41be6fSQu Wenruo /* 3505d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 3515d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 3525d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 3535d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 3545d41be6fSQu Wenruo */ 3555d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 3565d41be6fSQu Wenruo return 0; 35762fdaa52SQu Wenruo 35862fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 35962fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 36062fdaa52SQu Wenruo btrfs_err(fs_info, 36162fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 36262fdaa52SQu Wenruo eb->start); 36362fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 36462fdaa52SQu Wenruo return -EUCLEAN; 36562fdaa52SQu Wenruo } 36662fdaa52SQu Wenruo 367581c1760SQu Wenruo if (found_level) 368581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 369581c1760SQu Wenruo else 370581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 371581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 372581c1760SQu Wenruo 373581c1760SQu Wenruo if (ret) { 37463489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 37563489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 376581c1760SQu Wenruo btrfs_err(fs_info, 377ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 378ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 379ff76a864SLiu Bo first_key->type, first_key->offset, 380ff76a864SLiu Bo found_key.objectid, found_key.type, 381ff76a864SLiu Bo found_key.offset); 382581c1760SQu Wenruo } 383581c1760SQu Wenruo return ret; 384581c1760SQu Wenruo } 385581c1760SQu Wenruo 3861104a885SDavid Sterba /* 387d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 388d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 389581c1760SQu Wenruo * 390581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 391581c1760SQu Wenruo * @level: expected level, mandatory check 392581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 393d352ac68SChris Mason */ 3945ab12d1fSDavid Sterba static int btree_read_extent_buffer_pages(struct extent_buffer *eb, 395581c1760SQu Wenruo u64 parent_transid, int level, 396581c1760SQu Wenruo struct btrfs_key *first_key) 397f188591eSChris Mason { 3985ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 399f188591eSChris Mason struct extent_io_tree *io_tree; 400ea466794SJosef Bacik int failed = 0; 401f188591eSChris Mason int ret; 402f188591eSChris Mason int num_copies = 0; 403f188591eSChris Mason int mirror_num = 0; 404ea466794SJosef Bacik int failed_mirror = 0; 405f188591eSChris Mason 4060b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 407f188591eSChris Mason while (1) { 408f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 409c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); 410256dd1bbSStefan Behrens if (!ret) { 411581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 412b9fab919SChris Mason parent_transid, 0)) 413256dd1bbSStefan Behrens ret = -EIO; 414e064d5e9SDavid Sterba else if (btrfs_verify_level_key(eb, level, 415ff76a864SLiu Bo first_key, parent_transid)) 416581c1760SQu Wenruo ret = -EUCLEAN; 417581c1760SQu Wenruo else 418581c1760SQu Wenruo break; 419256dd1bbSStefan Behrens } 420d397712bSChris Mason 4210b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 422f188591eSChris Mason eb->start, eb->len); 4234235298eSChris Mason if (num_copies == 1) 424ea466794SJosef Bacik break; 4254235298eSChris Mason 4265cf1ab56SJosef Bacik if (!failed_mirror) { 4275cf1ab56SJosef Bacik failed = 1; 4285cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4295cf1ab56SJosef Bacik } 4305cf1ab56SJosef Bacik 431f188591eSChris Mason mirror_num++; 432ea466794SJosef Bacik if (mirror_num == failed_mirror) 433ea466794SJosef Bacik mirror_num++; 434ea466794SJosef Bacik 4354235298eSChris Mason if (mirror_num > num_copies) 436ea466794SJosef Bacik break; 437f188591eSChris Mason } 438ea466794SJosef Bacik 439c0901581SStefan Behrens if (failed && !ret && failed_mirror) 44020a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 441ea466794SJosef Bacik 442ea466794SJosef Bacik return ret; 443f188591eSChris Mason } 44419c00ddcSChris Mason 445d352ac68SChris Mason /* 446d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 447d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 448d352ac68SChris Mason */ 449d397712bSChris Mason 45001d58472SDaniel Dressler static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct page *page) 45119c00ddcSChris Mason { 4524eee4fa4SMiao Xie u64 start = page_offset(page); 45319c00ddcSChris Mason u64 found_start; 4542996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 4552996e1f8SJohannes Thumshirn u16 csum_size = btrfs_super_csum_size(fs_info->super_copy); 45619c00ddcSChris Mason struct extent_buffer *eb; 4578d47a0d8SQu Wenruo int ret; 458f188591eSChris Mason 4594f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4604f2de97aSJosef Bacik if (page != eb->pages[0]) 4614f2de97aSJosef Bacik return 0; 4620f805531SAlex Lyakas 46319c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 4640f805531SAlex Lyakas /* 4650f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 4660f805531SAlex Lyakas * It is useful to know what went wrong. 4670f805531SAlex Lyakas */ 4680f805531SAlex Lyakas if (WARN_ON(found_start != start)) 4690f805531SAlex Lyakas return -EUCLEAN; 4700f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 4710f805531SAlex Lyakas return -EUCLEAN; 4720f805531SAlex Lyakas 473de37aa51SNikolay Borisov ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 4749a8658e3SDavid Sterba offsetof(struct btrfs_header, fsid), 4759a8658e3SDavid Sterba BTRFS_FSID_SIZE) == 0); 4760f805531SAlex Lyakas 477c67b3892SDavid Sterba csum_tree_block(eb, result); 4782996e1f8SJohannes Thumshirn 4798d47a0d8SQu Wenruo if (btrfs_header_level(eb)) 4808d47a0d8SQu Wenruo ret = btrfs_check_node(eb); 4818d47a0d8SQu Wenruo else 4828d47a0d8SQu Wenruo ret = btrfs_check_leaf_full(eb); 4838d47a0d8SQu Wenruo 4848d47a0d8SQu Wenruo if (ret < 0) { 485c06631b0SQu Wenruo btrfs_print_tree(eb, 0); 4868d47a0d8SQu Wenruo btrfs_err(fs_info, 4878d47a0d8SQu Wenruo "block=%llu write time tree block corruption detected", 4888d47a0d8SQu Wenruo eb->start); 489c06631b0SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 4908d47a0d8SQu Wenruo return ret; 4918d47a0d8SQu Wenruo } 4922996e1f8SJohannes Thumshirn write_extent_buffer(eb, result, 0, csum_size); 4938d47a0d8SQu Wenruo 4942996e1f8SJohannes Thumshirn return 0; 49519c00ddcSChris Mason } 49619c00ddcSChris Mason 497b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 4982b82032cSYan Zheng { 499b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 500944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 50144880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 502944d3f9fSNikolay Borisov u8 *metadata_uuid; 5032b82032cSYan Zheng 5049a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 5059a8658e3SDavid Sterba BTRFS_FSID_SIZE); 5067239ff4bSNikolay Borisov /* 507944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 508944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 509944d3f9fSNikolay Borisov * ->fsid in this case is fine 5107239ff4bSNikolay Borisov */ 511944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 5127239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 5137239ff4bSNikolay Borisov else 5147239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 5157239ff4bSNikolay Borisov 516944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 517944d3f9fSNikolay Borisov return 0; 518944d3f9fSNikolay Borisov 519944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 520944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 521944d3f9fSNikolay Borisov return 0; 522944d3f9fSNikolay Borisov 523944d3f9fSNikolay Borisov return 1; 5242b82032cSYan Zheng } 5252b82032cSYan Zheng 5269a446d6aSNikolay Borisov int btrfs_validate_metadata_buffer(struct btrfs_io_bio *io_bio, u64 phy_offset, 5279a446d6aSNikolay Borisov struct page *page, u64 start, u64 end, 5289a446d6aSNikolay Borisov int mirror) 529ce9adaa5SChris Mason { 530ce9adaa5SChris Mason u64 found_start; 531ce9adaa5SChris Mason int found_level; 532ce9adaa5SChris Mason struct extent_buffer *eb; 53315b6e8a8SDavid Sterba struct btrfs_fs_info *fs_info; 53415b6e8a8SDavid Sterba u16 csum_size; 535f188591eSChris Mason int ret = 0; 5362996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 537727011e0SChris Mason int reads_done; 538ce9adaa5SChris Mason 539ce9adaa5SChris Mason if (!page->private) 540ce9adaa5SChris Mason goto out; 541d397712bSChris Mason 5424f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 54315b6e8a8SDavid Sterba fs_info = eb->fs_info; 54415b6e8a8SDavid Sterba csum_size = btrfs_super_csum_size(fs_info->super_copy); 545d397712bSChris Mason 5460b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 5470b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 5480b32f4bbSJosef Bacik */ 54967439dadSDavid Sterba atomic_inc(&eb->refs); 5500b32f4bbSJosef Bacik 5510b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 552727011e0SChris Mason if (!reads_done) 553727011e0SChris Mason goto err; 554f188591eSChris Mason 5555cf1ab56SJosef Bacik eb->read_mirror = mirror; 556656f30dbSFilipe Manana if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 557ea466794SJosef Bacik ret = -EIO; 558ea466794SJosef Bacik goto err; 559ea466794SJosef Bacik } 560ea466794SJosef Bacik 561ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 562727011e0SChris Mason if (found_start != eb->start) { 563893bf4b1SSu Yue btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", 564893bf4b1SSu Yue eb->start, found_start); 565f188591eSChris Mason ret = -EIO; 566ce9adaa5SChris Mason goto err; 567ce9adaa5SChris Mason } 568b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 56902873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 57094647322SDavid Sterba eb->start); 5711259ab75SChris Mason ret = -EIO; 5721259ab75SChris Mason goto err; 5731259ab75SChris Mason } 574ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 5751c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 576893bf4b1SSu Yue btrfs_err(fs_info, "bad tree block level %d on %llu", 577893bf4b1SSu Yue (int)btrfs_header_level(eb), eb->start); 5781c24c3ceSJosef Bacik ret = -EIO; 5791c24c3ceSJosef Bacik goto err; 5801c24c3ceSJosef Bacik } 581ce9adaa5SChris Mason 58285d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 58385d4e461SChris Mason eb, found_level); 5844008c04aSChris Mason 585c67b3892SDavid Sterba csum_tree_block(eb, result); 586a826d6dcSJosef Bacik 5872996e1f8SJohannes Thumshirn if (memcmp_extent_buffer(eb, result, 0, csum_size)) { 58835be8851SJohannes Thumshirn u8 val[BTRFS_CSUM_SIZE] = { 0 }; 5892996e1f8SJohannes Thumshirn 5902996e1f8SJohannes Thumshirn read_extent_buffer(eb, &val, 0, csum_size); 5912996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 59235be8851SJohannes Thumshirn "%s checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", 5932996e1f8SJohannes Thumshirn fs_info->sb->s_id, eb->start, 59435be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, val), 59535be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 59635be8851SJohannes Thumshirn btrfs_header_level(eb)); 5972996e1f8SJohannes Thumshirn ret = -EUCLEAN; 5982996e1f8SJohannes Thumshirn goto err; 5992996e1f8SJohannes Thumshirn } 6002996e1f8SJohannes Thumshirn 601a826d6dcSJosef Bacik /* 602a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 603a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 604a826d6dcSJosef Bacik * return -EIO. 605a826d6dcSJosef Bacik */ 6061c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 607a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 608a826d6dcSJosef Bacik ret = -EIO; 609a826d6dcSJosef Bacik } 610ce9adaa5SChris Mason 611813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 612053ab70fSLiu Bo ret = -EIO; 613053ab70fSLiu Bo 6140b32f4bbSJosef Bacik if (!ret) 6150b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 61675391f0dSQu Wenruo else 61775391f0dSQu Wenruo btrfs_err(fs_info, 61875391f0dSQu Wenruo "block=%llu read time tree block corruption detected", 61975391f0dSQu Wenruo eb->start); 620ce9adaa5SChris Mason err: 62179fb65a1SJosef Bacik if (reads_done && 62279fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 623d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 6244bb31e92SArne Jansen 62553b381b3SDavid Woodhouse if (ret) { 62653b381b3SDavid Woodhouse /* 62753b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 62853b381b3SDavid Woodhouse * again, we have to make sure it has something 62953b381b3SDavid Woodhouse * to decrement 63053b381b3SDavid Woodhouse */ 63153b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6320b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 63353b381b3SDavid Woodhouse } 6340b32f4bbSJosef Bacik free_extent_buffer(eb); 635ce9adaa5SChris Mason out: 636f188591eSChris Mason return ret; 637ce9adaa5SChris Mason } 638ce9adaa5SChris Mason 6394246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 640ce9adaa5SChris Mason { 64197eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 642ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 6439e0af237SLiu Bo struct btrfs_workqueue *wq; 644ce9adaa5SChris Mason 645ce9adaa5SChris Mason fs_info = end_io_wq->info; 6464e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 647d20f7043SChris Mason 64837226b21SMike Christie if (bio_op(bio) == REQ_OP_WRITE) { 649a0cac0ecSOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) 6509e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 651a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) 6529e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 653a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 6549e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 655a0cac0ecSOmar Sandoval else 6569e0af237SLiu Bo wq = fs_info->endio_write_workers; 6579e0af237SLiu Bo } else { 6585c047a69SOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 6599e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 660a0cac0ecSOmar Sandoval else if (end_io_wq->metadata) 6619e0af237SLiu Bo wq = fs_info->endio_meta_workers; 662a0cac0ecSOmar Sandoval else 6639e0af237SLiu Bo wq = fs_info->endio_workers; 6649e0af237SLiu Bo } 6659e0af237SLiu Bo 666a0cac0ecSOmar Sandoval btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); 6679e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 668ce9adaa5SChris Mason } 669ce9adaa5SChris Mason 6704e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 671bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 6720b86a832SChris Mason { 67397eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 6748b110e39SMiao Xie 67597eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 676ce9adaa5SChris Mason if (!end_io_wq) 6774e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 678ce9adaa5SChris Mason 679ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 680ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 68122c59948SChris Mason end_io_wq->info = info; 6824e4cbee9SChristoph Hellwig end_io_wq->status = 0; 683ce9adaa5SChris Mason end_io_wq->bio = bio; 68422c59948SChris Mason end_io_wq->metadata = metadata; 685ce9adaa5SChris Mason 686ce9adaa5SChris Mason bio->bi_private = end_io_wq; 687ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 68822c59948SChris Mason return 0; 68922c59948SChris Mason } 69022c59948SChris Mason 6914a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 6924a69a410SChris Mason { 6934a69a410SChris Mason struct async_submit_bio *async; 6944e4cbee9SChristoph Hellwig blk_status_t ret; 6954a69a410SChris Mason 6964a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 697c6100a4bSJosef Bacik ret = async->submit_bio_start(async->private_data, async->bio, 698eaf25d93SChris Mason async->bio_offset); 69979787eaaSJeff Mahoney if (ret) 7004e4cbee9SChristoph Hellwig async->status = ret; 7014a69a410SChris Mason } 7024a69a410SChris Mason 70306ea01b1SDavid Sterba /* 70406ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 70506ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 70606ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 70706ea01b1SDavid Sterba * 70806ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 70906ea01b1SDavid Sterba * inserted into the tree. 71006ea01b1SDavid Sterba */ 7114a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7128b712842SChris Mason { 7138b712842SChris Mason struct async_submit_bio *async; 71406ea01b1SDavid Sterba struct inode *inode; 71506ea01b1SDavid Sterba blk_status_t ret; 7168b712842SChris Mason 7178b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 71806ea01b1SDavid Sterba inode = async->private_data; 7194854ddd0SChris Mason 720bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 7214e4cbee9SChristoph Hellwig if (async->status) { 7224e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 7234246a0b6SChristoph Hellwig bio_endio(async->bio); 72479787eaaSJeff Mahoney return; 72579787eaaSJeff Mahoney } 72679787eaaSJeff Mahoney 727ec39f769SChris Mason /* 728ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 729ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 730ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 731ec39f769SChris Mason */ 732ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 73308635baeSChris Mason ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 73406ea01b1SDavid Sterba if (ret) { 73506ea01b1SDavid Sterba async->bio->bi_status = ret; 73606ea01b1SDavid Sterba bio_endio(async->bio); 73706ea01b1SDavid Sterba } 7384a69a410SChris Mason } 7394a69a410SChris Mason 7404a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7414a69a410SChris Mason { 7424a69a410SChris Mason struct async_submit_bio *async; 7434a69a410SChris Mason 7444a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7458b712842SChris Mason kfree(async); 7468b712842SChris Mason } 7478b712842SChris Mason 7488c27cb35SLinus Torvalds blk_status_t btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct bio *bio, 749c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 750c6100a4bSJosef Bacik u64 bio_offset, void *private_data, 751e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 75244b8bd7eSChris Mason { 75344b8bd7eSChris Mason struct async_submit_bio *async; 75444b8bd7eSChris Mason 75544b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 75644b8bd7eSChris Mason if (!async) 7574e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 75844b8bd7eSChris Mason 759c6100a4bSJosef Bacik async->private_data = private_data; 76044b8bd7eSChris Mason async->bio = bio; 76144b8bd7eSChris Mason async->mirror_num = mirror_num; 7624a69a410SChris Mason async->submit_bio_start = submit_bio_start; 7634a69a410SChris Mason 764a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 765a0cac0ecSOmar Sandoval run_one_async_free); 7664a69a410SChris Mason 767eaf25d93SChris Mason async->bio_offset = bio_offset; 7688c8bee1dSChris Mason 7694e4cbee9SChristoph Hellwig async->status = 0; 77079787eaaSJeff Mahoney 77167f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 7725cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 773d313d7a3SChris Mason 7745cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 77544b8bd7eSChris Mason return 0; 77644b8bd7eSChris Mason } 77744b8bd7eSChris Mason 7784e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 779ce3ed71aSChris Mason { 7802c30c71bSKent Overstreet struct bio_vec *bvec; 781ce3ed71aSChris Mason struct btrfs_root *root; 7822b070cfeSChristoph Hellwig int ret = 0; 7836dc4f100SMing Lei struct bvec_iter_all iter_all; 784ce3ed71aSChris Mason 785c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 7862b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 787ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 78801d58472SDaniel Dressler ret = csum_dirty_buffer(root->fs_info, bvec->bv_page); 78979787eaaSJeff Mahoney if (ret) 79079787eaaSJeff Mahoney break; 791ce3ed71aSChris Mason } 7922c30c71bSKent Overstreet 7934e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 794ce3ed71aSChris Mason } 795ce3ed71aSChris Mason 796d0ee3934SDavid Sterba static blk_status_t btree_submit_bio_start(void *private_data, struct bio *bio, 797eaf25d93SChris Mason u64 bio_offset) 79822c59948SChris Mason { 7998b712842SChris Mason /* 8008b712842SChris Mason * when we're called for a write, we're already in the async 8015443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8028b712842SChris Mason */ 80379787eaaSJeff Mahoney return btree_csum_one_bio(bio); 80422c59948SChris Mason } 80522c59948SChris Mason 8069b4e675aSDavid Sterba static int check_async_write(struct btrfs_fs_info *fs_info, 8079b4e675aSDavid Sterba struct btrfs_inode *bi) 808de0022b9SJosef Bacik { 8096300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 8106300463bSLiu Bo return 0; 8119b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 812de0022b9SJosef Bacik return 0; 813de0022b9SJosef Bacik return 1; 814de0022b9SJosef Bacik } 815de0022b9SJosef Bacik 8161b36294aSNikolay Borisov blk_status_t btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, 8171b36294aSNikolay Borisov int mirror_num, unsigned long bio_flags) 81844b8bd7eSChris Mason { 8190b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 8209b4e675aSDavid Sterba int async = check_async_write(fs_info, BTRFS_I(inode)); 8214e4cbee9SChristoph Hellwig blk_status_t ret; 822cad321adSChris Mason 82337226b21SMike Christie if (bio_op(bio) != REQ_OP_WRITE) { 824cad321adSChris Mason /* 825cad321adSChris Mason * called for a read, do the setup so that checksum validation 826cad321adSChris Mason * can happen in the async kernel threads 827cad321adSChris Mason */ 8280b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 8290b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 8301d4284bdSChris Mason if (ret) 83161891923SStefan Behrens goto out_w_error; 83208635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 833de0022b9SJosef Bacik } else if (!async) { 834de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 835de0022b9SJosef Bacik if (ret) 83661891923SStefan Behrens goto out_w_error; 83708635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 83861891923SStefan Behrens } else { 839cad321adSChris Mason /* 84061891923SStefan Behrens * kthread helpers are used to submit writes so that 84161891923SStefan Behrens * checksumming can happen in parallel across all CPUs 842cad321adSChris Mason */ 843c6100a4bSJosef Bacik ret = btrfs_wq_submit_bio(fs_info, bio, mirror_num, 0, 844e68f2ee7SNikolay Borisov 0, inode, btree_submit_bio_start); 84544b8bd7eSChris Mason } 84644b8bd7eSChris Mason 8474246a0b6SChristoph Hellwig if (ret) 8484246a0b6SChristoph Hellwig goto out_w_error; 8494246a0b6SChristoph Hellwig return 0; 8504246a0b6SChristoph Hellwig 85161891923SStefan Behrens out_w_error: 8524e4cbee9SChristoph Hellwig bio->bi_status = ret; 8534246a0b6SChristoph Hellwig bio_endio(bio); 85461891923SStefan Behrens return ret; 85561891923SStefan Behrens } 85661891923SStefan Behrens 8573dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 858784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 859a6bc32b8SMel Gorman struct page *newpage, struct page *page, 860a6bc32b8SMel Gorman enum migrate_mode mode) 861784b4e29SChris Mason { 862784b4e29SChris Mason /* 863784b4e29SChris Mason * we can't safely write a btree page from here, 864784b4e29SChris Mason * we haven't done the locking hook 865784b4e29SChris Mason */ 866784b4e29SChris Mason if (PageDirty(page)) 867784b4e29SChris Mason return -EAGAIN; 868784b4e29SChris Mason /* 869784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 870784b4e29SChris Mason * We must have no buffers or drop them. 871784b4e29SChris Mason */ 872784b4e29SChris Mason if (page_has_private(page) && 873784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 874784b4e29SChris Mason return -EAGAIN; 875a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 876784b4e29SChris Mason } 8773dd1462eSJan Beulich #endif 878784b4e29SChris Mason 8790da5468fSChris Mason 8800da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 8810da5468fSChris Mason struct writeback_control *wbc) 8820da5468fSChris Mason { 883e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 884e2d84521SMiao Xie int ret; 885e2d84521SMiao Xie 886d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 887448d640bSChris Mason 888448d640bSChris Mason if (wbc->for_kupdate) 889448d640bSChris Mason return 0; 890448d640bSChris Mason 891e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 892b9473439SChris Mason /* this is a bit racy, but that's ok */ 893d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 894d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 895d814a491SEthan Lien fs_info->dirty_metadata_batch); 896e2d84521SMiao Xie if (ret < 0) 897793955bcSChris Mason return 0; 898793955bcSChris Mason } 8990b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9000da5468fSChris Mason } 9010da5468fSChris Mason 90270dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 9035f39d397SChris Mason { 90498509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 90598509cfcSChris Mason return 0; 9060c4e538bSDavid Sterba 907f7a52a40SDavid Sterba return try_release_extent_buffer(page); 908d98237b3SChris Mason } 909d98237b3SChris Mason 910d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 911d47992f8SLukas Czerner unsigned int length) 912d98237b3SChris Mason { 913d1310b2eSChris Mason struct extent_io_tree *tree; 914d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9155f39d397SChris Mason extent_invalidatepage(tree, page, offset); 9165f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 9179ad6b7bcSChris Mason if (PagePrivate(page)) { 918efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 919efe120a0SFrank Holton "page private not zero on page %llu", 920efe120a0SFrank Holton (unsigned long long)page_offset(page)); 921d1b89bc0SGuoqing Jiang detach_page_private(page); 9229ad6b7bcSChris Mason } 923d98237b3SChris Mason } 924d98237b3SChris Mason 9250b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 9260b32f4bbSJosef Bacik { 927bb146eb2SJosef Bacik #ifdef DEBUG 9280b32f4bbSJosef Bacik struct extent_buffer *eb; 9290b32f4bbSJosef Bacik 9300b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 9310b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 9320b32f4bbSJosef Bacik BUG_ON(!eb); 9330b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 9340b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 9350b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 936bb146eb2SJosef Bacik #endif 9370b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 9380b32f4bbSJosef Bacik } 9390b32f4bbSJosef Bacik 9407f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 9410da5468fSChris Mason .writepages = btree_writepages, 9425f39d397SChris Mason .releasepage = btree_releasepage, 9435f39d397SChris Mason .invalidatepage = btree_invalidatepage, 9445a92bc88SChris Mason #ifdef CONFIG_MIGRATION 945784b4e29SChris Mason .migratepage = btree_migratepage, 9465a92bc88SChris Mason #endif 9470b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 948d98237b3SChris Mason }; 949123abc88SChris Mason 9502ff7e61eSJeff Mahoney void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr) 951090d1875SChris Mason { 9525f39d397SChris Mason struct extent_buffer *buf = NULL; 953537f38f0SNikolay Borisov int ret; 954090d1875SChris Mason 9552ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 956c871b0f2SLiu Bo if (IS_ERR(buf)) 9576197d86eSDavid Sterba return; 958537f38f0SNikolay Borisov 959c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(buf, WAIT_NONE, 0); 960537f38f0SNikolay Borisov if (ret < 0) 961537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 962537f38f0SNikolay Borisov else 9635f39d397SChris Mason free_extent_buffer(buf); 964090d1875SChris Mason } 965090d1875SChris Mason 9662ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 9672ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 9680308af44SDavid Sterba u64 bytenr) 9690999df54SChris Mason { 9700b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 9710b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 9720b246afaSJeff Mahoney return alloc_extent_buffer(fs_info, bytenr); 9730999df54SChris Mason } 9740999df54SChris Mason 975581c1760SQu Wenruo /* 976581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 977581c1760SQu Wenruo * verification. 978581c1760SQu Wenruo * 979581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 980581c1760SQu Wenruo * @level: expected level, mandatory check 981581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 982581c1760SQu Wenruo */ 9832ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 984581c1760SQu Wenruo u64 parent_transid, int level, 985581c1760SQu Wenruo struct btrfs_key *first_key) 986e20d96d6SChris Mason { 9875f39d397SChris Mason struct extent_buffer *buf = NULL; 98819c00ddcSChris Mason int ret; 98919c00ddcSChris Mason 9902ff7e61eSJeff Mahoney buf = btrfs_find_create_tree_block(fs_info, bytenr); 991c871b0f2SLiu Bo if (IS_ERR(buf)) 992c871b0f2SLiu Bo return buf; 993e4204dedSChris Mason 9945ab12d1fSDavid Sterba ret = btree_read_extent_buffer_pages(buf, parent_transid, 995581c1760SQu Wenruo level, first_key); 9960f0fe8f7SFilipe David Borba Manana if (ret) { 997537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 99864c043deSLiu Bo return ERR_PTR(ret); 9990f0fe8f7SFilipe David Borba Manana } 10005f39d397SChris Mason return buf; 1001ce9adaa5SChris Mason 1002eb60ceacSChris Mason } 1003eb60ceacSChris Mason 10046a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 1005ed2ff2cbSChris Mason { 10066a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 100755c69072SChris Mason if (btrfs_header_generation(buf) == 1008e2d84521SMiao Xie fs_info->running_transaction->transid) { 1009b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1010b4ce94deSChris Mason 1011b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1012104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1013e2d84521SMiao Xie -buf->len, 1014e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1015b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 10168bead258SDavid Sterba btrfs_set_lock_blocking_write(buf); 10170b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1018925baeddSChris Mason } 10195f39d397SChris Mason } 1020ed7b63ebSJosef Bacik } 10215f39d397SChris Mason 1022da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1023e20d96d6SChris Mason u64 objectid) 1024d97e63b6SChris Mason { 10257c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 102696dfcb46SJosef Bacik root->fs_info = fs_info; 1027cfaa7295SChris Mason root->node = NULL; 1028a28ec197SChris Mason root->commit_root = NULL; 102927cdeb70SMiao Xie root->state = 0; 1030d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 10310b86a832SChris Mason 10320f7d52f4SChris Mason root->last_trans = 0; 103313a8a7c8SYan, Zheng root->highest_objectid = 0; 1034eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1035199c2a9cSMiao Xie root->nr_ordered_extents = 0; 10366bef4d31SEric Paris root->inode_tree = RB_ROOT; 103716cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1038f0486c68SYan, Zheng root->block_rsv = NULL; 10390b86a832SChris Mason 10400b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 10415d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1042eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1043eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1044199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1045199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1046d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 10472ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 10482ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 10495d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1050eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1051199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1052f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 10532ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 10542ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 10558287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1056a2135011SChris Mason mutex_init(&root->objectid_mutex); 1057e02119d5SChris Mason mutex_init(&root->log_mutex); 105831f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1059573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1060c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 10617237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 10627237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 10637237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 10648b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 10658b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 10667237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 10677237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 10687237f183SYan Zheng atomic_set(&root->log_writers, 0); 10692ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 10700700cea7SElena Reshetova refcount_set(&root->refs, 1); 10718ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1072eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 10737237f183SYan Zheng root->log_transid = 0; 1074d1433debSMiao Xie root->log_transid_committed = -1; 1075257c62e1SChris Mason root->last_log_commit = 0; 1076e289f03eSFilipe Manana if (!dummy) { 107743eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 107843eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1079e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1080e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1081e289f03eSFilipe Manana } 1082017e5369SChris Mason 10833768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 10843768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 10856702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 10864d775673SChris Mason root->root_key.objectid = objectid; 10870ee5dc67SAl Viro root->anon_dev = 0; 10888ea05e3aSAlexander Block 10895f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1090370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1091bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1092bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1093bd647ce3SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 1094bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1095bd647ce3SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 1096bd647ce3SJosef Bacik #endif 10973768f368SChris Mason } 10983768f368SChris Mason 109974e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 110096dfcb46SJosef Bacik u64 objectid, gfp_t flags) 11016f07e42eSAl Viro { 110274e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 11036f07e42eSAl Viro if (root) 110496dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 11056f07e42eSAl Viro return root; 11066f07e42eSAl Viro } 11076f07e42eSAl Viro 110806ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 110906ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1110da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 111106ea65a3SJosef Bacik { 111206ea65a3SJosef Bacik struct btrfs_root *root; 111306ea65a3SJosef Bacik 11147c0260eeSJeff Mahoney if (!fs_info) 11157c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 11167c0260eeSJeff Mahoney 111796dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 111806ea65a3SJosef Bacik if (!root) 111906ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1120da17066cSJeff Mahoney 1121b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1122faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 112306ea65a3SJosef Bacik 112406ea65a3SJosef Bacik return root; 112506ea65a3SJosef Bacik } 112606ea65a3SJosef Bacik #endif 112706ea65a3SJosef Bacik 112820897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 112920897f5cSArne Jansen u64 objectid) 113020897f5cSArne Jansen { 11319b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 113220897f5cSArne Jansen struct extent_buffer *leaf; 113320897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 113420897f5cSArne Jansen struct btrfs_root *root; 113520897f5cSArne Jansen struct btrfs_key key; 1136b89f6d1fSFilipe Manana unsigned int nofs_flag; 113720897f5cSArne Jansen int ret = 0; 113820897f5cSArne Jansen 1139b89f6d1fSFilipe Manana /* 1140b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1141b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1142b89f6d1fSFilipe Manana */ 1143b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 114496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1145b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 114620897f5cSArne Jansen if (!root) 114720897f5cSArne Jansen return ERR_PTR(-ENOMEM); 114820897f5cSArne Jansen 114920897f5cSArne Jansen root->root_key.objectid = objectid; 115020897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 115120897f5cSArne Jansen root->root_key.offset = 0; 115220897f5cSArne Jansen 11539631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 11549631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 115520897f5cSArne Jansen if (IS_ERR(leaf)) { 115620897f5cSArne Jansen ret = PTR_ERR(leaf); 11571dd05682STsutomu Itoh leaf = NULL; 115820897f5cSArne Jansen goto fail; 115920897f5cSArne Jansen } 116020897f5cSArne Jansen 116120897f5cSArne Jansen root->node = leaf; 116220897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 116320897f5cSArne Jansen 116420897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 116527cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 116620897f5cSArne Jansen 116720897f5cSArne Jansen root->root_item.flags = 0; 116820897f5cSArne Jansen root->root_item.byte_limit = 0; 116920897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 117020897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 117120897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 117220897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 117320897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 117420897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 117520897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 117633d85fdaSQu Wenruo if (is_fstree(objectid)) 1177807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1178807fc790SAndy Shevchenko else 1179807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 118020897f5cSArne Jansen root->root_item.drop_level = 0; 118120897f5cSArne Jansen 118220897f5cSArne Jansen key.objectid = objectid; 118320897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 118420897f5cSArne Jansen key.offset = 0; 118520897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 118620897f5cSArne Jansen if (ret) 118720897f5cSArne Jansen goto fail; 118820897f5cSArne Jansen 118920897f5cSArne Jansen btrfs_tree_unlock(leaf); 119020897f5cSArne Jansen 119120897f5cSArne Jansen return root; 11921dd05682STsutomu Itoh 11931dd05682STsutomu Itoh fail: 11948c38938cSJosef Bacik if (leaf) 11951dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 119600246528SJosef Bacik btrfs_put_root(root); 11971dd05682STsutomu Itoh 11981dd05682STsutomu Itoh return ERR_PTR(ret); 119920897f5cSArne Jansen } 120020897f5cSArne Jansen 12017237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1202e02119d5SChris Mason struct btrfs_fs_info *fs_info) 12030f7d52f4SChris Mason { 12040f7d52f4SChris Mason struct btrfs_root *root; 12057237f183SYan Zheng struct extent_buffer *leaf; 1206e02119d5SChris Mason 120796dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1208e02119d5SChris Mason if (!root) 12097237f183SYan Zheng return ERR_PTR(-ENOMEM); 1210e02119d5SChris Mason 1211e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1212e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1213e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 121427cdeb70SMiao Xie 12157237f183SYan Zheng /* 121692a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 121727cdeb70SMiao Xie * 121892a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 121992a7cc42SQu Wenruo * and they go away before a real commit is actually done. 122092a7cc42SQu Wenruo * 122192a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 122292a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 12237237f183SYan Zheng */ 1224e02119d5SChris Mason 12254d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 12269631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 12277237f183SYan Zheng if (IS_ERR(leaf)) { 122800246528SJosef Bacik btrfs_put_root(root); 12297237f183SYan Zheng return ERR_CAST(leaf); 12307237f183SYan Zheng } 1231e02119d5SChris Mason 12327237f183SYan Zheng root->node = leaf; 1233e02119d5SChris Mason 1234e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1235e02119d5SChris Mason btrfs_tree_unlock(root->node); 12367237f183SYan Zheng return root; 12377237f183SYan Zheng } 12387237f183SYan Zheng 12397237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 12407237f183SYan Zheng struct btrfs_fs_info *fs_info) 12417237f183SYan Zheng { 12427237f183SYan Zheng struct btrfs_root *log_root; 12437237f183SYan Zheng 12447237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 12457237f183SYan Zheng if (IS_ERR(log_root)) 12467237f183SYan Zheng return PTR_ERR(log_root); 12477237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 12487237f183SYan Zheng fs_info->log_root_tree = log_root; 12497237f183SYan Zheng return 0; 12507237f183SYan Zheng } 12517237f183SYan Zheng 12527237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 12537237f183SYan Zheng struct btrfs_root *root) 12547237f183SYan Zheng { 12550b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 12567237f183SYan Zheng struct btrfs_root *log_root; 12577237f183SYan Zheng struct btrfs_inode_item *inode_item; 12587237f183SYan Zheng 12590b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 12607237f183SYan Zheng if (IS_ERR(log_root)) 12617237f183SYan Zheng return PTR_ERR(log_root); 12627237f183SYan Zheng 12637237f183SYan Zheng log_root->last_trans = trans->transid; 12647237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 12657237f183SYan Zheng 12667237f183SYan Zheng inode_item = &log_root->root_item.inode; 12673cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 12683cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 12693cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1270da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 12710b246afaSJeff Mahoney fs_info->nodesize); 12723cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 12737237f183SYan Zheng 12745d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 12757237f183SYan Zheng 12767237f183SYan Zheng WARN_ON(root->log_root); 12777237f183SYan Zheng root->log_root = log_root; 12787237f183SYan Zheng root->log_transid = 0; 1279d1433debSMiao Xie root->log_transid_committed = -1; 1280257c62e1SChris Mason root->last_log_commit = 0; 1281e02119d5SChris Mason return 0; 1282e02119d5SChris Mason } 1283e02119d5SChris Mason 128449d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 128549d11beaSJosef Bacik struct btrfs_path *path, 1286cb517eabSMiao Xie struct btrfs_key *key) 1287e02119d5SChris Mason { 1288e02119d5SChris Mason struct btrfs_root *root; 1289e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 129084234f3aSYan Zheng u64 generation; 1291cb517eabSMiao Xie int ret; 1292581c1760SQu Wenruo int level; 1293cb517eabSMiao Xie 129496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 129549d11beaSJosef Bacik if (!root) 129649d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 12970f7d52f4SChris Mason 1298cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1299cb517eabSMiao Xie &root->root_item, &root->root_key); 13000f7d52f4SChris Mason if (ret) { 130113a8a7c8SYan, Zheng if (ret > 0) 130213a8a7c8SYan, Zheng ret = -ENOENT; 130349d11beaSJosef Bacik goto fail; 13040f7d52f4SChris Mason } 130513a8a7c8SYan, Zheng 130684234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1307581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 13082ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 13092ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 1310581c1760SQu Wenruo generation, level, NULL); 131164c043deSLiu Bo if (IS_ERR(root->node)) { 131264c043deSLiu Bo ret = PTR_ERR(root->node); 13138c38938cSJosef Bacik root->node = NULL; 131449d11beaSJosef Bacik goto fail; 1315cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1316cb517eabSMiao Xie ret = -EIO; 131749d11beaSJosef Bacik goto fail; 1318416bc658SJosef Bacik } 13195d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1320cb517eabSMiao Xie return root; 132149d11beaSJosef Bacik fail: 132200246528SJosef Bacik btrfs_put_root(root); 132349d11beaSJosef Bacik return ERR_PTR(ret); 132449d11beaSJosef Bacik } 132549d11beaSJosef Bacik 132649d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 132749d11beaSJosef Bacik struct btrfs_key *key) 132849d11beaSJosef Bacik { 132949d11beaSJosef Bacik struct btrfs_root *root; 133049d11beaSJosef Bacik struct btrfs_path *path; 133149d11beaSJosef Bacik 133249d11beaSJosef Bacik path = btrfs_alloc_path(); 133349d11beaSJosef Bacik if (!path) 133449d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 133549d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 133649d11beaSJosef Bacik btrfs_free_path(path); 133749d11beaSJosef Bacik 133849d11beaSJosef Bacik return root; 1339cb517eabSMiao Xie } 1340cb517eabSMiao Xie 13412dfb1e43SQu Wenruo /* 13422dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 13432dfb1e43SQu Wenruo * 13442dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 13452dfb1e43SQu Wenruo */ 13462dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1347cb517eabSMiao Xie { 1348cb517eabSMiao Xie int ret; 1349dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1350cb517eabSMiao Xie 1351cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1352cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1353cb517eabSMiao Xie GFP_NOFS); 1354cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1355cb517eabSMiao Xie ret = -ENOMEM; 1356cb517eabSMiao Xie goto fail; 1357cb517eabSMiao Xie } 1358cb517eabSMiao Xie 1359dcc3eb96SNikolay Borisov /* 1360dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1361dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1362dcc3eb96SNikolay Borisov */ 1363dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1364dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1365dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1366dcc3eb96SNikolay Borisov if (ret) 13678257b2dcSMiao Xie goto fail; 13688257b2dcSMiao Xie 1369aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 1370aeb935a4SQu Wenruo root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) { 137192a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1372f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1373f39e4571SJosef Bacik } 1374f39e4571SJosef Bacik 1375cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 137657cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 137757cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1378cb517eabSMiao Xie 1379851fd730SQu Wenruo /* 1380851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1381851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1382851fd730SQu Wenruo */ 1383851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1384851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 13852dfb1e43SQu Wenruo if (!anon_dev) { 1386cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1387cb517eabSMiao Xie if (ret) 1388876d2cf1SLiu Bo goto fail; 13892dfb1e43SQu Wenruo } else { 13902dfb1e43SQu Wenruo root->anon_dev = anon_dev; 13912dfb1e43SQu Wenruo } 1392851fd730SQu Wenruo } 1393f32e48e9SChandan Rajendra 1394f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1395f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1396f32e48e9SChandan Rajendra &root->highest_objectid); 1397f32e48e9SChandan Rajendra if (ret) { 1398f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1399876d2cf1SLiu Bo goto fail; 1400f32e48e9SChandan Rajendra } 1401f32e48e9SChandan Rajendra 1402f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1403f32e48e9SChandan Rajendra 1404f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1405f32e48e9SChandan Rajendra 1406cb517eabSMiao Xie return 0; 1407cb517eabSMiao Xie fail: 140884db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1409cb517eabSMiao Xie return ret; 1410cb517eabSMiao Xie } 1411cb517eabSMiao Xie 1412a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1413cb517eabSMiao Xie u64 root_id) 1414cb517eabSMiao Xie { 1415cb517eabSMiao Xie struct btrfs_root *root; 1416cb517eabSMiao Xie 1417cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1418cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1419cb517eabSMiao Xie (unsigned long)root_id); 1420bc44d7c4SJosef Bacik if (root) 142100246528SJosef Bacik root = btrfs_grab_root(root); 1422cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1423cb517eabSMiao Xie return root; 1424cb517eabSMiao Xie } 1425cb517eabSMiao Xie 142649d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 142749d11beaSJosef Bacik u64 objectid) 142849d11beaSJosef Bacik { 142949d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 143049d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 143149d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 143249d11beaSJosef Bacik return btrfs_grab_root(fs_info->extent_root); 143349d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 143449d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 143549d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 143649d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 143749d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 143849d11beaSJosef Bacik return btrfs_grab_root(fs_info->csum_root); 143949d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 144049d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 144149d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 144249d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 144349d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 144449d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 144549d11beaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 144649d11beaSJosef Bacik return btrfs_grab_root(fs_info->free_space_root) ? 144749d11beaSJosef Bacik fs_info->free_space_root : ERR_PTR(-ENOENT); 144849d11beaSJosef Bacik return NULL; 144949d11beaSJosef Bacik } 145049d11beaSJosef Bacik 1451cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1452cb517eabSMiao Xie struct btrfs_root *root) 1453cb517eabSMiao Xie { 1454cb517eabSMiao Xie int ret; 1455cb517eabSMiao Xie 1456e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1457cb517eabSMiao Xie if (ret) 1458cb517eabSMiao Xie return ret; 1459cb517eabSMiao Xie 1460cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1461cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1462cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1463cb517eabSMiao Xie root); 1464af01d2e5SJosef Bacik if (ret == 0) { 146500246528SJosef Bacik btrfs_grab_root(root); 146627cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1467af01d2e5SJosef Bacik } 1468cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1469cb517eabSMiao Xie radix_tree_preload_end(); 1470cb517eabSMiao Xie 1471cb517eabSMiao Xie return ret; 1472cb517eabSMiao Xie } 1473cb517eabSMiao Xie 1474bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1475bd647ce3SJosef Bacik { 1476bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1477bd647ce3SJosef Bacik struct btrfs_root *root; 1478bd647ce3SJosef Bacik 1479bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1480457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1481457f1864SJosef Bacik 1482bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1483bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1484457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 1485457f1864SJosef Bacik btrfs_root_name(root->root_key.objectid, buf), 1486bd647ce3SJosef Bacik refcount_read(&root->refs)); 1487bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 148800246528SJosef Bacik btrfs_put_root(root); 148900246528SJosef Bacik btrfs_put_root(root); 1490bd647ce3SJosef Bacik } 1491bd647ce3SJosef Bacik #endif 1492bd647ce3SJosef Bacik } 1493bd647ce3SJosef Bacik 14940d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 14950d4b0463SJosef Bacik { 1496141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1497141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 1498141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dio_bytes); 1499141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1500141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1501141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1502141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 15030d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 15040d4b0463SJosef Bacik kfree(fs_info->delayed_root); 150500246528SJosef Bacik btrfs_put_root(fs_info->extent_root); 150600246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 150700246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 150800246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 150900246528SJosef Bacik btrfs_put_root(fs_info->csum_root); 151000246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 151100246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 151200246528SJosef Bacik btrfs_put_root(fs_info->free_space_root); 151300246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1514aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 1515bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 15163fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 15170d4b0463SJosef Bacik kfree(fs_info->super_copy); 15180d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 15190d4b0463SJosef Bacik kvfree(fs_info); 15200d4b0463SJosef Bacik } 15210d4b0463SJosef Bacik 15220d4b0463SJosef Bacik 15232dfb1e43SQu Wenruo /* 15242dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 15252dfb1e43SQu Wenruo * 15262dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 15272dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 15282dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 15292dfb1e43SQu Wenruo * 15302dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 15312dfb1e43SQu Wenruo * 15322dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 15332dfb1e43SQu Wenruo * same root objectid. 15342dfb1e43SQu Wenruo * 15352dfb1e43SQu Wenruo * @objectid: root id 15362dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 15372dfb1e43SQu Wenruo * pass 0 for new allocation. 15382dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 15392dfb1e43SQu Wenruo * for orphan roots 15402dfb1e43SQu Wenruo */ 15412dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 15422dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 15432dfb1e43SQu Wenruo bool check_ref) 15445eda7b5eSChris Mason { 15455eda7b5eSChris Mason struct btrfs_root *root; 1546381cf658SDavid Sterba struct btrfs_path *path; 15471d4c08e0SDavid Sterba struct btrfs_key key; 15485eda7b5eSChris Mason int ret; 15495eda7b5eSChris Mason 155049d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 155149d11beaSJosef Bacik if (root) 155249d11beaSJosef Bacik return root; 15534df27c4dSYan, Zheng again: 155456e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 155548475471SStefan Behrens if (root) { 15562dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 15572dfb1e43SQu Wenruo ASSERT(!anon_dev); 1558bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 155900246528SJosef Bacik btrfs_put_root(root); 156048475471SStefan Behrens return ERR_PTR(-ENOENT); 1561bc44d7c4SJosef Bacik } 15625eda7b5eSChris Mason return root; 156348475471SStefan Behrens } 15645eda7b5eSChris Mason 156556e9357aSDavid Sterba key.objectid = objectid; 156656e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 156756e9357aSDavid Sterba key.offset = (u64)-1; 156856e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 15695eda7b5eSChris Mason if (IS_ERR(root)) 15705eda7b5eSChris Mason return root; 15713394e160SChris Mason 1572c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1573d68fc57bSYan, Zheng ret = -ENOENT; 1574d68fc57bSYan, Zheng goto fail; 1575d68fc57bSYan, Zheng } 1576d68fc57bSYan, Zheng 15772dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1578cb517eabSMiao Xie if (ret) 1579cb517eabSMiao Xie goto fail; 1580cb517eabSMiao Xie 1581381cf658SDavid Sterba path = btrfs_alloc_path(); 1582381cf658SDavid Sterba if (!path) { 1583381cf658SDavid Sterba ret = -ENOMEM; 1584381cf658SDavid Sterba goto fail; 1585381cf658SDavid Sterba } 15861d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 15871d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 158856e9357aSDavid Sterba key.offset = objectid; 15891d4c08e0SDavid Sterba 15901d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1591381cf658SDavid Sterba btrfs_free_path(path); 1592d68fc57bSYan, Zheng if (ret < 0) 1593d68fc57bSYan, Zheng goto fail; 1594d68fc57bSYan, Zheng if (ret == 0) 159527cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1596d68fc57bSYan, Zheng 1597cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 15980f7d52f4SChris Mason if (ret) { 159900246528SJosef Bacik btrfs_put_root(root); 16004785e24fSJosef Bacik if (ret == -EEXIST) 16014df27c4dSYan, Zheng goto again; 16024df27c4dSYan, Zheng goto fail; 16034df27c4dSYan, Zheng } 1604edbd8d4eSChris Mason return root; 16054df27c4dSYan, Zheng fail: 16068c38938cSJosef Bacik btrfs_put_root(root); 16074df27c4dSYan, Zheng return ERR_PTR(ret); 1608edbd8d4eSChris Mason } 1609edbd8d4eSChris Mason 16102dfb1e43SQu Wenruo /* 16112dfb1e43SQu Wenruo * Get in-memory reference of a root structure 16122dfb1e43SQu Wenruo * 16132dfb1e43SQu Wenruo * @objectid: tree objectid 16142dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 16152dfb1e43SQu Wenruo * one reference 16162dfb1e43SQu Wenruo */ 16172dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 16182dfb1e43SQu Wenruo u64 objectid, bool check_ref) 16192dfb1e43SQu Wenruo { 16202dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 16212dfb1e43SQu Wenruo } 16222dfb1e43SQu Wenruo 16232dfb1e43SQu Wenruo /* 16242dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 16252dfb1e43SQu Wenruo * the anonymous block device id 16262dfb1e43SQu Wenruo * 16272dfb1e43SQu Wenruo * @objectid: tree objectid 16282dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 16292dfb1e43SQu Wenruo * parameter value 16302dfb1e43SQu Wenruo */ 16312dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 16322dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 16332dfb1e43SQu Wenruo { 16342dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 16352dfb1e43SQu Wenruo } 16362dfb1e43SQu Wenruo 16378b712842SChris Mason /* 163849d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 163949d11beaSJosef Bacik * @fs_info: the fs_info 164049d11beaSJosef Bacik * @objectid: the objectid we need to lookup 164149d11beaSJosef Bacik * 164249d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 164349d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 164449d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 164549d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 164649d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 164749d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 164849d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 164949d11beaSJosef Bacik * backref code is finished using the root. 165049d11beaSJosef Bacik */ 165149d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 165249d11beaSJosef Bacik struct btrfs_path *path, 165349d11beaSJosef Bacik u64 objectid) 165449d11beaSJosef Bacik { 165549d11beaSJosef Bacik struct btrfs_root *root; 165649d11beaSJosef Bacik struct btrfs_key key; 165749d11beaSJosef Bacik 165849d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 165949d11beaSJosef Bacik 166049d11beaSJosef Bacik /* 166149d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 166249d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 166349d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 166449d11beaSJosef Bacik * != NULL just return it. 166549d11beaSJosef Bacik */ 166649d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 166749d11beaSJosef Bacik if (root) 166849d11beaSJosef Bacik return root; 166949d11beaSJosef Bacik 167049d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 167149d11beaSJosef Bacik if (root) 167249d11beaSJosef Bacik return root; 167349d11beaSJosef Bacik 167449d11beaSJosef Bacik key.objectid = objectid; 167549d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 167649d11beaSJosef Bacik key.offset = (u64)-1; 167749d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 167849d11beaSJosef Bacik btrfs_release_path(path); 167949d11beaSJosef Bacik 168049d11beaSJosef Bacik return root; 168149d11beaSJosef Bacik } 168249d11beaSJosef Bacik 168349d11beaSJosef Bacik /* 16848b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 16858b712842SChris Mason * functions. This is where read checksum verification actually happens 16868b712842SChris Mason */ 16878b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1688ce9adaa5SChris Mason { 1689ce9adaa5SChris Mason struct bio *bio; 169097eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1691ce9adaa5SChris Mason 169297eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1693ce9adaa5SChris Mason bio = end_io_wq->bio; 16948b712842SChris Mason 16954e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1696ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1697ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 16984246a0b6SChristoph Hellwig bio_endio(bio); 16999be490f1SOmar Sandoval kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 1700ce9adaa5SChris Mason } 170144b8bd7eSChris Mason 1702a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1703a74a4b97SChris Mason { 1704a74a4b97SChris Mason struct btrfs_root *root = arg; 17050b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1706d0278245SMiao Xie int again; 1707a74a4b97SChris Mason 1708d6fd0ae2SOmar Sandoval while (1) { 1709d0278245SMiao Xie again = 0; 17109d1a2a3aSDavid Sterba 1711fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1712fd340d0fSJosef Bacik 1713d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 17142ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1715d0278245SMiao Xie goto sleep; 1716d0278245SMiao Xie 171790c711abSZygo Blaxell /* 171890c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 171990c711abSZygo Blaxell * before we have finished mounting the filesystem. 172090c711abSZygo Blaxell */ 17210b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 172290c711abSZygo Blaxell goto sleep; 172390c711abSZygo Blaxell 17240b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1725d0278245SMiao Xie goto sleep; 1726d0278245SMiao Xie 1727dc7f370cSMiao Xie /* 1728dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1729dc7f370cSMiao Xie * during the above check and trylock. 1730dc7f370cSMiao Xie */ 17312ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 17320b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1733dc7f370cSMiao Xie goto sleep; 1734dc7f370cSMiao Xie } 1735dc7f370cSMiao Xie 17362ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1737c2d6cb16SFilipe Manana 17389d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 17390b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1740a74a4b97SChris Mason 1741d0278245SMiao Xie /* 174205323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 174305323cd1SMiao Xie * needn't do anything special here. 1744d0278245SMiao Xie */ 17450b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 174667c5e7d4SFilipe Manana 174767c5e7d4SFilipe Manana /* 174867c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 174967c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 175067c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 175167c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 175267c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 175367c5e7d4SFilipe Manana * unused block groups. 175467c5e7d4SFilipe Manana */ 17550b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 1756d0278245SMiao Xie sleep: 1757fd340d0fSJosef Bacik clear_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1758d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1759d6fd0ae2SOmar Sandoval kthread_parkme(); 1760d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1761d6fd0ae2SOmar Sandoval return 0; 1762838fe188SJiri Kosina if (!again) { 1763a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1764a74a4b97SChris Mason schedule(); 1765a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1766a74a4b97SChris Mason } 1767da288d28SFilipe Manana } 1768a74a4b97SChris Mason } 1769a74a4b97SChris Mason 1770a74a4b97SChris Mason static int transaction_kthread(void *arg) 1771a74a4b97SChris Mason { 1772a74a4b97SChris Mason struct btrfs_root *root = arg; 17730b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1774a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1775a74a4b97SChris Mason struct btrfs_transaction *cur; 17768929ecfaSYan, Zheng u64 transid; 1777*643900beSNikolay Borisov time64_t delta; 1778a74a4b97SChris Mason unsigned long delay; 1779914b2007SJan Kara bool cannot_commit; 1780a74a4b97SChris Mason 1781a74a4b97SChris Mason do { 1782914b2007SJan Kara cannot_commit = false; 1783ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 17840b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1785a74a4b97SChris Mason 17860b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 17870b246afaSJeff Mahoney cur = fs_info->running_transaction; 1788a74a4b97SChris Mason if (!cur) { 17890b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1790a74a4b97SChris Mason goto sleep; 1791a74a4b97SChris Mason } 179231153d81SYan Zheng 1793*643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 17943296bf56SQu Wenruo if (cur->state < TRANS_STATE_COMMIT_START && 1795*643900beSNikolay Borisov delta < fs_info->commit_interval) { 17960b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1797ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(5000); 1798a74a4b97SChris Mason goto sleep; 1799a74a4b97SChris Mason } 18008929ecfaSYan, Zheng transid = cur->transid; 18010b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 180256bec294SChris Mason 180379787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1804354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1805914b2007SJan Kara if (IS_ERR(trans)) { 1806354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1807914b2007SJan Kara cannot_commit = true; 180879787eaaSJeff Mahoney goto sleep; 1809914b2007SJan Kara } 18108929ecfaSYan, Zheng if (transid == trans->transid) { 18113a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 18128929ecfaSYan, Zheng } else { 18133a45bb20SJeff Mahoney btrfs_end_transaction(trans); 18148929ecfaSYan, Zheng } 1815a74a4b97SChris Mason sleep: 18160b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 18170b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1818a74a4b97SChris Mason 18194e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 18200b246afaSJeff Mahoney &fs_info->fs_state))) 18212ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 18228929ecfaSYan, Zheng if (!kthread_should_stop() && 18230b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1824914b2007SJan Kara cannot_commit)) 1825bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1826a74a4b97SChris Mason } while (!kthread_should_stop()); 1827a74a4b97SChris Mason return 0; 1828a74a4b97SChris Mason } 1829a74a4b97SChris Mason 1830af31f5e5SChris Mason /* 183101f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 183201f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 183301f0f9daSNikolay Borisov * anything. 1834af31f5e5SChris Mason * 1835af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1836af31f5e5SChris Mason * generation of the latest root in the array with the generation 1837af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1838af31f5e5SChris Mason */ 183901f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1840af31f5e5SChris Mason { 184101f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1842af31f5e5SChris Mason u64 cur; 1843af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1844af31f5e5SChris Mason int i; 1845af31f5e5SChris Mason 1846af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1847af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1848af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1849af31f5e5SChris Mason if (cur == newest_gen) 185001f0f9daSNikolay Borisov return i; 1851af31f5e5SChris Mason } 1852af31f5e5SChris Mason 185301f0f9daSNikolay Borisov return -EINVAL; 1854af31f5e5SChris Mason } 1855af31f5e5SChris Mason 1856af31f5e5SChris Mason /* 1857af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1858af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1859af31f5e5SChris Mason * done 1860af31f5e5SChris Mason */ 1861af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1862af31f5e5SChris Mason { 18636ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1864af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1865af31f5e5SChris Mason 1866af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1867af31f5e5SChris Mason 1868af31f5e5SChris Mason /* 1869af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1870af31f5e5SChris Mason * regardless of which ones we use today 1871af31f5e5SChris Mason */ 1872af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1873af31f5e5SChris Mason 1874af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1875af31f5e5SChris Mason 1876af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1877af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1878af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1879af31f5e5SChris Mason 1880af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1881af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1882af31f5e5SChris Mason 1883af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1884af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1885af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1886af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1887af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1888af31f5e5SChris Mason 1889af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1890af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1891af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1892af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1893af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1894af31f5e5SChris Mason 18957c7e82a7SChris Mason /* 18967c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 18977c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 18987c7e82a7SChris Mason */ 18997c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 19007c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 19017c7e82a7SChris Mason info->fs_root->node->start); 1902af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1903af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1904af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1905af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19067c7e82a7SChris Mason } 1907af31f5e5SChris Mason 1908af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1909af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1910af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1911af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1912af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1913af31f5e5SChris Mason 1914af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1915af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1916af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1917af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1918af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1919af31f5e5SChris Mason 1920af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1921af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1922af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1923af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1924af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1925af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1926af31f5e5SChris Mason 1927af31f5e5SChris Mason /* 1928af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1929af31f5e5SChris Mason * for the next commit 1930af31f5e5SChris Mason */ 1931af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1932af31f5e5SChris Mason &info->super_for_commit->super_roots, 1933af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1934af31f5e5SChris Mason } 1935af31f5e5SChris Mason 1936af31f5e5SChris Mason /* 1937bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1938bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1939bd2336b2SNikolay Borisov * 1940bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1941bd2336b2SNikolay Borisov * priority - priority of backup root required 1942bd2336b2SNikolay Borisov * 1943bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1944bd2336b2SNikolay Borisov */ 1945bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1946bd2336b2SNikolay Borisov { 1947bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1948bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1949bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1950bd2336b2SNikolay Borisov 1951bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1952bd2336b2SNikolay Borisov if (priority == 0) 1953bd2336b2SNikolay Borisov return backup_index; 1954bd2336b2SNikolay Borisov 1955bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1956bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1957bd2336b2SNikolay Borisov } else { 1958bd2336b2SNikolay Borisov return -EINVAL; 1959bd2336b2SNikolay Borisov } 1960bd2336b2SNikolay Borisov 1961bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1962bd2336b2SNikolay Borisov 1963bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1964bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1965bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1966bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1967bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1968bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1969bd2336b2SNikolay Borisov 1970bd2336b2SNikolay Borisov /* 1971bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1972bd2336b2SNikolay Borisov * need a fsck 1973bd2336b2SNikolay Borisov */ 1974bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1975bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1976bd2336b2SNikolay Borisov 1977bd2336b2SNikolay Borisov return backup_index; 1978bd2336b2SNikolay Borisov } 1979bd2336b2SNikolay Borisov 19807abadb64SLiu Bo /* helper to cleanup workers */ 19817abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 19827abadb64SLiu Bo { 1983dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1984afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 19855cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1986fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 1987fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 1988d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 1989fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1990fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 19915b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1992e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1993736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 1994a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1995fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1996b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1997b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1998a9b9477dSFilipe Manana /* 1999a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2000a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2001a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2002a9b9477dSFilipe Manana */ 2003a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2004a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 20057abadb64SLiu Bo } 20067abadb64SLiu Bo 20072e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20082e9f5954SRashika { 20092e9f5954SRashika if (root) { 20102e9f5954SRashika free_extent_buffer(root->node); 20112e9f5954SRashika free_extent_buffer(root->commit_root); 20122e9f5954SRashika root->node = NULL; 20132e9f5954SRashika root->commit_root = NULL; 20142e9f5954SRashika } 20152e9f5954SRashika } 20162e9f5954SRashika 2017af31f5e5SChris Mason /* helper to cleanup tree roots */ 20184273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2019af31f5e5SChris Mason { 20202e9f5954SRashika free_root_extent_buffers(info->tree_root); 2021655b09feSJosef Bacik 20222e9f5954SRashika free_root_extent_buffers(info->dev_root); 20232e9f5954SRashika free_root_extent_buffers(info->extent_root); 20242e9f5954SRashika free_root_extent_buffers(info->csum_root); 20252e9f5954SRashika free_root_extent_buffers(info->quota_root); 20262e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20278c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2028aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 20294273eaffSAnand Jain if (free_chunk_root) 20302e9f5954SRashika free_root_extent_buffers(info->chunk_root); 203170f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2032af31f5e5SChris Mason } 2033af31f5e5SChris Mason 20348c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 20358c38938cSJosef Bacik { 20368c38938cSJosef Bacik if (!root) 20378c38938cSJosef Bacik return; 20388c38938cSJosef Bacik 20398c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 20408c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 20411dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 20428c38938cSJosef Bacik if (root->anon_dev) 20438c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 20448c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2045923eb523SJohannes Thumshirn free_root_extent_buffers(root); 20468c38938cSJosef Bacik kfree(root->free_ino_ctl); 20478c38938cSJosef Bacik kfree(root->free_ino_pinned); 20488c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 20498c38938cSJosef Bacik spin_lock(&root->fs_info->fs_roots_radix_lock); 20508c38938cSJosef Bacik list_del_init(&root->leak_list); 20518c38938cSJosef Bacik spin_unlock(&root->fs_info->fs_roots_radix_lock); 20528c38938cSJosef Bacik #endif 20538c38938cSJosef Bacik kfree(root); 20548c38938cSJosef Bacik } 20558c38938cSJosef Bacik } 20568c38938cSJosef Bacik 2057faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2058171f6537SJosef Bacik { 2059171f6537SJosef Bacik int ret; 2060171f6537SJosef Bacik struct btrfs_root *gang[8]; 2061171f6537SJosef Bacik int i; 2062171f6537SJosef Bacik 2063171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2064171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2065171f6537SJosef Bacik struct btrfs_root, root_list); 2066171f6537SJosef Bacik list_del(&gang[0]->root_list); 2067171f6537SJosef Bacik 20688c38938cSJosef Bacik if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2069cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 207000246528SJosef Bacik btrfs_put_root(gang[0]); 2071171f6537SJosef Bacik } 2072171f6537SJosef Bacik 2073171f6537SJosef Bacik while (1) { 2074171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2075171f6537SJosef Bacik (void **)gang, 0, 2076171f6537SJosef Bacik ARRAY_SIZE(gang)); 2077171f6537SJosef Bacik if (!ret) 2078171f6537SJosef Bacik break; 2079171f6537SJosef Bacik for (i = 0; i < ret; i++) 2080cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2081171f6537SJosef Bacik } 2082171f6537SJosef Bacik } 2083af31f5e5SChris Mason 2084638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2085638aa7edSEric Sandeen { 2086638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2087638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2088638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2089638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2090638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2091638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2092ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2093638aa7edSEric Sandeen } 2094638aa7edSEric Sandeen 2095779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2096779a65a4SEric Sandeen { 2097779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2098779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2099779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2100779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2101779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2102779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2103779a65a4SEric Sandeen } 2104779a65a4SEric Sandeen 21056bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2106f37938e0SEric Sandeen { 21072ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 21082ff7e61eSJeff Mahoney 21092ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 21102ff7e61eSJeff Mahoney set_nlink(inode, 1); 2111f37938e0SEric Sandeen /* 2112f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2113f37938e0SEric Sandeen * the real end of the address space is determined by all of 2114f37938e0SEric Sandeen * the devices in the system 2115f37938e0SEric Sandeen */ 21162ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 21172ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2118f37938e0SEric Sandeen 21192ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 212043eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 21212c53a14dSQu Wenruo IO_TREE_BTREE_INODE_IO, inode); 21227b439738SDavid Sterba BTRFS_I(inode)->io_tree.track_uptodate = false; 21232ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2124f37938e0SEric Sandeen 21255c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 21262ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 21272ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 21282ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2129f37938e0SEric Sandeen } 2130f37938e0SEric Sandeen 2131ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2132ad618368SEric Sandeen { 2133ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2134129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 21357f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2136ad618368SEric Sandeen } 2137ad618368SEric Sandeen 2138f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2139f9e92e40SEric Sandeen { 2140f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2141f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2142f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2143f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2144f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2145f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2146d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2147f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2148f9e92e40SEric Sandeen } 2149f9e92e40SEric Sandeen 21502a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 21512a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 21522a458198SEric Sandeen { 2153f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 21546f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 21552a458198SEric Sandeen 21562a458198SEric Sandeen fs_info->workers = 2157cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2158cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 21592a458198SEric Sandeen 21602a458198SEric Sandeen fs_info->delalloc_workers = 2161cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2162cb001095SJeff Mahoney flags, max_active, 2); 21632a458198SEric Sandeen 21642a458198SEric Sandeen fs_info->flush_workers = 2165cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2166cb001095SJeff Mahoney flags, max_active, 0); 21672a458198SEric Sandeen 21682a458198SEric Sandeen fs_info->caching_workers = 2169cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 21702a458198SEric Sandeen 21712a458198SEric Sandeen fs_info->fixup_workers = 2172cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 21732a458198SEric Sandeen 21742a458198SEric Sandeen /* 21752a458198SEric Sandeen * endios are largely parallel and should have a very 21762a458198SEric Sandeen * low idle thresh 21772a458198SEric Sandeen */ 21782a458198SEric Sandeen fs_info->endio_workers = 2179cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 21802a458198SEric Sandeen fs_info->endio_meta_workers = 2181cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2182cb001095SJeff Mahoney max_active, 4); 21832a458198SEric Sandeen fs_info->endio_meta_write_workers = 2184cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2185cb001095SJeff Mahoney max_active, 2); 21862a458198SEric Sandeen fs_info->endio_raid56_workers = 2187cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2188cb001095SJeff Mahoney max_active, 4); 21892a458198SEric Sandeen fs_info->rmw_workers = 2190cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 21912a458198SEric Sandeen fs_info->endio_write_workers = 2192cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2193cb001095SJeff Mahoney max_active, 2); 21942a458198SEric Sandeen fs_info->endio_freespace_worker = 2195cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2196cb001095SJeff Mahoney max_active, 0); 21972a458198SEric Sandeen fs_info->delayed_workers = 2198cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2199cb001095SJeff Mahoney max_active, 0); 22002a458198SEric Sandeen fs_info->readahead_workers = 2201cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2202cb001095SJeff Mahoney max_active, 2); 22032a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2204cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2205b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2206b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 22072a458198SEric Sandeen 22082a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 2209ba8a9d07SChris Mason fs_info->flush_workers && 22102a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 22112a458198SEric Sandeen fs_info->endio_meta_write_workers && 22122a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 22132a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 22142a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 22152a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 2216b0643e59SDennis Zhou fs_info->qgroup_rescan_workers && 2217b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 22182a458198SEric Sandeen return -ENOMEM; 22192a458198SEric Sandeen } 22202a458198SEric Sandeen 22212a458198SEric Sandeen return 0; 22222a458198SEric Sandeen } 22232a458198SEric Sandeen 22246d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 22256d97c6e3SJohannes Thumshirn { 22266d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2227b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 22286d97c6e3SJohannes Thumshirn 2229b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 22306d97c6e3SJohannes Thumshirn 22316d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 22326d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2233b4e967beSDavid Sterba csum_driver); 22346d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 22356d97c6e3SJohannes Thumshirn } 22366d97c6e3SJohannes Thumshirn 22376d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 22386d97c6e3SJohannes Thumshirn 22396d97c6e3SJohannes Thumshirn return 0; 22406d97c6e3SJohannes Thumshirn } 22416d97c6e3SJohannes Thumshirn 224263443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 224363443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 224463443bf5SEric Sandeen { 224563443bf5SEric Sandeen int ret; 224663443bf5SEric Sandeen struct btrfs_root *log_tree_root; 224763443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 224863443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2249581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 225063443bf5SEric Sandeen 225163443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2252f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 225363443bf5SEric Sandeen return -EIO; 225463443bf5SEric Sandeen } 225563443bf5SEric Sandeen 225696dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 225796dfcb46SJosef Bacik GFP_KERNEL); 225863443bf5SEric Sandeen if (!log_tree_root) 225963443bf5SEric Sandeen return -ENOMEM; 226063443bf5SEric Sandeen 22612ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 2262581c1760SQu Wenruo fs_info->generation + 1, 2263581c1760SQu Wenruo level, NULL); 226464c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2265f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 22660eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 22678c38938cSJosef Bacik log_tree_root->node = NULL; 226800246528SJosef Bacik btrfs_put_root(log_tree_root); 22690eeff236SLiu Bo return ret; 227064c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2271f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 227200246528SJosef Bacik btrfs_put_root(log_tree_root); 227363443bf5SEric Sandeen return -EIO; 227463443bf5SEric Sandeen } 227563443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 227663443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 227763443bf5SEric Sandeen if (ret) { 22780b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 227963443bf5SEric Sandeen "Failed to recover log tree"); 228000246528SJosef Bacik btrfs_put_root(log_tree_root); 228163443bf5SEric Sandeen return ret; 228263443bf5SEric Sandeen } 228363443bf5SEric Sandeen 2284bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 22856bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 228663443bf5SEric Sandeen if (ret) 228763443bf5SEric Sandeen return ret; 228863443bf5SEric Sandeen } 228963443bf5SEric Sandeen 229063443bf5SEric Sandeen return 0; 229163443bf5SEric Sandeen } 229263443bf5SEric Sandeen 22936bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 22944bbcaa64SEric Sandeen { 22956bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2296a4f3d2c4SDavid Sterba struct btrfs_root *root; 22974bbcaa64SEric Sandeen struct btrfs_key location; 22984bbcaa64SEric Sandeen int ret; 22994bbcaa64SEric Sandeen 23006bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 23016bccf3abSJeff Mahoney 23024bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 23034bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 23044bbcaa64SEric Sandeen location.offset = 0; 23054bbcaa64SEric Sandeen 2306a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2307f50f4353SLiu Bo if (IS_ERR(root)) { 2308f50f4353SLiu Bo ret = PTR_ERR(root); 2309f50f4353SLiu Bo goto out; 2310f50f4353SLiu Bo } 2311a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2312a4f3d2c4SDavid Sterba fs_info->extent_root = root; 23134bbcaa64SEric Sandeen 23144bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2315a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2316f50f4353SLiu Bo if (IS_ERR(root)) { 2317f50f4353SLiu Bo ret = PTR_ERR(root); 2318f50f4353SLiu Bo goto out; 2319f50f4353SLiu Bo } 2320a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2321a4f3d2c4SDavid Sterba fs_info->dev_root = root; 23224bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 23234bbcaa64SEric Sandeen 23244bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2325a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2326f50f4353SLiu Bo if (IS_ERR(root)) { 2327f50f4353SLiu Bo ret = PTR_ERR(root); 2328f50f4353SLiu Bo goto out; 2329f50f4353SLiu Bo } 2330a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2331a4f3d2c4SDavid Sterba fs_info->csum_root = root; 23324bbcaa64SEric Sandeen 2333aeb935a4SQu Wenruo /* 2334aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2335aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2336aeb935a4SQu Wenruo */ 233756e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 233856e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2339aeb935a4SQu Wenruo if (IS_ERR(root)) { 2340aeb935a4SQu Wenruo ret = PTR_ERR(root); 2341aeb935a4SQu Wenruo goto out; 2342aeb935a4SQu Wenruo } 2343aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2344aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 2345aeb935a4SQu Wenruo 23464bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2347a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2348a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2349a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2350afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2351a4f3d2c4SDavid Sterba fs_info->quota_root = root; 23524bbcaa64SEric Sandeen } 23534bbcaa64SEric Sandeen 23544bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2355a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2356a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 2357a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 23584bbcaa64SEric Sandeen if (ret != -ENOENT) 2359f50f4353SLiu Bo goto out; 23604bbcaa64SEric Sandeen } else { 2361a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2362a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 23634bbcaa64SEric Sandeen } 23644bbcaa64SEric Sandeen 236570f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 236670f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 236770f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 2368f50f4353SLiu Bo if (IS_ERR(root)) { 2369f50f4353SLiu Bo ret = PTR_ERR(root); 2370f50f4353SLiu Bo goto out; 2371f50f4353SLiu Bo } 237270f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 237370f6d82eSOmar Sandoval fs_info->free_space_root = root; 237470f6d82eSOmar Sandoval } 237570f6d82eSOmar Sandoval 23764bbcaa64SEric Sandeen return 0; 2377f50f4353SLiu Bo out: 2378f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2379f50f4353SLiu Bo location.objectid, ret); 2380f50f4353SLiu Bo return ret; 23814bbcaa64SEric Sandeen } 23824bbcaa64SEric Sandeen 2383069ec957SQu Wenruo /* 2384069ec957SQu Wenruo * Real super block validation 2385069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2386069ec957SQu Wenruo * 2387069ec957SQu Wenruo * @sb: super block to check 2388069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2389069ec957SQu Wenruo * 0 the primary (1st) sb 2390069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2391069ec957SQu Wenruo * -1 skip bytenr check 2392069ec957SQu Wenruo */ 2393069ec957SQu Wenruo static int validate_super(struct btrfs_fs_info *fs_info, 2394069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 239521a852b0SQu Wenruo { 239621a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 239721a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 239821a852b0SQu Wenruo int ret = 0; 239921a852b0SQu Wenruo 240021a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 240121a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 240221a852b0SQu Wenruo ret = -EINVAL; 240321a852b0SQu Wenruo } 240421a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 240521a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 240621a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 240721a852b0SQu Wenruo ret = -EINVAL; 240821a852b0SQu Wenruo } 240921a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 241021a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 241121a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 241221a852b0SQu Wenruo ret = -EINVAL; 241321a852b0SQu Wenruo } 241421a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 241521a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 241621a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 241721a852b0SQu Wenruo ret = -EINVAL; 241821a852b0SQu Wenruo } 241921a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 242021a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 242121a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 242221a852b0SQu Wenruo ret = -EINVAL; 242321a852b0SQu Wenruo } 242421a852b0SQu Wenruo 242521a852b0SQu Wenruo /* 242621a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 242721a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 242821a852b0SQu Wenruo */ 242921a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 243021a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 243121a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 243221a852b0SQu Wenruo ret = -EINVAL; 243321a852b0SQu Wenruo } 243421a852b0SQu Wenruo /* Only PAGE SIZE is supported yet */ 243521a852b0SQu Wenruo if (sectorsize != PAGE_SIZE) { 243621a852b0SQu Wenruo btrfs_err(fs_info, 243721a852b0SQu Wenruo "sectorsize %llu not supported yet, only support %lu", 243821a852b0SQu Wenruo sectorsize, PAGE_SIZE); 243921a852b0SQu Wenruo ret = -EINVAL; 244021a852b0SQu Wenruo } 244121a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 244221a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 244321a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 244421a852b0SQu Wenruo ret = -EINVAL; 244521a852b0SQu Wenruo } 244621a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 244721a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 244821a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 244921a852b0SQu Wenruo ret = -EINVAL; 245021a852b0SQu Wenruo } 245121a852b0SQu Wenruo 245221a852b0SQu Wenruo /* Root alignment check */ 245321a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 245421a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 245521a852b0SQu Wenruo btrfs_super_root(sb)); 245621a852b0SQu Wenruo ret = -EINVAL; 245721a852b0SQu Wenruo } 245821a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 245921a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 246021a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 246121a852b0SQu Wenruo ret = -EINVAL; 246221a852b0SQu Wenruo } 246321a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 246421a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 246521a852b0SQu Wenruo btrfs_super_log_root(sb)); 246621a852b0SQu Wenruo ret = -EINVAL; 246721a852b0SQu Wenruo } 246821a852b0SQu Wenruo 2469de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 24707239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 247121a852b0SQu Wenruo btrfs_err(fs_info, 24727239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2473de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 247421a852b0SQu Wenruo ret = -EINVAL; 247521a852b0SQu Wenruo } 247621a852b0SQu Wenruo 247721a852b0SQu Wenruo /* 247821a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 247921a852b0SQu Wenruo * done later 248021a852b0SQu Wenruo */ 248121a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 248221a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 248321a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 248421a852b0SQu Wenruo ret = -EINVAL; 248521a852b0SQu Wenruo } 248621a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 248721a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 248821a852b0SQu Wenruo btrfs_super_stripesize(sb)); 248921a852b0SQu Wenruo ret = -EINVAL; 249021a852b0SQu Wenruo } 249121a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 249221a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 249321a852b0SQu Wenruo btrfs_super_num_devices(sb)); 249421a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 249521a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 249621a852b0SQu Wenruo ret = -EINVAL; 249721a852b0SQu Wenruo } 249821a852b0SQu Wenruo 2499069ec957SQu Wenruo if (mirror_num >= 0 && 2500069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 250121a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 250221a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 250321a852b0SQu Wenruo ret = -EINVAL; 250421a852b0SQu Wenruo } 250521a852b0SQu Wenruo 250621a852b0SQu Wenruo /* 250721a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 250821a852b0SQu Wenruo * and one chunk 250921a852b0SQu Wenruo */ 251021a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 251121a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 251221a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 251321a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 251421a852b0SQu Wenruo ret = -EINVAL; 251521a852b0SQu Wenruo } 251621a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 251721a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 251821a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 251921a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 252021a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 252121a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 252221a852b0SQu Wenruo ret = -EINVAL; 252321a852b0SQu Wenruo } 252421a852b0SQu Wenruo 252521a852b0SQu Wenruo /* 252621a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 252721a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 252821a852b0SQu Wenruo */ 252921a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 253021a852b0SQu Wenruo btrfs_warn(fs_info, 253121a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 253221a852b0SQu Wenruo btrfs_super_generation(sb), 253321a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 253421a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 253521a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 253621a852b0SQu Wenruo btrfs_warn(fs_info, 253721a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 253821a852b0SQu Wenruo btrfs_super_generation(sb), 253921a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 254021a852b0SQu Wenruo 254121a852b0SQu Wenruo return ret; 254221a852b0SQu Wenruo } 254321a852b0SQu Wenruo 2544069ec957SQu Wenruo /* 2545069ec957SQu Wenruo * Validation of super block at mount time. 2546069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2547069ec957SQu Wenruo * flags will be skipped. 2548069ec957SQu Wenruo */ 2549069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2550069ec957SQu Wenruo { 2551069ec957SQu Wenruo return validate_super(fs_info, fs_info->super_copy, 0); 2552069ec957SQu Wenruo } 2553069ec957SQu Wenruo 255475cb857dSQu Wenruo /* 255575cb857dSQu Wenruo * Validation of super block at write time. 255675cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 255775cb857dSQu Wenruo * overwritten soon. 255875cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 255975cb857dSQu Wenruo */ 256075cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 256175cb857dSQu Wenruo struct btrfs_super_block *sb) 256275cb857dSQu Wenruo { 256375cb857dSQu Wenruo int ret; 256475cb857dSQu Wenruo 256575cb857dSQu Wenruo ret = validate_super(fs_info, sb, -1); 256675cb857dSQu Wenruo if (ret < 0) 256775cb857dSQu Wenruo goto out; 2568e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 256975cb857dSQu Wenruo ret = -EUCLEAN; 257075cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 257175cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 257275cb857dSQu Wenruo goto out; 257375cb857dSQu Wenruo } 257475cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 257575cb857dSQu Wenruo ret = -EUCLEAN; 257675cb857dSQu Wenruo btrfs_err(fs_info, 257775cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 257875cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 257975cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 258075cb857dSQu Wenruo goto out; 258175cb857dSQu Wenruo } 258275cb857dSQu Wenruo out: 258375cb857dSQu Wenruo if (ret < 0) 258475cb857dSQu Wenruo btrfs_err(fs_info, 258575cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 258675cb857dSQu Wenruo return ret; 258775cb857dSQu Wenruo } 258875cb857dSQu Wenruo 25896ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2590b8522a1eSNikolay Borisov { 25916ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2592b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2593b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2594b8522a1eSNikolay Borisov bool handle_error = false; 2595b8522a1eSNikolay Borisov int ret = 0; 2596b8522a1eSNikolay Borisov int i; 2597b8522a1eSNikolay Borisov 2598b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2599b8522a1eSNikolay Borisov u64 generation; 2600b8522a1eSNikolay Borisov int level; 2601b8522a1eSNikolay Borisov 2602b8522a1eSNikolay Borisov if (handle_error) { 2603b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2604b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2605b8522a1eSNikolay Borisov tree_root->node = NULL; 2606b8522a1eSNikolay Borisov 2607b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2608b8522a1eSNikolay Borisov break; 2609b8522a1eSNikolay Borisov 2610b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2611b8522a1eSNikolay Borisov 2612b8522a1eSNikolay Borisov /* 2613b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2614b8522a1eSNikolay Borisov * valid 2615b8522a1eSNikolay Borisov */ 2616b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2617b8522a1eSNikolay Borisov 2618b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2619b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2620b8522a1eSNikolay Borisov 2621b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26226ef108ddSNikolay Borisov backup_index = ret; 2623b8522a1eSNikolay Borisov if (ret < 0) 2624b8522a1eSNikolay Borisov return ret; 2625b8522a1eSNikolay Borisov } 2626b8522a1eSNikolay Borisov generation = btrfs_super_generation(sb); 2627b8522a1eSNikolay Borisov level = btrfs_super_root_level(sb); 2628b8522a1eSNikolay Borisov tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), 2629b8522a1eSNikolay Borisov generation, level, NULL); 26300465337cSJosef Bacik if (IS_ERR(tree_root->node)) { 2631217f5004SNikolay Borisov handle_error = true; 2632b8522a1eSNikolay Borisov ret = PTR_ERR(tree_root->node); 26330465337cSJosef Bacik tree_root->node = NULL; 2634217f5004SNikolay Borisov btrfs_warn(fs_info, "couldn't read tree root"); 2635217f5004SNikolay Borisov continue; 2636b8522a1eSNikolay Borisov 2637217f5004SNikolay Borisov } else if (!extent_buffer_uptodate(tree_root->node)) { 2638217f5004SNikolay Borisov handle_error = true; 2639217f5004SNikolay Borisov ret = -EIO; 2640217f5004SNikolay Borisov btrfs_warn(fs_info, "error while reading tree root"); 2641b8522a1eSNikolay Borisov continue; 2642b8522a1eSNikolay Borisov } 2643b8522a1eSNikolay Borisov 2644b8522a1eSNikolay Borisov btrfs_set_root_node(&tree_root->root_item, tree_root->node); 2645b8522a1eSNikolay Borisov tree_root->commit_root = btrfs_root_node(tree_root); 2646b8522a1eSNikolay Borisov btrfs_set_root_refs(&tree_root->root_item, 1); 2647b8522a1eSNikolay Borisov 2648336a0d8dSNikolay Borisov /* 2649336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2650336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2651336a0d8dSNikolay Borisov */ 2652b8522a1eSNikolay Borisov ret = btrfs_find_highest_objectid(tree_root, 2653b8522a1eSNikolay Borisov &tree_root->highest_objectid); 2654b8522a1eSNikolay Borisov if (ret < 0) { 2655b8522a1eSNikolay Borisov handle_error = true; 2656b8522a1eSNikolay Borisov continue; 2657b8522a1eSNikolay Borisov } 2658b8522a1eSNikolay Borisov 2659b8522a1eSNikolay Borisov ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 2660b8522a1eSNikolay Borisov 2661b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2662b8522a1eSNikolay Borisov if (ret < 0) { 2663b8522a1eSNikolay Borisov handle_error = true; 2664b8522a1eSNikolay Borisov continue; 2665b8522a1eSNikolay Borisov } 2666b8522a1eSNikolay Borisov 2667b8522a1eSNikolay Borisov /* All successful */ 2668b8522a1eSNikolay Borisov fs_info->generation = generation; 2669b8522a1eSNikolay Borisov fs_info->last_trans_committed = generation; 26706ef108ddSNikolay Borisov 26716ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26726ef108ddSNikolay Borisov if (backup_index < 0) { 26736ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26746ef108ddSNikolay Borisov } else { 26756ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26766ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26776ef108ddSNikolay Borisov } 2678b8522a1eSNikolay Borisov break; 2679b8522a1eSNikolay Borisov } 2680b8522a1eSNikolay Borisov 2681b8522a1eSNikolay Borisov return ret; 2682b8522a1eSNikolay Borisov } 2683b8522a1eSNikolay Borisov 26848260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2685eb60ceacSChris Mason { 268676dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2687f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26888fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2689facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 269024bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2691eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 269211833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2693eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2694a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 269576dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 269624bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26974cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2698ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2699f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 270047ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2701f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2702d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 270367c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 27047585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2705573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2706de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 270719c00ddcSChris Mason 27080b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 27096324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2710f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 271147ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 2712bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2713bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 27143fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27153fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2716bd647ce3SJosef Bacik #endif 2717c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 271866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 271966d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 272066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 272166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 272266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 272366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 272466d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2725ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2726ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2727ba2c4d4eSJosef Bacik 2728771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27294cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 27302fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2731034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2732fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 273395ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27349ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27354cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2736a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2737f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27388b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2739f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 274090519d66SArne Jansen /* readahead state */ 2741d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 274290519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2743fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2744c8b97818SChris Mason 2745b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2746b34b086cSChris Mason num_online_cpus() + 2, 8); 27470afbaf8cSChris Mason 2748199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2749199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 275069fe2d75SJosef Bacik 2751638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 275221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 275321adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 275421adbd5cSStefan Behrens #endif 2755779a65a4SEric Sandeen btrfs_init_balance(fs_info); 275657056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2757a2de733cSArne Jansen 27580f9dd46cSJosef Bacik spin_lock_init(&fs_info->block_group_cache_lock); 27596bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2760a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 27610f9dd46cSJosef Bacik 2762fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 2763fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 2764afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 276539279cc3SChris Mason 27665a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2767e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2768925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2769a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2770a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27711bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27729e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2773c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 277476dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2775803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2776fa9c0d79SChris Mason 2777ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2778f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2779b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2780416ac51dSArne Jansen 2781fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2782fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2783fa9c0d79SChris Mason 2784e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2785f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2786bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27874854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2788034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 27893768f368SChris Mason 2790da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2791da17066cSJeff Mahoney fs_info->nodesize = 4096; 2792da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2793da17066cSJeff Mahoney fs_info->stripesize = 4096; 2794da17066cSJeff Mahoney 2795eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2796eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2797eede2bf3SOmar Sandoval 27989e967495SFilipe Manana fs_info->send_in_progress = 0; 27998260edbaSJosef Bacik } 28008260edbaSJosef Bacik 28018260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 28028260edbaSJosef Bacik { 28038260edbaSJosef Bacik int ret; 28048260edbaSJosef Bacik 28058260edbaSJosef Bacik fs_info->sb = sb; 28068260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 28078260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 28089e967495SFilipe Manana 2809c75e8394SJosef Bacik ret = percpu_counter_init(&fs_info->dio_bytes, 0, GFP_KERNEL); 2810ae18c37aSJosef Bacik if (ret) 2811ae18c37aSJosef Bacik return ret; 2812ae18c37aSJosef Bacik 2813ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2814ae18c37aSJosef Bacik if (ret) 2815c75e8394SJosef Bacik return ret; 2816ae18c37aSJosef Bacik 2817ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2818ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2819ae18c37aSJosef Bacik 2820ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2821ae18c37aSJosef Bacik if (ret) 2822c75e8394SJosef Bacik return ret; 2823ae18c37aSJosef Bacik 2824ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2825ae18c37aSJosef Bacik GFP_KERNEL); 2826ae18c37aSJosef Bacik if (ret) 2827c75e8394SJosef Bacik return ret; 2828ae18c37aSJosef Bacik 2829ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2830ae18c37aSJosef Bacik GFP_KERNEL); 2831c75e8394SJosef Bacik if (!fs_info->delayed_root) 2832c75e8394SJosef Bacik return -ENOMEM; 2833ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2834ae18c37aSJosef Bacik 2835c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2836ae18c37aSJosef Bacik } 2837ae18c37aSJosef Bacik 283897f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 283997f4dd09SNikolay Borisov { 284097f4dd09SNikolay Borisov struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; 284197f4dd09SNikolay Borisov int ret; 284297f4dd09SNikolay Borisov 284397f4dd09SNikolay Borisov /* 284497f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 284597f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 284697f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 284797f4dd09SNikolay Borisov */ 284897f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 284997f4dd09SNikolay Borisov if (ret < 0) { 2850c94bec2cSJosef Bacik if (ret != -EINTR) 2851c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2852c94bec2cSJosef Bacik ret); 285397f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 285497f4dd09SNikolay Borisov return ret; 285597f4dd09SNikolay Borisov } 285697f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 285797f4dd09SNikolay Borisov } 285897f4dd09SNikolay Borisov 285997f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 286097f4dd09SNikolay Borisov { 286197f4dd09SNikolay Borisov struct task_struct *task; 286297f4dd09SNikolay Borisov 286397f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 286497f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 286597f4dd09SNikolay Borisov if (IS_ERR(task)) { 286697f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 286797f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 286897f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 286997f4dd09SNikolay Borisov return PTR_ERR(task); 287097f4dd09SNikolay Borisov } 287197f4dd09SNikolay Borisov 287297f4dd09SNikolay Borisov return 0; 287397f4dd09SNikolay Borisov } 287497f4dd09SNikolay Borisov 2875ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 2876ae18c37aSJosef Bacik char *options) 2877ae18c37aSJosef Bacik { 2878ae18c37aSJosef Bacik u32 sectorsize; 2879ae18c37aSJosef Bacik u32 nodesize; 2880ae18c37aSJosef Bacik u32 stripesize; 2881ae18c37aSJosef Bacik u64 generation; 2882ae18c37aSJosef Bacik u64 features; 2883ae18c37aSJosef Bacik u16 csum_type; 2884ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 2885ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2886ae18c37aSJosef Bacik struct btrfs_root *tree_root; 2887ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 2888ae18c37aSJosef Bacik int ret; 2889ae18c37aSJosef Bacik int err = -EINVAL; 2890ae18c37aSJosef Bacik int clear_free_space_tree = 0; 2891ae18c37aSJosef Bacik int level; 2892ae18c37aSJosef Bacik 28938260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 289453b381b3SDavid Woodhouse if (ret) { 289583c8266aSDavid Sterba err = ret; 2896ae18c37aSJosef Bacik goto fail; 289753b381b3SDavid Woodhouse } 289853b381b3SDavid Woodhouse 2899ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 2900ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 2901ae18c37aSJosef Bacik GFP_KERNEL); 2902ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 2903ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 2904ae18c37aSJosef Bacik GFP_KERNEL); 2905ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 2906ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 2907ae18c37aSJosef Bacik err = -ENOMEM; 2908c75e8394SJosef Bacik goto fail; 2909ae18c37aSJosef Bacik } 2910ae18c37aSJosef Bacik 2911ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 2912ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 2913ae18c37aSJosef Bacik err = -ENOMEM; 2914c75e8394SJosef Bacik goto fail; 2915ae18c37aSJosef Bacik } 2916ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 2917ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 2918ae18c37aSJosef Bacik 29193c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 29201104a885SDavid Sterba 29211104a885SDavid Sterba /* 29221104a885SDavid Sterba * Read super block and check the signature bytes only 29231104a885SDavid Sterba */ 29248f32380dSJohannes Thumshirn disk_super = btrfs_read_dev_super(fs_devices->latest_bdev); 29258f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 29268f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 292716cdcec7SMiao Xie goto fail_alloc; 292820b45077SDave Young } 292939279cc3SChris Mason 29301104a885SDavid Sterba /* 2931260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 29328dc3f22cSJohannes Thumshirn * corrupted, we'll find out 29338dc3f22cSJohannes Thumshirn */ 29348f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 293551bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 29368dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 293751bce6c9SJohannes Thumshirn csum_type); 29388dc3f22cSJohannes Thumshirn err = -EINVAL; 29398f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 29408dc3f22cSJohannes Thumshirn goto fail_alloc; 29418dc3f22cSJohannes Thumshirn } 29428dc3f22cSJohannes Thumshirn 29436d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 29446d97c6e3SJohannes Thumshirn if (ret) { 29456d97c6e3SJohannes Thumshirn err = ret; 29468f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 29476d97c6e3SJohannes Thumshirn goto fail_alloc; 29486d97c6e3SJohannes Thumshirn } 29496d97c6e3SJohannes Thumshirn 29508dc3f22cSJohannes Thumshirn /* 29511104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 29521104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 29531104a885SDavid Sterba */ 29548f32380dSJohannes Thumshirn if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { 295505135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 29561104a885SDavid Sterba err = -EINVAL; 29578f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 2958141386e1SJosef Bacik goto fail_alloc; 29591104a885SDavid Sterba } 29601104a885SDavid Sterba 29611104a885SDavid Sterba /* 29621104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 29631104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 29641104a885SDavid Sterba * the whole block of INFO_SIZE 29651104a885SDavid Sterba */ 29668f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 29678f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 29685f39d397SChris Mason 2969fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 2970fbc6feaeSNikolay Borisov 2971de37aa51SNikolay Borisov ASSERT(!memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2972de37aa51SNikolay Borisov BTRFS_FSID_SIZE)); 2973de37aa51SNikolay Borisov 29747239ff4bSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) { 2975de37aa51SNikolay Borisov ASSERT(!memcmp(fs_info->fs_devices->metadata_uuid, 2976de37aa51SNikolay Borisov fs_info->super_copy->metadata_uuid, 2977de37aa51SNikolay Borisov BTRFS_FSID_SIZE)); 29787239ff4bSNikolay Borisov } 29790b86a832SChris Mason 2980fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 2981fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 2982fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 2983fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 2984fbc6feaeSNikolay Borisov btrfs_info(fs_info, 2985fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 2986fbc6feaeSNikolay Borisov } 2987fbc6feaeSNikolay Borisov 2988fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 2989fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 2990de37aa51SNikolay Borisov 2991069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 29921104a885SDavid Sterba if (ret) { 299305135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 29941104a885SDavid Sterba err = -EINVAL; 2995141386e1SJosef Bacik goto fail_alloc; 29961104a885SDavid Sterba } 29971104a885SDavid Sterba 29980f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 2999141386e1SJosef Bacik goto fail_alloc; 30000f7d52f4SChris Mason 3001acce952bSliubo /* check FS state, whether FS is broken. */ 300287533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 300387533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3004acce952bSliubo 300575e7cb7fSLiu Bo /* 300675e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 300775e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 300875e7cb7fSLiu Bo */ 300975e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 301075e7cb7fSLiu Bo 30112ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 30122b82032cSYan Zheng if (ret) { 30132b82032cSYan Zheng err = ret; 3014141386e1SJosef Bacik goto fail_alloc; 30152b82032cSYan Zheng } 3016dfe25020SChris Mason 3017f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 3018f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 3019f2b636e8SJosef Bacik if (features) { 302005135f59SDavid Sterba btrfs_err(fs_info, 302105135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 3022c1c9ff7cSGeert Uytterhoeven features); 3023f2b636e8SJosef Bacik err = -EINVAL; 3024141386e1SJosef Bacik goto fail_alloc; 3025f2b636e8SJosef Bacik } 3026f2b636e8SJosef Bacik 30275d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 30285d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 30290b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3030a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 30315c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 30325c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3033727011e0SChris Mason 30343173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 303505135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 30363173a18fSJosef Bacik 3037727011e0SChris Mason /* 3038727011e0SChris Mason * flag our filesystem as having big metadata blocks if 3039727011e0SChris Mason * they are bigger than the page size 3040727011e0SChris Mason */ 304109cbfeafSKirill A. Shutemov if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 3042727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 304305135f59SDavid Sterba btrfs_info(fs_info, 304405135f59SDavid Sterba "flagging fs with big metadata feature"); 3045727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3046727011e0SChris Mason } 3047727011e0SChris Mason 3048bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 3049bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 3050b7f67055SChandan Rajendra stripesize = sectorsize; 3051707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 3052963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 3053bc3f116fSChris Mason 3054da17066cSJeff Mahoney /* Cache block sizes */ 3055da17066cSJeff Mahoney fs_info->nodesize = nodesize; 3056da17066cSJeff Mahoney fs_info->sectorsize = sectorsize; 3057da17066cSJeff Mahoney fs_info->stripesize = stripesize; 3058da17066cSJeff Mahoney 3059bc3f116fSChris Mason /* 3060bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 3061bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 3062bc3f116fSChris Mason */ 3063bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3064707e8a07SDavid Sterba (sectorsize != nodesize)) { 306505135f59SDavid Sterba btrfs_err(fs_info, 306605135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 306705135f59SDavid Sterba nodesize, sectorsize); 3068141386e1SJosef Bacik goto fail_alloc; 3069bc3f116fSChris Mason } 3070bc3f116fSChris Mason 3071ceda0864SMiao Xie /* 3072ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 3073ceda0864SMiao Xie * update the flag. 3074ceda0864SMiao Xie */ 30755d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 30765d4f98a2SYan Zheng 3077f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 3078f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 3079bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 308005135f59SDavid Sterba btrfs_err(fs_info, 308105135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 3082c1c9ff7cSGeert Uytterhoeven features); 3083f2b636e8SJosef Bacik err = -EINVAL; 3084141386e1SJosef Bacik goto fail_alloc; 3085f2b636e8SJosef Bacik } 308661d92c32SChris Mason 30872a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 30882a458198SEric Sandeen if (ret) { 30892a458198SEric Sandeen err = ret; 30900dc3b84aSJosef Bacik goto fail_sb_buffer; 30910dc3b84aSJosef Bacik } 30924543df7eSChris Mason 30939e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 30949e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 30954575c9ccSChris Mason 3096a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3097a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3098de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3099db94535dSChris Mason 3100925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 31016bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3102925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 310384eed90fSChris Mason if (ret) { 310405135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 31055d4f98a2SYan Zheng goto fail_sb_buffer; 310684eed90fSChris Mason } 31070b86a832SChris Mason 310884234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3109581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 31100b86a832SChris Mason 31112ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 31120b86a832SChris Mason btrfs_super_chunk_root(disk_super), 3113581c1760SQu Wenruo generation, level, NULL); 311464c043deSLiu Bo if (IS_ERR(chunk_root->node) || 311564c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 311605135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3117e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 3118e5fffbacSchandan free_extent_buffer(chunk_root->node); 311995ab1f64SZhao Lei chunk_root->node = NULL; 3120af31f5e5SChris Mason goto fail_tree_roots; 312183121942SDavid Woodhouse } 31225d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 31235d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 31240b86a832SChris Mason 3125e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3126c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3127c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3128e17cade2SChris Mason 31295b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 31302b82032cSYan Zheng if (ret) { 313105135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3132af31f5e5SChris Mason goto fail_tree_roots; 31332b82032cSYan Zheng } 31340b86a832SChris Mason 31358dabb742SStefan Behrens /* 31369b99b115SAnand Jain * Keep the devid that is marked to be the target device for the 31379b99b115SAnand Jain * device replace procedure 31388dabb742SStefan Behrens */ 31399b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 0); 3140dfe25020SChris Mason 3141a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 314205135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3143a6b0d5c8SChris Mason goto fail_tree_roots; 3144a6b0d5c8SChris Mason } 3145a6b0d5c8SChris Mason 3146b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 31474bbcaa64SEric Sandeen if (ret) 3148b8522a1eSNikolay Borisov goto fail_tree_roots; 31498929ecfaSYan, Zheng 315075ec1db8SJosef Bacik /* 315175ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 315275ec1db8SJosef Bacik * check the generation here so we can set the 315375ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 315475ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 315575ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 315675ec1db8SJosef Bacik * even though it was perfectly fine. 315775ec1db8SJosef Bacik */ 315875ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 315975ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 316075ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 316175ec1db8SJosef Bacik 3162cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3163cf90d884SQu Wenruo if (ret) { 3164cf90d884SQu Wenruo btrfs_err(fs_info, 3165cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3166cf90d884SQu Wenruo ret); 3167cf90d884SQu Wenruo goto fail_block_groups; 3168cf90d884SQu Wenruo } 316968310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 317068310a5eSIlya Dryomov if (ret) { 317105135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 317268310a5eSIlya Dryomov goto fail_block_groups; 317368310a5eSIlya Dryomov } 317468310a5eSIlya Dryomov 3175733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3176733f4fbbSStefan Behrens if (ret) { 317705135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3178733f4fbbSStefan Behrens goto fail_block_groups; 3179733f4fbbSStefan Behrens } 3180733f4fbbSStefan Behrens 31818dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 31828dabb742SStefan Behrens if (ret) { 318305135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 31848dabb742SStefan Behrens goto fail_block_groups; 31858dabb742SStefan Behrens } 31868dabb742SStefan Behrens 31879b99b115SAnand Jain btrfs_free_extra_devids(fs_devices, 1); 31888dabb742SStefan Behrens 3189c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3190b7c35e81SAnand Jain if (ret) { 319105135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 319205135f59SDavid Sterba ret); 3193b7c35e81SAnand Jain goto fail_block_groups; 3194b7c35e81SAnand Jain } 3195b7c35e81SAnand Jain 319696f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 31975ac1d209SJeff Mahoney if (ret) { 319805135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3199b7c35e81SAnand Jain goto fail_fsdev_sysfs; 32005ac1d209SJeff Mahoney } 32015ac1d209SJeff Mahoney 3202c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3203c59021f8Sliubo if (ret) { 320405135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 32052365dd3cSAnand Jain goto fail_sysfs; 3206c59021f8Sliubo } 3207c59021f8Sliubo 32085b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 32091b1d1f66SJosef Bacik if (ret) { 321005135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 32112365dd3cSAnand Jain goto fail_sysfs; 32121b1d1f66SJosef Bacik } 32134330e183SQu Wenruo 32146528b99dSAnand Jain if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { 321505135f59SDavid Sterba btrfs_warn(fs_info, 321652042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 32172365dd3cSAnand Jain goto fail_sysfs; 3218292fd7fcSStefan Behrens } 32199078a3e1SChris Mason 3220a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 3221a74a4b97SChris Mason "btrfs-cleaner"); 322257506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 32232365dd3cSAnand Jain goto fail_sysfs; 3224a74a4b97SChris Mason 3225a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3226a74a4b97SChris Mason tree_root, 3227a74a4b97SChris Mason "btrfs-transaction"); 322857506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 32293f157a2fSChris Mason goto fail_cleaner; 3230a74a4b97SChris Mason 3231583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3232c289811cSChris Mason !fs_info->fs_devices->rotating) { 3233583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3234c289811cSChris Mason } 3235c289811cSChris Mason 3236572d9ab7SDavid Sterba /* 323701327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 3238572d9ab7SDavid Sterba * not have to wait for transaction commit 3239572d9ab7SDavid Sterba */ 3240572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 32413818aea2SQu Wenruo 324221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 32430b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 32442ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 32450b246afaSJeff Mahoney btrfs_test_opt(fs_info, 324621adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 324721adbd5cSStefan Behrens 1 : 0, 324821adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 324921adbd5cSStefan Behrens if (ret) 325005135f59SDavid Sterba btrfs_warn(fs_info, 325105135f59SDavid Sterba "failed to initialize integrity check module: %d", 325205135f59SDavid Sterba ret); 325321adbd5cSStefan Behrens } 325421adbd5cSStefan Behrens #endif 3255bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3256bcef60f2SArne Jansen if (ret) 3257bcef60f2SArne Jansen goto fail_trans_kthread; 325821adbd5cSStefan Behrens 3259fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3260fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3261fd708b81SJosef Bacik 326296da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 326396da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 32640b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3265e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 326663443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 326779787eaaSJeff Mahoney if (ret) { 326863443bf5SEric Sandeen err = ret; 326928c16cbbSWang Shilong goto fail_qgroup; 3270e556ce2cSYan Zheng } 3271e02119d5SChris Mason } 32721a40e23bSZheng Yan 32736bccf3abSJeff Mahoney ret = btrfs_find_orphan_roots(fs_info); 327479787eaaSJeff Mahoney if (ret) 327528c16cbbSWang Shilong goto fail_qgroup; 327676dda93cSYan, Zheng 3277bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 3278d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 327944c44af2SIlya Dryomov if (ret) 328028c16cbbSWang Shilong goto fail_qgroup; 328190c711abSZygo Blaxell 328290c711abSZygo Blaxell mutex_lock(&fs_info->cleaner_mutex); 32835d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 328490c711abSZygo Blaxell mutex_unlock(&fs_info->cleaner_mutex); 3285d7ce5843SMiao Xie if (ret < 0) { 328605135f59SDavid Sterba btrfs_warn(fs_info, "failed to recover relocation: %d", 328705135f59SDavid Sterba ret); 3288d7ce5843SMiao Xie err = -EINVAL; 3289bcef60f2SArne Jansen goto fail_qgroup; 3290d7ce5843SMiao Xie } 32917c2ca468SChris Mason } 32921a40e23bSZheng Yan 329356e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 32943140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 32953140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3296f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3297315bf8efSJosef Bacik fs_info->fs_root = NULL; 3298bcef60f2SArne Jansen goto fail_qgroup; 32993140c9a3SDan Carpenter } 3300c289811cSChris Mason 3301bc98a42cSDavid Howells if (sb_rdonly(sb)) 33022b6ba629SIlya Dryomov return 0; 33032b6ba629SIlya Dryomov 3304f8d468a1SOmar Sandoval if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 3305f8d468a1SOmar Sandoval btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 33066675df31SOmar Sandoval clear_free_space_tree = 1; 33076675df31SOmar Sandoval } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 33086675df31SOmar Sandoval !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 33096675df31SOmar Sandoval btrfs_warn(fs_info, "free space tree is invalid"); 33106675df31SOmar Sandoval clear_free_space_tree = 1; 33116675df31SOmar Sandoval } 33126675df31SOmar Sandoval 33136675df31SOmar Sandoval if (clear_free_space_tree) { 3314f8d468a1SOmar Sandoval btrfs_info(fs_info, "clearing free space tree"); 3315f8d468a1SOmar Sandoval ret = btrfs_clear_free_space_tree(fs_info); 3316f8d468a1SOmar Sandoval if (ret) { 3317f8d468a1SOmar Sandoval btrfs_warn(fs_info, 3318f8d468a1SOmar Sandoval "failed to clear free space tree: %d", ret); 33196bccf3abSJeff Mahoney close_ctree(fs_info); 3320f8d468a1SOmar Sandoval return ret; 3321f8d468a1SOmar Sandoval } 3322f8d468a1SOmar Sandoval } 3323f8d468a1SOmar Sandoval 33240b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 3325511711afSChris Mason !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 332605135f59SDavid Sterba btrfs_info(fs_info, "creating free space tree"); 3327511711afSChris Mason ret = btrfs_create_free_space_tree(fs_info); 3328511711afSChris Mason if (ret) { 332905135f59SDavid Sterba btrfs_warn(fs_info, 333005135f59SDavid Sterba "failed to create free space tree: %d", ret); 33316bccf3abSJeff Mahoney close_ctree(fs_info); 3332511711afSChris Mason return ret; 3333511711afSChris Mason } 3334511711afSChris Mason } 3335511711afSChris Mason 3336e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 33372b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 33382b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 33392b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 33406bccf3abSJeff Mahoney close_ctree(fs_info); 33412b6ba629SIlya Dryomov return ret; 33422b6ba629SIlya Dryomov } 3343e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 334459641015SIlya Dryomov 33452b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 33462b6ba629SIlya Dryomov if (ret) { 334705135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume balance: %d", ret); 33486bccf3abSJeff Mahoney close_ctree(fs_info); 33492b6ba629SIlya Dryomov return ret; 3350e3acc2a6SJosef Bacik } 3351e3acc2a6SJosef Bacik 33528dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 33538dabb742SStefan Behrens if (ret) { 335405135f59SDavid Sterba btrfs_warn(fs_info, "failed to resume device replace: %d", ret); 33556bccf3abSJeff Mahoney close_ctree(fs_info); 33568dabb742SStefan Behrens return ret; 33578dabb742SStefan Behrens } 33588dabb742SStefan Behrens 3359b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 3360b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3361b382a324SJan Schmidt 33624bbcaa64SEric Sandeen if (!fs_info->uuid_root) { 336305135f59SDavid Sterba btrfs_info(fs_info, "creating UUID tree"); 3364f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 3365f7a81ea4SStefan Behrens if (ret) { 336605135f59SDavid Sterba btrfs_warn(fs_info, 336705135f59SDavid Sterba "failed to create the UUID tree: %d", ret); 33686bccf3abSJeff Mahoney close_ctree(fs_info); 3369f7a81ea4SStefan Behrens return ret; 3370f7a81ea4SStefan Behrens } 33710b246afaSJeff Mahoney } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 33724bbcaa64SEric Sandeen fs_info->generation != 33734bbcaa64SEric Sandeen btrfs_super_uuid_tree_generation(disk_super)) { 337405135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 337570f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 337670f80175SStefan Behrens if (ret) { 337705135f59SDavid Sterba btrfs_warn(fs_info, 337805135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 33796bccf3abSJeff Mahoney close_ctree(fs_info); 338070f80175SStefan Behrens return ret; 338170f80175SStefan Behrens } 3382f7a81ea4SStefan Behrens } 3383afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 338447ab2a6cSJosef Bacik 33858dcddfa0SQu Wenruo /* 33868dcddfa0SQu Wenruo * backuproot only affect mount behavior, and if open_ctree succeeded, 33878dcddfa0SQu Wenruo * no need to keep the flag 33888dcddfa0SQu Wenruo */ 33898dcddfa0SQu Wenruo btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 33908dcddfa0SQu Wenruo 3391ad2b2c80SAl Viro return 0; 339239279cc3SChris Mason 3393bcef60f2SArne Jansen fail_qgroup: 3394bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 33957c2ca468SChris Mason fail_trans_kthread: 33967c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 33972ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3398faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 33993f157a2fSChris Mason fail_cleaner: 3400a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 34017c2ca468SChris Mason 34027c2ca468SChris Mason /* 34037c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 34047c2ca468SChris Mason * kthreads 34057c2ca468SChris Mason */ 34067c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 34077c2ca468SChris Mason 34082365dd3cSAnand Jain fail_sysfs: 34096618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 34102365dd3cSAnand Jain 3411b7c35e81SAnand Jain fail_fsdev_sysfs: 3412b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3413b7c35e81SAnand Jain 34141b1d1f66SJosef Bacik fail_block_groups: 341554067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3416af31f5e5SChris Mason 3417af31f5e5SChris Mason fail_tree_roots: 34189e3aa805SJosef Bacik if (fs_info->data_reloc_root) 34199e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 34204273eaffSAnand Jain free_root_pointers(fs_info, true); 34212b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3422af31f5e5SChris Mason 342339279cc3SChris Mason fail_sb_buffer: 34247abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 34255cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 342616cdcec7SMiao Xie fail_alloc: 3427586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3428586e46e2SIlya Dryomov 34294543df7eSChris Mason iput(fs_info->btree_inode); 34307e662854SQinghuang Feng fail: 3431586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3432ad2b2c80SAl Viro return err; 3433eb60ceacSChris Mason } 3434663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3435eb60ceacSChris Mason 3436314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3437f2984462SChris Mason { 3438314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3439314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3440314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3441314b6dd0SJohannes Thumshirn struct page *page; 3442442a4f63SStefan Behrens 3443314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3444314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3445314b6dd0SJohannes Thumshirn 3446314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3447fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3448314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3449314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3450314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3451314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3452314b6dd0SJohannes Thumshirn SetPageError(page); 3453314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3454314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3455314b6dd0SJohannes Thumshirn } else { 3456314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3457f2984462SChris Mason } 3458314b6dd0SJohannes Thumshirn 3459314b6dd0SJohannes Thumshirn put_page(page); 3460314b6dd0SJohannes Thumshirn unlock_page(page); 3461314b6dd0SJohannes Thumshirn } 3462314b6dd0SJohannes Thumshirn 3463314b6dd0SJohannes Thumshirn bio_put(bio); 3464f2984462SChris Mason } 3465f2984462SChris Mason 34668f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 34678f32380dSJohannes Thumshirn int copy_num) 346829c36d72SAnand Jain { 346929c36d72SAnand Jain struct btrfs_super_block *super; 34708f32380dSJohannes Thumshirn struct page *page; 347129c36d72SAnand Jain u64 bytenr; 34728f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 347329c36d72SAnand Jain 347429c36d72SAnand Jain bytenr = btrfs_sb_offset(copy_num); 347529c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 34768f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 347729c36d72SAnand Jain 34788f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 34798f32380dSJohannes Thumshirn if (IS_ERR(page)) 34808f32380dSJohannes Thumshirn return ERR_CAST(page); 348129c36d72SAnand Jain 34828f32380dSJohannes Thumshirn super = page_address(page); 348396c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 348496c2e067SAnand Jain btrfs_release_disk_super(super); 348596c2e067SAnand Jain return ERR_PTR(-ENODATA); 348696c2e067SAnand Jain } 348796c2e067SAnand Jain 348896c2e067SAnand Jain if (btrfs_super_bytenr(super) != bytenr) { 34898f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 34908f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 349129c36d72SAnand Jain } 349229c36d72SAnand Jain 34938f32380dSJohannes Thumshirn return super; 349429c36d72SAnand Jain } 349529c36d72SAnand Jain 349629c36d72SAnand Jain 34978f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3498a512bbf8SYan Zheng { 34998f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3500a512bbf8SYan Zheng int i; 3501a512bbf8SYan Zheng u64 transid = 0; 3502a512bbf8SYan Zheng 3503a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3504a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3505a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3506a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3507a512bbf8SYan Zheng */ 3508a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 35098f32380dSJohannes Thumshirn super = btrfs_read_dev_one_super(bdev, i); 35108f32380dSJohannes Thumshirn if (IS_ERR(super)) 3511a512bbf8SYan Zheng continue; 3512a512bbf8SYan Zheng 3513a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 35148f32380dSJohannes Thumshirn if (latest) 35158f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 35168f32380dSJohannes Thumshirn 35178f32380dSJohannes Thumshirn latest = super; 3518a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3519a512bbf8SYan Zheng } 3520a512bbf8SYan Zheng } 352192fc03fbSAnand Jain 35228f32380dSJohannes Thumshirn return super; 3523a512bbf8SYan Zheng } 3524a512bbf8SYan Zheng 35254eedeb75SHisashi Hifumi /* 3526abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3527314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 35284eedeb75SHisashi Hifumi * 3529abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3530abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3531abbb3b8eSDavid Sterba * same for write and wait phases. 35324eedeb75SHisashi Hifumi * 3533314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 35344eedeb75SHisashi Hifumi */ 3535a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3536abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3537a512bbf8SYan Zheng { 3538d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3539314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3540d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3541a512bbf8SYan Zheng int i; 3542a512bbf8SYan Zheng int errors = 0; 3543a512bbf8SYan Zheng u64 bytenr; 3544a512bbf8SYan Zheng 3545a512bbf8SYan Zheng if (max_mirrors == 0) 3546a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3547a512bbf8SYan Zheng 3548d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3549d5178578SJohannes Thumshirn 3550a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3551314b6dd0SJohannes Thumshirn struct page *page; 3552314b6dd0SJohannes Thumshirn struct bio *bio; 3553314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3554314b6dd0SJohannes Thumshirn 3555a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3556935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3557935e5cc9SMiao Xie device->commit_total_bytes) 3558a512bbf8SYan Zheng break; 3559a512bbf8SYan Zheng 3560a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3561a512bbf8SYan Zheng 3562fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3563fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3564fd08001fSEric Biggers sb->csum); 3565a512bbf8SYan Zheng 3566314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3567314b6dd0SJohannes Thumshirn GFP_NOFS); 3568314b6dd0SJohannes Thumshirn if (!page) { 3569fb456252SJeff Mahoney btrfs_err(device->fs_info, 3570314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3571f14d104dSDavid Sterba bytenr); 3572634554dcSJosef Bacik errors++; 3573634554dcSJosef Bacik continue; 3574634554dcSJosef Bacik } 3575634554dcSJosef Bacik 3576314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3577314b6dd0SJohannes Thumshirn get_page(page); 3578a512bbf8SYan Zheng 3579314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3580314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3581a512bbf8SYan Zheng 3582387125fcSChris Mason /* 3583314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3584314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3585314b6dd0SJohannes Thumshirn * checking. 3586387125fcSChris Mason */ 3587314b6dd0SJohannes Thumshirn bio = bio_alloc(GFP_NOFS, 1); 3588314b6dd0SJohannes Thumshirn bio_set_dev(bio, device->bdev); 3589314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3590314b6dd0SJohannes Thumshirn bio->bi_private = device; 3591314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3592314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3593314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3594314b6dd0SJohannes Thumshirn 3595314b6dd0SJohannes Thumshirn /* 3596314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3597314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3598314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3599314b6dd0SJohannes Thumshirn */ 3600314b6dd0SJohannes Thumshirn bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; 36011b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3602314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3603314b6dd0SJohannes Thumshirn 3604314b6dd0SJohannes Thumshirn btrfsic_submit_bio(bio); 3605a512bbf8SYan Zheng } 3606a512bbf8SYan Zheng return errors < i ? 0 : -1; 3607a512bbf8SYan Zheng } 3608a512bbf8SYan Zheng 3609387125fcSChris Mason /* 3610abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3611abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3612abbb3b8eSDavid Sterba * 3613314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3614abbb3b8eSDavid Sterba * date. 3615abbb3b8eSDavid Sterba */ 3616abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3617abbb3b8eSDavid Sterba { 3618abbb3b8eSDavid Sterba int i; 3619abbb3b8eSDavid Sterba int errors = 0; 3620b6a535faSHoward McLauchlan bool primary_failed = false; 3621abbb3b8eSDavid Sterba u64 bytenr; 3622abbb3b8eSDavid Sterba 3623abbb3b8eSDavid Sterba if (max_mirrors == 0) 3624abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3625abbb3b8eSDavid Sterba 3626abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3627314b6dd0SJohannes Thumshirn struct page *page; 3628314b6dd0SJohannes Thumshirn 3629abbb3b8eSDavid Sterba bytenr = btrfs_sb_offset(i); 3630abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3631abbb3b8eSDavid Sterba device->commit_total_bytes) 3632abbb3b8eSDavid Sterba break; 3633abbb3b8eSDavid Sterba 3634314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3635314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3636314b6dd0SJohannes Thumshirn if (!page) { 3637abbb3b8eSDavid Sterba errors++; 3638b6a535faSHoward McLauchlan if (i == 0) 3639b6a535faSHoward McLauchlan primary_failed = true; 3640abbb3b8eSDavid Sterba continue; 3641abbb3b8eSDavid Sterba } 3642314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3643314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3644314b6dd0SJohannes Thumshirn if (PageError(page)) { 3645abbb3b8eSDavid Sterba errors++; 3646b6a535faSHoward McLauchlan if (i == 0) 3647b6a535faSHoward McLauchlan primary_failed = true; 3648b6a535faSHoward McLauchlan } 3649abbb3b8eSDavid Sterba 3650314b6dd0SJohannes Thumshirn /* Drop our reference */ 3651314b6dd0SJohannes Thumshirn put_page(page); 3652abbb3b8eSDavid Sterba 3653314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3654314b6dd0SJohannes Thumshirn put_page(page); 3655abbb3b8eSDavid Sterba } 3656abbb3b8eSDavid Sterba 3657b6a535faSHoward McLauchlan /* log error, force error return */ 3658b6a535faSHoward McLauchlan if (primary_failed) { 3659b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3660b6a535faSHoward McLauchlan device->devid); 3661b6a535faSHoward McLauchlan return -1; 3662b6a535faSHoward McLauchlan } 3663b6a535faSHoward McLauchlan 3664abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3665abbb3b8eSDavid Sterba } 3666abbb3b8eSDavid Sterba 3667abbb3b8eSDavid Sterba /* 3668387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3669387125fcSChris Mason * for the barrier when it is done 3670387125fcSChris Mason */ 36714246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3672387125fcSChris Mason { 3673387125fcSChris Mason complete(bio->bi_private); 3674387125fcSChris Mason } 3675387125fcSChris Mason 3676387125fcSChris Mason /* 36774fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 36784fc6441aSAnand Jain * done in the waiting counterpart. 3679387125fcSChris Mason */ 36804fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3681387125fcSChris Mason { 3682c2a9c7abSAnand Jain struct request_queue *q = bdev_get_queue(device->bdev); 3683e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3684387125fcSChris Mason 3685c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 36864fc6441aSAnand Jain return; 3687387125fcSChris Mason 3688e0ae9994SDavid Sterba bio_reset(bio); 3689387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 369074d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 36918d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 3692387125fcSChris Mason init_completion(&device->flush_wait); 3693387125fcSChris Mason bio->bi_private = &device->flush_wait; 3694387125fcSChris Mason 369543a01111SLu Fengqi btrfsic_submit_bio(bio); 36961c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 36974fc6441aSAnand Jain } 3698387125fcSChris Mason 36994fc6441aSAnand Jain /* 37004fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 37014fc6441aSAnand Jain */ 37028c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 37034fc6441aSAnand Jain { 37044fc6441aSAnand Jain struct bio *bio = device->flush_bio; 37054fc6441aSAnand Jain 37061c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 370758efbc9fSOmar Sandoval return BLK_STS_OK; 37084fc6441aSAnand Jain 37091c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 37102980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 37114fc6441aSAnand Jain 37128c27cb35SLinus Torvalds return bio->bi_status; 3713387125fcSChris Mason } 3714387125fcSChris Mason 3715d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 3716401b41e5SAnand Jain { 37176528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 3718401b41e5SAnand Jain return -EIO; 3719387125fcSChris Mason return 0; 3720387125fcSChris Mason } 3721387125fcSChris Mason 3722387125fcSChris Mason /* 3723387125fcSChris Mason * send an empty flush down to each device in parallel, 3724387125fcSChris Mason * then wait for them 3725387125fcSChris Mason */ 3726387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3727387125fcSChris Mason { 3728387125fcSChris Mason struct list_head *head; 3729387125fcSChris Mason struct btrfs_device *dev; 37305af3e8ccSStefan Behrens int errors_wait = 0; 37314e4cbee9SChristoph Hellwig blk_status_t ret; 3732387125fcSChris Mason 37331538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3734387125fcSChris Mason /* send down all the barriers */ 3735387125fcSChris Mason head = &info->fs_devices->devices; 37361538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3737e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3738f88ba6a2SHidetoshi Seto continue; 3739cea7c8bfSAnand Jain if (!dev->bdev) 3740387125fcSChris Mason continue; 3741e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3742ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3743387125fcSChris Mason continue; 3744387125fcSChris Mason 37454fc6441aSAnand Jain write_dev_flush(dev); 374658efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 3747387125fcSChris Mason } 3748387125fcSChris Mason 3749387125fcSChris Mason /* wait for all the barriers */ 37501538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3751e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3752f88ba6a2SHidetoshi Seto continue; 3753387125fcSChris Mason if (!dev->bdev) { 37545af3e8ccSStefan Behrens errors_wait++; 3755387125fcSChris Mason continue; 3756387125fcSChris Mason } 3757e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3758ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3759387125fcSChris Mason continue; 3760387125fcSChris Mason 37614fc6441aSAnand Jain ret = wait_dev_flush(dev); 3762401b41e5SAnand Jain if (ret) { 3763401b41e5SAnand Jain dev->last_flush_error = ret; 376466b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 376566b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 37665af3e8ccSStefan Behrens errors_wait++; 3767387125fcSChris Mason } 3768401b41e5SAnand Jain } 3769401b41e5SAnand Jain 3770cea7c8bfSAnand Jain if (errors_wait) { 3771401b41e5SAnand Jain /* 3772401b41e5SAnand Jain * At some point we need the status of all disks 3773401b41e5SAnand Jain * to arrive at the volume status. So error checking 3774401b41e5SAnand Jain * is being pushed to a separate loop. 3775401b41e5SAnand Jain */ 3776d10b82feSQu Wenruo return check_barrier_error(info); 3777401b41e5SAnand Jain } 3778387125fcSChris Mason return 0; 3779387125fcSChris Mason } 3780387125fcSChris Mason 3781943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3782943c6e99SZhao Lei { 37838789f4feSZhao Lei int raid_type; 37848789f4feSZhao Lei int min_tolerated = INT_MAX; 3785943c6e99SZhao Lei 37868789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 37878789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 37888c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 37898789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 37908789f4feSZhao Lei tolerated_failures); 3791943c6e99SZhao Lei 37928789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 37938789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 37948789f4feSZhao Lei continue; 379541a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 37968789f4feSZhao Lei continue; 37978c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 37988789f4feSZhao Lei btrfs_raid_array[raid_type]. 37998789f4feSZhao Lei tolerated_failures); 38008789f4feSZhao Lei } 3801943c6e99SZhao Lei 38028789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3803ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 38048789f4feSZhao Lei min_tolerated = 0; 38058789f4feSZhao Lei } 38068789f4feSZhao Lei 38078789f4feSZhao Lei return min_tolerated; 3808943c6e99SZhao Lei } 3809943c6e99SZhao Lei 3810eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3811f2984462SChris Mason { 3812e5e9a520SChris Mason struct list_head *head; 3813f2984462SChris Mason struct btrfs_device *dev; 3814a061fc8dSChris Mason struct btrfs_super_block *sb; 3815f2984462SChris Mason struct btrfs_dev_item *dev_item; 3816f2984462SChris Mason int ret; 3817f2984462SChris Mason int do_barriers; 3818a236aed1SChris Mason int max_errors; 3819a236aed1SChris Mason int total_errors = 0; 3820a061fc8dSChris Mason u64 flags; 3821f2984462SChris Mason 38220b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 3823fed3b381SLiu Bo 3824fed3b381SLiu Bo /* 3825fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 3826fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 3827fed3b381SLiu Bo * been consistent on disk. 3828fed3b381SLiu Bo */ 3829fed3b381SLiu Bo if (max_mirrors == 0) 38300b246afaSJeff Mahoney backup_super_roots(fs_info); 3831f2984462SChris Mason 38320b246afaSJeff Mahoney sb = fs_info->super_for_commit; 3833a061fc8dSChris Mason dev_item = &sb->dev_item; 3834e5e9a520SChris Mason 38350b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 38360b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 38370b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 3838387125fcSChris Mason 38395af3e8ccSStefan Behrens if (do_barriers) { 38400b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 38415af3e8ccSStefan Behrens if (ret) { 38425af3e8ccSStefan Behrens mutex_unlock( 38430b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 38440b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 38455af3e8ccSStefan Behrens "errors while submitting device barriers."); 38465af3e8ccSStefan Behrens return ret; 38475af3e8ccSStefan Behrens } 38485af3e8ccSStefan Behrens } 3849387125fcSChris Mason 38501538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3851dfe25020SChris Mason if (!dev->bdev) { 3852dfe25020SChris Mason total_errors++; 3853dfe25020SChris Mason continue; 3854dfe25020SChris Mason } 3855e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3856ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3857dfe25020SChris Mason continue; 3858dfe25020SChris Mason 38592b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3860a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3861a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 38627df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3863935e5cc9SMiao Xie dev->commit_total_bytes); 3864ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3865ce7213c7SMiao Xie dev->commit_bytes_used); 3866a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3867a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3868a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3869a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 38707239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 38717239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 3872a512bbf8SYan Zheng 3873a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3874a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3875f2984462SChris Mason 387675cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 387775cb857dSQu Wenruo if (ret < 0) { 387875cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 387975cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 388075cb857dSQu Wenruo "unexpected superblock corruption detected"); 388175cb857dSQu Wenruo return -EUCLEAN; 388275cb857dSQu Wenruo } 388375cb857dSQu Wenruo 3884abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 3885a236aed1SChris Mason if (ret) 3886a236aed1SChris Mason total_errors++; 3887f2984462SChris Mason } 3888a236aed1SChris Mason if (total_errors > max_errors) { 38890b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 3890d397712bSChris Mason total_errors); 38910b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 389279787eaaSJeff Mahoney 38939d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 38940b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 38950b246afaSJeff Mahoney "%d errors while writing supers", 38960b246afaSJeff Mahoney total_errors); 38979d565ba4SStefan Behrens return -EIO; 3898a236aed1SChris Mason } 3899f2984462SChris Mason 3900a512bbf8SYan Zheng total_errors = 0; 39011538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3902dfe25020SChris Mason if (!dev->bdev) 3903dfe25020SChris Mason continue; 3904e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3905ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3906dfe25020SChris Mason continue; 3907dfe25020SChris Mason 3908abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 3909a512bbf8SYan Zheng if (ret) 39101259ab75SChris Mason total_errors++; 3911f2984462SChris Mason } 39120b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 3913a236aed1SChris Mason if (total_errors > max_errors) { 39140b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 39150b246afaSJeff Mahoney "%d errors while writing supers", 39160b246afaSJeff Mahoney total_errors); 391779787eaaSJeff Mahoney return -EIO; 3918a236aed1SChris Mason } 3919f2984462SChris Mason return 0; 3920f2984462SChris Mason } 3921f2984462SChris Mason 3922cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3923cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3924cb517eabSMiao Xie struct btrfs_root *root) 39252619ba1fSChris Mason { 39264785e24fSJosef Bacik bool drop_ref = false; 39274785e24fSJosef Bacik 39284df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 39292619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 39302619ba1fSChris Mason (unsigned long)root->root_key.objectid); 3931af01d2e5SJosef Bacik if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 39324785e24fSJosef Bacik drop_ref = true; 39334df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 393476dda93cSYan, Zheng 39351c1ea4f7SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 3936ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 39371c1ea4f7SLiu Bo if (root->reloc_root) { 393800246528SJosef Bacik btrfs_put_root(root->reloc_root); 39391c1ea4f7SLiu Bo root->reloc_root = NULL; 39401c1ea4f7SLiu Bo } 39411c1ea4f7SLiu Bo } 39423321719eSLiu Bo 3943faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3944581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3945faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3946581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 39470e996e7fSJosef Bacik if (root->ino_cache_inode) { 39480e996e7fSJosef Bacik iput(root->ino_cache_inode); 39490e996e7fSJosef Bacik root->ino_cache_inode = NULL; 39500e996e7fSJosef Bacik } 39514785e24fSJosef Bacik if (drop_ref) 395200246528SJosef Bacik btrfs_put_root(root); 39532619ba1fSChris Mason } 39542619ba1fSChris Mason 3955c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3956c146afadSYan Zheng { 3957c146afadSYan Zheng u64 root_objectid = 0; 3958c146afadSYan Zheng struct btrfs_root *gang[8]; 395965d33fd7SQu Wenruo int i = 0; 396065d33fd7SQu Wenruo int err = 0; 396165d33fd7SQu Wenruo unsigned int ret = 0; 3962c146afadSYan Zheng 3963c146afadSYan Zheng while (1) { 3964efc34534SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 3965c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3966c146afadSYan Zheng (void **)gang, root_objectid, 3967c146afadSYan Zheng ARRAY_SIZE(gang)); 396865d33fd7SQu Wenruo if (!ret) { 3969efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 3970c146afadSYan Zheng break; 397165d33fd7SQu Wenruo } 39725d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 397366b4ffd1SJosef Bacik 397465d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 397565d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 397665d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 397765d33fd7SQu Wenruo gang[i] = NULL; 397865d33fd7SQu Wenruo continue; 397965d33fd7SQu Wenruo } 398065d33fd7SQu Wenruo /* grab all the search result for later use */ 398100246528SJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 398265d33fd7SQu Wenruo } 3983efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 398465d33fd7SQu Wenruo 398565d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 398665d33fd7SQu Wenruo if (!gang[i]) 398765d33fd7SQu Wenruo continue; 3988c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 398966b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 399066b4ffd1SJosef Bacik if (err) 399165d33fd7SQu Wenruo break; 399200246528SJosef Bacik btrfs_put_root(gang[i]); 3993c146afadSYan Zheng } 3994c146afadSYan Zheng root_objectid++; 3995c146afadSYan Zheng } 399665d33fd7SQu Wenruo 399765d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 399865d33fd7SQu Wenruo for (; i < ret; i++) { 399965d33fd7SQu Wenruo if (gang[i]) 400000246528SJosef Bacik btrfs_put_root(gang[i]); 400165d33fd7SQu Wenruo } 400265d33fd7SQu Wenruo return err; 4003c146afadSYan Zheng } 4004c146afadSYan Zheng 40056bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4006c146afadSYan Zheng { 40076bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4008c146afadSYan Zheng struct btrfs_trans_handle *trans; 4009c146afadSYan Zheng 40100b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 40112ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 40120b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 40130b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4014c71bf099SYan, Zheng 4015c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 40160b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 40170b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4018c71bf099SYan, Zheng 40197a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 40203612b495STsutomu Itoh if (IS_ERR(trans)) 40213612b495STsutomu Itoh return PTR_ERR(trans); 40223a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4023c146afadSYan Zheng } 4024c146afadSYan Zheng 4025b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4026eb60ceacSChris Mason { 4027c146afadSYan Zheng int ret; 4028e089f05cSChris Mason 4029afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 4030d6fd0ae2SOmar Sandoval /* 4031d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4032d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4033d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4034d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4035d6fd0ae2SOmar Sandoval */ 4036d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4037a2135011SChris Mason 40387343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4039d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 40407343dd61SJustin Maggard 4041803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4042803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4043803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4044803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4045803b2f54SStefan Behrens 4046837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4047aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4048837d5b6eSIlya Dryomov 40498dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 40508dabb742SStefan Behrens 4051aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 40524cb5300bSChris Mason 40534cb5300bSChris Mason /* wait for any defraggers to finish */ 40544cb5300bSChris Mason wait_event(fs_info->transaction_wait, 40554cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 40564cb5300bSChris Mason 40574cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 405826176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 40594cb5300bSChris Mason 406021c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 406157056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 406221c7e756SMiao Xie 4063b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4064b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4065b0643e59SDennis Zhou 4066bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4067e44163e1SJeff Mahoney /* 4068d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4069d6fd0ae2SOmar Sandoval * unused block groups. 4070e44163e1SJeff Mahoney */ 40710b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4072e44163e1SJeff Mahoney 4073f0cc2cd7SFilipe Manana /* 4074f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4075f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4076f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4077f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4078f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4079f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4080f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4081f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4082f0cc2cd7SFilipe Manana * This is a very rare case. 4083f0cc2cd7SFilipe Manana */ 4084f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4085f0cc2cd7SFilipe Manana 40866bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4087d397712bSChris Mason if (ret) 408804892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4089c146afadSYan Zheng } 4090ed2ff2cbSChris Mason 4091af722733SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || 4092af722733SLiu Bo test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) 40932ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4094acce952bSliubo 4095e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4096e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 40978929ecfaSYan, Zheng 4098e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4099afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4100f25784b3SYan Zheng 41015958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 41025958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 41035958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 41045958253cSQu Wenruo } 41055958253cSQu Wenruo 410604892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4107fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4108bcef60f2SArne Jansen 4109963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 411004892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4111963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4112b0c68f8bSChris Mason } 411331153d81SYan Zheng 41144297ff84SJosef Bacik if (percpu_counter_sum(&fs_info->dio_bytes)) 41154297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 41164297ff84SJosef Bacik percpu_counter_sum(&fs_info->dio_bytes)); 41174297ff84SJosef Bacik 41186618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4119b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 41205ac1d209SJeff Mahoney 41211a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 41221a4319ccSLiu Bo 4123de348ee0SWang Shilong /* 4124de348ee0SWang Shilong * we must make sure there is not any read request to 4125de348ee0SWang Shilong * submit after we stopping all workers. 4126de348ee0SWang Shilong */ 4127de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 412896192499SJosef Bacik btrfs_stop_all_workers(fs_info); 412996192499SJosef Bacik 4130afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 41314273eaffSAnand Jain free_root_pointers(fs_info, true); 41328c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 41339ad6b7bcSChris Mason 41344e19443dSJosef Bacik /* 41354e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 41364e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 41374e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 41384e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 41394e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 41404e19443dSJosef Bacik */ 41414e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 41424e19443dSJosef Bacik 414313e6c37bSJosef Bacik iput(fs_info->btree_inode); 4144d6bfde87SChris Mason 414521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 41460b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 41472ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 414821adbd5cSStefan Behrens #endif 414921adbd5cSStefan Behrens 41500b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 415168c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4152eb60ceacSChris Mason } 4153eb60ceacSChris Mason 4154b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4155b9fab919SChris Mason int atomic) 4156ccd467d6SChris Mason { 41571259ab75SChris Mason int ret; 4158727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 41591259ab75SChris Mason 41600b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 41611259ab75SChris Mason if (!ret) 41621259ab75SChris Mason return ret; 41631259ab75SChris Mason 41641259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4165b9fab919SChris Mason parent_transid, atomic); 4166b9fab919SChris Mason if (ret == -EAGAIN) 4167b9fab919SChris Mason return ret; 41681259ab75SChris Mason return !ret; 41695f39d397SChris Mason } 41706702ed49SChris Mason 41715f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 41725f39d397SChris Mason { 41730b246afaSJeff Mahoney struct btrfs_fs_info *fs_info; 417406ea65a3SJosef Bacik struct btrfs_root *root; 41755f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4176b9473439SChris Mason int was_dirty; 4177b4ce94deSChris Mason 417806ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 417906ea65a3SJosef Bacik /* 418006ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 418152042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 418206ea65a3SJosef Bacik * outside of the sanity tests. 418306ea65a3SJosef Bacik */ 4184b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 418506ea65a3SJosef Bacik return; 418606ea65a3SJosef Bacik #endif 418706ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 41880b246afaSJeff Mahoney fs_info = root->fs_info; 4189b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 41900b246afaSJeff Mahoney if (transid != fs_info->generation) 41915d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 41920b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 41930b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4194e2d84521SMiao Xie if (!was_dirty) 4195104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4196e2d84521SMiao Xie buf->len, 41970b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 41981f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 419969fc6cbbSQu Wenruo /* 420069fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 420169fc6cbbSQu Wenruo * but item data not updated. 420269fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 420369fc6cbbSQu Wenruo */ 420469fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4205cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4206a4f78750SDavid Sterba btrfs_print_leaf(buf); 42071f21ef0aSFilipe Manana ASSERT(0); 42081f21ef0aSFilipe Manana } 42091f21ef0aSFilipe Manana #endif 4210eb60ceacSChris Mason } 4211eb60ceacSChris Mason 42122ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4213b53d3f5dSLiu Bo int flush_delayed) 421435b7e476SChris Mason { 4215188de649SChris Mason /* 4216188de649SChris Mason * looks as though older kernels can get into trouble with 4217188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4218188de649SChris Mason */ 4219e2d84521SMiao Xie int ret; 4220d6bfde87SChris Mason 42216933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4222d6bfde87SChris Mason return; 4223d6bfde87SChris Mason 4224b53d3f5dSLiu Bo if (flush_delayed) 42252ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 422616cdcec7SMiao Xie 4227d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4228d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4229d814a491SEthan Lien fs_info->dirty_metadata_batch); 4230e2d84521SMiao Xie if (ret > 0) { 42310b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 423216cdcec7SMiao Xie } 423316cdcec7SMiao Xie } 423416cdcec7SMiao Xie 42352ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 423616cdcec7SMiao Xie { 42372ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 423835b7e476SChris Mason } 4239b53d3f5dSLiu Bo 42402ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4241b53d3f5dSLiu Bo { 42422ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4243d6bfde87SChris Mason } 42446b80053dSChris Mason 4245581c1760SQu Wenruo int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, 4246581c1760SQu Wenruo struct btrfs_key *first_key) 42476b80053dSChris Mason { 42485ab12d1fSDavid Sterba return btree_read_extent_buffer_pages(buf, parent_transid, 4249581c1760SQu Wenruo level, first_key); 42506b80053dSChris Mason } 42510da5468fSChris Mason 42522ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4253acce952bSliubo { 4254fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4255fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4256fe816d0fSNikolay Borisov 42570b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 42582ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 42590b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4260acce952bSliubo 42610b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 42620b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4263acce952bSliubo } 4264acce952bSliubo 4265ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4266ef67963dSJosef Bacik { 4267ef67963dSJosef Bacik struct btrfs_root *gang[8]; 4268ef67963dSJosef Bacik u64 root_objectid = 0; 4269ef67963dSJosef Bacik int ret; 4270ef67963dSJosef Bacik 4271ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4272ef67963dSJosef Bacik while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4273ef67963dSJosef Bacik (void **)gang, root_objectid, 4274ef67963dSJosef Bacik ARRAY_SIZE(gang))) != 0) { 4275ef67963dSJosef Bacik int i; 4276ef67963dSJosef Bacik 4277ef67963dSJosef Bacik for (i = 0; i < ret; i++) 4278ef67963dSJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 4279ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4280ef67963dSJosef Bacik 4281ef67963dSJosef Bacik for (i = 0; i < ret; i++) { 4282ef67963dSJosef Bacik if (!gang[i]) 4283ef67963dSJosef Bacik continue; 4284ef67963dSJosef Bacik root_objectid = gang[i]->root_key.objectid; 4285ef67963dSJosef Bacik btrfs_free_log(NULL, gang[i]); 4286ef67963dSJosef Bacik btrfs_put_root(gang[i]); 4287ef67963dSJosef Bacik } 4288ef67963dSJosef Bacik root_objectid++; 4289ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4290ef67963dSJosef Bacik } 4291ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4292ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4293ef67963dSJosef Bacik } 4294ef67963dSJosef Bacik 4295143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4296acce952bSliubo { 4297acce952bSliubo struct btrfs_ordered_extent *ordered; 4298acce952bSliubo 4299199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4300779880efSJosef Bacik /* 4301779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4302779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4303779880efSJosef Bacik */ 4304199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4305779880efSJosef Bacik root_extent_list) 4306779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4307199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4308199c2a9cSMiao Xie } 4309199c2a9cSMiao Xie 4310199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4311199c2a9cSMiao Xie { 4312199c2a9cSMiao Xie struct btrfs_root *root; 4313199c2a9cSMiao Xie struct list_head splice; 4314199c2a9cSMiao Xie 4315199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4316199c2a9cSMiao Xie 4317199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4318199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4319199c2a9cSMiao Xie while (!list_empty(&splice)) { 4320199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4321199c2a9cSMiao Xie ordered_root); 43221de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 43231de2cfdeSJosef Bacik &fs_info->ordered_roots); 4324199c2a9cSMiao Xie 43252a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4326199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4327199c2a9cSMiao Xie 43282a85d9caSLiu Bo cond_resched(); 43292a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4330199c2a9cSMiao Xie } 4331199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 433274d5d229SJosef Bacik 433374d5d229SJosef Bacik /* 433474d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 433574d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 433674d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 433774d5d229SJosef Bacik * actually get run and error out properly. 433874d5d229SJosef Bacik */ 433974d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4340acce952bSliubo } 4341acce952bSliubo 434235a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 43432ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4344acce952bSliubo { 4345acce952bSliubo struct rb_node *node; 4346acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4347acce952bSliubo struct btrfs_delayed_ref_node *ref; 4348acce952bSliubo int ret = 0; 4349acce952bSliubo 4350acce952bSliubo delayed_refs = &trans->delayed_refs; 4351acce952bSliubo 4352acce952bSliubo spin_lock(&delayed_refs->lock); 4353d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4354cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4355b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4356acce952bSliubo return ret; 4357acce952bSliubo } 4358acce952bSliubo 43595c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4360d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 43610e0adbcfSJosef Bacik struct rb_node *n; 4362e78417d1SJosef Bacik bool pin_bytes = false; 4363acce952bSliubo 4364d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4365d7df2c79SJosef Bacik href_node); 43663069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4367b939d1abSJosef Bacik continue; 43683069bd26SJosef Bacik 4369d7df2c79SJosef Bacik spin_lock(&head->lock); 4370e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 43710e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 43720e0adbcfSJosef Bacik ref_node); 4373d7df2c79SJosef Bacik ref->in_tree = 0; 4374e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 43750e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 43761d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 43771d57ee94SWang Xiaoguang list_del(&ref->add_list); 4378d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4379d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4380d7df2c79SJosef Bacik } 438154067ae9SJosef Bacik if (head->must_insert_reserved) 4382e78417d1SJosef Bacik pin_bytes = true; 438378a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4384fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4385d7df2c79SJosef Bacik spin_unlock(&head->lock); 4386acce952bSliubo spin_unlock(&delayed_refs->lock); 4387e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4388acce952bSliubo 4389f603bb94SNikolay Borisov if (pin_bytes) { 4390f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4391f603bb94SNikolay Borisov 4392f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4393f603bb94SNikolay Borisov BUG_ON(!cache); 4394f603bb94SNikolay Borisov 4395f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4396f603bb94SNikolay Borisov spin_lock(&cache->lock); 4397f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4398f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4399f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4400f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4401f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4402f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4403f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4404f603bb94SNikolay Borisov percpu_counter_add_batch( 4405f603bb94SNikolay Borisov &cache->space_info->total_bytes_pinned, 4406f603bb94SNikolay Borisov head->num_bytes, BTRFS_TOTAL_BYTES_PINNED_BATCH); 4407f603bb94SNikolay Borisov 4408f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4409f603bb94SNikolay Borisov 4410f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4411f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4412f603bb94SNikolay Borisov } 441331890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4414d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4415acce952bSliubo cond_resched(); 4416acce952bSliubo spin_lock(&delayed_refs->lock); 4417acce952bSliubo } 441881f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4419acce952bSliubo 4420acce952bSliubo spin_unlock(&delayed_refs->lock); 4421acce952bSliubo 4422acce952bSliubo return ret; 4423acce952bSliubo } 4424acce952bSliubo 4425143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4426acce952bSliubo { 4427acce952bSliubo struct btrfs_inode *btrfs_inode; 4428acce952bSliubo struct list_head splice; 4429acce952bSliubo 4430acce952bSliubo INIT_LIST_HEAD(&splice); 4431acce952bSliubo 4432eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4433eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4434acce952bSliubo 4435acce952bSliubo while (!list_empty(&splice)) { 4436fe816d0fSNikolay Borisov struct inode *inode = NULL; 4437eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4438acce952bSliubo delalloc_inodes); 4439fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4440eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4441acce952bSliubo 4442fe816d0fSNikolay Borisov /* 4443fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4444fe816d0fSNikolay Borisov * meanwhile. 4445fe816d0fSNikolay Borisov */ 4446fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4447fe816d0fSNikolay Borisov if (inode) { 4448fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4449fe816d0fSNikolay Borisov iput(inode); 4450fe816d0fSNikolay Borisov } 4451eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4452acce952bSliubo } 4453eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4454eb73c1b7SMiao Xie } 4455eb73c1b7SMiao Xie 4456eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4457eb73c1b7SMiao Xie { 4458eb73c1b7SMiao Xie struct btrfs_root *root; 4459eb73c1b7SMiao Xie struct list_head splice; 4460eb73c1b7SMiao Xie 4461eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4462eb73c1b7SMiao Xie 4463eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4464eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4465eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4466eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4467eb73c1b7SMiao Xie delalloc_root); 446800246528SJosef Bacik root = btrfs_grab_root(root); 4469eb73c1b7SMiao Xie BUG_ON(!root); 4470eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4471eb73c1b7SMiao Xie 4472eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 447300246528SJosef Bacik btrfs_put_root(root); 4474eb73c1b7SMiao Xie 4475eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4476eb73c1b7SMiao Xie } 4477eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4478acce952bSliubo } 4479acce952bSliubo 44802ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4481acce952bSliubo struct extent_io_tree *dirty_pages, 4482acce952bSliubo int mark) 4483acce952bSliubo { 4484acce952bSliubo int ret; 4485acce952bSliubo struct extent_buffer *eb; 4486acce952bSliubo u64 start = 0; 4487acce952bSliubo u64 end; 4488acce952bSliubo 4489acce952bSliubo while (1) { 4490acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4491e6138876SJosef Bacik mark, NULL); 4492acce952bSliubo if (ret) 4493acce952bSliubo break; 4494acce952bSliubo 449591166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4496acce952bSliubo while (start <= end) { 44970b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 44980b246afaSJeff Mahoney start += fs_info->nodesize; 4499fd8b2b61SJosef Bacik if (!eb) 4500acce952bSliubo continue; 4501fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4502acce952bSliubo 4503fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4504fd8b2b61SJosef Bacik &eb->bflags)) 4505fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4506fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4507acce952bSliubo } 4508acce952bSliubo } 4509acce952bSliubo 4510acce952bSliubo return ret; 4511acce952bSliubo } 4512acce952bSliubo 45132ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4514fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4515acce952bSliubo { 4516acce952bSliubo u64 start; 4517acce952bSliubo u64 end; 4518acce952bSliubo int ret; 4519acce952bSliubo 4520acce952bSliubo while (1) { 45210e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 45220e6ec385SFilipe Manana 4523fcd5e742SLu Fengqi /* 4524fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4525fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4526fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4527fcd5e742SLu Fengqi * the same extent range. 4528fcd5e742SLu Fengqi */ 4529fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4530acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 45310e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4532fcd5e742SLu Fengqi if (ret) { 4533fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4534acce952bSliubo break; 4535fcd5e742SLu Fengqi } 4536acce952bSliubo 45370e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 45380e6ec385SFilipe Manana free_extent_state(cached_state); 45392ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4540fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4541acce952bSliubo cond_resched(); 4542acce952bSliubo } 4543acce952bSliubo 4544acce952bSliubo return 0; 4545acce952bSliubo } 4546acce952bSliubo 454732da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4548c79a1751SLiu Bo { 4549c79a1751SLiu Bo struct inode *inode; 4550c79a1751SLiu Bo 4551c79a1751SLiu Bo inode = cache->io_ctl.inode; 4552c79a1751SLiu Bo if (inode) { 4553c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4554c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4555c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4556c79a1751SLiu Bo iput(inode); 4557c79a1751SLiu Bo } 4558bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4559c79a1751SLiu Bo btrfs_put_block_group(cache); 4560c79a1751SLiu Bo } 4561c79a1751SLiu Bo 4562c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 45632ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4564c79a1751SLiu Bo { 456532da5386SDavid Sterba struct btrfs_block_group *cache; 4566c79a1751SLiu Bo 4567c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4568c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4569c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 457032da5386SDavid Sterba struct btrfs_block_group, 4571c79a1751SLiu Bo dirty_list); 4572c79a1751SLiu Bo 4573c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4574c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4575c79a1751SLiu Bo list_del_init(&cache->io_list); 4576c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4577c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4578c79a1751SLiu Bo } 4579c79a1751SLiu Bo 4580c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4581c79a1751SLiu Bo spin_lock(&cache->lock); 4582c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4583c79a1751SLiu Bo spin_unlock(&cache->lock); 4584c79a1751SLiu Bo 4585c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4586c79a1751SLiu Bo btrfs_put_block_group(cache); 4587ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4588c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4589c79a1751SLiu Bo } 4590c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4591c79a1751SLiu Bo 459245ae2c18SNikolay Borisov /* 459345ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 459445ae2c18SNikolay Borisov * to use it without any locking 459545ae2c18SNikolay Borisov */ 4596c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4597c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 459832da5386SDavid Sterba struct btrfs_block_group, 4599c79a1751SLiu Bo io_list); 4600c79a1751SLiu Bo 4601c79a1751SLiu Bo list_del_init(&cache->io_list); 4602c79a1751SLiu Bo spin_lock(&cache->lock); 4603c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4604c79a1751SLiu Bo spin_unlock(&cache->lock); 4605c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4606c79a1751SLiu Bo } 4607c79a1751SLiu Bo } 4608c79a1751SLiu Bo 460949b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 46102ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 461149b25e05SJeff Mahoney { 4612bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4613bbbf7243SNikolay Borisov 46142ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4615c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4616c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4617c79a1751SLiu Bo 4618bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4619bbbf7243SNikolay Borisov post_commit_list) { 4620bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4621bbbf7243SNikolay Borisov } 4622bbbf7243SNikolay Borisov 46232ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 462449b25e05SJeff Mahoney 46254a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 46260b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 462749b25e05SJeff Mahoney 46284a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 46290b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 463049b25e05SJeff Mahoney 4631ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 463267cde344SMiao Xie 46332ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 463449b25e05SJeff Mahoney EXTENT_DIRTY); 4635fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 463649b25e05SJeff Mahoney 46374a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 46384a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 463949b25e05SJeff Mahoney } 464049b25e05SJeff Mahoney 46412ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4642acce952bSliubo { 4643acce952bSliubo struct btrfs_transaction *t; 4644acce952bSliubo 46450b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4646acce952bSliubo 46470b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 46480b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 46490b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4650724e2315SJosef Bacik struct btrfs_transaction, list); 4651724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 46529b64f57dSElena Reshetova refcount_inc(&t->use_count); 46530b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 46542ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4655724e2315SJosef Bacik btrfs_put_transaction(t); 46560b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4657724e2315SJosef Bacik continue; 4658724e2315SJosef Bacik } 46590b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4660724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 46610b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4662724e2315SJosef Bacik /* 4663724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4664724e2315SJosef Bacik * handle open currently for this transaction. 4665724e2315SJosef Bacik */ 4666724e2315SJosef Bacik wait_event(t->writer_wait, 4667724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4668724e2315SJosef Bacik } else { 46690b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4670724e2315SJosef Bacik } 46712ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4672724e2315SJosef Bacik 46730b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 46740b246afaSJeff Mahoney if (t == fs_info->running_transaction) 46750b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4676724e2315SJosef Bacik list_del_init(&t->list); 46770b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4678a4abeea4SJosef Bacik 4679724e2315SJosef Bacik btrfs_put_transaction(t); 46802ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 46810b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4682724e2315SJosef Bacik } 46830b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 46840b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4685ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4686ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 46870b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4688ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 46890b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4690acce952bSliubo 4691acce952bSliubo return 0; 4692acce952bSliubo } 4693