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" 3221adbd5cSStefan Behrens #include "check-integrity.h" 33606686eeSJosef Bacik #include "rcu-string.h" 348dabb742SStefan Behrens #include "dev-replace.h" 3553b381b3SDavid Woodhouse #include "raid56.h" 365ac1d209SJeff Mahoney #include "sysfs.h" 37fcebe456SJosef Bacik #include "qgroup.h" 38ebb8765bSAnand Jain #include "compression.h" 39557ea5ddSQu Wenruo #include "tree-checker.h" 40fd708b81SJosef Bacik #include "ref-verify.h" 41aac0023cSJosef Bacik #include "block-group.h" 42b0643e59SDennis Zhou #include "discard.h" 43f603bb94SNikolay Borisov #include "space-info.h" 44b70f5097SNaohiro Aota #include "zoned.h" 45139e8cd3SQu Wenruo #include "subpage.h" 46eb60ceacSChris Mason 47319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 48319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 49319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 50319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 51e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 52e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 53319e4d06SQu Wenruo 548b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 55143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 56acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 572ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 58143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 592ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 60acce952bSliubo struct extent_io_tree *dirty_pages, 61acce952bSliubo int mark); 622ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 63acce952bSliubo struct extent_io_tree *pinned_extents); 642ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 652ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 66ce9adaa5SChris Mason 67d352ac68SChris Mason /* 6897eb6b69SDavid Sterba * btrfs_end_io_wq structs are used to do processing in task context when an IO 6997eb6b69SDavid Sterba * is complete. This is used during reads to verify checksums, and it is used 70d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 71d352ac68SChris Mason */ 7297eb6b69SDavid Sterba struct btrfs_end_io_wq { 73ce9adaa5SChris Mason struct bio *bio; 74ce9adaa5SChris Mason bio_end_io_t *end_io; 75ce9adaa5SChris Mason void *private; 76ce9adaa5SChris Mason struct btrfs_fs_info *info; 774e4cbee9SChristoph Hellwig blk_status_t status; 78bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata; 798b712842SChris Mason struct btrfs_work work; 80ce9adaa5SChris Mason }; 810da5468fSChris Mason 8297eb6b69SDavid Sterba static struct kmem_cache *btrfs_end_io_wq_cache; 8397eb6b69SDavid Sterba 8497eb6b69SDavid Sterba int __init btrfs_end_io_wq_init(void) 8597eb6b69SDavid Sterba { 8697eb6b69SDavid Sterba btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", 8797eb6b69SDavid Sterba sizeof(struct btrfs_end_io_wq), 8897eb6b69SDavid Sterba 0, 89fba4b697SNikolay Borisov SLAB_MEM_SPREAD, 9097eb6b69SDavid Sterba NULL); 9197eb6b69SDavid Sterba if (!btrfs_end_io_wq_cache) 9297eb6b69SDavid Sterba return -ENOMEM; 9397eb6b69SDavid Sterba return 0; 9497eb6b69SDavid Sterba } 9597eb6b69SDavid Sterba 96e67c718bSDavid Sterba void __cold btrfs_end_io_wq_exit(void) 9797eb6b69SDavid Sterba { 9897eb6b69SDavid Sterba kmem_cache_destroy(btrfs_end_io_wq_cache); 9997eb6b69SDavid Sterba } 10097eb6b69SDavid Sterba 101141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 102141386e1SJosef Bacik { 103141386e1SJosef Bacik if (fs_info->csum_shash) 104141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 105141386e1SJosef Bacik } 106141386e1SJosef Bacik 107d352ac68SChris Mason /* 108d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 109d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 110d352ac68SChris Mason * just before they are sent down the IO stack. 111d352ac68SChris Mason */ 11244b8bd7eSChris Mason struct async_submit_bio { 1138896a08dSQu Wenruo struct inode *inode; 11444b8bd7eSChris Mason struct bio *bio; 115a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 11644b8bd7eSChris Mason int mirror_num; 1171941b64bSQu Wenruo 1181941b64bSQu Wenruo /* Optional parameter for submit_bio_start used by direct io */ 1191941b64bSQu Wenruo u64 dio_file_offset; 1208b712842SChris Mason struct btrfs_work work; 1214e4cbee9SChristoph Hellwig blk_status_t status; 12244b8bd7eSChris Mason }; 12344b8bd7eSChris Mason 12485d4e461SChris Mason /* 12585d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 12685d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 12785d4e461SChris Mason * the level the eb occupies in the tree. 1284008c04aSChris Mason * 12985d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 13085d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 13185d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 1324fd786e6SMisono Tomohiro * by btrfs_root->root_key.objectid. This ensures that all special purpose 1334fd786e6SMisono Tomohiro * roots have separate keysets. 1344008c04aSChris Mason * 13585d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 13685d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 13785d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1384008c04aSChris Mason * 13985d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 14085d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 14185d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 14285d4e461SChris Mason * 14385d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 14485d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 14585d4e461SChris Mason * needs update as well. 1464008c04aSChris Mason */ 1474008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1484008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1494008c04aSChris Mason # error 1504008c04aSChris Mason # endif 15185d4e461SChris Mason 152ab1405aaSDavid Sterba #define DEFINE_LEVEL(stem, level) \ 153ab1405aaSDavid Sterba .names[level] = "btrfs-" stem "-0" #level, 154ab1405aaSDavid Sterba 155ab1405aaSDavid Sterba #define DEFINE_NAME(stem) \ 156ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 0) \ 157ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 1) \ 158ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 2) \ 159ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 3) \ 160ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 4) \ 161ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 5) \ 162ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 6) \ 163ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 7) 164ab1405aaSDavid Sterba 16585d4e461SChris Mason static struct btrfs_lockdep_keyset { 16685d4e461SChris Mason u64 id; /* root objectid */ 167ab1405aaSDavid Sterba /* Longest entry: btrfs-free-space-00 */ 168387824afSDavid Sterba char names[BTRFS_MAX_LEVEL][20]; 169387824afSDavid Sterba struct lock_class_key keys[BTRFS_MAX_LEVEL]; 17085d4e461SChris Mason } btrfs_lockdep_keysets[] = { 171ab1405aaSDavid Sterba { .id = BTRFS_ROOT_TREE_OBJECTID, DEFINE_NAME("root") }, 172ab1405aaSDavid Sterba { .id = BTRFS_EXTENT_TREE_OBJECTID, DEFINE_NAME("extent") }, 173ab1405aaSDavid Sterba { .id = BTRFS_CHUNK_TREE_OBJECTID, DEFINE_NAME("chunk") }, 174ab1405aaSDavid Sterba { .id = BTRFS_DEV_TREE_OBJECTID, DEFINE_NAME("dev") }, 175ab1405aaSDavid Sterba { .id = BTRFS_CSUM_TREE_OBJECTID, DEFINE_NAME("csum") }, 176ab1405aaSDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, DEFINE_NAME("quota") }, 177ab1405aaSDavid Sterba { .id = BTRFS_TREE_LOG_OBJECTID, DEFINE_NAME("log") }, 178ab1405aaSDavid Sterba { .id = BTRFS_TREE_RELOC_OBJECTID, DEFINE_NAME("treloc") }, 179ab1405aaSDavid Sterba { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, DEFINE_NAME("dreloc") }, 180ab1405aaSDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, DEFINE_NAME("uuid") }, 181ab1405aaSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, DEFINE_NAME("free-space") }, 182ab1405aaSDavid Sterba { .id = 0, DEFINE_NAME("tree") }, 1834008c04aSChris Mason }; 18485d4e461SChris Mason 185ab1405aaSDavid Sterba #undef DEFINE_LEVEL 186ab1405aaSDavid Sterba #undef DEFINE_NAME 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; 2127280305eSDavid Sterba const int num_pages = num_extent_pages(buf); 213a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 214d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 21519c00ddcSChris Mason char *kaddr; 216e9be5a30SDavid Sterba int i; 217d5178578SJohannes Thumshirn 218d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 219d5178578SJohannes Thumshirn crypto_shash_init(shash); 220a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 221e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 222a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 22319c00ddcSChris Mason 224e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 225e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 226e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 22719c00ddcSChris Mason } 22871a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 229d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 23019c00ddcSChris Mason } 23119c00ddcSChris Mason 232d352ac68SChris Mason /* 233d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 234d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 235d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 236d352ac68SChris Mason * in the wrong place. 237d352ac68SChris Mason */ 2381259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 239b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 240b9fab919SChris Mason int atomic) 2411259ab75SChris Mason { 2422ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 2431259ab75SChris Mason int ret; 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 2512ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 252ff13db41SDavid Sterba &cached_state); 2530b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 2541259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 2551259ab75SChris Mason ret = 0; 2561259ab75SChris Mason goto out; 2571259ab75SChris Mason } 25894647322SDavid Sterba btrfs_err_rl(eb->fs_info, 25994647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 26094647322SDavid Sterba eb->start, 26129549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 2621259ab75SChris Mason ret = 1; 2630b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 26433958dc6SChris Mason out: 2652ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 266e43bbe5eSDavid Sterba &cached_state); 2671259ab75SChris Mason return ret; 2681259ab75SChris Mason } 2691259ab75SChris Mason 270e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 271e7e16f48SJohannes Thumshirn { 272e7e16f48SJohannes Thumshirn switch (csum_type) { 273e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 2743951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 2753831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 276352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 277e7e16f48SJohannes Thumshirn return true; 278e7e16f48SJohannes Thumshirn default: 279e7e16f48SJohannes Thumshirn return false; 280e7e16f48SJohannes Thumshirn } 281e7e16f48SJohannes Thumshirn } 282e7e16f48SJohannes Thumshirn 283d352ac68SChris Mason /* 2841104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 2851104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 2861104a885SDavid Sterba */ 287ab8d0fc4SJeff Mahoney static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 288ab8d0fc4SJeff Mahoney char *raw_disk_sb) 2891104a885SDavid Sterba { 2901104a885SDavid Sterba struct btrfs_super_block *disk_sb = 2911104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 29251bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 293d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 294d5178578SJohannes Thumshirn 295d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 2961104a885SDavid Sterba 2971104a885SDavid Sterba /* 2981104a885SDavid Sterba * The super_block structure does not span the whole 29951bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 30051bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 3011104a885SDavid Sterba */ 302fd08001fSEric Biggers crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, 303fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 3041104a885SDavid Sterba 30555fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 306e7e16f48SJohannes Thumshirn return 1; 3071104a885SDavid Sterba 308e7e16f48SJohannes Thumshirn return 0; 3091104a885SDavid Sterba } 3101104a885SDavid Sterba 311e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 312ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 313581c1760SQu Wenruo { 314e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 315581c1760SQu Wenruo int found_level; 316581c1760SQu Wenruo struct btrfs_key found_key; 317581c1760SQu Wenruo int ret; 318581c1760SQu Wenruo 319581c1760SQu Wenruo found_level = btrfs_header_level(eb); 320581c1760SQu Wenruo if (found_level != level) { 32163489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 32263489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 323581c1760SQu Wenruo btrfs_err(fs_info, 324581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 325581c1760SQu Wenruo eb->start, level, found_level); 326581c1760SQu Wenruo return -EIO; 327581c1760SQu Wenruo } 328581c1760SQu Wenruo 329581c1760SQu Wenruo if (!first_key) 330581c1760SQu Wenruo return 0; 331581c1760SQu Wenruo 3325d41be6fSQu Wenruo /* 3335d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 3345d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 3355d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 3365d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 3375d41be6fSQu Wenruo */ 3385d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 3395d41be6fSQu Wenruo return 0; 34062fdaa52SQu Wenruo 34162fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 34262fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 34362fdaa52SQu Wenruo btrfs_err(fs_info, 34462fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 34562fdaa52SQu Wenruo eb->start); 34662fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 34762fdaa52SQu Wenruo return -EUCLEAN; 34862fdaa52SQu Wenruo } 34962fdaa52SQu Wenruo 350581c1760SQu Wenruo if (found_level) 351581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 352581c1760SQu Wenruo else 353581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 354581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 355581c1760SQu Wenruo 356581c1760SQu Wenruo if (ret) { 35763489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 35863489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 359581c1760SQu Wenruo btrfs_err(fs_info, 360ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 361ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 362ff76a864SLiu Bo first_key->type, first_key->offset, 363ff76a864SLiu Bo found_key.objectid, found_key.type, 364ff76a864SLiu Bo found_key.offset); 365581c1760SQu Wenruo } 366581c1760SQu Wenruo return ret; 367581c1760SQu Wenruo } 368581c1760SQu Wenruo 3691104a885SDavid Sterba /* 370d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 371d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 372581c1760SQu Wenruo * 373581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 374581c1760SQu Wenruo * @level: expected level, mandatory check 375581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 376d352ac68SChris Mason */ 3775ab12d1fSDavid Sterba static int btree_read_extent_buffer_pages(struct extent_buffer *eb, 378581c1760SQu Wenruo u64 parent_transid, int level, 379581c1760SQu Wenruo struct btrfs_key *first_key) 380f188591eSChris Mason { 3815ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 382f188591eSChris Mason struct extent_io_tree *io_tree; 383ea466794SJosef Bacik int failed = 0; 384f188591eSChris Mason int ret; 385f188591eSChris Mason int num_copies = 0; 386f188591eSChris Mason int mirror_num = 0; 387ea466794SJosef Bacik int failed_mirror = 0; 388f188591eSChris Mason 3890b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 390f188591eSChris Mason while (1) { 391f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 392c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); 393256dd1bbSStefan Behrens if (!ret) { 394581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 395b9fab919SChris Mason parent_transid, 0)) 396256dd1bbSStefan Behrens ret = -EIO; 397e064d5e9SDavid Sterba else if (btrfs_verify_level_key(eb, level, 398ff76a864SLiu Bo first_key, parent_transid)) 399581c1760SQu Wenruo ret = -EUCLEAN; 400581c1760SQu Wenruo else 401581c1760SQu Wenruo break; 402256dd1bbSStefan Behrens } 403d397712bSChris Mason 4040b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 405f188591eSChris Mason eb->start, eb->len); 4064235298eSChris Mason if (num_copies == 1) 407ea466794SJosef Bacik break; 4084235298eSChris Mason 4095cf1ab56SJosef Bacik if (!failed_mirror) { 4105cf1ab56SJosef Bacik failed = 1; 4115cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4125cf1ab56SJosef Bacik } 4135cf1ab56SJosef Bacik 414f188591eSChris Mason mirror_num++; 415ea466794SJosef Bacik if (mirror_num == failed_mirror) 416ea466794SJosef Bacik mirror_num++; 417ea466794SJosef Bacik 4184235298eSChris Mason if (mirror_num > num_copies) 419ea466794SJosef Bacik break; 420f188591eSChris Mason } 421ea466794SJosef Bacik 422c0901581SStefan Behrens if (failed && !ret && failed_mirror) 42320a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 424ea466794SJosef Bacik 425ea466794SJosef Bacik return ret; 426f188591eSChris Mason } 42719c00ddcSChris Mason 428eca0f6f6SQu Wenruo static int csum_one_extent_buffer(struct extent_buffer *eb) 429eca0f6f6SQu Wenruo { 430eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 431eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 432eca0f6f6SQu Wenruo int ret; 433eca0f6f6SQu Wenruo 434eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 435eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 436eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 437eca0f6f6SQu Wenruo csum_tree_block(eb, result); 438eca0f6f6SQu Wenruo 439eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 440eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 441eca0f6f6SQu Wenruo else 442eca0f6f6SQu Wenruo ret = btrfs_check_leaf_full(eb); 443eca0f6f6SQu Wenruo 444eca0f6f6SQu Wenruo if (ret < 0) { 445eca0f6f6SQu Wenruo btrfs_print_tree(eb, 0); 446eca0f6f6SQu Wenruo btrfs_err(fs_info, 447eca0f6f6SQu Wenruo "block=%llu write time tree block corruption detected", 448eca0f6f6SQu Wenruo eb->start); 449eca0f6f6SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 450eca0f6f6SQu Wenruo return ret; 451eca0f6f6SQu Wenruo } 452eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 453eca0f6f6SQu Wenruo 454eca0f6f6SQu Wenruo return 0; 455eca0f6f6SQu Wenruo } 456eca0f6f6SQu Wenruo 457eca0f6f6SQu Wenruo /* Checksum all dirty extent buffers in one bio_vec */ 458eca0f6f6SQu Wenruo static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, 459eca0f6f6SQu Wenruo struct bio_vec *bvec) 460eca0f6f6SQu Wenruo { 461eca0f6f6SQu Wenruo struct page *page = bvec->bv_page; 462eca0f6f6SQu Wenruo u64 bvec_start = page_offset(page) + bvec->bv_offset; 463eca0f6f6SQu Wenruo u64 cur; 464eca0f6f6SQu Wenruo int ret = 0; 465eca0f6f6SQu Wenruo 466eca0f6f6SQu Wenruo for (cur = bvec_start; cur < bvec_start + bvec->bv_len; 467eca0f6f6SQu Wenruo cur += fs_info->nodesize) { 468eca0f6f6SQu Wenruo struct extent_buffer *eb; 469eca0f6f6SQu Wenruo bool uptodate; 470eca0f6f6SQu Wenruo 471eca0f6f6SQu Wenruo eb = find_extent_buffer(fs_info, cur); 472eca0f6f6SQu Wenruo uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, 473eca0f6f6SQu Wenruo fs_info->nodesize); 474eca0f6f6SQu Wenruo 475eca0f6f6SQu Wenruo /* A dirty eb shouldn't disappear from buffer_radix */ 476eca0f6f6SQu Wenruo if (WARN_ON(!eb)) 477eca0f6f6SQu Wenruo return -EUCLEAN; 478eca0f6f6SQu Wenruo 479eca0f6f6SQu Wenruo if (WARN_ON(cur != btrfs_header_bytenr(eb))) { 480eca0f6f6SQu Wenruo free_extent_buffer(eb); 481eca0f6f6SQu Wenruo return -EUCLEAN; 482eca0f6f6SQu Wenruo } 483eca0f6f6SQu Wenruo if (WARN_ON(!uptodate)) { 484eca0f6f6SQu Wenruo free_extent_buffer(eb); 485eca0f6f6SQu Wenruo return -EUCLEAN; 486eca0f6f6SQu Wenruo } 487eca0f6f6SQu Wenruo 488eca0f6f6SQu Wenruo ret = csum_one_extent_buffer(eb); 489eca0f6f6SQu Wenruo free_extent_buffer(eb); 490eca0f6f6SQu Wenruo if (ret < 0) 491eca0f6f6SQu Wenruo return ret; 492eca0f6f6SQu Wenruo } 493eca0f6f6SQu Wenruo return ret; 494eca0f6f6SQu Wenruo } 495eca0f6f6SQu Wenruo 496d352ac68SChris Mason /* 497ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 498ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 499ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 500d352ac68SChris Mason */ 501ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 50219c00ddcSChris Mason { 503ac303b69SQu Wenruo struct page *page = bvec->bv_page; 5044eee4fa4SMiao Xie u64 start = page_offset(page); 50519c00ddcSChris Mason u64 found_start; 50619c00ddcSChris Mason struct extent_buffer *eb; 507eca0f6f6SQu Wenruo 508eca0f6f6SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE) 509eca0f6f6SQu Wenruo return csum_dirty_subpage_buffers(fs_info, bvec); 510f188591eSChris Mason 5114f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5124f2de97aSJosef Bacik if (page != eb->pages[0]) 5134f2de97aSJosef Bacik return 0; 5140f805531SAlex Lyakas 51519c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 516d3575156SNaohiro Aota 517d3575156SNaohiro Aota if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 518d3575156SNaohiro Aota WARN_ON(found_start != 0); 519d3575156SNaohiro Aota return 0; 520d3575156SNaohiro Aota } 521d3575156SNaohiro Aota 5220f805531SAlex Lyakas /* 5230f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5240f805531SAlex Lyakas * It is useful to know what went wrong. 5250f805531SAlex Lyakas */ 5260f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5270f805531SAlex Lyakas return -EUCLEAN; 5280f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5290f805531SAlex Lyakas return -EUCLEAN; 5300f805531SAlex Lyakas 531eca0f6f6SQu Wenruo return csum_one_extent_buffer(eb); 53219c00ddcSChris Mason } 53319c00ddcSChris Mason 534b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 5352b82032cSYan Zheng { 536b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 537944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 53844880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 539944d3f9fSNikolay Borisov u8 *metadata_uuid; 5402b82032cSYan Zheng 5419a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 5429a8658e3SDavid Sterba BTRFS_FSID_SIZE); 5437239ff4bSNikolay Borisov /* 544944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 545944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 546944d3f9fSNikolay Borisov * ->fsid in this case is fine 5477239ff4bSNikolay Borisov */ 548944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 5497239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 5507239ff4bSNikolay Borisov else 5517239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 5527239ff4bSNikolay Borisov 553944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 554944d3f9fSNikolay Borisov return 0; 555944d3f9fSNikolay Borisov 556944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 557944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 558944d3f9fSNikolay Borisov return 0; 559944d3f9fSNikolay Borisov 560944d3f9fSNikolay Borisov return 1; 5612b82032cSYan Zheng } 5622b82032cSYan Zheng 56377bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 56477bf40a2SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb) 565ce9adaa5SChris Mason { 56677bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 567ce9adaa5SChris Mason u64 found_start; 56877bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 56977bf40a2SQu Wenruo u8 found_level; 5702996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 571dfd29eedSDavid Sterba const u8 *header_csum; 57277bf40a2SQu Wenruo int ret = 0; 573ea466794SJosef Bacik 574ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 575727011e0SChris Mason if (found_start != eb->start) { 576893bf4b1SSu Yue btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", 577893bf4b1SSu Yue eb->start, found_start); 578f188591eSChris Mason ret = -EIO; 57977bf40a2SQu Wenruo goto out; 580ce9adaa5SChris Mason } 581b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 58202873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 58394647322SDavid Sterba eb->start); 5841259ab75SChris Mason ret = -EIO; 58577bf40a2SQu Wenruo goto out; 5861259ab75SChris Mason } 587ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 5881c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 589893bf4b1SSu Yue btrfs_err(fs_info, "bad tree block level %d on %llu", 590893bf4b1SSu Yue (int)btrfs_header_level(eb), eb->start); 5911c24c3ceSJosef Bacik ret = -EIO; 59277bf40a2SQu Wenruo goto out; 5931c24c3ceSJosef Bacik } 594ce9adaa5SChris Mason 595c67b3892SDavid Sterba csum_tree_block(eb, result); 596dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 597dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 598a826d6dcSJosef Bacik 599dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 6002996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 601ff14aa79SDavid Sterba "checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", 602ff14aa79SDavid Sterba eb->start, 603dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 60435be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 60535be8851SJohannes Thumshirn btrfs_header_level(eb)); 6062996e1f8SJohannes Thumshirn ret = -EUCLEAN; 60777bf40a2SQu Wenruo goto out; 6082996e1f8SJohannes Thumshirn } 6092996e1f8SJohannes Thumshirn 610a826d6dcSJosef Bacik /* 611a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 612a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 613a826d6dcSJosef Bacik * return -EIO. 614a826d6dcSJosef Bacik */ 6151c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 616a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 617a826d6dcSJosef Bacik ret = -EIO; 618a826d6dcSJosef Bacik } 619ce9adaa5SChris Mason 620813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 621053ab70fSLiu Bo ret = -EIO; 622053ab70fSLiu Bo 6230b32f4bbSJosef Bacik if (!ret) 6240b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 62575391f0dSQu Wenruo else 62675391f0dSQu Wenruo btrfs_err(fs_info, 62775391f0dSQu Wenruo "block=%llu read time tree block corruption detected", 62875391f0dSQu Wenruo eb->start); 62977bf40a2SQu Wenruo out: 63077bf40a2SQu Wenruo return ret; 63177bf40a2SQu Wenruo } 63277bf40a2SQu Wenruo 633371cdc07SQu Wenruo static int validate_subpage_buffer(struct page *page, u64 start, u64 end, 634371cdc07SQu Wenruo int mirror) 635371cdc07SQu Wenruo { 636371cdc07SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 637371cdc07SQu Wenruo struct extent_buffer *eb; 638371cdc07SQu Wenruo bool reads_done; 639371cdc07SQu Wenruo int ret = 0; 640371cdc07SQu Wenruo 641371cdc07SQu Wenruo /* 642371cdc07SQu Wenruo * We don't allow bio merge for subpage metadata read, so we should 643371cdc07SQu Wenruo * only get one eb for each endio hook. 644371cdc07SQu Wenruo */ 645371cdc07SQu Wenruo ASSERT(end == start + fs_info->nodesize - 1); 646371cdc07SQu Wenruo ASSERT(PagePrivate(page)); 647371cdc07SQu Wenruo 648371cdc07SQu Wenruo eb = find_extent_buffer(fs_info, start); 649371cdc07SQu Wenruo /* 650371cdc07SQu Wenruo * When we are reading one tree block, eb must have been inserted into 651371cdc07SQu Wenruo * the radix tree. If not, something is wrong. 652371cdc07SQu Wenruo */ 653371cdc07SQu Wenruo ASSERT(eb); 654371cdc07SQu Wenruo 655371cdc07SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 656371cdc07SQu Wenruo /* Subpage read must finish in page read */ 657371cdc07SQu Wenruo ASSERT(reads_done); 658371cdc07SQu Wenruo 659371cdc07SQu Wenruo eb->read_mirror = mirror; 660371cdc07SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 661371cdc07SQu Wenruo ret = -EIO; 662371cdc07SQu Wenruo goto err; 663371cdc07SQu Wenruo } 664371cdc07SQu Wenruo ret = validate_extent_buffer(eb); 665371cdc07SQu Wenruo if (ret < 0) 666371cdc07SQu Wenruo goto err; 667371cdc07SQu Wenruo 668371cdc07SQu Wenruo if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 669371cdc07SQu Wenruo btree_readahead_hook(eb, ret); 670371cdc07SQu Wenruo 671371cdc07SQu Wenruo set_extent_buffer_uptodate(eb); 672371cdc07SQu Wenruo 673371cdc07SQu Wenruo free_extent_buffer(eb); 674371cdc07SQu Wenruo return ret; 675371cdc07SQu Wenruo err: 676371cdc07SQu Wenruo /* 677371cdc07SQu Wenruo * end_bio_extent_readpage decrements io_pages in case of error, 678371cdc07SQu Wenruo * make sure it has something to decrement. 679371cdc07SQu Wenruo */ 680371cdc07SQu Wenruo atomic_inc(&eb->io_pages); 681371cdc07SQu Wenruo clear_extent_buffer_uptodate(eb); 682371cdc07SQu Wenruo free_extent_buffer(eb); 683371cdc07SQu Wenruo return ret; 684371cdc07SQu Wenruo } 685371cdc07SQu Wenruo 686c3a3b19bSQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, 68777bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 68877bf40a2SQu Wenruo int mirror) 68977bf40a2SQu Wenruo { 69077bf40a2SQu Wenruo struct extent_buffer *eb; 69177bf40a2SQu Wenruo int ret = 0; 69277bf40a2SQu Wenruo int reads_done; 69377bf40a2SQu Wenruo 69477bf40a2SQu Wenruo ASSERT(page->private); 695371cdc07SQu Wenruo 696371cdc07SQu Wenruo if (btrfs_sb(page->mapping->host->i_sb)->sectorsize < PAGE_SIZE) 697371cdc07SQu Wenruo return validate_subpage_buffer(page, start, end, mirror); 698371cdc07SQu Wenruo 69977bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 70077bf40a2SQu Wenruo 70177bf40a2SQu Wenruo /* 70277bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 70377bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 70477bf40a2SQu Wenruo */ 70577bf40a2SQu Wenruo atomic_inc(&eb->refs); 70677bf40a2SQu Wenruo 70777bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 70877bf40a2SQu Wenruo if (!reads_done) 70977bf40a2SQu Wenruo goto err; 71077bf40a2SQu Wenruo 71177bf40a2SQu Wenruo eb->read_mirror = mirror; 71277bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 71377bf40a2SQu Wenruo ret = -EIO; 71477bf40a2SQu Wenruo goto err; 71577bf40a2SQu Wenruo } 71677bf40a2SQu Wenruo ret = validate_extent_buffer(eb); 717ce9adaa5SChris Mason err: 71879fb65a1SJosef Bacik if (reads_done && 71979fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 720d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 7214bb31e92SArne Jansen 72253b381b3SDavid Woodhouse if (ret) { 72353b381b3SDavid Woodhouse /* 72453b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 72553b381b3SDavid Woodhouse * again, we have to make sure it has something 72653b381b3SDavid Woodhouse * to decrement 72753b381b3SDavid Woodhouse */ 72853b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 7290b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 73053b381b3SDavid Woodhouse } 7310b32f4bbSJosef Bacik free_extent_buffer(eb); 73277bf40a2SQu Wenruo 733f188591eSChris Mason return ret; 734ce9adaa5SChris Mason } 735ce9adaa5SChris Mason 7364246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 737ce9adaa5SChris Mason { 73897eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 739ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 7409e0af237SLiu Bo struct btrfs_workqueue *wq; 741ce9adaa5SChris Mason 742ce9adaa5SChris Mason fs_info = end_io_wq->info; 7434e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 744d20f7043SChris Mason 745cfe94440SNaohiro Aota if (btrfs_op(bio) == BTRFS_MAP_WRITE) { 746a0cac0ecSOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) 7479e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 748a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) 7499e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 750a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 7519e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 752a0cac0ecSOmar Sandoval else 7539e0af237SLiu Bo wq = fs_info->endio_write_workers; 7549e0af237SLiu Bo } else { 7555c047a69SOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 7569e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 757a0cac0ecSOmar Sandoval else if (end_io_wq->metadata) 7589e0af237SLiu Bo wq = fs_info->endio_meta_workers; 759a0cac0ecSOmar Sandoval else 7609e0af237SLiu Bo wq = fs_info->endio_workers; 7619e0af237SLiu Bo } 7629e0af237SLiu Bo 763a0cac0ecSOmar Sandoval btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); 7649e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 765ce9adaa5SChris Mason } 766ce9adaa5SChris Mason 7674e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 768bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7690b86a832SChris Mason { 77097eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7718b110e39SMiao Xie 77297eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 773ce9adaa5SChris Mason if (!end_io_wq) 7744e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 775ce9adaa5SChris Mason 776ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 777ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 77822c59948SChris Mason end_io_wq->info = info; 7794e4cbee9SChristoph Hellwig end_io_wq->status = 0; 780ce9adaa5SChris Mason end_io_wq->bio = bio; 78122c59948SChris Mason end_io_wq->metadata = metadata; 782ce9adaa5SChris Mason 783ce9adaa5SChris Mason bio->bi_private = end_io_wq; 784ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 78522c59948SChris Mason return 0; 78622c59948SChris Mason } 78722c59948SChris Mason 7884a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7894a69a410SChris Mason { 7904a69a410SChris Mason struct async_submit_bio *async; 7914e4cbee9SChristoph Hellwig blk_status_t ret; 7924a69a410SChris Mason 7934a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7941941b64bSQu Wenruo ret = async->submit_bio_start(async->inode, async->bio, 7951941b64bSQu Wenruo async->dio_file_offset); 79679787eaaSJeff Mahoney if (ret) 7974e4cbee9SChristoph Hellwig async->status = ret; 7984a69a410SChris Mason } 7994a69a410SChris Mason 80006ea01b1SDavid Sterba /* 80106ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 80206ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 80306ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 80406ea01b1SDavid Sterba * 80506ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 80606ea01b1SDavid Sterba * inserted into the tree. 80706ea01b1SDavid Sterba */ 8084a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 8098b712842SChris Mason { 8108b712842SChris Mason struct async_submit_bio *async; 81106ea01b1SDavid Sterba struct inode *inode; 81206ea01b1SDavid Sterba blk_status_t ret; 8138b712842SChris Mason 8148b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 8158896a08dSQu Wenruo inode = async->inode; 8164854ddd0SChris Mason 817bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 8184e4cbee9SChristoph Hellwig if (async->status) { 8194e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 8204246a0b6SChristoph Hellwig bio_endio(async->bio); 82179787eaaSJeff Mahoney return; 82279787eaaSJeff Mahoney } 82379787eaaSJeff Mahoney 824ec39f769SChris Mason /* 825ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 826ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 827ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 828ec39f769SChris Mason */ 829ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 83008635baeSChris Mason ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 83106ea01b1SDavid Sterba if (ret) { 83206ea01b1SDavid Sterba async->bio->bi_status = ret; 83306ea01b1SDavid Sterba bio_endio(async->bio); 83406ea01b1SDavid Sterba } 8354a69a410SChris Mason } 8364a69a410SChris Mason 8374a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 8384a69a410SChris Mason { 8394a69a410SChris Mason struct async_submit_bio *async; 8404a69a410SChris Mason 8414a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 8428b712842SChris Mason kfree(async); 8438b712842SChris Mason } 8448b712842SChris Mason 8458896a08dSQu Wenruo blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, 846c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 8471941b64bSQu Wenruo u64 dio_file_offset, 848e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 84944b8bd7eSChris Mason { 8508896a08dSQu Wenruo struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 85144b8bd7eSChris Mason struct async_submit_bio *async; 85244b8bd7eSChris Mason 85344b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 85444b8bd7eSChris Mason if (!async) 8554e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 85644b8bd7eSChris Mason 8578896a08dSQu Wenruo async->inode = inode; 85844b8bd7eSChris Mason async->bio = bio; 85944b8bd7eSChris Mason async->mirror_num = mirror_num; 8604a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8614a69a410SChris Mason 862a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 863a0cac0ecSOmar Sandoval run_one_async_free); 8644a69a410SChris Mason 8651941b64bSQu Wenruo async->dio_file_offset = dio_file_offset; 8668c8bee1dSChris Mason 8674e4cbee9SChristoph Hellwig async->status = 0; 86879787eaaSJeff Mahoney 86967f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 8705cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 871d313d7a3SChris Mason 8725cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 87344b8bd7eSChris Mason return 0; 87444b8bd7eSChris Mason } 87544b8bd7eSChris Mason 8764e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 877ce3ed71aSChris Mason { 8782c30c71bSKent Overstreet struct bio_vec *bvec; 879ce3ed71aSChris Mason struct btrfs_root *root; 8802b070cfeSChristoph Hellwig int ret = 0; 8816dc4f100SMing Lei struct bvec_iter_all iter_all; 882ce3ed71aSChris Mason 883c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 8842b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 885ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 886ac303b69SQu Wenruo ret = csum_dirty_buffer(root->fs_info, bvec); 88779787eaaSJeff Mahoney if (ret) 88879787eaaSJeff Mahoney break; 889ce3ed71aSChris Mason } 8902c30c71bSKent Overstreet 8914e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 892ce3ed71aSChris Mason } 893ce3ed71aSChris Mason 8948896a08dSQu Wenruo static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, 8951941b64bSQu Wenruo u64 dio_file_offset) 89622c59948SChris Mason { 8978b712842SChris Mason /* 8988b712842SChris Mason * when we're called for a write, we're already in the async 8995443be45SChris Mason * submission context. Just jump into btrfs_map_bio 9008b712842SChris Mason */ 90179787eaaSJeff Mahoney return btree_csum_one_bio(bio); 90222c59948SChris Mason } 90322c59948SChris Mason 904f4dcfb30SJohannes Thumshirn static bool should_async_write(struct btrfs_fs_info *fs_info, 9059b4e675aSDavid Sterba struct btrfs_inode *bi) 906de0022b9SJosef Bacik { 9074eef29efSNaohiro Aota if (btrfs_is_zoned(fs_info)) 908f4dcfb30SJohannes Thumshirn return false; 9096300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 910f4dcfb30SJohannes Thumshirn return false; 9119b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 912f4dcfb30SJohannes Thumshirn return false; 913f4dcfb30SJohannes Thumshirn return true; 914de0022b9SJosef Bacik } 915de0022b9SJosef Bacik 9161b36294aSNikolay Borisov blk_status_t btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, 9171b36294aSNikolay Borisov int mirror_num, unsigned long bio_flags) 91844b8bd7eSChris Mason { 9190b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 9204e4cbee9SChristoph Hellwig blk_status_t ret; 921cad321adSChris Mason 922cfe94440SNaohiro Aota if (btrfs_op(bio) != BTRFS_MAP_WRITE) { 923cad321adSChris Mason /* 924cad321adSChris Mason * called for a read, do the setup so that checksum validation 925cad321adSChris Mason * can happen in the async kernel threads 926cad321adSChris Mason */ 9270b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 9280b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 9291d4284bdSChris Mason if (ret) 93061891923SStefan Behrens goto out_w_error; 93108635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 932f4dcfb30SJohannes Thumshirn } else if (!should_async_write(fs_info, BTRFS_I(inode))) { 933de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 934de0022b9SJosef Bacik if (ret) 93561891923SStefan Behrens goto out_w_error; 93608635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 93761891923SStefan Behrens } else { 938cad321adSChris Mason /* 93961891923SStefan Behrens * kthread helpers are used to submit writes so that 94061891923SStefan Behrens * checksumming can happen in parallel across all CPUs 941cad321adSChris Mason */ 9428896a08dSQu Wenruo ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, 9438896a08dSQu Wenruo 0, btree_submit_bio_start); 94444b8bd7eSChris Mason } 94544b8bd7eSChris Mason 9464246a0b6SChristoph Hellwig if (ret) 9474246a0b6SChristoph Hellwig goto out_w_error; 9484246a0b6SChristoph Hellwig return 0; 9494246a0b6SChristoph Hellwig 95061891923SStefan Behrens out_w_error: 9514e4cbee9SChristoph Hellwig bio->bi_status = ret; 9524246a0b6SChristoph Hellwig bio_endio(bio); 95361891923SStefan Behrens return ret; 95461891923SStefan Behrens } 95561891923SStefan Behrens 9563dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 957784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 958a6bc32b8SMel Gorman struct page *newpage, struct page *page, 959a6bc32b8SMel Gorman enum migrate_mode mode) 960784b4e29SChris Mason { 961784b4e29SChris Mason /* 962784b4e29SChris Mason * we can't safely write a btree page from here, 963784b4e29SChris Mason * we haven't done the locking hook 964784b4e29SChris Mason */ 965784b4e29SChris Mason if (PageDirty(page)) 966784b4e29SChris Mason return -EAGAIN; 967784b4e29SChris Mason /* 968784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 969784b4e29SChris Mason * We must have no buffers or drop them. 970784b4e29SChris Mason */ 971784b4e29SChris Mason if (page_has_private(page) && 972784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 973784b4e29SChris Mason return -EAGAIN; 974a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 975784b4e29SChris Mason } 9763dd1462eSJan Beulich #endif 977784b4e29SChris Mason 9780da5468fSChris Mason 9790da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9800da5468fSChris Mason struct writeback_control *wbc) 9810da5468fSChris Mason { 982e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 983e2d84521SMiao Xie int ret; 984e2d84521SMiao Xie 985d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 986448d640bSChris Mason 987448d640bSChris Mason if (wbc->for_kupdate) 988448d640bSChris Mason return 0; 989448d640bSChris Mason 990e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 991b9473439SChris Mason /* this is a bit racy, but that's ok */ 992d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 993d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 994d814a491SEthan Lien fs_info->dirty_metadata_batch); 995e2d84521SMiao Xie if (ret < 0) 996793955bcSChris Mason return 0; 997793955bcSChris Mason } 9980b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9990da5468fSChris Mason } 10000da5468fSChris Mason 100170dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 10025f39d397SChris Mason { 100398509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 100498509cfcSChris Mason return 0; 10050c4e538bSDavid Sterba 1006f7a52a40SDavid Sterba return try_release_extent_buffer(page); 1007d98237b3SChris Mason } 1008d98237b3SChris Mason 1009d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 1010d47992f8SLukas Czerner unsigned int length) 1011d98237b3SChris Mason { 1012d1310b2eSChris Mason struct extent_io_tree *tree; 1013d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10145f39d397SChris Mason extent_invalidatepage(tree, page, offset); 10155f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 10169ad6b7bcSChris Mason if (PagePrivate(page)) { 1017efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 1018efe120a0SFrank Holton "page private not zero on page %llu", 1019efe120a0SFrank Holton (unsigned long long)page_offset(page)); 1020d1b89bc0SGuoqing Jiang detach_page_private(page); 10219ad6b7bcSChris Mason } 1022d98237b3SChris Mason } 1023d98237b3SChris Mason 10240b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 10250b32f4bbSJosef Bacik { 1026bb146eb2SJosef Bacik #ifdef DEBUG 1027139e8cd3SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 1028139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 10290b32f4bbSJosef Bacik struct extent_buffer *eb; 1030139e8cd3SQu Wenruo int cur_bit = 0; 1031139e8cd3SQu Wenruo u64 page_start = page_offset(page); 10320b32f4bbSJosef Bacik 1033139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 10340b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 10350b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 10360b32f4bbSJosef Bacik BUG_ON(!eb); 10370b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10380b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 103949d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 1040139e8cd3SQu Wenruo return __set_page_dirty_nobuffers(page); 1041139e8cd3SQu Wenruo } 1042139e8cd3SQu Wenruo ASSERT(PagePrivate(page) && page->private); 1043139e8cd3SQu Wenruo subpage = (struct btrfs_subpage *)page->private; 1044139e8cd3SQu Wenruo 1045139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 1046139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 1047139e8cd3SQu Wenruo unsigned long flags; 1048139e8cd3SQu Wenruo u64 cur; 1049139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 1050139e8cd3SQu Wenruo 1051139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 1052139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 1053139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 1054139e8cd3SQu Wenruo cur_bit++; 1055139e8cd3SQu Wenruo continue; 1056139e8cd3SQu Wenruo } 1057139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 1058139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 1059139e8cd3SQu Wenruo 1060139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 1061139e8cd3SQu Wenruo ASSERT(eb); 1062139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 1063139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 106449d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 1065139e8cd3SQu Wenruo free_extent_buffer(eb); 1066139e8cd3SQu Wenruo 1067139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 1068139e8cd3SQu Wenruo } 1069bb146eb2SJosef Bacik #endif 10700b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10710b32f4bbSJosef Bacik } 10720b32f4bbSJosef Bacik 10737f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 10740da5468fSChris Mason .writepages = btree_writepages, 10755f39d397SChris Mason .releasepage = btree_releasepage, 10765f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10775a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1078784b4e29SChris Mason .migratepage = btree_migratepage, 10795a92bc88SChris Mason #endif 10800b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1081d98237b3SChris Mason }; 1082123abc88SChris Mason 10832ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 10842ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 10853fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 10863fbaf258SJosef Bacik int level) 10870999df54SChris Mason { 10880b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 10890b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 10903fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 10910999df54SChris Mason } 10920999df54SChris Mason 1093581c1760SQu Wenruo /* 1094581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 1095581c1760SQu Wenruo * verification. 1096581c1760SQu Wenruo * 10971b7ec85eSJosef Bacik * @owner_root: the objectid of the root owner for this block. 1098581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 1099581c1760SQu Wenruo * @level: expected level, mandatory check 1100581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 1101581c1760SQu Wenruo */ 11022ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 11031b7ec85eSJosef Bacik u64 owner_root, u64 parent_transid, 11041b7ec85eSJosef Bacik int level, struct btrfs_key *first_key) 1105e20d96d6SChris Mason { 11065f39d397SChris Mason struct extent_buffer *buf = NULL; 110719c00ddcSChris Mason int ret; 110819c00ddcSChris Mason 11093fbaf258SJosef Bacik buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); 1110c871b0f2SLiu Bo if (IS_ERR(buf)) 1111c871b0f2SLiu Bo return buf; 1112e4204dedSChris Mason 11135ab12d1fSDavid Sterba ret = btree_read_extent_buffer_pages(buf, parent_transid, 1114581c1760SQu Wenruo level, first_key); 11150f0fe8f7SFilipe David Borba Manana if (ret) { 1116537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 111764c043deSLiu Bo return ERR_PTR(ret); 11180f0fe8f7SFilipe David Borba Manana } 11195f39d397SChris Mason return buf; 1120ce9adaa5SChris Mason 1121eb60ceacSChris Mason } 1122eb60ceacSChris Mason 11236a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 1124ed2ff2cbSChris Mason { 11256a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 112655c69072SChris Mason if (btrfs_header_generation(buf) == 1127e2d84521SMiao Xie fs_info->running_transaction->transid) { 112849d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 1129b4ce94deSChris Mason 1130b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1131104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1132e2d84521SMiao Xie -buf->len, 1133e2d84521SMiao Xie fs_info->dirty_metadata_batch); 11340b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1135925baeddSChris Mason } 11365f39d397SChris Mason } 1137ed7b63ebSJosef Bacik } 11385f39d397SChris Mason 1139da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1140e20d96d6SChris Mason u64 objectid) 1141d97e63b6SChris Mason { 11427c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 114396dfcb46SJosef Bacik root->fs_info = fs_info; 1144cfaa7295SChris Mason root->node = NULL; 1145a28ec197SChris Mason root->commit_root = NULL; 114627cdeb70SMiao Xie root->state = 0; 1147d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 11480b86a832SChris Mason 11490f7d52f4SChris Mason root->last_trans = 0; 11506b8fad57SNikolay Borisov root->free_objectid = 0; 1151eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1152199c2a9cSMiao Xie root->nr_ordered_extents = 0; 11536bef4d31SEric Paris root->inode_tree = RB_ROOT; 115416cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1155f0486c68SYan, Zheng root->block_rsv = NULL; 11560b86a832SChris Mason 11570b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 11585d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1159eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1160eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1161199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1162199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1163d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 11642ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 11652ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 11665d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1167eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1168199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1169f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 11702ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 11712ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 11728287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1173a2135011SChris Mason mutex_init(&root->objectid_mutex); 1174e02119d5SChris Mason mutex_init(&root->log_mutex); 117531f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1176573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1177c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 11787237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 11797237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 11807237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 11818b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 11828b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 11837237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 11847237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 11857237f183SYan Zheng atomic_set(&root->log_writers, 0); 11862ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11870700cea7SElena Reshetova refcount_set(&root->refs, 1); 11888ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1189eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 11907237f183SYan Zheng root->log_transid = 0; 1191d1433debSMiao Xie root->log_transid_committed = -1; 1192257c62e1SChris Mason root->last_log_commit = 0; 1193e289f03eSFilipe Manana if (!dummy) { 119443eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 119543eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1196e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1197e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1198e289f03eSFilipe Manana } 1199017e5369SChris Mason 12003768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 12013768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 12026702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 12034d775673SChris Mason root->root_key.objectid = objectid; 12040ee5dc67SAl Viro root->anon_dev = 0; 12058ea05e3aSAlexander Block 12065f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1207370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1208bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1209bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1210bd647ce3SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 1211bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1212bd647ce3SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 1213bd647ce3SJosef Bacik #endif 12143768f368SChris Mason } 12153768f368SChris Mason 121674e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 121796dfcb46SJosef Bacik u64 objectid, gfp_t flags) 12186f07e42eSAl Viro { 121974e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 12206f07e42eSAl Viro if (root) 122196dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 12226f07e42eSAl Viro return root; 12236f07e42eSAl Viro } 12246f07e42eSAl Viro 122506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 122606ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1227da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 122806ea65a3SJosef Bacik { 122906ea65a3SJosef Bacik struct btrfs_root *root; 123006ea65a3SJosef Bacik 12317c0260eeSJeff Mahoney if (!fs_info) 12327c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 12337c0260eeSJeff Mahoney 123496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 123506ea65a3SJosef Bacik if (!root) 123606ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1237da17066cSJeff Mahoney 1238b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1239faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 124006ea65a3SJosef Bacik 124106ea65a3SJosef Bacik return root; 124206ea65a3SJosef Bacik } 124306ea65a3SJosef Bacik #endif 124406ea65a3SJosef Bacik 124520897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 124620897f5cSArne Jansen u64 objectid) 124720897f5cSArne Jansen { 12489b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 124920897f5cSArne Jansen struct extent_buffer *leaf; 125020897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 125120897f5cSArne Jansen struct btrfs_root *root; 125220897f5cSArne Jansen struct btrfs_key key; 1253b89f6d1fSFilipe Manana unsigned int nofs_flag; 125420897f5cSArne Jansen int ret = 0; 125520897f5cSArne Jansen 1256b89f6d1fSFilipe Manana /* 1257b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1258b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1259b89f6d1fSFilipe Manana */ 1260b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 126196dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1262b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 126320897f5cSArne Jansen if (!root) 126420897f5cSArne Jansen return ERR_PTR(-ENOMEM); 126520897f5cSArne Jansen 126620897f5cSArne Jansen root->root_key.objectid = objectid; 126720897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 126820897f5cSArne Jansen root->root_key.offset = 0; 126920897f5cSArne Jansen 12709631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 12719631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 127220897f5cSArne Jansen if (IS_ERR(leaf)) { 127320897f5cSArne Jansen ret = PTR_ERR(leaf); 12741dd05682STsutomu Itoh leaf = NULL; 12758a6a87cdSBoris Burkov goto fail_unlock; 127620897f5cSArne Jansen } 127720897f5cSArne Jansen 127820897f5cSArne Jansen root->node = leaf; 127920897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 128020897f5cSArne Jansen 128120897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 128227cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 128320897f5cSArne Jansen 1284f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1285f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 128620897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 128720897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 128820897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 128920897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 129020897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 129120897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 129220897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 129333d85fdaSQu Wenruo if (is_fstree(objectid)) 1294807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1295807fc790SAndy Shevchenko else 1296807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1297c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 129820897f5cSArne Jansen 12998a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 13008a6a87cdSBoris Burkov 130120897f5cSArne Jansen key.objectid = objectid; 130220897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 130320897f5cSArne Jansen key.offset = 0; 130420897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 130520897f5cSArne Jansen if (ret) 130620897f5cSArne Jansen goto fail; 130720897f5cSArne Jansen 130820897f5cSArne Jansen return root; 13091dd05682STsutomu Itoh 13108a6a87cdSBoris Burkov fail_unlock: 13118c38938cSJosef Bacik if (leaf) 13121dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 13138a6a87cdSBoris Burkov fail: 131400246528SJosef Bacik btrfs_put_root(root); 13151dd05682STsutomu Itoh 13161dd05682STsutomu Itoh return ERR_PTR(ret); 131720897f5cSArne Jansen } 131820897f5cSArne Jansen 13197237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1320e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13210f7d52f4SChris Mason { 13220f7d52f4SChris Mason struct btrfs_root *root; 1323e02119d5SChris Mason 132496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1325e02119d5SChris Mason if (!root) 13267237f183SYan Zheng return ERR_PTR(-ENOMEM); 1327e02119d5SChris Mason 1328e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1329e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1330e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 133127cdeb70SMiao Xie 13326ab6ebb7SNaohiro Aota return root; 13336ab6ebb7SNaohiro Aota } 13346ab6ebb7SNaohiro Aota 13356ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 13366ab6ebb7SNaohiro Aota struct btrfs_root *root) 13376ab6ebb7SNaohiro Aota { 13386ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 13396ab6ebb7SNaohiro Aota 13407237f183SYan Zheng /* 134192a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 134227cdeb70SMiao Xie * 134392a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 134492a7cc42SQu Wenruo * and they go away before a real commit is actually done. 134592a7cc42SQu Wenruo * 134692a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 134792a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 13487237f183SYan Zheng */ 1349e02119d5SChris Mason 13504d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 13519631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 13526ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 13536ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 1354e02119d5SChris Mason 13557237f183SYan Zheng root->node = leaf; 1356e02119d5SChris Mason 1357e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1358e02119d5SChris Mason btrfs_tree_unlock(root->node); 13596ab6ebb7SNaohiro Aota 13606ab6ebb7SNaohiro Aota return 0; 13617237f183SYan Zheng } 13627237f183SYan Zheng 13637237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13647237f183SYan Zheng struct btrfs_fs_info *fs_info) 13657237f183SYan Zheng { 13667237f183SYan Zheng struct btrfs_root *log_root; 13677237f183SYan Zheng 13687237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13697237f183SYan Zheng if (IS_ERR(log_root)) 13707237f183SYan Zheng return PTR_ERR(log_root); 13716ab6ebb7SNaohiro Aota 13723ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 13733ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 13743ddebf27SNaohiro Aota 13756ab6ebb7SNaohiro Aota if (ret) { 13766ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13776ab6ebb7SNaohiro Aota return ret; 13786ab6ebb7SNaohiro Aota } 13793ddebf27SNaohiro Aota } 13806ab6ebb7SNaohiro Aota 13817237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13827237f183SYan Zheng fs_info->log_root_tree = log_root; 13837237f183SYan Zheng return 0; 13847237f183SYan Zheng } 13857237f183SYan Zheng 13867237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13877237f183SYan Zheng struct btrfs_root *root) 13887237f183SYan Zheng { 13890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13907237f183SYan Zheng struct btrfs_root *log_root; 13917237f183SYan Zheng struct btrfs_inode_item *inode_item; 13926ab6ebb7SNaohiro Aota int ret; 13937237f183SYan Zheng 13940b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13957237f183SYan Zheng if (IS_ERR(log_root)) 13967237f183SYan Zheng return PTR_ERR(log_root); 13977237f183SYan Zheng 13986ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 13996ab6ebb7SNaohiro Aota if (ret) { 14006ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 14016ab6ebb7SNaohiro Aota return ret; 14026ab6ebb7SNaohiro Aota } 14036ab6ebb7SNaohiro Aota 14047237f183SYan Zheng log_root->last_trans = trans->transid; 14057237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14067237f183SYan Zheng 14077237f183SYan Zheng inode_item = &log_root->root_item.inode; 14083cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14093cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14103cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1411da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 14120b246afaSJeff Mahoney fs_info->nodesize); 14133cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14147237f183SYan Zheng 14155d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14167237f183SYan Zheng 14177237f183SYan Zheng WARN_ON(root->log_root); 14187237f183SYan Zheng root->log_root = log_root; 14197237f183SYan Zheng root->log_transid = 0; 1420d1433debSMiao Xie root->log_transid_committed = -1; 1421257c62e1SChris Mason root->last_log_commit = 0; 1422e02119d5SChris Mason return 0; 1423e02119d5SChris Mason } 1424e02119d5SChris Mason 142549d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 142649d11beaSJosef Bacik struct btrfs_path *path, 1427cb517eabSMiao Xie struct btrfs_key *key) 1428e02119d5SChris Mason { 1429e02119d5SChris Mason struct btrfs_root *root; 1430e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 143184234f3aSYan Zheng u64 generation; 1432cb517eabSMiao Xie int ret; 1433581c1760SQu Wenruo int level; 1434cb517eabSMiao Xie 143596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 143649d11beaSJosef Bacik if (!root) 143749d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 14380f7d52f4SChris Mason 1439cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1440cb517eabSMiao Xie &root->root_item, &root->root_key); 14410f7d52f4SChris Mason if (ret) { 144213a8a7c8SYan, Zheng if (ret > 0) 144313a8a7c8SYan, Zheng ret = -ENOENT; 144449d11beaSJosef Bacik goto fail; 14450f7d52f4SChris Mason } 144613a8a7c8SYan, Zheng 144784234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1448581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 14492ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 14502ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 14511b7ec85eSJosef Bacik key->objectid, generation, level, NULL); 145264c043deSLiu Bo if (IS_ERR(root->node)) { 145364c043deSLiu Bo ret = PTR_ERR(root->node); 14548c38938cSJosef Bacik root->node = NULL; 145549d11beaSJosef Bacik goto fail; 1456cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1457cb517eabSMiao Xie ret = -EIO; 145849d11beaSJosef Bacik goto fail; 1459416bc658SJosef Bacik } 14605d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1461cb517eabSMiao Xie return root; 146249d11beaSJosef Bacik fail: 146300246528SJosef Bacik btrfs_put_root(root); 146449d11beaSJosef Bacik return ERR_PTR(ret); 146549d11beaSJosef Bacik } 146649d11beaSJosef Bacik 146749d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 146849d11beaSJosef Bacik struct btrfs_key *key) 146949d11beaSJosef Bacik { 147049d11beaSJosef Bacik struct btrfs_root *root; 147149d11beaSJosef Bacik struct btrfs_path *path; 147249d11beaSJosef Bacik 147349d11beaSJosef Bacik path = btrfs_alloc_path(); 147449d11beaSJosef Bacik if (!path) 147549d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 147649d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 147749d11beaSJosef Bacik btrfs_free_path(path); 147849d11beaSJosef Bacik 147949d11beaSJosef Bacik return root; 1480cb517eabSMiao Xie } 1481cb517eabSMiao Xie 14822dfb1e43SQu Wenruo /* 14832dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 14842dfb1e43SQu Wenruo * 14852dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 14862dfb1e43SQu Wenruo */ 14872dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1488cb517eabSMiao Xie { 1489cb517eabSMiao Xie int ret; 1490dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1491cb517eabSMiao Xie 1492dcc3eb96SNikolay Borisov /* 1493dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1494dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1495dcc3eb96SNikolay Borisov */ 1496dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1497dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1498dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1499dcc3eb96SNikolay Borisov if (ret) 15008257b2dcSMiao Xie goto fail; 15018257b2dcSMiao Xie 1502aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 150337f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 150492a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1505f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1506f39e4571SJosef Bacik } 1507f39e4571SJosef Bacik 1508851fd730SQu Wenruo /* 1509851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1510851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1511851fd730SQu Wenruo */ 1512851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1513851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 15142dfb1e43SQu Wenruo if (!anon_dev) { 1515cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1516cb517eabSMiao Xie if (ret) 1517876d2cf1SLiu Bo goto fail; 15182dfb1e43SQu Wenruo } else { 15192dfb1e43SQu Wenruo root->anon_dev = anon_dev; 15202dfb1e43SQu Wenruo } 1521851fd730SQu Wenruo } 1522f32e48e9SChandan Rajendra 1523f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1524453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1525f32e48e9SChandan Rajendra if (ret) { 1526f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1527876d2cf1SLiu Bo goto fail; 1528f32e48e9SChandan Rajendra } 1529f32e48e9SChandan Rajendra 15306b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1531f32e48e9SChandan Rajendra 1532f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1533f32e48e9SChandan Rajendra 1534cb517eabSMiao Xie return 0; 1535cb517eabSMiao Xie fail: 153684db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1537cb517eabSMiao Xie return ret; 1538cb517eabSMiao Xie } 1539cb517eabSMiao Xie 1540a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1541cb517eabSMiao Xie u64 root_id) 1542cb517eabSMiao Xie { 1543cb517eabSMiao Xie struct btrfs_root *root; 1544cb517eabSMiao Xie 1545cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1546cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1547cb517eabSMiao Xie (unsigned long)root_id); 1548bc44d7c4SJosef Bacik if (root) 154900246528SJosef Bacik root = btrfs_grab_root(root); 1550cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1551cb517eabSMiao Xie return root; 1552cb517eabSMiao Xie } 1553cb517eabSMiao Xie 155449d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 155549d11beaSJosef Bacik u64 objectid) 155649d11beaSJosef Bacik { 155749d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 155849d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 155949d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 156049d11beaSJosef Bacik return btrfs_grab_root(fs_info->extent_root); 156149d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 156249d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 156349d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 156449d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 156549d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 156649d11beaSJosef Bacik return btrfs_grab_root(fs_info->csum_root); 156749d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 156849d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 156949d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 157049d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 157149d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 157249d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 157349d11beaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 157449d11beaSJosef Bacik return btrfs_grab_root(fs_info->free_space_root) ? 157549d11beaSJosef Bacik fs_info->free_space_root : ERR_PTR(-ENOENT); 157649d11beaSJosef Bacik return NULL; 157749d11beaSJosef Bacik } 157849d11beaSJosef Bacik 1579cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1580cb517eabSMiao Xie struct btrfs_root *root) 1581cb517eabSMiao Xie { 1582cb517eabSMiao Xie int ret; 1583cb517eabSMiao Xie 1584e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1585cb517eabSMiao Xie if (ret) 1586cb517eabSMiao Xie return ret; 1587cb517eabSMiao Xie 1588cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1589cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1590cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1591cb517eabSMiao Xie root); 1592af01d2e5SJosef Bacik if (ret == 0) { 159300246528SJosef Bacik btrfs_grab_root(root); 159427cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1595af01d2e5SJosef Bacik } 1596cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1597cb517eabSMiao Xie radix_tree_preload_end(); 1598cb517eabSMiao Xie 1599cb517eabSMiao Xie return ret; 1600cb517eabSMiao Xie } 1601cb517eabSMiao Xie 1602bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1603bd647ce3SJosef Bacik { 1604bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1605bd647ce3SJosef Bacik struct btrfs_root *root; 1606bd647ce3SJosef Bacik 1607bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1608457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1609457f1864SJosef Bacik 1610bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1611bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1612457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 161371008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1614bd647ce3SJosef Bacik refcount_read(&root->refs)); 1615bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 161600246528SJosef Bacik btrfs_put_root(root); 161700246528SJosef Bacik btrfs_put_root(root); 1618bd647ce3SJosef Bacik } 1619bd647ce3SJosef Bacik #endif 1620bd647ce3SJosef Bacik } 1621bd647ce3SJosef Bacik 16220d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 16230d4b0463SJosef Bacik { 1624141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1625141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 16265deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1627141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1628141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1629141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1630141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 16310d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 16320d4b0463SJosef Bacik kfree(fs_info->delayed_root); 163300246528SJosef Bacik btrfs_put_root(fs_info->extent_root); 163400246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 163500246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 163600246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 163700246528SJosef Bacik btrfs_put_root(fs_info->csum_root); 163800246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 163900246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 164000246528SJosef Bacik btrfs_put_root(fs_info->free_space_root); 164100246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1642aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 1643bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 16443fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 16450d4b0463SJosef Bacik kfree(fs_info->super_copy); 16460d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 16478481dd80SQu Wenruo kfree(fs_info->subpage_info); 16480d4b0463SJosef Bacik kvfree(fs_info); 16490d4b0463SJosef Bacik } 16500d4b0463SJosef Bacik 16510d4b0463SJosef Bacik 16522dfb1e43SQu Wenruo /* 16532dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 16542dfb1e43SQu Wenruo * 16552dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 16562dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 16572dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 16582dfb1e43SQu Wenruo * 16592dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 16602dfb1e43SQu Wenruo * 16612dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 16622dfb1e43SQu Wenruo * same root objectid. 16632dfb1e43SQu Wenruo * 16642dfb1e43SQu Wenruo * @objectid: root id 16652dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 16662dfb1e43SQu Wenruo * pass 0 for new allocation. 16672dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 16682dfb1e43SQu Wenruo * for orphan roots 16692dfb1e43SQu Wenruo */ 16702dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 16712dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 16722dfb1e43SQu Wenruo bool check_ref) 16735eda7b5eSChris Mason { 16745eda7b5eSChris Mason struct btrfs_root *root; 1675381cf658SDavid Sterba struct btrfs_path *path; 16761d4c08e0SDavid Sterba struct btrfs_key key; 16775eda7b5eSChris Mason int ret; 16785eda7b5eSChris Mason 167949d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 168049d11beaSJosef Bacik if (root) 168149d11beaSJosef Bacik return root; 16824df27c4dSYan, Zheng again: 168356e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 168448475471SStefan Behrens if (root) { 16852dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 16862dfb1e43SQu Wenruo ASSERT(!anon_dev); 1687bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 168800246528SJosef Bacik btrfs_put_root(root); 168948475471SStefan Behrens return ERR_PTR(-ENOENT); 1690bc44d7c4SJosef Bacik } 16915eda7b5eSChris Mason return root; 169248475471SStefan Behrens } 16935eda7b5eSChris Mason 169456e9357aSDavid Sterba key.objectid = objectid; 169556e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 169656e9357aSDavid Sterba key.offset = (u64)-1; 169756e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 16985eda7b5eSChris Mason if (IS_ERR(root)) 16995eda7b5eSChris Mason return root; 17003394e160SChris Mason 1701c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1702d68fc57bSYan, Zheng ret = -ENOENT; 1703d68fc57bSYan, Zheng goto fail; 1704d68fc57bSYan, Zheng } 1705d68fc57bSYan, Zheng 17062dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1707cb517eabSMiao Xie if (ret) 1708cb517eabSMiao Xie goto fail; 1709cb517eabSMiao Xie 1710381cf658SDavid Sterba path = btrfs_alloc_path(); 1711381cf658SDavid Sterba if (!path) { 1712381cf658SDavid Sterba ret = -ENOMEM; 1713381cf658SDavid Sterba goto fail; 1714381cf658SDavid Sterba } 17151d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 17161d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 171756e9357aSDavid Sterba key.offset = objectid; 17181d4c08e0SDavid Sterba 17191d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1720381cf658SDavid Sterba btrfs_free_path(path); 1721d68fc57bSYan, Zheng if (ret < 0) 1722d68fc57bSYan, Zheng goto fail; 1723d68fc57bSYan, Zheng if (ret == 0) 172427cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1725d68fc57bSYan, Zheng 1726cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 17270f7d52f4SChris Mason if (ret) { 172800246528SJosef Bacik btrfs_put_root(root); 17294785e24fSJosef Bacik if (ret == -EEXIST) 17304df27c4dSYan, Zheng goto again; 17314df27c4dSYan, Zheng goto fail; 17324df27c4dSYan, Zheng } 1733edbd8d4eSChris Mason return root; 17344df27c4dSYan, Zheng fail: 173533fab972SFilipe Manana /* 173633fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 173733fab972SFilipe Manana * responsability to free it in case we fail. So we have to set our 173833fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 173933fab972SFilipe Manana * and once again by our caller. 174033fab972SFilipe Manana */ 174133fab972SFilipe Manana if (anon_dev) 174233fab972SFilipe Manana root->anon_dev = 0; 17438c38938cSJosef Bacik btrfs_put_root(root); 17444df27c4dSYan, Zheng return ERR_PTR(ret); 1745edbd8d4eSChris Mason } 1746edbd8d4eSChris Mason 17472dfb1e43SQu Wenruo /* 17482dfb1e43SQu Wenruo * Get in-memory reference of a root structure 17492dfb1e43SQu Wenruo * 17502dfb1e43SQu Wenruo * @objectid: tree objectid 17512dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 17522dfb1e43SQu Wenruo * one reference 17532dfb1e43SQu Wenruo */ 17542dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 17552dfb1e43SQu Wenruo u64 objectid, bool check_ref) 17562dfb1e43SQu Wenruo { 17572dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 17582dfb1e43SQu Wenruo } 17592dfb1e43SQu Wenruo 17602dfb1e43SQu Wenruo /* 17612dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 17622dfb1e43SQu Wenruo * the anonymous block device id 17632dfb1e43SQu Wenruo * 17642dfb1e43SQu Wenruo * @objectid: tree objectid 17652dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 17662dfb1e43SQu Wenruo * parameter value 17672dfb1e43SQu Wenruo */ 17682dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 17692dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 17702dfb1e43SQu Wenruo { 17712dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 17722dfb1e43SQu Wenruo } 17732dfb1e43SQu Wenruo 17748b712842SChris Mason /* 177549d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 177649d11beaSJosef Bacik * @fs_info: the fs_info 177749d11beaSJosef Bacik * @objectid: the objectid we need to lookup 177849d11beaSJosef Bacik * 177949d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 178049d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 178149d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 178249d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 178349d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 178449d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 178549d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 178649d11beaSJosef Bacik * backref code is finished using the root. 178749d11beaSJosef Bacik */ 178849d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 178949d11beaSJosef Bacik struct btrfs_path *path, 179049d11beaSJosef Bacik u64 objectid) 179149d11beaSJosef Bacik { 179249d11beaSJosef Bacik struct btrfs_root *root; 179349d11beaSJosef Bacik struct btrfs_key key; 179449d11beaSJosef Bacik 179549d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 179649d11beaSJosef Bacik 179749d11beaSJosef Bacik /* 179849d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 179949d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 180049d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 180149d11beaSJosef Bacik * != NULL just return it. 180249d11beaSJosef Bacik */ 180349d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 180449d11beaSJosef Bacik if (root) 180549d11beaSJosef Bacik return root; 180649d11beaSJosef Bacik 180749d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 180849d11beaSJosef Bacik if (root) 180949d11beaSJosef Bacik return root; 181049d11beaSJosef Bacik 181149d11beaSJosef Bacik key.objectid = objectid; 181249d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 181349d11beaSJosef Bacik key.offset = (u64)-1; 181449d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 181549d11beaSJosef Bacik btrfs_release_path(path); 181649d11beaSJosef Bacik 181749d11beaSJosef Bacik return root; 181849d11beaSJosef Bacik } 181949d11beaSJosef Bacik 182049d11beaSJosef Bacik /* 18218b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 18228b712842SChris Mason * functions. This is where read checksum verification actually happens 18238b712842SChris Mason */ 18248b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1825ce9adaa5SChris Mason { 1826ce9adaa5SChris Mason struct bio *bio; 182797eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1828ce9adaa5SChris Mason 182997eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1830ce9adaa5SChris Mason bio = end_io_wq->bio; 18318b712842SChris Mason 18324e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1833ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1834ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 18354246a0b6SChristoph Hellwig bio_endio(bio); 18369be490f1SOmar Sandoval kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 1837ce9adaa5SChris Mason } 183844b8bd7eSChris Mason 1839a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1840a74a4b97SChris Mason { 1841a74a4b97SChris Mason struct btrfs_root *root = arg; 18420b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1843d0278245SMiao Xie int again; 1844a74a4b97SChris Mason 1845d6fd0ae2SOmar Sandoval while (1) { 1846d0278245SMiao Xie again = 0; 18479d1a2a3aSDavid Sterba 1848fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1849fd340d0fSJosef Bacik 1850d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 18512ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1852d0278245SMiao Xie goto sleep; 1853d0278245SMiao Xie 185490c711abSZygo Blaxell /* 185590c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 185690c711abSZygo Blaxell * before we have finished mounting the filesystem. 185790c711abSZygo Blaxell */ 18580b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 185990c711abSZygo Blaxell goto sleep; 186090c711abSZygo Blaxell 18610b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1862d0278245SMiao Xie goto sleep; 1863d0278245SMiao Xie 1864dc7f370cSMiao Xie /* 1865dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1866dc7f370cSMiao Xie * during the above check and trylock. 1867dc7f370cSMiao Xie */ 18682ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 18690b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1870dc7f370cSMiao Xie goto sleep; 1871dc7f370cSMiao Xie } 1872dc7f370cSMiao Xie 18732ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1874c2d6cb16SFilipe Manana 18759d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 18760b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1877a74a4b97SChris Mason 1878d0278245SMiao Xie /* 187905323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 188005323cd1SMiao Xie * needn't do anything special here. 1881d0278245SMiao Xie */ 18820b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 188367c5e7d4SFilipe Manana 188467c5e7d4SFilipe Manana /* 1885f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 188667c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 188767c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1888f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 188967c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 189067c5e7d4SFilipe Manana * unused block groups. 189167c5e7d4SFilipe Manana */ 18920b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 189318bb8bbfSJohannes Thumshirn 189418bb8bbfSJohannes Thumshirn /* 189518bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 189618bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 189718bb8bbfSJohannes Thumshirn * space. 189818bb8bbfSJohannes Thumshirn */ 189918bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1900d0278245SMiao Xie sleep: 1901a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1902d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1903d6fd0ae2SOmar Sandoval kthread_parkme(); 1904d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1905d6fd0ae2SOmar Sandoval return 0; 1906838fe188SJiri Kosina if (!again) { 1907a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1908a74a4b97SChris Mason schedule(); 1909a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1910a74a4b97SChris Mason } 1911da288d28SFilipe Manana } 1912a74a4b97SChris Mason } 1913a74a4b97SChris Mason 1914a74a4b97SChris Mason static int transaction_kthread(void *arg) 1915a74a4b97SChris Mason { 1916a74a4b97SChris Mason struct btrfs_root *root = arg; 19170b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1918a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1919a74a4b97SChris Mason struct btrfs_transaction *cur; 19208929ecfaSYan, Zheng u64 transid; 1921643900beSNikolay Borisov time64_t delta; 1922a74a4b97SChris Mason unsigned long delay; 1923914b2007SJan Kara bool cannot_commit; 1924a74a4b97SChris Mason 1925a74a4b97SChris Mason do { 1926914b2007SJan Kara cannot_commit = false; 1927ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 19280b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1929a74a4b97SChris Mason 19300b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 19310b246afaSJeff Mahoney cur = fs_info->running_transaction; 1932a74a4b97SChris Mason if (!cur) { 19330b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1934a74a4b97SChris Mason goto sleep; 1935a74a4b97SChris Mason } 193631153d81SYan Zheng 1937643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 19383296bf56SQu Wenruo if (cur->state < TRANS_STATE_COMMIT_START && 1939643900beSNikolay Borisov delta < fs_info->commit_interval) { 19400b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1941fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1942fb8a7e94SNikolay Borisov delay = min(delay, 1943fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1944a74a4b97SChris Mason goto sleep; 1945a74a4b97SChris Mason } 19468929ecfaSYan, Zheng transid = cur->transid; 19470b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 194856bec294SChris Mason 194979787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1950354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1951914b2007SJan Kara if (IS_ERR(trans)) { 1952354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1953914b2007SJan Kara cannot_commit = true; 195479787eaaSJeff Mahoney goto sleep; 1955914b2007SJan Kara } 19568929ecfaSYan, Zheng if (transid == trans->transid) { 19573a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 19588929ecfaSYan, Zheng } else { 19593a45bb20SJeff Mahoney btrfs_end_transaction(trans); 19608929ecfaSYan, Zheng } 1961a74a4b97SChris Mason sleep: 19620b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 19630b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1964a74a4b97SChris Mason 196584961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 19662ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 19678929ecfaSYan, Zheng if (!kthread_should_stop() && 19680b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1969914b2007SJan Kara cannot_commit)) 1970bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1971a74a4b97SChris Mason } while (!kthread_should_stop()); 1972a74a4b97SChris Mason return 0; 1973a74a4b97SChris Mason } 1974a74a4b97SChris Mason 1975af31f5e5SChris Mason /* 197601f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 197701f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 197801f0f9daSNikolay Borisov * anything. 1979af31f5e5SChris Mason * 1980af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1981af31f5e5SChris Mason * generation of the latest root in the array with the generation 1982af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1983af31f5e5SChris Mason */ 198401f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1985af31f5e5SChris Mason { 198601f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1987af31f5e5SChris Mason u64 cur; 1988af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1989af31f5e5SChris Mason int i; 1990af31f5e5SChris Mason 1991af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1992af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1993af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1994af31f5e5SChris Mason if (cur == newest_gen) 199501f0f9daSNikolay Borisov return i; 1996af31f5e5SChris Mason } 1997af31f5e5SChris Mason 199801f0f9daSNikolay Borisov return -EINVAL; 1999af31f5e5SChris Mason } 2000af31f5e5SChris Mason 2001af31f5e5SChris Mason /* 2002af31f5e5SChris Mason * copy all the root pointers into the super backup array. 2003af31f5e5SChris Mason * this will bump the backup pointer by one when it is 2004af31f5e5SChris Mason * done 2005af31f5e5SChris Mason */ 2006af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 2007af31f5e5SChris Mason { 20086ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 2009af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2010af31f5e5SChris Mason 2011af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 2012af31f5e5SChris Mason 2013af31f5e5SChris Mason /* 2014af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 2015af31f5e5SChris Mason * regardless of which ones we use today 2016af31f5e5SChris Mason */ 2017af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 2018af31f5e5SChris Mason 2019af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 2020af31f5e5SChris Mason 2021af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 2022af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 2023af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 2024af31f5e5SChris Mason 2025af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 2026af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 2027af31f5e5SChris Mason 2028af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 2029af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 2030af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 2031af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 2032af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 2033af31f5e5SChris Mason 2034af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 2035af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 2036af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 2037af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 2038af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 2039af31f5e5SChris Mason 20407c7e82a7SChris Mason /* 20417c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 20427c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 20437c7e82a7SChris Mason */ 20447c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 20457c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 20467c7e82a7SChris Mason info->fs_root->node->start); 2047af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2048af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2049af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2050af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 20517c7e82a7SChris Mason } 2052af31f5e5SChris Mason 2053af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2054af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2055af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2056af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2057af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2058af31f5e5SChris Mason 2059af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 2060af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 2061af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 2062af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 2063af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 2064af31f5e5SChris Mason 2065af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2066af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2067af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2068af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2069af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2070af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2071af31f5e5SChris Mason 2072af31f5e5SChris Mason /* 2073af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2074af31f5e5SChris Mason * for the next commit 2075af31f5e5SChris Mason */ 2076af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2077af31f5e5SChris Mason &info->super_for_commit->super_roots, 2078af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2079af31f5e5SChris Mason } 2080af31f5e5SChris Mason 2081af31f5e5SChris Mason /* 2082bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 2083bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 2084bd2336b2SNikolay Borisov * 2085bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 2086bd2336b2SNikolay Borisov * priority - priority of backup root required 2087bd2336b2SNikolay Borisov * 2088bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 2089bd2336b2SNikolay Borisov */ 2090bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 2091bd2336b2SNikolay Borisov { 2092bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2093bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 2094bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 2095bd2336b2SNikolay Borisov 2096bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 2097bd2336b2SNikolay Borisov if (priority == 0) 2098bd2336b2SNikolay Borisov return backup_index; 2099bd2336b2SNikolay Borisov 2100bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 2101bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 2102bd2336b2SNikolay Borisov } else { 2103bd2336b2SNikolay Borisov return -EINVAL; 2104bd2336b2SNikolay Borisov } 2105bd2336b2SNikolay Borisov 2106bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 2107bd2336b2SNikolay Borisov 2108bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 2109bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 2110bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2111bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 2112bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 2113bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2114bd2336b2SNikolay Borisov 2115bd2336b2SNikolay Borisov /* 2116bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 2117bd2336b2SNikolay Borisov * need a fsck 2118bd2336b2SNikolay Borisov */ 2119bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2120bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2121bd2336b2SNikolay Borisov 2122bd2336b2SNikolay Borisov return backup_index; 2123bd2336b2SNikolay Borisov } 2124bd2336b2SNikolay Borisov 21257abadb64SLiu Bo /* helper to cleanup workers */ 21267abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 21277abadb64SLiu Bo { 2128dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2129afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 21305cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2131fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2132fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 2133d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2134fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2135fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 21365b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2137e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2138736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2139a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2140fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2141b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 2142b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 2143a9b9477dSFilipe Manana /* 2144a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2145a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2146a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2147a9b9477dSFilipe Manana */ 2148a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2149a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 21507abadb64SLiu Bo } 21517abadb64SLiu Bo 21522e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 21532e9f5954SRashika { 21542e9f5954SRashika if (root) { 21552e9f5954SRashika free_extent_buffer(root->node); 21562e9f5954SRashika free_extent_buffer(root->commit_root); 21572e9f5954SRashika root->node = NULL; 21582e9f5954SRashika root->commit_root = NULL; 21592e9f5954SRashika } 21602e9f5954SRashika } 21612e9f5954SRashika 2162af31f5e5SChris Mason /* helper to cleanup tree roots */ 21634273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2164af31f5e5SChris Mason { 21652e9f5954SRashika free_root_extent_buffers(info->tree_root); 2166655b09feSJosef Bacik 21672e9f5954SRashika free_root_extent_buffers(info->dev_root); 21682e9f5954SRashika free_root_extent_buffers(info->extent_root); 21692e9f5954SRashika free_root_extent_buffers(info->csum_root); 21702e9f5954SRashika free_root_extent_buffers(info->quota_root); 21712e9f5954SRashika free_root_extent_buffers(info->uuid_root); 21728c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2173aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 21744273eaffSAnand Jain if (free_chunk_root) 21752e9f5954SRashika free_root_extent_buffers(info->chunk_root); 217670f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2177af31f5e5SChris Mason } 2178af31f5e5SChris Mason 21798c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 21808c38938cSJosef Bacik { 21818c38938cSJosef Bacik if (!root) 21828c38938cSJosef Bacik return; 21838c38938cSJosef Bacik 21848c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 21858c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 21861dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 21878c38938cSJosef Bacik if (root->anon_dev) 21888c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 21898c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2190923eb523SJohannes Thumshirn free_root_extent_buffers(root); 21918c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 21928c38938cSJosef Bacik spin_lock(&root->fs_info->fs_roots_radix_lock); 21938c38938cSJosef Bacik list_del_init(&root->leak_list); 21948c38938cSJosef Bacik spin_unlock(&root->fs_info->fs_roots_radix_lock); 21958c38938cSJosef Bacik #endif 21968c38938cSJosef Bacik kfree(root); 21978c38938cSJosef Bacik } 21988c38938cSJosef Bacik } 21998c38938cSJosef Bacik 2200faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2201171f6537SJosef Bacik { 2202171f6537SJosef Bacik int ret; 2203171f6537SJosef Bacik struct btrfs_root *gang[8]; 2204171f6537SJosef Bacik int i; 2205171f6537SJosef Bacik 2206171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2207171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2208171f6537SJosef Bacik struct btrfs_root, root_list); 2209171f6537SJosef Bacik list_del(&gang[0]->root_list); 2210171f6537SJosef Bacik 22118c38938cSJosef Bacik if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2212cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 221300246528SJosef Bacik btrfs_put_root(gang[0]); 2214171f6537SJosef Bacik } 2215171f6537SJosef Bacik 2216171f6537SJosef Bacik while (1) { 2217171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2218171f6537SJosef Bacik (void **)gang, 0, 2219171f6537SJosef Bacik ARRAY_SIZE(gang)); 2220171f6537SJosef Bacik if (!ret) 2221171f6537SJosef Bacik break; 2222171f6537SJosef Bacik for (i = 0; i < ret; i++) 2223cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2224171f6537SJosef Bacik } 2225171f6537SJosef Bacik } 2226af31f5e5SChris Mason 2227638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2228638aa7edSEric Sandeen { 2229638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2230638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2231638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2232638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2233638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2234638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2235ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2236638aa7edSEric Sandeen } 2237638aa7edSEric Sandeen 2238779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2239779a65a4SEric Sandeen { 2240779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2241779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2242779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2243779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2244779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2245779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2246907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 2247779a65a4SEric Sandeen } 2248779a65a4SEric Sandeen 22496bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2250f37938e0SEric Sandeen { 22512ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 22522ff7e61eSJeff Mahoney 22532ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 22542ff7e61eSJeff Mahoney set_nlink(inode, 1); 2255f37938e0SEric Sandeen /* 2256f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2257f37938e0SEric Sandeen * the real end of the address space is determined by all of 2258f37938e0SEric Sandeen * the devices in the system 2259f37938e0SEric Sandeen */ 22602ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 22612ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2262f37938e0SEric Sandeen 22632ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 226443eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 22652c53a14dSQu Wenruo IO_TREE_BTREE_INODE_IO, inode); 22667b439738SDavid Sterba BTRFS_I(inode)->io_tree.track_uptodate = false; 22672ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2268f37938e0SEric Sandeen 22695c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 22702ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 22712ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 22722ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2273f37938e0SEric Sandeen } 2274f37938e0SEric Sandeen 2275ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2276ad618368SEric Sandeen { 2277ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2278129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 22797f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2280ad618368SEric Sandeen } 2281ad618368SEric Sandeen 2282f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2283f9e92e40SEric Sandeen { 2284f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2285f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2286f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2287f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2288f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2289f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2290d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2291f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2292f9e92e40SEric Sandeen } 2293f9e92e40SEric Sandeen 2294d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 22952a458198SEric Sandeen { 2296f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 22976f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 22982a458198SEric Sandeen 22992a458198SEric Sandeen fs_info->workers = 2300cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2301cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 23022a458198SEric Sandeen 23032a458198SEric Sandeen fs_info->delalloc_workers = 2304cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2305cb001095SJeff Mahoney flags, max_active, 2); 23062a458198SEric Sandeen 23072a458198SEric Sandeen fs_info->flush_workers = 2308cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2309cb001095SJeff Mahoney flags, max_active, 0); 23102a458198SEric Sandeen 23112a458198SEric Sandeen fs_info->caching_workers = 2312cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 23132a458198SEric Sandeen 23142a458198SEric Sandeen fs_info->fixup_workers = 2315cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 23162a458198SEric Sandeen 23172a458198SEric Sandeen /* 23182a458198SEric Sandeen * endios are largely parallel and should have a very 23192a458198SEric Sandeen * low idle thresh 23202a458198SEric Sandeen */ 23212a458198SEric Sandeen fs_info->endio_workers = 2322cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 23232a458198SEric Sandeen fs_info->endio_meta_workers = 2324cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2325cb001095SJeff Mahoney max_active, 4); 23262a458198SEric Sandeen fs_info->endio_meta_write_workers = 2327cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2328cb001095SJeff Mahoney max_active, 2); 23292a458198SEric Sandeen fs_info->endio_raid56_workers = 2330cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2331cb001095SJeff Mahoney max_active, 4); 23322a458198SEric Sandeen fs_info->rmw_workers = 2333cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 23342a458198SEric Sandeen fs_info->endio_write_workers = 2335cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2336cb001095SJeff Mahoney max_active, 2); 23372a458198SEric Sandeen fs_info->endio_freespace_worker = 2338cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2339cb001095SJeff Mahoney max_active, 0); 23402a458198SEric Sandeen fs_info->delayed_workers = 2341cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2342cb001095SJeff Mahoney max_active, 0); 23432a458198SEric Sandeen fs_info->readahead_workers = 2344cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2345cb001095SJeff Mahoney max_active, 2); 23462a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2347cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2348b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2349b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 23502a458198SEric Sandeen 23512a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 2352ba8a9d07SChris Mason fs_info->flush_workers && 23532a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 23542a458198SEric Sandeen fs_info->endio_meta_write_workers && 23552a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 23562a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 23572a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 23582a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 2359b0643e59SDennis Zhou fs_info->qgroup_rescan_workers && 2360b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 23612a458198SEric Sandeen return -ENOMEM; 23622a458198SEric Sandeen } 23632a458198SEric Sandeen 23642a458198SEric Sandeen return 0; 23652a458198SEric Sandeen } 23662a458198SEric Sandeen 23676d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 23686d97c6e3SJohannes Thumshirn { 23696d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2370b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 23716d97c6e3SJohannes Thumshirn 2372b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 23736d97c6e3SJohannes Thumshirn 23746d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 23756d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2376b4e967beSDavid Sterba csum_driver); 23776d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 23786d97c6e3SJohannes Thumshirn } 23796d97c6e3SJohannes Thumshirn 23806d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 23816d97c6e3SJohannes Thumshirn 23826d97c6e3SJohannes Thumshirn return 0; 23836d97c6e3SJohannes Thumshirn } 23846d97c6e3SJohannes Thumshirn 238563443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 238663443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 238763443bf5SEric Sandeen { 238863443bf5SEric Sandeen int ret; 238963443bf5SEric Sandeen struct btrfs_root *log_tree_root; 239063443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 239163443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2392581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 239363443bf5SEric Sandeen 239463443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2395f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 239663443bf5SEric Sandeen return -EIO; 239763443bf5SEric Sandeen } 239863443bf5SEric Sandeen 239996dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 240096dfcb46SJosef Bacik GFP_KERNEL); 240163443bf5SEric Sandeen if (!log_tree_root) 240263443bf5SEric Sandeen return -ENOMEM; 240363443bf5SEric Sandeen 24042ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 24051b7ec85eSJosef Bacik BTRFS_TREE_LOG_OBJECTID, 24061b7ec85eSJosef Bacik fs_info->generation + 1, level, 24071b7ec85eSJosef Bacik NULL); 240864c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2409f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 24100eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 24118c38938cSJosef Bacik log_tree_root->node = NULL; 241200246528SJosef Bacik btrfs_put_root(log_tree_root); 24130eeff236SLiu Bo return ret; 241464c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2415f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 241600246528SJosef Bacik btrfs_put_root(log_tree_root); 241763443bf5SEric Sandeen return -EIO; 241863443bf5SEric Sandeen } 241963443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 242063443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 242163443bf5SEric Sandeen if (ret) { 24220b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 242363443bf5SEric Sandeen "Failed to recover log tree"); 242400246528SJosef Bacik btrfs_put_root(log_tree_root); 242563443bf5SEric Sandeen return ret; 242663443bf5SEric Sandeen } 242763443bf5SEric Sandeen 2428bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 24296bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 243063443bf5SEric Sandeen if (ret) 243163443bf5SEric Sandeen return ret; 243263443bf5SEric Sandeen } 243363443bf5SEric Sandeen 243463443bf5SEric Sandeen return 0; 243563443bf5SEric Sandeen } 243663443bf5SEric Sandeen 24376bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 24384bbcaa64SEric Sandeen { 24396bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2440a4f3d2c4SDavid Sterba struct btrfs_root *root; 24414bbcaa64SEric Sandeen struct btrfs_key location; 24424bbcaa64SEric Sandeen int ret; 24434bbcaa64SEric Sandeen 24446bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 24456bccf3abSJeff Mahoney 24464bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 24474bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 24484bbcaa64SEric Sandeen location.offset = 0; 24494bbcaa64SEric Sandeen 2450a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2451f50f4353SLiu Bo if (IS_ERR(root)) { 245242437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2453f50f4353SLiu Bo ret = PTR_ERR(root); 2454f50f4353SLiu Bo goto out; 2455f50f4353SLiu Bo } 245642437a63SJosef Bacik } else { 2457a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2458a4f3d2c4SDavid Sterba fs_info->extent_root = root; 245942437a63SJosef Bacik } 24604bbcaa64SEric Sandeen 24614bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2462a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2463f50f4353SLiu Bo if (IS_ERR(root)) { 246442437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2465f50f4353SLiu Bo ret = PTR_ERR(root); 2466f50f4353SLiu Bo goto out; 2467f50f4353SLiu Bo } 246842437a63SJosef Bacik } else { 2469a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2470a4f3d2c4SDavid Sterba fs_info->dev_root = root; 247142437a63SJosef Bacik } 2472820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2473820a49daSJosef Bacik btrfs_init_devices_late(fs_info); 24744bbcaa64SEric Sandeen 2475882dbe0cSJosef Bacik /* If IGNOREDATACSUMS is set don't bother reading the csum root. */ 2476882dbe0cSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 24774bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2478a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2479f50f4353SLiu Bo if (IS_ERR(root)) { 248042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2481f50f4353SLiu Bo ret = PTR_ERR(root); 2482f50f4353SLiu Bo goto out; 2483f50f4353SLiu Bo } 248442437a63SJosef Bacik } else { 2485a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2486a4f3d2c4SDavid Sterba fs_info->csum_root = root; 248742437a63SJosef Bacik } 2488882dbe0cSJosef Bacik } 24894bbcaa64SEric Sandeen 2490aeb935a4SQu Wenruo /* 2491aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2492aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2493aeb935a4SQu Wenruo */ 249456e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 249556e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2496aeb935a4SQu Wenruo if (IS_ERR(root)) { 249742437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2498aeb935a4SQu Wenruo ret = PTR_ERR(root); 2499aeb935a4SQu Wenruo goto out; 2500aeb935a4SQu Wenruo } 250142437a63SJosef Bacik } else { 2502aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2503aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 250442437a63SJosef Bacik } 2505aeb935a4SQu Wenruo 25064bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2507a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2508a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2509a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2510afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2511a4f3d2c4SDavid Sterba fs_info->quota_root = root; 25124bbcaa64SEric Sandeen } 25134bbcaa64SEric Sandeen 25144bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2515a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2516a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 251742437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2518a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 25194bbcaa64SEric Sandeen if (ret != -ENOENT) 2520f50f4353SLiu Bo goto out; 252142437a63SJosef Bacik } 25224bbcaa64SEric Sandeen } else { 2523a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2524a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 25254bbcaa64SEric Sandeen } 25264bbcaa64SEric Sandeen 252770f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 252870f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 252970f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 2530f50f4353SLiu Bo if (IS_ERR(root)) { 253142437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2532f50f4353SLiu Bo ret = PTR_ERR(root); 2533f50f4353SLiu Bo goto out; 2534f50f4353SLiu Bo } 253542437a63SJosef Bacik } else { 253670f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 253770f6d82eSOmar Sandoval fs_info->free_space_root = root; 253870f6d82eSOmar Sandoval } 253942437a63SJosef Bacik } 254070f6d82eSOmar Sandoval 25414bbcaa64SEric Sandeen return 0; 2542f50f4353SLiu Bo out: 2543f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2544f50f4353SLiu Bo location.objectid, ret); 2545f50f4353SLiu Bo return ret; 25464bbcaa64SEric Sandeen } 25474bbcaa64SEric Sandeen 2548069ec957SQu Wenruo /* 2549069ec957SQu Wenruo * Real super block validation 2550069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2551069ec957SQu Wenruo * 2552069ec957SQu Wenruo * @sb: super block to check 2553069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2554069ec957SQu Wenruo * 0 the primary (1st) sb 2555069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2556069ec957SQu Wenruo * -1 skip bytenr check 2557069ec957SQu Wenruo */ 2558069ec957SQu Wenruo static int validate_super(struct btrfs_fs_info *fs_info, 2559069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 256021a852b0SQu Wenruo { 256121a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 256221a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 256321a852b0SQu Wenruo int ret = 0; 256421a852b0SQu Wenruo 256521a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 256621a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 256721a852b0SQu Wenruo ret = -EINVAL; 256821a852b0SQu Wenruo } 256921a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 257021a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 257121a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 257221a852b0SQu Wenruo ret = -EINVAL; 257321a852b0SQu Wenruo } 257421a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 257521a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 257621a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 257721a852b0SQu Wenruo ret = -EINVAL; 257821a852b0SQu Wenruo } 257921a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 258021a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 258121a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 258221a852b0SQu Wenruo ret = -EINVAL; 258321a852b0SQu Wenruo } 258421a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 258521a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 258621a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 258721a852b0SQu Wenruo ret = -EINVAL; 258821a852b0SQu Wenruo } 258921a852b0SQu Wenruo 259021a852b0SQu Wenruo /* 259121a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 259221a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 259321a852b0SQu Wenruo */ 259421a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 259521a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 259621a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 259721a852b0SQu Wenruo ret = -EINVAL; 259821a852b0SQu Wenruo } 25990bb3eb3eSQu Wenruo 26000bb3eb3eSQu Wenruo /* 26010bb3eb3eSQu Wenruo * For 4K page size, we only support 4K sector size. 260250780d9bSAnand Jain * For 64K page size, we support 64K and 4K sector sizes. 26030bb3eb3eSQu Wenruo */ 26040bb3eb3eSQu Wenruo if ((PAGE_SIZE == SZ_4K && sectorsize != PAGE_SIZE) || 26050bb3eb3eSQu Wenruo (PAGE_SIZE == SZ_64K && (sectorsize != SZ_4K && 26060bb3eb3eSQu Wenruo sectorsize != SZ_64K))) { 260721a852b0SQu Wenruo btrfs_err(fs_info, 26080bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 260921a852b0SQu Wenruo sectorsize, PAGE_SIZE); 261021a852b0SQu Wenruo ret = -EINVAL; 261121a852b0SQu Wenruo } 26120bb3eb3eSQu Wenruo 261321a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 261421a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 261521a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 261621a852b0SQu Wenruo ret = -EINVAL; 261721a852b0SQu Wenruo } 261821a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 261921a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 262021a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 262121a852b0SQu Wenruo ret = -EINVAL; 262221a852b0SQu Wenruo } 262321a852b0SQu Wenruo 262421a852b0SQu Wenruo /* Root alignment check */ 262521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 262621a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 262721a852b0SQu Wenruo btrfs_super_root(sb)); 262821a852b0SQu Wenruo ret = -EINVAL; 262921a852b0SQu Wenruo } 263021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 263121a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 263221a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 263321a852b0SQu Wenruo ret = -EINVAL; 263421a852b0SQu Wenruo } 263521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 263621a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 263721a852b0SQu Wenruo btrfs_super_log_root(sb)); 263821a852b0SQu Wenruo ret = -EINVAL; 263921a852b0SQu Wenruo } 264021a852b0SQu Wenruo 2641aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2642aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2643aefd7f70SNikolay Borisov btrfs_err(fs_info, 2644aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2645aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2646aefd7f70SNikolay Borisov ret = -EINVAL; 2647aefd7f70SNikolay Borisov } 2648aefd7f70SNikolay Borisov 2649aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2650aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2651aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2652aefd7f70SNikolay Borisov btrfs_err(fs_info, 2653aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2654aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2655aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2656aefd7f70SNikolay Borisov ret = -EINVAL; 2657aefd7f70SNikolay Borisov } 2658aefd7f70SNikolay Borisov 2659de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 26607239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 266121a852b0SQu Wenruo btrfs_err(fs_info, 26627239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2663de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 266421a852b0SQu Wenruo ret = -EINVAL; 266521a852b0SQu Wenruo } 266621a852b0SQu Wenruo 266721a852b0SQu Wenruo /* 266821a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 266921a852b0SQu Wenruo * done later 267021a852b0SQu Wenruo */ 267121a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 267221a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 267321a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 267421a852b0SQu Wenruo ret = -EINVAL; 267521a852b0SQu Wenruo } 267621a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 267721a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 267821a852b0SQu Wenruo btrfs_super_stripesize(sb)); 267921a852b0SQu Wenruo ret = -EINVAL; 268021a852b0SQu Wenruo } 268121a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 268221a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 268321a852b0SQu Wenruo btrfs_super_num_devices(sb)); 268421a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 268521a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 268621a852b0SQu Wenruo ret = -EINVAL; 268721a852b0SQu Wenruo } 268821a852b0SQu Wenruo 2689069ec957SQu Wenruo if (mirror_num >= 0 && 2690069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 269121a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 269221a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 269321a852b0SQu Wenruo ret = -EINVAL; 269421a852b0SQu Wenruo } 269521a852b0SQu Wenruo 269621a852b0SQu Wenruo /* 269721a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 269821a852b0SQu Wenruo * and one chunk 269921a852b0SQu Wenruo */ 270021a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 270121a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 270221a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 270321a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 270421a852b0SQu Wenruo ret = -EINVAL; 270521a852b0SQu Wenruo } 270621a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 270721a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 270821a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 270921a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 271021a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 271121a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 271221a852b0SQu Wenruo ret = -EINVAL; 271321a852b0SQu Wenruo } 271421a852b0SQu Wenruo 271521a852b0SQu Wenruo /* 271621a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 271721a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 271821a852b0SQu Wenruo */ 271921a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 272021a852b0SQu Wenruo btrfs_warn(fs_info, 272121a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 272221a852b0SQu Wenruo btrfs_super_generation(sb), 272321a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 272421a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 272521a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 272621a852b0SQu Wenruo btrfs_warn(fs_info, 272721a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 272821a852b0SQu Wenruo btrfs_super_generation(sb), 272921a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 273021a852b0SQu Wenruo 273121a852b0SQu Wenruo return ret; 273221a852b0SQu Wenruo } 273321a852b0SQu Wenruo 2734069ec957SQu Wenruo /* 2735069ec957SQu Wenruo * Validation of super block at mount time. 2736069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2737069ec957SQu Wenruo * flags will be skipped. 2738069ec957SQu Wenruo */ 2739069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2740069ec957SQu Wenruo { 2741069ec957SQu Wenruo return validate_super(fs_info, fs_info->super_copy, 0); 2742069ec957SQu Wenruo } 2743069ec957SQu Wenruo 274475cb857dSQu Wenruo /* 274575cb857dSQu Wenruo * Validation of super block at write time. 274675cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 274775cb857dSQu Wenruo * overwritten soon. 274875cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 274975cb857dSQu Wenruo */ 275075cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 275175cb857dSQu Wenruo struct btrfs_super_block *sb) 275275cb857dSQu Wenruo { 275375cb857dSQu Wenruo int ret; 275475cb857dSQu Wenruo 275575cb857dSQu Wenruo ret = validate_super(fs_info, sb, -1); 275675cb857dSQu Wenruo if (ret < 0) 275775cb857dSQu Wenruo goto out; 2758e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 275975cb857dSQu Wenruo ret = -EUCLEAN; 276075cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 276175cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 276275cb857dSQu Wenruo goto out; 276375cb857dSQu Wenruo } 276475cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 276575cb857dSQu Wenruo ret = -EUCLEAN; 276675cb857dSQu Wenruo btrfs_err(fs_info, 276775cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 276875cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 276975cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 277075cb857dSQu Wenruo goto out; 277175cb857dSQu Wenruo } 277275cb857dSQu Wenruo out: 277375cb857dSQu Wenruo if (ret < 0) 277475cb857dSQu Wenruo btrfs_err(fs_info, 277575cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 277675cb857dSQu Wenruo return ret; 277775cb857dSQu Wenruo } 277875cb857dSQu Wenruo 27796ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2780b8522a1eSNikolay Borisov { 27816ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2782b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2783b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2784b8522a1eSNikolay Borisov bool handle_error = false; 2785b8522a1eSNikolay Borisov int ret = 0; 2786b8522a1eSNikolay Borisov int i; 2787b8522a1eSNikolay Borisov 2788b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2789b8522a1eSNikolay Borisov u64 generation; 2790b8522a1eSNikolay Borisov int level; 2791b8522a1eSNikolay Borisov 2792b8522a1eSNikolay Borisov if (handle_error) { 2793b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2794b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2795b8522a1eSNikolay Borisov tree_root->node = NULL; 2796b8522a1eSNikolay Borisov 2797b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2798b8522a1eSNikolay Borisov break; 2799b8522a1eSNikolay Borisov 2800b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2801b8522a1eSNikolay Borisov 2802b8522a1eSNikolay Borisov /* 2803b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2804b8522a1eSNikolay Borisov * valid 2805b8522a1eSNikolay Borisov */ 2806b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2807b8522a1eSNikolay Borisov 2808b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2809b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2810b8522a1eSNikolay Borisov 2811b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 28126ef108ddSNikolay Borisov backup_index = ret; 2813b8522a1eSNikolay Borisov if (ret < 0) 2814b8522a1eSNikolay Borisov return ret; 2815b8522a1eSNikolay Borisov } 2816b8522a1eSNikolay Borisov generation = btrfs_super_generation(sb); 2817b8522a1eSNikolay Borisov level = btrfs_super_root_level(sb); 2818b8522a1eSNikolay Borisov tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), 28191b7ec85eSJosef Bacik BTRFS_ROOT_TREE_OBJECTID, 2820b8522a1eSNikolay Borisov generation, level, NULL); 28210465337cSJosef Bacik if (IS_ERR(tree_root->node)) { 2822217f5004SNikolay Borisov handle_error = true; 2823b8522a1eSNikolay Borisov ret = PTR_ERR(tree_root->node); 28240465337cSJosef Bacik tree_root->node = NULL; 2825217f5004SNikolay Borisov btrfs_warn(fs_info, "couldn't read tree root"); 2826217f5004SNikolay Borisov continue; 2827b8522a1eSNikolay Borisov 2828217f5004SNikolay Borisov } else if (!extent_buffer_uptodate(tree_root->node)) { 2829217f5004SNikolay Borisov handle_error = true; 2830217f5004SNikolay Borisov ret = -EIO; 2831217f5004SNikolay Borisov btrfs_warn(fs_info, "error while reading tree root"); 2832b8522a1eSNikolay Borisov continue; 2833b8522a1eSNikolay Borisov } 2834b8522a1eSNikolay Borisov 2835b8522a1eSNikolay Borisov btrfs_set_root_node(&tree_root->root_item, tree_root->node); 2836b8522a1eSNikolay Borisov tree_root->commit_root = btrfs_root_node(tree_root); 2837b8522a1eSNikolay Borisov btrfs_set_root_refs(&tree_root->root_item, 1); 2838b8522a1eSNikolay Borisov 2839336a0d8dSNikolay Borisov /* 2840336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2841336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2842336a0d8dSNikolay Borisov */ 2843453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2844b8522a1eSNikolay Borisov if (ret < 0) { 2845b8522a1eSNikolay Borisov handle_error = true; 2846b8522a1eSNikolay Borisov continue; 2847b8522a1eSNikolay Borisov } 2848b8522a1eSNikolay Borisov 28496b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2850b8522a1eSNikolay Borisov 2851b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2852b8522a1eSNikolay Borisov if (ret < 0) { 2853b8522a1eSNikolay Borisov handle_error = true; 2854b8522a1eSNikolay Borisov continue; 2855b8522a1eSNikolay Borisov } 2856b8522a1eSNikolay Borisov 2857b8522a1eSNikolay Borisov /* All successful */ 2858b8522a1eSNikolay Borisov fs_info->generation = generation; 2859b8522a1eSNikolay Borisov fs_info->last_trans_committed = generation; 28606ef108ddSNikolay Borisov 28616ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 28626ef108ddSNikolay Borisov if (backup_index < 0) { 28636ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 28646ef108ddSNikolay Borisov } else { 28656ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 28666ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 28676ef108ddSNikolay Borisov } 2868b8522a1eSNikolay Borisov break; 2869b8522a1eSNikolay Borisov } 2870b8522a1eSNikolay Borisov 2871b8522a1eSNikolay Borisov return ret; 2872b8522a1eSNikolay Borisov } 2873b8522a1eSNikolay Borisov 28748260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2875eb60ceacSChris Mason { 287676dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2877f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 28788fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2879facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 288024bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2881eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 288211833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2883eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2884a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 288576dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 288624bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 28874cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2888ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2889f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 289047ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 289140ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2892afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2893c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2894f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2895d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2896f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 28977585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2898573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 28990bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 2900de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 290119c00ddcSChris Mason 29020b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 29036324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2904f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 290547ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 290618bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2907afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2908bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2909bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 29103fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 29113fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2912bd647ce3SJosef Bacik #endif 2913c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 291466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 291566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 291666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 291766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 291866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 291966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 292066d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2921ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2922ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2923ba2c4d4eSJosef Bacik 2924771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 29254cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 29262fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2927034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2928fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 292995ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 29309ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 29314cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2932a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2933f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 29348b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2935f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 293690519d66SArne Jansen /* readahead state */ 2937d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 293890519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2939fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2940c8b97818SChris Mason 2941b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2942b34b086cSChris Mason num_online_cpus() + 2, 8); 29430afbaf8cSChris Mason 2944199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2945199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 294669fe2d75SJosef Bacik 2947638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 294821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 294921adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 295021adbd5cSStefan Behrens #endif 2951779a65a4SEric Sandeen btrfs_init_balance(fs_info); 295257056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2953a2de733cSArne Jansen 29540f9dd46cSJosef Bacik spin_lock_init(&fs_info->block_group_cache_lock); 29556bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2956a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 29570f9dd46cSJosef Bacik 2958fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 2959fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 2960afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 296139279cc3SChris Mason 29625a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2963e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2964925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2965a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2966a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 29671bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 29689e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2969c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 297076dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2971803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2972fa9c0d79SChris Mason 2973ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2974f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2975b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2976416ac51dSArne Jansen 2977fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2978fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2979fa9c0d79SChris Mason 2980e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2981f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2982bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 29834854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2984034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 29853768f368SChris Mason 2986da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2987da17066cSJeff Mahoney fs_info->nodesize = 4096; 2988da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2989ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2990da17066cSJeff Mahoney fs_info->stripesize = 4096; 2991da17066cSJeff Mahoney 2992eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2993eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2994eede2bf3SOmar Sandoval 29951cea5cf0SFilipe Manana spin_lock_init(&fs_info->send_reloc_lock); 29969e967495SFilipe Manana fs_info->send_in_progress = 0; 299718bb8bbfSJohannes Thumshirn 299818bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 299918bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 30008260edbaSJosef Bacik } 30018260edbaSJosef Bacik 30028260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 30038260edbaSJosef Bacik { 30048260edbaSJosef Bacik int ret; 30058260edbaSJosef Bacik 30068260edbaSJosef Bacik fs_info->sb = sb; 30078260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 30088260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 30099e967495SFilipe Manana 30105deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 3011ae18c37aSJosef Bacik if (ret) 3012ae18c37aSJosef Bacik return ret; 3013ae18c37aSJosef Bacik 3014ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 3015ae18c37aSJosef Bacik if (ret) 3016c75e8394SJosef Bacik return ret; 3017ae18c37aSJosef Bacik 3018ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 3019ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 3020ae18c37aSJosef Bacik 3021ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 3022ae18c37aSJosef Bacik if (ret) 3023c75e8394SJosef Bacik return ret; 3024ae18c37aSJosef Bacik 3025ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 3026ae18c37aSJosef Bacik GFP_KERNEL); 3027ae18c37aSJosef Bacik if (ret) 3028c75e8394SJosef Bacik return ret; 3029ae18c37aSJosef Bacik 3030ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 3031ae18c37aSJosef Bacik GFP_KERNEL); 3032c75e8394SJosef Bacik if (!fs_info->delayed_root) 3033c75e8394SJosef Bacik return -ENOMEM; 3034ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 3035ae18c37aSJosef Bacik 3036a0a1db70SFilipe Manana if (sb_rdonly(sb)) 3037a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 3038a0a1db70SFilipe Manana 3039c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 3040ae18c37aSJosef Bacik } 3041ae18c37aSJosef Bacik 304297f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 304397f4dd09SNikolay Borisov { 304497f4dd09SNikolay Borisov struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; 304597f4dd09SNikolay Borisov int ret; 304697f4dd09SNikolay Borisov 304797f4dd09SNikolay Borisov /* 304897f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 304997f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 305097f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 305197f4dd09SNikolay Borisov */ 305297f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 305397f4dd09SNikolay Borisov if (ret < 0) { 3054c94bec2cSJosef Bacik if (ret != -EINTR) 3055c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 3056c94bec2cSJosef Bacik ret); 305797f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 305897f4dd09SNikolay Borisov return ret; 305997f4dd09SNikolay Borisov } 306097f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 306197f4dd09SNikolay Borisov } 306297f4dd09SNikolay Borisov 306397f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 306497f4dd09SNikolay Borisov { 306597f4dd09SNikolay Borisov struct task_struct *task; 306697f4dd09SNikolay Borisov 306797f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 306897f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 306997f4dd09SNikolay Borisov if (IS_ERR(task)) { 307097f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 307197f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 307297f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 307397f4dd09SNikolay Borisov return PTR_ERR(task); 307497f4dd09SNikolay Borisov } 307597f4dd09SNikolay Borisov 307697f4dd09SNikolay Borisov return 0; 307797f4dd09SNikolay Borisov } 307897f4dd09SNikolay Borisov 307944c0ca21SBoris Burkov /* 30808cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 30818cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 30828cd29088SBoris Burkov * code paths. 30838cd29088SBoris Burkov */ 30848cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 30858cd29088SBoris Burkov { 30868cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 30878b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 30888cd29088SBoris Burkov } 30898cd29088SBoris Burkov 30908cd29088SBoris Burkov /* 309144c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 309244c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 309344c0ca21SBoris Burkov */ 309444c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 309544c0ca21SBoris Burkov { 309644c0ca21SBoris Burkov int ret; 309794846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 30988b228324SBoris Burkov bool clear_free_space_tree = false; 30998b228324SBoris Burkov 31008b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 31018b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 31028b228324SBoris Burkov clear_free_space_tree = true; 31038b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 31048b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 31058b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 31068b228324SBoris Burkov clear_free_space_tree = true; 31078b228324SBoris Burkov } 31088b228324SBoris Burkov 31098b228324SBoris Burkov if (clear_free_space_tree) { 31108b228324SBoris Burkov btrfs_info(fs_info, "clearing free space tree"); 31118b228324SBoris Burkov ret = btrfs_clear_free_space_tree(fs_info); 31128b228324SBoris Burkov if (ret) { 31138b228324SBoris Burkov btrfs_warn(fs_info, 31148b228324SBoris Burkov "failed to clear free space tree: %d", ret); 31158b228324SBoris Burkov goto out; 31168b228324SBoris Burkov } 31178b228324SBoris Burkov } 311844c0ca21SBoris Burkov 31198d488a8cSFilipe Manana /* 31208d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 31218d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 31228d488a8cSFilipe Manana * them into the fs_info->fs_roots_radix tree. This must be done before 31238d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 31248d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 31258d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 31268d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 31278d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 31288d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 31298d488a8cSFilipe Manana */ 31308d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 31318d488a8cSFilipe Manana if (ret) 31328d488a8cSFilipe Manana goto out; 31338d488a8cSFilipe Manana 313444c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 313544c0ca21SBoris Burkov if (ret) 313644c0ca21SBoris Burkov goto out; 313744c0ca21SBoris Burkov 31388f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 31398f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 31408f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 31418f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 31428f1c21d7SBoris Burkov goto out; 31438f1c21d7SBoris Burkov } 31448f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 31458f1c21d7SBoris Burkov 314644c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 314744c0ca21SBoris Burkov ret = btrfs_recover_relocation(fs_info->tree_root); 314844c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 314944c0ca21SBoris Burkov if (ret < 0) { 315044c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 315144c0ca21SBoris Burkov goto out; 315244c0ca21SBoris Burkov } 315344c0ca21SBoris Burkov 31545011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 31555011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 31565011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 31575011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 31585011139aSBoris Burkov if (ret) { 31595011139aSBoris Burkov btrfs_warn(fs_info, 31605011139aSBoris Burkov "failed to create free space tree: %d", ret); 31615011139aSBoris Burkov goto out; 31625011139aSBoris Burkov } 31635011139aSBoris Burkov } 31645011139aSBoris Burkov 316594846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 316694846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 316794846229SBoris Burkov if (ret) 316894846229SBoris Burkov goto out; 316994846229SBoris Burkov } 317094846229SBoris Burkov 317144c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 317244c0ca21SBoris Burkov if (ret) 317344c0ca21SBoris Burkov goto out; 317444c0ca21SBoris Burkov 317544c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 317644c0ca21SBoris Burkov if (ret) { 317744c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 317844c0ca21SBoris Burkov goto out; 317944c0ca21SBoris Burkov } 318044c0ca21SBoris Burkov 318144c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 318244c0ca21SBoris Burkov 318344c0ca21SBoris Burkov if (!fs_info->uuid_root) { 318444c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 318544c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 318644c0ca21SBoris Burkov if (ret) { 318744c0ca21SBoris Burkov btrfs_warn(fs_info, 318844c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 318944c0ca21SBoris Burkov goto out; 319044c0ca21SBoris Burkov } 319144c0ca21SBoris Burkov } 319244c0ca21SBoris Burkov 319344c0ca21SBoris Burkov out: 319444c0ca21SBoris Burkov return ret; 319544c0ca21SBoris Burkov } 319644c0ca21SBoris Burkov 3197ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3198ae18c37aSJosef Bacik char *options) 3199ae18c37aSJosef Bacik { 3200ae18c37aSJosef Bacik u32 sectorsize; 3201ae18c37aSJosef Bacik u32 nodesize; 3202ae18c37aSJosef Bacik u32 stripesize; 3203ae18c37aSJosef Bacik u64 generation; 3204ae18c37aSJosef Bacik u64 features; 3205ae18c37aSJosef Bacik u16 csum_type; 3206ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3207ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3208ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3209ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3210ae18c37aSJosef Bacik int ret; 3211ae18c37aSJosef Bacik int err = -EINVAL; 3212ae18c37aSJosef Bacik int level; 3213ae18c37aSJosef Bacik 32148260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 321553b381b3SDavid Woodhouse if (ret) { 321683c8266aSDavid Sterba err = ret; 3217ae18c37aSJosef Bacik goto fail; 321853b381b3SDavid Woodhouse } 321953b381b3SDavid Woodhouse 3220ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3221ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3222ae18c37aSJosef Bacik GFP_KERNEL); 3223ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3224ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3225ae18c37aSJosef Bacik GFP_KERNEL); 3226ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3227ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 3228ae18c37aSJosef Bacik err = -ENOMEM; 3229c75e8394SJosef Bacik goto fail; 3230ae18c37aSJosef Bacik } 3231ae18c37aSJosef Bacik 3232ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 3233ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 3234ae18c37aSJosef Bacik err = -ENOMEM; 3235c75e8394SJosef Bacik goto fail; 3236ae18c37aSJosef Bacik } 3237ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 3238ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 3239ae18c37aSJosef Bacik 3240d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 32411104a885SDavid Sterba 32421104a885SDavid Sterba /* 32431104a885SDavid Sterba * Read super block and check the signature bytes only 32441104a885SDavid Sterba */ 3245d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 32468f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 32478f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 324816cdcec7SMiao Xie goto fail_alloc; 324920b45077SDave Young } 325039279cc3SChris Mason 32511104a885SDavid Sterba /* 3252260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 32538dc3f22cSJohannes Thumshirn * corrupted, we'll find out 32548dc3f22cSJohannes Thumshirn */ 32558f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 325651bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 32578dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 325851bce6c9SJohannes Thumshirn csum_type); 32598dc3f22cSJohannes Thumshirn err = -EINVAL; 32608f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32618dc3f22cSJohannes Thumshirn goto fail_alloc; 32628dc3f22cSJohannes Thumshirn } 32638dc3f22cSJohannes Thumshirn 326483c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 326583c68bbcSSu Yue 32666d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 32676d97c6e3SJohannes Thumshirn if (ret) { 32686d97c6e3SJohannes Thumshirn err = ret; 32698f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32706d97c6e3SJohannes Thumshirn goto fail_alloc; 32716d97c6e3SJohannes Thumshirn } 32726d97c6e3SJohannes Thumshirn 32738dc3f22cSJohannes Thumshirn /* 32741104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 32751104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 32761104a885SDavid Sterba */ 32778f32380dSJohannes Thumshirn if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { 327805135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 32791104a885SDavid Sterba err = -EINVAL; 32808f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3281141386e1SJosef Bacik goto fail_alloc; 32821104a885SDavid Sterba } 32831104a885SDavid Sterba 32841104a885SDavid Sterba /* 32851104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 32861104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 32871104a885SDavid Sterba * the whole block of INFO_SIZE 32881104a885SDavid Sterba */ 32898f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 32908f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32915f39d397SChris Mason 3292fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3293fbc6feaeSNikolay Borisov 32940b86a832SChris Mason 3295fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3296fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3297fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3298fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3299fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3300fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3301fbc6feaeSNikolay Borisov } 3302fbc6feaeSNikolay Borisov 3303fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3304fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3305de37aa51SNikolay Borisov 3306069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 33071104a885SDavid Sterba if (ret) { 330805135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 33091104a885SDavid Sterba err = -EINVAL; 3310141386e1SJosef Bacik goto fail_alloc; 33111104a885SDavid Sterba } 33121104a885SDavid Sterba 33130f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3314141386e1SJosef Bacik goto fail_alloc; 33150f7d52f4SChris Mason 3316acce952bSliubo /* check FS state, whether FS is broken. */ 331787533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 331887533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3319acce952bSliubo 332075e7cb7fSLiu Bo /* 332175e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 332275e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 332375e7cb7fSLiu Bo */ 332475e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 332575e7cb7fSLiu Bo 33266f93e834SAnand Jain /* 33276f93e834SAnand Jain * Flag our filesystem as having big metadata blocks if they are bigger 33286f93e834SAnand Jain * than the page size. 33296f93e834SAnand Jain */ 33306f93e834SAnand Jain if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 33316f93e834SAnand Jain if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 33326f93e834SAnand Jain btrfs_info(fs_info, 33336f93e834SAnand Jain "flagging fs with big metadata feature"); 33346f93e834SAnand Jain features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 33356f93e834SAnand Jain } 33366f93e834SAnand Jain 33376f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 33386f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 33396f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 33406f93e834SAnand Jain stripesize = sectorsize; 33416f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 33426f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 33436f93e834SAnand Jain 33446f93e834SAnand Jain fs_info->nodesize = nodesize; 33456f93e834SAnand Jain fs_info->sectorsize = sectorsize; 33466f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 33476f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 33486f93e834SAnand Jain fs_info->stripesize = stripesize; 33496f93e834SAnand Jain 33502ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 33512b82032cSYan Zheng if (ret) { 33522b82032cSYan Zheng err = ret; 3353141386e1SJosef Bacik goto fail_alloc; 33542b82032cSYan Zheng } 3355dfe25020SChris Mason 3356f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 3357f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 3358f2b636e8SJosef Bacik if (features) { 335905135f59SDavid Sterba btrfs_err(fs_info, 336005135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 3361c1c9ff7cSGeert Uytterhoeven features); 3362f2b636e8SJosef Bacik err = -EINVAL; 3363141386e1SJosef Bacik goto fail_alloc; 3364f2b636e8SJosef Bacik } 3365f2b636e8SJosef Bacik 33665d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 33675d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 33680b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3369a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 33705c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 33715c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3372727011e0SChris Mason 33733173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 337405135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 33753173a18fSJosef Bacik 3376727011e0SChris Mason /* 3377bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 3378bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 3379bc3f116fSChris Mason */ 3380bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3381707e8a07SDavid Sterba (sectorsize != nodesize)) { 338205135f59SDavid Sterba btrfs_err(fs_info, 338305135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 338405135f59SDavid Sterba nodesize, sectorsize); 3385141386e1SJosef Bacik goto fail_alloc; 3386bc3f116fSChris Mason } 3387bc3f116fSChris Mason 3388ceda0864SMiao Xie /* 3389ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 3390ceda0864SMiao Xie * update the flag. 3391ceda0864SMiao Xie */ 33925d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 33935d4f98a2SYan Zheng 3394f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 3395f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 3396bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 339705135f59SDavid Sterba btrfs_err(fs_info, 339805135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 3399c1c9ff7cSGeert Uytterhoeven features); 3400f2b636e8SJosef Bacik err = -EINVAL; 3401141386e1SJosef Bacik goto fail_alloc; 3402f2b636e8SJosef Bacik } 340361d92c32SChris Mason 34048481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 34058481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 34068481dd80SQu Wenruo 340795ea0486SQu Wenruo btrfs_warn(fs_info, 340895ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 34090bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 3410c8050b3bSQu Wenruo if (btrfs_super_incompat_flags(fs_info->super_copy) & 3411c8050b3bSQu Wenruo BTRFS_FEATURE_INCOMPAT_RAID56) { 3412c8050b3bSQu Wenruo btrfs_err(fs_info, 3413c8050b3bSQu Wenruo "RAID56 is not yet supported for sector size %u with page size %lu", 3414c8050b3bSQu Wenruo sectorsize, PAGE_SIZE); 3415c8050b3bSQu Wenruo err = -EINVAL; 3416c8050b3bSQu Wenruo goto fail_alloc; 3417c8050b3bSQu Wenruo } 34188481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 34198481dd80SQu Wenruo if (!subpage_info) 34208481dd80SQu Wenruo goto fail_alloc; 34218481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 34228481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3423c8050b3bSQu Wenruo } 34240bb3eb3eSQu Wenruo 3425d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 34262a458198SEric Sandeen if (ret) { 34272a458198SEric Sandeen err = ret; 34280dc3b84aSJosef Bacik goto fail_sb_buffer; 34290dc3b84aSJosef Bacik } 34304543df7eSChris Mason 34319e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 34329e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 34334575c9ccSChris Mason 3434a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3435a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3436de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3437db94535dSChris Mason 3438925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 34396bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3440925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 344184eed90fSChris Mason if (ret) { 344205135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 34435d4f98a2SYan Zheng goto fail_sb_buffer; 344484eed90fSChris Mason } 34450b86a832SChris Mason 344684234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3447581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 34480b86a832SChris Mason 34492ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 34500b86a832SChris Mason btrfs_super_chunk_root(disk_super), 34511b7ec85eSJosef Bacik BTRFS_CHUNK_TREE_OBJECTID, 3452581c1760SQu Wenruo generation, level, NULL); 345364c043deSLiu Bo if (IS_ERR(chunk_root->node) || 345464c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 345505135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3456e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 3457e5fffbacSchandan free_extent_buffer(chunk_root->node); 345895ab1f64SZhao Lei chunk_root->node = NULL; 3459af31f5e5SChris Mason goto fail_tree_roots; 346083121942SDavid Woodhouse } 34615d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 34625d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 34630b86a832SChris Mason 3464e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3465c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3466c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3467e17cade2SChris Mason 34685b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 34692b82032cSYan Zheng if (ret) { 347005135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3471af31f5e5SChris Mason goto fail_tree_roots; 34722b82032cSYan Zheng } 34730b86a832SChris Mason 34748dabb742SStefan Behrens /* 3475bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3476bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3477bacce86aSAnand Jain * target is required. So free devices that are not part of this 34781a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3479bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 34808dabb742SStefan Behrens */ 3481bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3482d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 348305135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3484a6b0d5c8SChris Mason goto fail_tree_roots; 3485a6b0d5c8SChris Mason } 3486a6b0d5c8SChris Mason 3487b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 34884bbcaa64SEric Sandeen if (ret) 3489b8522a1eSNikolay Borisov goto fail_tree_roots; 34908929ecfaSYan, Zheng 349175ec1db8SJosef Bacik /* 349273651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 349373651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 349473651042SNaohiro Aota * zoned incompat feature flag set. 349573651042SNaohiro Aota */ 349673651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 349773651042SNaohiro Aota if (ret) { 349873651042SNaohiro Aota btrfs_err(fs_info, 349973651042SNaohiro Aota "zoned: failed to read device zone info: %d", 350073651042SNaohiro Aota ret); 350173651042SNaohiro Aota goto fail_block_groups; 350273651042SNaohiro Aota } 350373651042SNaohiro Aota 350473651042SNaohiro Aota /* 350575ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 350675ec1db8SJosef Bacik * check the generation here so we can set the 350775ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 350875ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 350975ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 351075ec1db8SJosef Bacik * even though it was perfectly fine. 351175ec1db8SJosef Bacik */ 351275ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 351375ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 351475ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 351575ec1db8SJosef Bacik 3516cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3517cf90d884SQu Wenruo if (ret) { 3518cf90d884SQu Wenruo btrfs_err(fs_info, 3519cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3520cf90d884SQu Wenruo ret); 3521cf90d884SQu Wenruo goto fail_block_groups; 3522cf90d884SQu Wenruo } 352368310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 352468310a5eSIlya Dryomov if (ret) { 352505135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 352668310a5eSIlya Dryomov goto fail_block_groups; 352768310a5eSIlya Dryomov } 352868310a5eSIlya Dryomov 3529733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3530733f4fbbSStefan Behrens if (ret) { 353105135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3532733f4fbbSStefan Behrens goto fail_block_groups; 3533733f4fbbSStefan Behrens } 3534733f4fbbSStefan Behrens 35358dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 35368dabb742SStefan Behrens if (ret) { 353705135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 35388dabb742SStefan Behrens goto fail_block_groups; 35398dabb742SStefan Behrens } 35408dabb742SStefan Behrens 3541b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3542b70f5097SNaohiro Aota if (ret) { 3543b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3544b70f5097SNaohiro Aota ret); 3545b70f5097SNaohiro Aota goto fail_block_groups; 3546b70f5097SNaohiro Aota } 3547b70f5097SNaohiro Aota 3548c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3549b7c35e81SAnand Jain if (ret) { 355005135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 355105135f59SDavid Sterba ret); 3552b7c35e81SAnand Jain goto fail_block_groups; 3553b7c35e81SAnand Jain } 3554b7c35e81SAnand Jain 355596f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 35565ac1d209SJeff Mahoney if (ret) { 355705135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3558b7c35e81SAnand Jain goto fail_fsdev_sysfs; 35595ac1d209SJeff Mahoney } 35605ac1d209SJeff Mahoney 3561c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3562c59021f8Sliubo if (ret) { 356305135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 35642365dd3cSAnand Jain goto fail_sysfs; 3565c59021f8Sliubo } 3566c59021f8Sliubo 35675b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 35681b1d1f66SJosef Bacik if (ret) { 356905135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 35702365dd3cSAnand Jain goto fail_sysfs; 35711b1d1f66SJosef Bacik } 35724330e183SQu Wenruo 3573*16beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 3574*16beac87SNaohiro Aota 35755c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 35765c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 357705135f59SDavid Sterba btrfs_warn(fs_info, 357852042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 35792365dd3cSAnand Jain goto fail_sysfs; 3580292fd7fcSStefan Behrens } 35819078a3e1SChris Mason 3582a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 3583a74a4b97SChris Mason "btrfs-cleaner"); 358457506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 35852365dd3cSAnand Jain goto fail_sysfs; 3586a74a4b97SChris Mason 3587a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3588a74a4b97SChris Mason tree_root, 3589a74a4b97SChris Mason "btrfs-transaction"); 359057506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 35913f157a2fSChris Mason goto fail_cleaner; 3592a74a4b97SChris Mason 3593583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3594c289811cSChris Mason !fs_info->fs_devices->rotating) { 3595583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3596c289811cSChris Mason } 3597c289811cSChris Mason 3598572d9ab7SDavid Sterba /* 359901327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 3600572d9ab7SDavid Sterba * not have to wait for transaction commit 3601572d9ab7SDavid Sterba */ 3602572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 36033818aea2SQu Wenruo 360421adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 36050b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 36062ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 36070b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3608cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 360921adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 361021adbd5cSStefan Behrens if (ret) 361105135f59SDavid Sterba btrfs_warn(fs_info, 361205135f59SDavid Sterba "failed to initialize integrity check module: %d", 361305135f59SDavid Sterba ret); 361421adbd5cSStefan Behrens } 361521adbd5cSStefan Behrens #endif 3616bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3617bcef60f2SArne Jansen if (ret) 3618bcef60f2SArne Jansen goto fail_trans_kthread; 361921adbd5cSStefan Behrens 3620fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3621fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3622fd708b81SJosef Bacik 362396da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 362496da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 36250b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3626e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 362763443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 362879787eaaSJeff Mahoney if (ret) { 362963443bf5SEric Sandeen err = ret; 363028c16cbbSWang Shilong goto fail_qgroup; 3631e556ce2cSYan Zheng } 3632e02119d5SChris Mason } 36331a40e23bSZheng Yan 363456e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 36353140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 36363140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3637f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3638315bf8efSJosef Bacik fs_info->fs_root = NULL; 3639bcef60f2SArne Jansen goto fail_qgroup; 36403140c9a3SDan Carpenter } 3641c289811cSChris Mason 3642bc98a42cSDavid Howells if (sb_rdonly(sb)) 36438cd29088SBoris Burkov goto clear_oneshot; 36442b6ba629SIlya Dryomov 364544c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 36462b6ba629SIlya Dryomov if (ret) { 36476bccf3abSJeff Mahoney close_ctree(fs_info); 36482b6ba629SIlya Dryomov return ret; 3649e3acc2a6SJosef Bacik } 3650b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3651b382a324SJan Schmidt 365244c0ca21SBoris Burkov if (fs_info->uuid_root && 365344c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 365444c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 365505135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 365670f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 365770f80175SStefan Behrens if (ret) { 365805135f59SDavid Sterba btrfs_warn(fs_info, 365905135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 36606bccf3abSJeff Mahoney close_ctree(fs_info); 366170f80175SStefan Behrens return ret; 366270f80175SStefan Behrens } 3663f7a81ea4SStefan Behrens } 366494846229SBoris Burkov 3665afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 366647ab2a6cSJosef Bacik 36678cd29088SBoris Burkov clear_oneshot: 36688cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3669ad2b2c80SAl Viro return 0; 367039279cc3SChris Mason 3671bcef60f2SArne Jansen fail_qgroup: 3672bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 36737c2ca468SChris Mason fail_trans_kthread: 36747c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 36752ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3676faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 36773f157a2fSChris Mason fail_cleaner: 3678a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 36797c2ca468SChris Mason 36807c2ca468SChris Mason /* 36817c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 36827c2ca468SChris Mason * kthreads 36837c2ca468SChris Mason */ 36847c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 36857c2ca468SChris Mason 36862365dd3cSAnand Jain fail_sysfs: 36876618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 36882365dd3cSAnand Jain 3689b7c35e81SAnand Jain fail_fsdev_sysfs: 3690b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3691b7c35e81SAnand Jain 36921b1d1f66SJosef Bacik fail_block_groups: 369354067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3694af31f5e5SChris Mason 3695af31f5e5SChris Mason fail_tree_roots: 36969e3aa805SJosef Bacik if (fs_info->data_reloc_root) 36979e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 36984273eaffSAnand Jain free_root_pointers(fs_info, true); 36992b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3700af31f5e5SChris Mason 370139279cc3SChris Mason fail_sb_buffer: 37027abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 37035cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 370416cdcec7SMiao Xie fail_alloc: 3705586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3706586e46e2SIlya Dryomov 37074543df7eSChris Mason iput(fs_info->btree_inode); 37087e662854SQinghuang Feng fail: 3709586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3710ad2b2c80SAl Viro return err; 3711eb60ceacSChris Mason } 3712663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3713eb60ceacSChris Mason 3714314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3715f2984462SChris Mason { 3716314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3717314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3718314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3719314b6dd0SJohannes Thumshirn struct page *page; 3720442a4f63SStefan Behrens 3721314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3722314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3723314b6dd0SJohannes Thumshirn 3724314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3725fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3726314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3727314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3728314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3729314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3730314b6dd0SJohannes Thumshirn SetPageError(page); 3731314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3732314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3733314b6dd0SJohannes Thumshirn } else { 3734314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3735f2984462SChris Mason } 3736314b6dd0SJohannes Thumshirn 3737314b6dd0SJohannes Thumshirn put_page(page); 3738314b6dd0SJohannes Thumshirn unlock_page(page); 3739314b6dd0SJohannes Thumshirn } 3740314b6dd0SJohannes Thumshirn 3741314b6dd0SJohannes Thumshirn bio_put(bio); 3742f2984462SChris Mason } 3743f2984462SChris Mason 37448f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 37458f32380dSJohannes Thumshirn int copy_num) 374629c36d72SAnand Jain { 374729c36d72SAnand Jain struct btrfs_super_block *super; 37488f32380dSJohannes Thumshirn struct page *page; 374912659251SNaohiro Aota u64 bytenr, bytenr_orig; 37508f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 375112659251SNaohiro Aota int ret; 375229c36d72SAnand Jain 375312659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 375412659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 375512659251SNaohiro Aota if (ret == -ENOENT) 375612659251SNaohiro Aota return ERR_PTR(-EINVAL); 375712659251SNaohiro Aota else if (ret) 375812659251SNaohiro Aota return ERR_PTR(ret); 375912659251SNaohiro Aota 3760cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 37618f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 376229c36d72SAnand Jain 37638f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 37648f32380dSJohannes Thumshirn if (IS_ERR(page)) 37658f32380dSJohannes Thumshirn return ERR_CAST(page); 376629c36d72SAnand Jain 37678f32380dSJohannes Thumshirn super = page_address(page); 376896c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 376996c2e067SAnand Jain btrfs_release_disk_super(super); 377096c2e067SAnand Jain return ERR_PTR(-ENODATA); 377196c2e067SAnand Jain } 377296c2e067SAnand Jain 377312659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 37748f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 37758f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 377629c36d72SAnand Jain } 377729c36d72SAnand Jain 37788f32380dSJohannes Thumshirn return super; 377929c36d72SAnand Jain } 378029c36d72SAnand Jain 378129c36d72SAnand Jain 37828f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3783a512bbf8SYan Zheng { 37848f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3785a512bbf8SYan Zheng int i; 3786a512bbf8SYan Zheng u64 transid = 0; 3787a512bbf8SYan Zheng 3788a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3789a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3790a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3791a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3792a512bbf8SYan Zheng */ 3793a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 37948f32380dSJohannes Thumshirn super = btrfs_read_dev_one_super(bdev, i); 37958f32380dSJohannes Thumshirn if (IS_ERR(super)) 3796a512bbf8SYan Zheng continue; 3797a512bbf8SYan Zheng 3798a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 37998f32380dSJohannes Thumshirn if (latest) 38008f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 38018f32380dSJohannes Thumshirn 38028f32380dSJohannes Thumshirn latest = super; 3803a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3804a512bbf8SYan Zheng } 3805a512bbf8SYan Zheng } 380692fc03fbSAnand Jain 38078f32380dSJohannes Thumshirn return super; 3808a512bbf8SYan Zheng } 3809a512bbf8SYan Zheng 38104eedeb75SHisashi Hifumi /* 3811abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3812314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 38134eedeb75SHisashi Hifumi * 3814abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3815abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3816abbb3b8eSDavid Sterba * same for write and wait phases. 38174eedeb75SHisashi Hifumi * 3818314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 38194eedeb75SHisashi Hifumi */ 3820a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3821abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3822a512bbf8SYan Zheng { 3823d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3824314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3825d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3826a512bbf8SYan Zheng int i; 3827a512bbf8SYan Zheng int errors = 0; 382812659251SNaohiro Aota int ret; 382912659251SNaohiro Aota u64 bytenr, bytenr_orig; 3830a512bbf8SYan Zheng 3831a512bbf8SYan Zheng if (max_mirrors == 0) 3832a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3833a512bbf8SYan Zheng 3834d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3835d5178578SJohannes Thumshirn 3836a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3837314b6dd0SJohannes Thumshirn struct page *page; 3838314b6dd0SJohannes Thumshirn struct bio *bio; 3839314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3840314b6dd0SJohannes Thumshirn 384112659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 384212659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 384312659251SNaohiro Aota if (ret == -ENOENT) { 384412659251SNaohiro Aota continue; 384512659251SNaohiro Aota } else if (ret < 0) { 384612659251SNaohiro Aota btrfs_err(device->fs_info, 384712659251SNaohiro Aota "couldn't get super block location for mirror %d", 384812659251SNaohiro Aota i); 384912659251SNaohiro Aota errors++; 385012659251SNaohiro Aota continue; 385112659251SNaohiro Aota } 3852935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3853935e5cc9SMiao Xie device->commit_total_bytes) 3854a512bbf8SYan Zheng break; 3855a512bbf8SYan Zheng 385612659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3857a512bbf8SYan Zheng 3858fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3859fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3860fd08001fSEric Biggers sb->csum); 3861a512bbf8SYan Zheng 3862314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3863314b6dd0SJohannes Thumshirn GFP_NOFS); 3864314b6dd0SJohannes Thumshirn if (!page) { 3865fb456252SJeff Mahoney btrfs_err(device->fs_info, 3866314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3867f14d104dSDavid Sterba bytenr); 3868634554dcSJosef Bacik errors++; 3869634554dcSJosef Bacik continue; 3870634554dcSJosef Bacik } 3871634554dcSJosef Bacik 3872314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3873314b6dd0SJohannes Thumshirn get_page(page); 3874a512bbf8SYan Zheng 3875314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3876314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3877a512bbf8SYan Zheng 3878387125fcSChris Mason /* 3879314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3880314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3881314b6dd0SJohannes Thumshirn * checking. 3882387125fcSChris Mason */ 3883314b6dd0SJohannes Thumshirn bio = bio_alloc(GFP_NOFS, 1); 3884314b6dd0SJohannes Thumshirn bio_set_dev(bio, device->bdev); 3885314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3886314b6dd0SJohannes Thumshirn bio->bi_private = device; 3887314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3888314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3889314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3890314b6dd0SJohannes Thumshirn 3891314b6dd0SJohannes Thumshirn /* 3892314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3893314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3894314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3895314b6dd0SJohannes Thumshirn */ 3896314b6dd0SJohannes Thumshirn bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; 38971b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3898314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3899314b6dd0SJohannes Thumshirn 3900314b6dd0SJohannes Thumshirn btrfsic_submit_bio(bio); 39018376d9e1SNaohiro Aota 39028376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 39038376d9e1SNaohiro Aota errors++; 3904a512bbf8SYan Zheng } 3905a512bbf8SYan Zheng return errors < i ? 0 : -1; 3906a512bbf8SYan Zheng } 3907a512bbf8SYan Zheng 3908387125fcSChris Mason /* 3909abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3910abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3911abbb3b8eSDavid Sterba * 3912314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3913abbb3b8eSDavid Sterba * date. 3914abbb3b8eSDavid Sterba */ 3915abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3916abbb3b8eSDavid Sterba { 3917abbb3b8eSDavid Sterba int i; 3918abbb3b8eSDavid Sterba int errors = 0; 3919b6a535faSHoward McLauchlan bool primary_failed = false; 392012659251SNaohiro Aota int ret; 3921abbb3b8eSDavid Sterba u64 bytenr; 3922abbb3b8eSDavid Sterba 3923abbb3b8eSDavid Sterba if (max_mirrors == 0) 3924abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3925abbb3b8eSDavid Sterba 3926abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3927314b6dd0SJohannes Thumshirn struct page *page; 3928314b6dd0SJohannes Thumshirn 392912659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 393012659251SNaohiro Aota if (ret == -ENOENT) { 393112659251SNaohiro Aota break; 393212659251SNaohiro Aota } else if (ret < 0) { 393312659251SNaohiro Aota errors++; 393412659251SNaohiro Aota if (i == 0) 393512659251SNaohiro Aota primary_failed = true; 393612659251SNaohiro Aota continue; 393712659251SNaohiro Aota } 3938abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3939abbb3b8eSDavid Sterba device->commit_total_bytes) 3940abbb3b8eSDavid Sterba break; 3941abbb3b8eSDavid Sterba 3942314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3943314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3944314b6dd0SJohannes Thumshirn if (!page) { 3945abbb3b8eSDavid Sterba errors++; 3946b6a535faSHoward McLauchlan if (i == 0) 3947b6a535faSHoward McLauchlan primary_failed = true; 3948abbb3b8eSDavid Sterba continue; 3949abbb3b8eSDavid Sterba } 3950314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3951314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3952314b6dd0SJohannes Thumshirn if (PageError(page)) { 3953abbb3b8eSDavid Sterba errors++; 3954b6a535faSHoward McLauchlan if (i == 0) 3955b6a535faSHoward McLauchlan primary_failed = true; 3956b6a535faSHoward McLauchlan } 3957abbb3b8eSDavid Sterba 3958314b6dd0SJohannes Thumshirn /* Drop our reference */ 3959314b6dd0SJohannes Thumshirn put_page(page); 3960abbb3b8eSDavid Sterba 3961314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3962314b6dd0SJohannes Thumshirn put_page(page); 3963abbb3b8eSDavid Sterba } 3964abbb3b8eSDavid Sterba 3965b6a535faSHoward McLauchlan /* log error, force error return */ 3966b6a535faSHoward McLauchlan if (primary_failed) { 3967b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3968b6a535faSHoward McLauchlan device->devid); 3969b6a535faSHoward McLauchlan return -1; 3970b6a535faSHoward McLauchlan } 3971b6a535faSHoward McLauchlan 3972abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3973abbb3b8eSDavid Sterba } 3974abbb3b8eSDavid Sterba 3975abbb3b8eSDavid Sterba /* 3976387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3977387125fcSChris Mason * for the barrier when it is done 3978387125fcSChris Mason */ 39794246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3980387125fcSChris Mason { 3981387125fcSChris Mason complete(bio->bi_private); 3982387125fcSChris Mason } 3983387125fcSChris Mason 3984387125fcSChris Mason /* 39854fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 39864fc6441aSAnand Jain * done in the waiting counterpart. 3987387125fcSChris Mason */ 39884fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3989387125fcSChris Mason { 3990e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3991387125fcSChris Mason 3992a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3993a91cf0ffSWang Yugui /* 3994a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 3995a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 3996a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 3997a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 3998a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 3999a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 4000a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 4001a91cf0ffSWang Yugui * non-debug builds. 4002a91cf0ffSWang Yugui */ 4003a91cf0ffSWang Yugui struct request_queue *q = bdev_get_queue(device->bdev); 4004c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 40054fc6441aSAnand Jain return; 4006a91cf0ffSWang Yugui #endif 4007387125fcSChris Mason 4008e0ae9994SDavid Sterba bio_reset(bio); 4009387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 401074d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 40118d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 4012387125fcSChris Mason init_completion(&device->flush_wait); 4013387125fcSChris Mason bio->bi_private = &device->flush_wait; 4014387125fcSChris Mason 401543a01111SLu Fengqi btrfsic_submit_bio(bio); 40161c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 40174fc6441aSAnand Jain } 4018387125fcSChris Mason 40194fc6441aSAnand Jain /* 40204fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 40214fc6441aSAnand Jain */ 40228c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 40234fc6441aSAnand Jain { 40244fc6441aSAnand Jain struct bio *bio = device->flush_bio; 40254fc6441aSAnand Jain 40261c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 402758efbc9fSOmar Sandoval return BLK_STS_OK; 40284fc6441aSAnand Jain 40291c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 40302980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 40314fc6441aSAnand Jain 40328c27cb35SLinus Torvalds return bio->bi_status; 4033387125fcSChris Mason } 4034387125fcSChris Mason 4035d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 4036401b41e5SAnand Jain { 40376528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 4038401b41e5SAnand Jain return -EIO; 4039387125fcSChris Mason return 0; 4040387125fcSChris Mason } 4041387125fcSChris Mason 4042387125fcSChris Mason /* 4043387125fcSChris Mason * send an empty flush down to each device in parallel, 4044387125fcSChris Mason * then wait for them 4045387125fcSChris Mason */ 4046387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 4047387125fcSChris Mason { 4048387125fcSChris Mason struct list_head *head; 4049387125fcSChris Mason struct btrfs_device *dev; 40505af3e8ccSStefan Behrens int errors_wait = 0; 40514e4cbee9SChristoph Hellwig blk_status_t ret; 4052387125fcSChris Mason 40531538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 4054387125fcSChris Mason /* send down all the barriers */ 4055387125fcSChris Mason head = &info->fs_devices->devices; 40561538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4057e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4058f88ba6a2SHidetoshi Seto continue; 4059cea7c8bfSAnand Jain if (!dev->bdev) 4060387125fcSChris Mason continue; 4061e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4062ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4063387125fcSChris Mason continue; 4064387125fcSChris Mason 40654fc6441aSAnand Jain write_dev_flush(dev); 406658efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 4067387125fcSChris Mason } 4068387125fcSChris Mason 4069387125fcSChris Mason /* wait for all the barriers */ 40701538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4071e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4072f88ba6a2SHidetoshi Seto continue; 4073387125fcSChris Mason if (!dev->bdev) { 40745af3e8ccSStefan Behrens errors_wait++; 4075387125fcSChris Mason continue; 4076387125fcSChris Mason } 4077e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4078ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4079387125fcSChris Mason continue; 4080387125fcSChris Mason 40814fc6441aSAnand Jain ret = wait_dev_flush(dev); 4082401b41e5SAnand Jain if (ret) { 4083401b41e5SAnand Jain dev->last_flush_error = ret; 408466b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 408566b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 40865af3e8ccSStefan Behrens errors_wait++; 4087387125fcSChris Mason } 4088401b41e5SAnand Jain } 4089401b41e5SAnand Jain 4090cea7c8bfSAnand Jain if (errors_wait) { 4091401b41e5SAnand Jain /* 4092401b41e5SAnand Jain * At some point we need the status of all disks 4093401b41e5SAnand Jain * to arrive at the volume status. So error checking 4094401b41e5SAnand Jain * is being pushed to a separate loop. 4095401b41e5SAnand Jain */ 4096d10b82feSQu Wenruo return check_barrier_error(info); 4097401b41e5SAnand Jain } 4098387125fcSChris Mason return 0; 4099387125fcSChris Mason } 4100387125fcSChris Mason 4101943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4102943c6e99SZhao Lei { 41038789f4feSZhao Lei int raid_type; 41048789f4feSZhao Lei int min_tolerated = INT_MAX; 4105943c6e99SZhao Lei 41068789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 41078789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 41088c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 41098789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 41108789f4feSZhao Lei tolerated_failures); 4111943c6e99SZhao Lei 41128789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 41138789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 41148789f4feSZhao Lei continue; 411541a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 41168789f4feSZhao Lei continue; 41178c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 41188789f4feSZhao Lei btrfs_raid_array[raid_type]. 41198789f4feSZhao Lei tolerated_failures); 41208789f4feSZhao Lei } 4121943c6e99SZhao Lei 41228789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4123ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 41248789f4feSZhao Lei min_tolerated = 0; 41258789f4feSZhao Lei } 41268789f4feSZhao Lei 41278789f4feSZhao Lei return min_tolerated; 4128943c6e99SZhao Lei } 4129943c6e99SZhao Lei 4130eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4131f2984462SChris Mason { 4132e5e9a520SChris Mason struct list_head *head; 4133f2984462SChris Mason struct btrfs_device *dev; 4134a061fc8dSChris Mason struct btrfs_super_block *sb; 4135f2984462SChris Mason struct btrfs_dev_item *dev_item; 4136f2984462SChris Mason int ret; 4137f2984462SChris Mason int do_barriers; 4138a236aed1SChris Mason int max_errors; 4139a236aed1SChris Mason int total_errors = 0; 4140a061fc8dSChris Mason u64 flags; 4141f2984462SChris Mason 41420b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4143fed3b381SLiu Bo 4144fed3b381SLiu Bo /* 4145fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4146fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4147fed3b381SLiu Bo * been consistent on disk. 4148fed3b381SLiu Bo */ 4149fed3b381SLiu Bo if (max_mirrors == 0) 41500b246afaSJeff Mahoney backup_super_roots(fs_info); 4151f2984462SChris Mason 41520b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4153a061fc8dSChris Mason dev_item = &sb->dev_item; 4154e5e9a520SChris Mason 41550b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 41560b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 41570b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4158387125fcSChris Mason 41595af3e8ccSStefan Behrens if (do_barriers) { 41600b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 41615af3e8ccSStefan Behrens if (ret) { 41625af3e8ccSStefan Behrens mutex_unlock( 41630b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 41640b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 41655af3e8ccSStefan Behrens "errors while submitting device barriers."); 41665af3e8ccSStefan Behrens return ret; 41675af3e8ccSStefan Behrens } 41685af3e8ccSStefan Behrens } 4169387125fcSChris Mason 41701538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4171dfe25020SChris Mason if (!dev->bdev) { 4172dfe25020SChris Mason total_errors++; 4173dfe25020SChris Mason continue; 4174dfe25020SChris Mason } 4175e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4176ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4177dfe25020SChris Mason continue; 4178dfe25020SChris Mason 41792b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4180a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4181a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 41827df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4183935e5cc9SMiao Xie dev->commit_total_bytes); 4184ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4185ce7213c7SMiao Xie dev->commit_bytes_used); 4186a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4187a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4188a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4189a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 41907239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 41917239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4192a512bbf8SYan Zheng 4193a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4194a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4195f2984462SChris Mason 419675cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 419775cb857dSQu Wenruo if (ret < 0) { 419875cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 419975cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 420075cb857dSQu Wenruo "unexpected superblock corruption detected"); 420175cb857dSQu Wenruo return -EUCLEAN; 420275cb857dSQu Wenruo } 420375cb857dSQu Wenruo 4204abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4205a236aed1SChris Mason if (ret) 4206a236aed1SChris Mason total_errors++; 4207f2984462SChris Mason } 4208a236aed1SChris Mason if (total_errors > max_errors) { 42090b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4210d397712bSChris Mason total_errors); 42110b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 421279787eaaSJeff Mahoney 42139d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 42140b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 42150b246afaSJeff Mahoney "%d errors while writing supers", 42160b246afaSJeff Mahoney total_errors); 42179d565ba4SStefan Behrens return -EIO; 4218a236aed1SChris Mason } 4219f2984462SChris Mason 4220a512bbf8SYan Zheng total_errors = 0; 42211538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4222dfe25020SChris Mason if (!dev->bdev) 4223dfe25020SChris Mason continue; 4224e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4225ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4226dfe25020SChris Mason continue; 4227dfe25020SChris Mason 4228abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4229a512bbf8SYan Zheng if (ret) 42301259ab75SChris Mason total_errors++; 4231f2984462SChris Mason } 42320b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4233a236aed1SChris Mason if (total_errors > max_errors) { 42340b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 42350b246afaSJeff Mahoney "%d errors while writing supers", 42360b246afaSJeff Mahoney total_errors); 423779787eaaSJeff Mahoney return -EIO; 4238a236aed1SChris Mason } 4239f2984462SChris Mason return 0; 4240f2984462SChris Mason } 4241f2984462SChris Mason 4242cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4243cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4244cb517eabSMiao Xie struct btrfs_root *root) 42452619ba1fSChris Mason { 42464785e24fSJosef Bacik bool drop_ref = false; 42474785e24fSJosef Bacik 42484df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 42492619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 42502619ba1fSChris Mason (unsigned long)root->root_key.objectid); 4251af01d2e5SJosef Bacik if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 42524785e24fSJosef Bacik drop_ref = true; 42534df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 425476dda93cSYan, Zheng 425584961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4256ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 42571c1ea4f7SLiu Bo if (root->reloc_root) { 425800246528SJosef Bacik btrfs_put_root(root->reloc_root); 42591c1ea4f7SLiu Bo root->reloc_root = NULL; 42601c1ea4f7SLiu Bo } 42611c1ea4f7SLiu Bo } 42623321719eSLiu Bo 42634785e24fSJosef Bacik if (drop_ref) 426400246528SJosef Bacik btrfs_put_root(root); 42652619ba1fSChris Mason } 42662619ba1fSChris Mason 4267c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4268c146afadSYan Zheng { 4269c146afadSYan Zheng u64 root_objectid = 0; 4270c146afadSYan Zheng struct btrfs_root *gang[8]; 427165d33fd7SQu Wenruo int i = 0; 427265d33fd7SQu Wenruo int err = 0; 427365d33fd7SQu Wenruo unsigned int ret = 0; 4274c146afadSYan Zheng 4275c146afadSYan Zheng while (1) { 4276efc34534SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4277c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4278c146afadSYan Zheng (void **)gang, root_objectid, 4279c146afadSYan Zheng ARRAY_SIZE(gang)); 428065d33fd7SQu Wenruo if (!ret) { 4281efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4282c146afadSYan Zheng break; 428365d33fd7SQu Wenruo } 42845d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 428566b4ffd1SJosef Bacik 428665d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 428765d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 428865d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 428965d33fd7SQu Wenruo gang[i] = NULL; 429065d33fd7SQu Wenruo continue; 429165d33fd7SQu Wenruo } 429265d33fd7SQu Wenruo /* grab all the search result for later use */ 429300246528SJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 429465d33fd7SQu Wenruo } 4295efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 429665d33fd7SQu Wenruo 429765d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 429865d33fd7SQu Wenruo if (!gang[i]) 429965d33fd7SQu Wenruo continue; 4300c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 430166b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 430266b4ffd1SJosef Bacik if (err) 430365d33fd7SQu Wenruo break; 430400246528SJosef Bacik btrfs_put_root(gang[i]); 4305c146afadSYan Zheng } 4306c146afadSYan Zheng root_objectid++; 4307c146afadSYan Zheng } 430865d33fd7SQu Wenruo 430965d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 431065d33fd7SQu Wenruo for (; i < ret; i++) { 431165d33fd7SQu Wenruo if (gang[i]) 431200246528SJosef Bacik btrfs_put_root(gang[i]); 431365d33fd7SQu Wenruo } 431465d33fd7SQu Wenruo return err; 4315c146afadSYan Zheng } 4316c146afadSYan Zheng 43176bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4318c146afadSYan Zheng { 43196bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4320c146afadSYan Zheng struct btrfs_trans_handle *trans; 4321c146afadSYan Zheng 43220b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 43232ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 43240b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 43250b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4326c71bf099SYan, Zheng 4327c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 43280b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 43290b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4330c71bf099SYan, Zheng 43317a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 43323612b495STsutomu Itoh if (IS_ERR(trans)) 43333612b495STsutomu Itoh return PTR_ERR(trans); 43343a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4335c146afadSYan Zheng } 4336c146afadSYan Zheng 4337b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4338eb60ceacSChris Mason { 4339c146afadSYan Zheng int ret; 4340e089f05cSChris Mason 4341afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 4342d6fd0ae2SOmar Sandoval /* 4343d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4344d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4345d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4346d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4347d6fd0ae2SOmar Sandoval */ 4348d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4349a2135011SChris Mason 43507343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4351d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 43527343dd61SJustin Maggard 4353803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4354803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4355803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4356803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4357803b2f54SStefan Behrens 4358837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4359aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4360837d5b6eSIlya Dryomov 43618dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 43628dabb742SStefan Behrens 4363aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 43644cb5300bSChris Mason 43654cb5300bSChris Mason /* wait for any defraggers to finish */ 43664cb5300bSChris Mason wait_event(fs_info->transaction_wait, 43674cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 43684cb5300bSChris Mason 43694cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 437026176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 43714cb5300bSChris Mason 437221c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 437357056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4374576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 437521c7e756SMiao Xie 437618bb8bbfSJohannes Thumshirn cancel_work_sync(&fs_info->reclaim_bgs_work); 437718bb8bbfSJohannes Thumshirn 4378b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4379b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4380b0643e59SDennis Zhou 4381bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4382e44163e1SJeff Mahoney /* 4383d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4384d6fd0ae2SOmar Sandoval * unused block groups. 4385e44163e1SJeff Mahoney */ 43860b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4387e44163e1SJeff Mahoney 4388f0cc2cd7SFilipe Manana /* 4389f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4390f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4391f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4392f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4393f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4394f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4395f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4396f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4397f0cc2cd7SFilipe Manana * This is a very rare case. 4398f0cc2cd7SFilipe Manana */ 4399f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4400f0cc2cd7SFilipe Manana 44016bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4402d397712bSChris Mason if (ret) 440304892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4404c146afadSYan Zheng } 4405ed2ff2cbSChris Mason 440684961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 44072ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4408acce952bSliubo 4409e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4410e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 44118929ecfaSYan, Zheng 4412e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4413afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4414f25784b3SYan Zheng 44155958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 44165958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 44175958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 44185958253cSQu Wenruo } 44195958253cSQu Wenruo 442004892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4421fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4422bcef60f2SArne Jansen 4423963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 442404892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4425963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4426b0c68f8bSChris Mason } 442731153d81SYan Zheng 44285deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 44294297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 44305deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 44314297ff84SJosef Bacik 44326618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4433b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 44345ac1d209SJeff Mahoney 44351a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 44361a4319ccSLiu Bo 4437de348ee0SWang Shilong /* 4438de348ee0SWang Shilong * we must make sure there is not any read request to 4439de348ee0SWang Shilong * submit after we stopping all workers. 4440de348ee0SWang Shilong */ 4441de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 444296192499SJosef Bacik btrfs_stop_all_workers(fs_info); 444396192499SJosef Bacik 44440a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 44450a31daa4SFilipe Manana ASSERT(list_empty(&fs_info->trans_list)); 44460a31daa4SFilipe Manana 4447afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 44484273eaffSAnand Jain free_root_pointers(fs_info, true); 44498c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 44509ad6b7bcSChris Mason 44514e19443dSJosef Bacik /* 44524e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 44534e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 44544e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 44554e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 44564e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 44574e19443dSJosef Bacik */ 44584e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 44594e19443dSJosef Bacik 446013e6c37bSJosef Bacik iput(fs_info->btree_inode); 4461d6bfde87SChris Mason 446221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 44630b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 44642ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 446521adbd5cSStefan Behrens #endif 446621adbd5cSStefan Behrens 44670b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 446868c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4469eb60ceacSChris Mason } 4470eb60ceacSChris Mason 4471b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4472b9fab919SChris Mason int atomic) 4473ccd467d6SChris Mason { 44741259ab75SChris Mason int ret; 4475727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 44761259ab75SChris Mason 44770b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 44781259ab75SChris Mason if (!ret) 44791259ab75SChris Mason return ret; 44801259ab75SChris Mason 44811259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4482b9fab919SChris Mason parent_transid, atomic); 4483b9fab919SChris Mason if (ret == -EAGAIN) 4484b9fab919SChris Mason return ret; 44851259ab75SChris Mason return !ret; 44865f39d397SChris Mason } 44876702ed49SChris Mason 44885f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 44895f39d397SChris Mason { 44902f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 44915f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4492b9473439SChris Mason int was_dirty; 4493b4ce94deSChris Mason 449406ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 449506ea65a3SJosef Bacik /* 449606ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 449752042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 449806ea65a3SJosef Bacik * outside of the sanity tests. 449906ea65a3SJosef Bacik */ 4500b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 450106ea65a3SJosef Bacik return; 450206ea65a3SJosef Bacik #endif 450349d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 45040b246afaSJeff Mahoney if (transid != fs_info->generation) 45055d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 45060b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 45070b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4508e2d84521SMiao Xie if (!was_dirty) 4509104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4510e2d84521SMiao Xie buf->len, 45110b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 45121f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 451369fc6cbbSQu Wenruo /* 451469fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 451569fc6cbbSQu Wenruo * but item data not updated. 451669fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 451769fc6cbbSQu Wenruo */ 451869fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4519cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4520a4f78750SDavid Sterba btrfs_print_leaf(buf); 45211f21ef0aSFilipe Manana ASSERT(0); 45221f21ef0aSFilipe Manana } 45231f21ef0aSFilipe Manana #endif 4524eb60ceacSChris Mason } 4525eb60ceacSChris Mason 45262ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4527b53d3f5dSLiu Bo int flush_delayed) 452835b7e476SChris Mason { 4529188de649SChris Mason /* 4530188de649SChris Mason * looks as though older kernels can get into trouble with 4531188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4532188de649SChris Mason */ 4533e2d84521SMiao Xie int ret; 4534d6bfde87SChris Mason 45356933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4536d6bfde87SChris Mason return; 4537d6bfde87SChris Mason 4538b53d3f5dSLiu Bo if (flush_delayed) 45392ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 454016cdcec7SMiao Xie 4541d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4542d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4543d814a491SEthan Lien fs_info->dirty_metadata_batch); 4544e2d84521SMiao Xie if (ret > 0) { 45450b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 454616cdcec7SMiao Xie } 454716cdcec7SMiao Xie } 454816cdcec7SMiao Xie 45492ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 455016cdcec7SMiao Xie { 45512ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 455235b7e476SChris Mason } 4553b53d3f5dSLiu Bo 45542ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4555b53d3f5dSLiu Bo { 45562ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4557d6bfde87SChris Mason } 45586b80053dSChris Mason 4559581c1760SQu Wenruo int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, 4560581c1760SQu Wenruo struct btrfs_key *first_key) 45616b80053dSChris Mason { 45625ab12d1fSDavid Sterba return btree_read_extent_buffer_pages(buf, parent_transid, 4563581c1760SQu Wenruo level, first_key); 45646b80053dSChris Mason } 45650da5468fSChris Mason 45662ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4567acce952bSliubo { 4568fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4569fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4570fe816d0fSNikolay Borisov 45710b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45722ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45730b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4574acce952bSliubo 45750b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45760b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4577acce952bSliubo } 4578acce952bSliubo 4579ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4580ef67963dSJosef Bacik { 4581ef67963dSJosef Bacik struct btrfs_root *gang[8]; 4582ef67963dSJosef Bacik u64 root_objectid = 0; 4583ef67963dSJosef Bacik int ret; 4584ef67963dSJosef Bacik 4585ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4586ef67963dSJosef Bacik while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4587ef67963dSJosef Bacik (void **)gang, root_objectid, 4588ef67963dSJosef Bacik ARRAY_SIZE(gang))) != 0) { 4589ef67963dSJosef Bacik int i; 4590ef67963dSJosef Bacik 4591ef67963dSJosef Bacik for (i = 0; i < ret; i++) 4592ef67963dSJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 4593ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4594ef67963dSJosef Bacik 4595ef67963dSJosef Bacik for (i = 0; i < ret; i++) { 4596ef67963dSJosef Bacik if (!gang[i]) 4597ef67963dSJosef Bacik continue; 4598ef67963dSJosef Bacik root_objectid = gang[i]->root_key.objectid; 4599ef67963dSJosef Bacik btrfs_free_log(NULL, gang[i]); 4600ef67963dSJosef Bacik btrfs_put_root(gang[i]); 4601ef67963dSJosef Bacik } 4602ef67963dSJosef Bacik root_objectid++; 4603ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4604ef67963dSJosef Bacik } 4605ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4606ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4607ef67963dSJosef Bacik } 4608ef67963dSJosef Bacik 4609143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4610acce952bSliubo { 4611acce952bSliubo struct btrfs_ordered_extent *ordered; 4612acce952bSliubo 4613199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4614779880efSJosef Bacik /* 4615779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4616779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4617779880efSJosef Bacik */ 4618199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4619779880efSJosef Bacik root_extent_list) 4620779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4621199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4622199c2a9cSMiao Xie } 4623199c2a9cSMiao Xie 4624199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4625199c2a9cSMiao Xie { 4626199c2a9cSMiao Xie struct btrfs_root *root; 4627199c2a9cSMiao Xie struct list_head splice; 4628199c2a9cSMiao Xie 4629199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4630199c2a9cSMiao Xie 4631199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4632199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4633199c2a9cSMiao Xie while (!list_empty(&splice)) { 4634199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4635199c2a9cSMiao Xie ordered_root); 46361de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 46371de2cfdeSJosef Bacik &fs_info->ordered_roots); 4638199c2a9cSMiao Xie 46392a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4640199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4641199c2a9cSMiao Xie 46422a85d9caSLiu Bo cond_resched(); 46432a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4644199c2a9cSMiao Xie } 4645199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 464674d5d229SJosef Bacik 464774d5d229SJosef Bacik /* 464874d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 464974d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 465074d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 465174d5d229SJosef Bacik * actually get run and error out properly. 465274d5d229SJosef Bacik */ 465374d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4654acce952bSliubo } 4655acce952bSliubo 465635a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 46572ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4658acce952bSliubo { 4659acce952bSliubo struct rb_node *node; 4660acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4661acce952bSliubo struct btrfs_delayed_ref_node *ref; 4662acce952bSliubo int ret = 0; 4663acce952bSliubo 4664acce952bSliubo delayed_refs = &trans->delayed_refs; 4665acce952bSliubo 4666acce952bSliubo spin_lock(&delayed_refs->lock); 4667d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4668cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4669b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4670acce952bSliubo return ret; 4671acce952bSliubo } 4672acce952bSliubo 46735c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4674d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 46750e0adbcfSJosef Bacik struct rb_node *n; 4676e78417d1SJosef Bacik bool pin_bytes = false; 4677acce952bSliubo 4678d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4679d7df2c79SJosef Bacik href_node); 46803069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4681b939d1abSJosef Bacik continue; 46823069bd26SJosef Bacik 4683d7df2c79SJosef Bacik spin_lock(&head->lock); 4684e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 46850e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 46860e0adbcfSJosef Bacik ref_node); 4687d7df2c79SJosef Bacik ref->in_tree = 0; 4688e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 46890e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 46901d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 46911d57ee94SWang Xiaoguang list_del(&ref->add_list); 4692d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4693d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4694d7df2c79SJosef Bacik } 469554067ae9SJosef Bacik if (head->must_insert_reserved) 4696e78417d1SJosef Bacik pin_bytes = true; 469778a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4698fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4699d7df2c79SJosef Bacik spin_unlock(&head->lock); 4700acce952bSliubo spin_unlock(&delayed_refs->lock); 4701e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4702acce952bSliubo 4703f603bb94SNikolay Borisov if (pin_bytes) { 4704f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4705f603bb94SNikolay Borisov 4706f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4707f603bb94SNikolay Borisov BUG_ON(!cache); 4708f603bb94SNikolay Borisov 4709f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4710f603bb94SNikolay Borisov spin_lock(&cache->lock); 4711f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4712f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4713f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4714f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4715f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4716f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4717f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4718f603bb94SNikolay Borisov 4719f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4720f603bb94SNikolay Borisov 4721f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4722f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4723f603bb94SNikolay Borisov } 472431890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4725d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4726acce952bSliubo cond_resched(); 4727acce952bSliubo spin_lock(&delayed_refs->lock); 4728acce952bSliubo } 472981f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4730acce952bSliubo 4731acce952bSliubo spin_unlock(&delayed_refs->lock); 4732acce952bSliubo 4733acce952bSliubo return ret; 4734acce952bSliubo } 4735acce952bSliubo 4736143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4737acce952bSliubo { 4738acce952bSliubo struct btrfs_inode *btrfs_inode; 4739acce952bSliubo struct list_head splice; 4740acce952bSliubo 4741acce952bSliubo INIT_LIST_HEAD(&splice); 4742acce952bSliubo 4743eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4744eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4745acce952bSliubo 4746acce952bSliubo while (!list_empty(&splice)) { 4747fe816d0fSNikolay Borisov struct inode *inode = NULL; 4748eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4749acce952bSliubo delalloc_inodes); 4750fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4751eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4752acce952bSliubo 4753fe816d0fSNikolay Borisov /* 4754fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4755fe816d0fSNikolay Borisov * meanwhile. 4756fe816d0fSNikolay Borisov */ 4757fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4758fe816d0fSNikolay Borisov if (inode) { 4759fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4760fe816d0fSNikolay Borisov iput(inode); 4761fe816d0fSNikolay Borisov } 4762eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4763acce952bSliubo } 4764eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4765eb73c1b7SMiao Xie } 4766eb73c1b7SMiao Xie 4767eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4768eb73c1b7SMiao Xie { 4769eb73c1b7SMiao Xie struct btrfs_root *root; 4770eb73c1b7SMiao Xie struct list_head splice; 4771eb73c1b7SMiao Xie 4772eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4773eb73c1b7SMiao Xie 4774eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4775eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4776eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4777eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4778eb73c1b7SMiao Xie delalloc_root); 477900246528SJosef Bacik root = btrfs_grab_root(root); 4780eb73c1b7SMiao Xie BUG_ON(!root); 4781eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4782eb73c1b7SMiao Xie 4783eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 478400246528SJosef Bacik btrfs_put_root(root); 4785eb73c1b7SMiao Xie 4786eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4787eb73c1b7SMiao Xie } 4788eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4789acce952bSliubo } 4790acce952bSliubo 47912ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4792acce952bSliubo struct extent_io_tree *dirty_pages, 4793acce952bSliubo int mark) 4794acce952bSliubo { 4795acce952bSliubo int ret; 4796acce952bSliubo struct extent_buffer *eb; 4797acce952bSliubo u64 start = 0; 4798acce952bSliubo u64 end; 4799acce952bSliubo 4800acce952bSliubo while (1) { 4801acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4802e6138876SJosef Bacik mark, NULL); 4803acce952bSliubo if (ret) 4804acce952bSliubo break; 4805acce952bSliubo 480691166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4807acce952bSliubo while (start <= end) { 48080b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 48090b246afaSJeff Mahoney start += fs_info->nodesize; 4810fd8b2b61SJosef Bacik if (!eb) 4811acce952bSliubo continue; 4812fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4813acce952bSliubo 4814fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4815fd8b2b61SJosef Bacik &eb->bflags)) 4816fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4817fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4818acce952bSliubo } 4819acce952bSliubo } 4820acce952bSliubo 4821acce952bSliubo return ret; 4822acce952bSliubo } 4823acce952bSliubo 48242ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4825fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4826acce952bSliubo { 4827acce952bSliubo u64 start; 4828acce952bSliubo u64 end; 4829acce952bSliubo int ret; 4830acce952bSliubo 4831acce952bSliubo while (1) { 48320e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 48330e6ec385SFilipe Manana 4834fcd5e742SLu Fengqi /* 4835fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4836fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4837fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4838fcd5e742SLu Fengqi * the same extent range. 4839fcd5e742SLu Fengqi */ 4840fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4841acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 48420e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4843fcd5e742SLu Fengqi if (ret) { 4844fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4845acce952bSliubo break; 4846fcd5e742SLu Fengqi } 4847acce952bSliubo 48480e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 48490e6ec385SFilipe Manana free_extent_state(cached_state); 48502ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4851fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4852acce952bSliubo cond_resched(); 4853acce952bSliubo } 4854acce952bSliubo 4855acce952bSliubo return 0; 4856acce952bSliubo } 4857acce952bSliubo 485832da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4859c79a1751SLiu Bo { 4860c79a1751SLiu Bo struct inode *inode; 4861c79a1751SLiu Bo 4862c79a1751SLiu Bo inode = cache->io_ctl.inode; 4863c79a1751SLiu Bo if (inode) { 4864c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4865c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4866c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4867c79a1751SLiu Bo iput(inode); 4868c79a1751SLiu Bo } 4869bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4870c79a1751SLiu Bo btrfs_put_block_group(cache); 4871c79a1751SLiu Bo } 4872c79a1751SLiu Bo 4873c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 48742ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4875c79a1751SLiu Bo { 487632da5386SDavid Sterba struct btrfs_block_group *cache; 4877c79a1751SLiu Bo 4878c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4879c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4880c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 488132da5386SDavid Sterba struct btrfs_block_group, 4882c79a1751SLiu Bo dirty_list); 4883c79a1751SLiu Bo 4884c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4885c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4886c79a1751SLiu Bo list_del_init(&cache->io_list); 4887c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4888c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4889c79a1751SLiu Bo } 4890c79a1751SLiu Bo 4891c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4892c79a1751SLiu Bo spin_lock(&cache->lock); 4893c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4894c79a1751SLiu Bo spin_unlock(&cache->lock); 4895c79a1751SLiu Bo 4896c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4897c79a1751SLiu Bo btrfs_put_block_group(cache); 4898ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4899c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4900c79a1751SLiu Bo } 4901c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4902c79a1751SLiu Bo 490345ae2c18SNikolay Borisov /* 490445ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 490545ae2c18SNikolay Borisov * to use it without any locking 490645ae2c18SNikolay Borisov */ 4907c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4908c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 490932da5386SDavid Sterba struct btrfs_block_group, 4910c79a1751SLiu Bo io_list); 4911c79a1751SLiu Bo 4912c79a1751SLiu Bo list_del_init(&cache->io_list); 4913c79a1751SLiu Bo spin_lock(&cache->lock); 4914c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4915c79a1751SLiu Bo spin_unlock(&cache->lock); 4916c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4917c79a1751SLiu Bo } 4918c79a1751SLiu Bo } 4919c79a1751SLiu Bo 492049b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 49212ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 492249b25e05SJeff Mahoney { 4923bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4924bbbf7243SNikolay Borisov 49252ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4926c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4927c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4928c79a1751SLiu Bo 4929bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4930bbbf7243SNikolay Borisov post_commit_list) { 4931bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4932bbbf7243SNikolay Borisov } 4933bbbf7243SNikolay Borisov 49342ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 493549b25e05SJeff Mahoney 49364a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 49370b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 493849b25e05SJeff Mahoney 49394a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 49400b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 494149b25e05SJeff Mahoney 4942ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 494367cde344SMiao Xie 49442ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 494549b25e05SJeff Mahoney EXTENT_DIRTY); 4946fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 494749b25e05SJeff Mahoney 4948d3575156SNaohiro Aota btrfs_free_redirty_list(cur_trans); 4949d3575156SNaohiro Aota 49504a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 49514a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 495249b25e05SJeff Mahoney } 495349b25e05SJeff Mahoney 49542ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4955acce952bSliubo { 4956acce952bSliubo struct btrfs_transaction *t; 4957acce952bSliubo 49580b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4959acce952bSliubo 49600b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49610b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 49620b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4963724e2315SJosef Bacik struct btrfs_transaction, list); 4964724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 49659b64f57dSElena Reshetova refcount_inc(&t->use_count); 49660b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49672ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4968724e2315SJosef Bacik btrfs_put_transaction(t); 49690b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4970724e2315SJosef Bacik continue; 4971724e2315SJosef Bacik } 49720b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4973724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 49740b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4975724e2315SJosef Bacik /* 4976724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4977724e2315SJosef Bacik * handle open currently for this transaction. 4978724e2315SJosef Bacik */ 4979724e2315SJosef Bacik wait_event(t->writer_wait, 4980724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4981724e2315SJosef Bacik } else { 49820b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4983724e2315SJosef Bacik } 49842ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4985724e2315SJosef Bacik 49860b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49870b246afaSJeff Mahoney if (t == fs_info->running_transaction) 49880b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4989724e2315SJosef Bacik list_del_init(&t->list); 49900b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4991a4abeea4SJosef Bacik 4992724e2315SJosef Bacik btrfs_put_transaction(t); 49932ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 49940b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4995724e2315SJosef Bacik } 49960b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49970b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4998ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4999ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 50000b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 5001ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 50020b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 5003acce952bSliubo 5004acce952bSliubo return 0; 5005acce952bSliubo } 5006ec7d6dfdSNikolay Borisov 5007453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 5008ec7d6dfdSNikolay Borisov { 5009ec7d6dfdSNikolay Borisov struct btrfs_path *path; 5010ec7d6dfdSNikolay Borisov int ret; 5011ec7d6dfdSNikolay Borisov struct extent_buffer *l; 5012ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 5013ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 5014ec7d6dfdSNikolay Borisov int slot; 5015ec7d6dfdSNikolay Borisov 5016ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 5017ec7d6dfdSNikolay Borisov if (!path) 5018ec7d6dfdSNikolay Borisov return -ENOMEM; 5019ec7d6dfdSNikolay Borisov 5020ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 5021ec7d6dfdSNikolay Borisov search_key.type = -1; 5022ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 5023ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 5024ec7d6dfdSNikolay Borisov if (ret < 0) 5025ec7d6dfdSNikolay Borisov goto error; 5026ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 5027ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 5028ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 5029ec7d6dfdSNikolay Borisov l = path->nodes[0]; 5030ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 503123125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 503223125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 5033ec7d6dfdSNikolay Borisov } else { 503423125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 5035ec7d6dfdSNikolay Borisov } 5036ec7d6dfdSNikolay Borisov ret = 0; 5037ec7d6dfdSNikolay Borisov error: 5038ec7d6dfdSNikolay Borisov btrfs_free_path(path); 5039ec7d6dfdSNikolay Borisov return ret; 5040ec7d6dfdSNikolay Borisov } 5041ec7d6dfdSNikolay Borisov 5042543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 5043ec7d6dfdSNikolay Borisov { 5044ec7d6dfdSNikolay Borisov int ret; 5045ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5046ec7d6dfdSNikolay Borisov 50476b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5048ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5049ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5050ec7d6dfdSNikolay Borisov root->root_key.objectid); 5051ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5052ec7d6dfdSNikolay Borisov goto out; 5053ec7d6dfdSNikolay Borisov } 5054ec7d6dfdSNikolay Borisov 505523125104SNikolay Borisov *objectid = root->free_objectid++; 5056ec7d6dfdSNikolay Borisov ret = 0; 5057ec7d6dfdSNikolay Borisov out: 5058ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5059ec7d6dfdSNikolay Borisov return ret; 5060ec7d6dfdSNikolay Borisov } 5061