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> 8fc7cbcd4SDavid Sterba #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 */ 3776a2e9dc4SFilipe Manana int btrfs_read_extent_buffer(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 4443777369fSQu Wenruo if (ret < 0) 4453777369fSQu Wenruo goto error; 4463777369fSQu Wenruo 4473777369fSQu Wenruo /* 4483777369fSQu Wenruo * Also check the generation, the eb reached here must be newer than 4493777369fSQu Wenruo * last committed. Or something seriously wrong happened. 4503777369fSQu Wenruo */ 4513777369fSQu Wenruo if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { 4523777369fSQu Wenruo ret = -EUCLEAN; 453eca0f6f6SQu Wenruo btrfs_err(fs_info, 4543777369fSQu Wenruo "block=%llu bad generation, have %llu expect > %llu", 4553777369fSQu Wenruo eb->start, btrfs_header_generation(eb), 4563777369fSQu Wenruo fs_info->last_trans_committed); 4573777369fSQu Wenruo goto error; 458eca0f6f6SQu Wenruo } 459eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 460eca0f6f6SQu Wenruo 461eca0f6f6SQu Wenruo return 0; 4623777369fSQu Wenruo 4633777369fSQu Wenruo error: 4643777369fSQu Wenruo btrfs_print_tree(eb, 0); 4653777369fSQu Wenruo btrfs_err(fs_info, "block=%llu write time tree block corruption detected", 4663777369fSQu Wenruo eb->start); 4673777369fSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 4683777369fSQu Wenruo return ret; 469eca0f6f6SQu Wenruo } 470eca0f6f6SQu Wenruo 471eca0f6f6SQu Wenruo /* Checksum all dirty extent buffers in one bio_vec */ 472eca0f6f6SQu Wenruo static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, 473eca0f6f6SQu Wenruo struct bio_vec *bvec) 474eca0f6f6SQu Wenruo { 475eca0f6f6SQu Wenruo struct page *page = bvec->bv_page; 476eca0f6f6SQu Wenruo u64 bvec_start = page_offset(page) + bvec->bv_offset; 477eca0f6f6SQu Wenruo u64 cur; 478eca0f6f6SQu Wenruo int ret = 0; 479eca0f6f6SQu Wenruo 480eca0f6f6SQu Wenruo for (cur = bvec_start; cur < bvec_start + bvec->bv_len; 481eca0f6f6SQu Wenruo cur += fs_info->nodesize) { 482eca0f6f6SQu Wenruo struct extent_buffer *eb; 483eca0f6f6SQu Wenruo bool uptodate; 484eca0f6f6SQu Wenruo 485eca0f6f6SQu Wenruo eb = find_extent_buffer(fs_info, cur); 486eca0f6f6SQu Wenruo uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, 487eca0f6f6SQu Wenruo fs_info->nodesize); 488eca0f6f6SQu Wenruo 48901cd3909SDavid Sterba /* A dirty eb shouldn't disappear from buffer_radix */ 490eca0f6f6SQu Wenruo if (WARN_ON(!eb)) 491eca0f6f6SQu Wenruo return -EUCLEAN; 492eca0f6f6SQu Wenruo 493eca0f6f6SQu Wenruo if (WARN_ON(cur != btrfs_header_bytenr(eb))) { 494eca0f6f6SQu Wenruo free_extent_buffer(eb); 495eca0f6f6SQu Wenruo return -EUCLEAN; 496eca0f6f6SQu Wenruo } 497eca0f6f6SQu Wenruo if (WARN_ON(!uptodate)) { 498eca0f6f6SQu Wenruo free_extent_buffer(eb); 499eca0f6f6SQu Wenruo return -EUCLEAN; 500eca0f6f6SQu Wenruo } 501eca0f6f6SQu Wenruo 502eca0f6f6SQu Wenruo ret = csum_one_extent_buffer(eb); 503eca0f6f6SQu Wenruo free_extent_buffer(eb); 504eca0f6f6SQu Wenruo if (ret < 0) 505eca0f6f6SQu Wenruo return ret; 506eca0f6f6SQu Wenruo } 507eca0f6f6SQu Wenruo return ret; 508eca0f6f6SQu Wenruo } 509eca0f6f6SQu Wenruo 510d352ac68SChris Mason /* 511ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 512ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 513ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 514d352ac68SChris Mason */ 515ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 51619c00ddcSChris Mason { 517ac303b69SQu Wenruo struct page *page = bvec->bv_page; 5184eee4fa4SMiao Xie u64 start = page_offset(page); 51919c00ddcSChris Mason u64 found_start; 52019c00ddcSChris Mason struct extent_buffer *eb; 521eca0f6f6SQu Wenruo 522fbca46ebSQu Wenruo if (fs_info->nodesize < PAGE_SIZE) 523eca0f6f6SQu Wenruo return csum_dirty_subpage_buffers(fs_info, bvec); 524f188591eSChris Mason 5254f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5264f2de97aSJosef Bacik if (page != eb->pages[0]) 5274f2de97aSJosef Bacik return 0; 5280f805531SAlex Lyakas 52919c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 530d3575156SNaohiro Aota 531d3575156SNaohiro Aota if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 532d3575156SNaohiro Aota WARN_ON(found_start != 0); 533d3575156SNaohiro Aota return 0; 534d3575156SNaohiro Aota } 535d3575156SNaohiro Aota 5360f805531SAlex Lyakas /* 5370f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5380f805531SAlex Lyakas * It is useful to know what went wrong. 5390f805531SAlex Lyakas */ 5400f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5410f805531SAlex Lyakas return -EUCLEAN; 5420f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5430f805531SAlex Lyakas return -EUCLEAN; 5440f805531SAlex Lyakas 545eca0f6f6SQu Wenruo return csum_one_extent_buffer(eb); 54619c00ddcSChris Mason } 54719c00ddcSChris Mason 548b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 5492b82032cSYan Zheng { 550b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 551944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 55244880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 553944d3f9fSNikolay Borisov u8 *metadata_uuid; 5542b82032cSYan Zheng 5559a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 5569a8658e3SDavid Sterba BTRFS_FSID_SIZE); 5577239ff4bSNikolay Borisov /* 558944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 559944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 560944d3f9fSNikolay Borisov * ->fsid in this case is fine 5617239ff4bSNikolay Borisov */ 562944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 5637239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 5647239ff4bSNikolay Borisov else 5657239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 5667239ff4bSNikolay Borisov 567944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 568944d3f9fSNikolay Borisov return 0; 569944d3f9fSNikolay Borisov 570944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 571944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 572944d3f9fSNikolay Borisov return 0; 573944d3f9fSNikolay Borisov 574944d3f9fSNikolay Borisov return 1; 5752b82032cSYan Zheng } 5762b82032cSYan Zheng 57777bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 57877bf40a2SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb) 579ce9adaa5SChris Mason { 58077bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 581ce9adaa5SChris Mason u64 found_start; 58277bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 58377bf40a2SQu Wenruo u8 found_level; 5842996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 585dfd29eedSDavid Sterba const u8 *header_csum; 58677bf40a2SQu Wenruo int ret = 0; 587ea466794SJosef Bacik 588ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 589727011e0SChris Mason if (found_start != eb->start) { 590893bf4b1SSu Yue btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", 591893bf4b1SSu Yue eb->start, found_start); 592f188591eSChris Mason ret = -EIO; 59377bf40a2SQu Wenruo goto out; 594ce9adaa5SChris Mason } 595b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 59602873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 59794647322SDavid Sterba eb->start); 5981259ab75SChris Mason ret = -EIO; 59977bf40a2SQu Wenruo goto out; 6001259ab75SChris Mason } 601ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6021c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 603893bf4b1SSu Yue btrfs_err(fs_info, "bad tree block level %d on %llu", 604893bf4b1SSu Yue (int)btrfs_header_level(eb), eb->start); 6051c24c3ceSJosef Bacik ret = -EIO; 60677bf40a2SQu Wenruo goto out; 6071c24c3ceSJosef Bacik } 608ce9adaa5SChris Mason 609c67b3892SDavid Sterba csum_tree_block(eb, result); 610dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 611dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 612a826d6dcSJosef Bacik 613dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 6142996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 615ff14aa79SDavid Sterba "checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", 616ff14aa79SDavid Sterba eb->start, 617dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 61835be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 61935be8851SJohannes Thumshirn btrfs_header_level(eb)); 6202996e1f8SJohannes Thumshirn ret = -EUCLEAN; 62177bf40a2SQu Wenruo goto out; 6222996e1f8SJohannes Thumshirn } 6232996e1f8SJohannes Thumshirn 624a826d6dcSJosef Bacik /* 625a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 626a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 627a826d6dcSJosef Bacik * return -EIO. 628a826d6dcSJosef Bacik */ 6291c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 630a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 631a826d6dcSJosef Bacik ret = -EIO; 632a826d6dcSJosef Bacik } 633ce9adaa5SChris Mason 634813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 635053ab70fSLiu Bo ret = -EIO; 636053ab70fSLiu Bo 6370b32f4bbSJosef Bacik if (!ret) 6380b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 63975391f0dSQu Wenruo else 64075391f0dSQu Wenruo btrfs_err(fs_info, 64175391f0dSQu Wenruo "block=%llu read time tree block corruption detected", 64275391f0dSQu Wenruo eb->start); 64377bf40a2SQu Wenruo out: 64477bf40a2SQu Wenruo return ret; 64577bf40a2SQu Wenruo } 64677bf40a2SQu Wenruo 647371cdc07SQu Wenruo static int validate_subpage_buffer(struct page *page, u64 start, u64 end, 648371cdc07SQu Wenruo int mirror) 649371cdc07SQu Wenruo { 650371cdc07SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 651371cdc07SQu Wenruo struct extent_buffer *eb; 652371cdc07SQu Wenruo bool reads_done; 653371cdc07SQu Wenruo int ret = 0; 654371cdc07SQu Wenruo 655371cdc07SQu Wenruo /* 656371cdc07SQu Wenruo * We don't allow bio merge for subpage metadata read, so we should 657371cdc07SQu Wenruo * only get one eb for each endio hook. 658371cdc07SQu Wenruo */ 659371cdc07SQu Wenruo ASSERT(end == start + fs_info->nodesize - 1); 660371cdc07SQu Wenruo ASSERT(PagePrivate(page)); 661371cdc07SQu Wenruo 662371cdc07SQu Wenruo eb = find_extent_buffer(fs_info, start); 663371cdc07SQu Wenruo /* 664371cdc07SQu Wenruo * When we are reading one tree block, eb must have been inserted into 665371cdc07SQu Wenruo * the radix tree. If not, something is wrong. 666371cdc07SQu Wenruo */ 667371cdc07SQu Wenruo ASSERT(eb); 668371cdc07SQu Wenruo 669371cdc07SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 670371cdc07SQu Wenruo /* Subpage read must finish in page read */ 671371cdc07SQu Wenruo ASSERT(reads_done); 672371cdc07SQu Wenruo 673371cdc07SQu Wenruo eb->read_mirror = mirror; 674371cdc07SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 675371cdc07SQu Wenruo ret = -EIO; 676371cdc07SQu Wenruo goto err; 677371cdc07SQu Wenruo } 678371cdc07SQu Wenruo ret = validate_extent_buffer(eb); 679371cdc07SQu Wenruo if (ret < 0) 680371cdc07SQu Wenruo goto err; 681371cdc07SQu Wenruo 682371cdc07SQu Wenruo set_extent_buffer_uptodate(eb); 683371cdc07SQu Wenruo 684371cdc07SQu Wenruo free_extent_buffer(eb); 685371cdc07SQu Wenruo return ret; 686371cdc07SQu Wenruo err: 687371cdc07SQu Wenruo /* 688371cdc07SQu Wenruo * end_bio_extent_readpage decrements io_pages in case of error, 689371cdc07SQu Wenruo * make sure it has something to decrement. 690371cdc07SQu Wenruo */ 691371cdc07SQu Wenruo atomic_inc(&eb->io_pages); 692371cdc07SQu Wenruo clear_extent_buffer_uptodate(eb); 693371cdc07SQu Wenruo free_extent_buffer(eb); 694371cdc07SQu Wenruo return ret; 695371cdc07SQu Wenruo } 696371cdc07SQu Wenruo 697c3a3b19bSQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, 69877bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 69977bf40a2SQu Wenruo int mirror) 70077bf40a2SQu Wenruo { 70177bf40a2SQu Wenruo struct extent_buffer *eb; 70277bf40a2SQu Wenruo int ret = 0; 70377bf40a2SQu Wenruo int reads_done; 70477bf40a2SQu Wenruo 70577bf40a2SQu Wenruo ASSERT(page->private); 706371cdc07SQu Wenruo 707fbca46ebSQu Wenruo if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) 708371cdc07SQu Wenruo return validate_subpage_buffer(page, start, end, mirror); 709371cdc07SQu Wenruo 71077bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 71177bf40a2SQu Wenruo 71277bf40a2SQu Wenruo /* 71377bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 71477bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 71577bf40a2SQu Wenruo */ 71677bf40a2SQu Wenruo atomic_inc(&eb->refs); 71777bf40a2SQu Wenruo 71877bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 71977bf40a2SQu Wenruo if (!reads_done) 72077bf40a2SQu Wenruo goto err; 72177bf40a2SQu Wenruo 72277bf40a2SQu Wenruo eb->read_mirror = mirror; 72377bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 72477bf40a2SQu Wenruo ret = -EIO; 72577bf40a2SQu Wenruo goto err; 72677bf40a2SQu Wenruo } 72777bf40a2SQu Wenruo ret = validate_extent_buffer(eb); 728ce9adaa5SChris Mason err: 72953b381b3SDavid Woodhouse if (ret) { 73053b381b3SDavid Woodhouse /* 73153b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 73253b381b3SDavid Woodhouse * again, we have to make sure it has something 73353b381b3SDavid Woodhouse * to decrement 73453b381b3SDavid Woodhouse */ 73553b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 7360b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 73753b381b3SDavid Woodhouse } 7380b32f4bbSJosef Bacik free_extent_buffer(eb); 73977bf40a2SQu Wenruo 740f188591eSChris Mason return ret; 741ce9adaa5SChris Mason } 742ce9adaa5SChris Mason 7434246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 744ce9adaa5SChris Mason { 74597eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 746ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 7479e0af237SLiu Bo struct btrfs_workqueue *wq; 748ce9adaa5SChris Mason 749ce9adaa5SChris Mason fs_info = end_io_wq->info; 7504e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 751d20f7043SChris Mason 752cfe94440SNaohiro Aota if (btrfs_op(bio) == BTRFS_MAP_WRITE) { 753a0cac0ecSOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) 7549e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 755a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) 7569e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 757a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 7589e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 759a0cac0ecSOmar Sandoval else 7609e0af237SLiu Bo wq = fs_info->endio_write_workers; 7619e0af237SLiu Bo } else { 7625c047a69SOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 7639e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 764a0cac0ecSOmar Sandoval else if (end_io_wq->metadata) 7659e0af237SLiu Bo wq = fs_info->endio_meta_workers; 766a0cac0ecSOmar Sandoval else 7679e0af237SLiu Bo wq = fs_info->endio_workers; 7689e0af237SLiu Bo } 7699e0af237SLiu Bo 770a0cac0ecSOmar Sandoval btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); 7719e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 772ce9adaa5SChris Mason } 773ce9adaa5SChris Mason 7744e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 775bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7760b86a832SChris Mason { 77797eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7788b110e39SMiao Xie 77997eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 780ce9adaa5SChris Mason if (!end_io_wq) 7814e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 782ce9adaa5SChris Mason 783ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 784ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 78522c59948SChris Mason end_io_wq->info = info; 7864e4cbee9SChristoph Hellwig end_io_wq->status = 0; 787ce9adaa5SChris Mason end_io_wq->bio = bio; 78822c59948SChris Mason end_io_wq->metadata = metadata; 789ce9adaa5SChris Mason 790ce9adaa5SChris Mason bio->bi_private = end_io_wq; 791ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 79222c59948SChris Mason return 0; 79322c59948SChris Mason } 79422c59948SChris Mason 7954a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7964a69a410SChris Mason { 7974a69a410SChris Mason struct async_submit_bio *async; 7984e4cbee9SChristoph Hellwig blk_status_t ret; 7994a69a410SChris Mason 8004a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 8011941b64bSQu Wenruo ret = async->submit_bio_start(async->inode, async->bio, 8021941b64bSQu Wenruo async->dio_file_offset); 80379787eaaSJeff Mahoney if (ret) 8044e4cbee9SChristoph Hellwig async->status = ret; 8054a69a410SChris Mason } 8064a69a410SChris Mason 80706ea01b1SDavid Sterba /* 80806ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 80906ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 81006ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 81106ea01b1SDavid Sterba * 81206ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 81306ea01b1SDavid Sterba * inserted into the tree. 81406ea01b1SDavid Sterba */ 8154a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 8168b712842SChris Mason { 8178b712842SChris Mason struct async_submit_bio *async; 81806ea01b1SDavid Sterba struct inode *inode; 81906ea01b1SDavid Sterba blk_status_t ret; 8208b712842SChris Mason 8218b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 8228896a08dSQu Wenruo inode = async->inode; 8234854ddd0SChris Mason 824bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 8254e4cbee9SChristoph Hellwig if (async->status) { 8264e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 8274246a0b6SChristoph Hellwig bio_endio(async->bio); 82879787eaaSJeff Mahoney return; 82979787eaaSJeff Mahoney } 83079787eaaSJeff Mahoney 831ec39f769SChris Mason /* 832ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 833ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 834ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 835ec39f769SChris Mason */ 836ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 83708635baeSChris Mason ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 83806ea01b1SDavid Sterba if (ret) { 83906ea01b1SDavid Sterba async->bio->bi_status = ret; 84006ea01b1SDavid Sterba bio_endio(async->bio); 84106ea01b1SDavid Sterba } 8424a69a410SChris Mason } 8434a69a410SChris Mason 8444a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 8454a69a410SChris Mason { 8464a69a410SChris Mason struct async_submit_bio *async; 8474a69a410SChris Mason 8484a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 8498b712842SChris Mason kfree(async); 8508b712842SChris Mason } 8518b712842SChris Mason 8528896a08dSQu Wenruo blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, 853a6f5e39eSDavid Sterba int mirror_num, u64 dio_file_offset, 854e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 85544b8bd7eSChris Mason { 8568896a08dSQu Wenruo struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 85744b8bd7eSChris Mason struct async_submit_bio *async; 85844b8bd7eSChris Mason 85944b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 86044b8bd7eSChris Mason if (!async) 8614e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 86244b8bd7eSChris Mason 8638896a08dSQu Wenruo async->inode = inode; 86444b8bd7eSChris Mason async->bio = bio; 86544b8bd7eSChris Mason async->mirror_num = mirror_num; 8664a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8674a69a410SChris Mason 868a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 869a0cac0ecSOmar Sandoval run_one_async_free); 8704a69a410SChris Mason 8711941b64bSQu Wenruo async->dio_file_offset = dio_file_offset; 8728c8bee1dSChris Mason 8734e4cbee9SChristoph Hellwig async->status = 0; 87479787eaaSJeff Mahoney 87567f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 876a31b4a43SChristoph Hellwig btrfs_queue_work(fs_info->hipri_workers, &async->work); 877a31b4a43SChristoph Hellwig else 8785cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 87944b8bd7eSChris Mason return 0; 88044b8bd7eSChris Mason } 88144b8bd7eSChris Mason 8824e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 883ce3ed71aSChris Mason { 8842c30c71bSKent Overstreet struct bio_vec *bvec; 885ce3ed71aSChris Mason struct btrfs_root *root; 8862b070cfeSChristoph Hellwig int ret = 0; 8876dc4f100SMing Lei struct bvec_iter_all iter_all; 888ce3ed71aSChris Mason 889c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 8902b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 891ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 892ac303b69SQu Wenruo ret = csum_dirty_buffer(root->fs_info, bvec); 89379787eaaSJeff Mahoney if (ret) 89479787eaaSJeff Mahoney break; 895ce3ed71aSChris Mason } 8962c30c71bSKent Overstreet 8974e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 898ce3ed71aSChris Mason } 899ce3ed71aSChris Mason 9008896a08dSQu Wenruo static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, 9011941b64bSQu Wenruo u64 dio_file_offset) 90222c59948SChris Mason { 9038b712842SChris Mason /* 9048b712842SChris Mason * when we're called for a write, we're already in the async 9055443be45SChris Mason * submission context. Just jump into btrfs_map_bio 9068b712842SChris Mason */ 90779787eaaSJeff Mahoney return btree_csum_one_bio(bio); 90822c59948SChris Mason } 90922c59948SChris Mason 910f4dcfb30SJohannes Thumshirn static bool should_async_write(struct btrfs_fs_info *fs_info, 9119b4e675aSDavid Sterba struct btrfs_inode *bi) 912de0022b9SJosef Bacik { 9134eef29efSNaohiro Aota if (btrfs_is_zoned(fs_info)) 914f4dcfb30SJohannes Thumshirn return false; 9156300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 916f4dcfb30SJohannes Thumshirn return false; 9179b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 918f4dcfb30SJohannes Thumshirn return false; 919f4dcfb30SJohannes Thumshirn return true; 920de0022b9SJosef Bacik } 921de0022b9SJosef Bacik 92294d9e11bSChristoph Hellwig void btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, int mirror_num) 92344b8bd7eSChris Mason { 9240b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 9254e4cbee9SChristoph Hellwig blk_status_t ret; 926cad321adSChris Mason 927cfe94440SNaohiro Aota if (btrfs_op(bio) != BTRFS_MAP_WRITE) { 928cad321adSChris Mason /* 929cad321adSChris Mason * called for a read, do the setup so that checksum validation 930cad321adSChris Mason * can happen in the async kernel threads 931cad321adSChris Mason */ 9320b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 9330b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 93494d9e11bSChristoph Hellwig if (!ret) 93508635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 936f4dcfb30SJohannes Thumshirn } else if (!should_async_write(fs_info, BTRFS_I(inode))) { 937de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 93894d9e11bSChristoph Hellwig if (!ret) 93908635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 94061891923SStefan Behrens } else { 941cad321adSChris Mason /* 94261891923SStefan Behrens * kthread helpers are used to submit writes so that 94361891923SStefan Behrens * checksumming can happen in parallel across all CPUs 944cad321adSChris Mason */ 9458896a08dSQu Wenruo ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, 946a6f5e39eSDavid Sterba btree_submit_bio_start); 94744b8bd7eSChris Mason } 94844b8bd7eSChris Mason 94994d9e11bSChristoph Hellwig if (ret) { 9504e4cbee9SChristoph Hellwig bio->bi_status = ret; 9514246a0b6SChristoph Hellwig bio_endio(bio); 95294d9e11bSChristoph Hellwig } 95361891923SStefan Behrens } 95461891923SStefan Behrens 9553dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 956784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 957a6bc32b8SMel Gorman struct page *newpage, struct page *page, 958a6bc32b8SMel Gorman enum migrate_mode mode) 959784b4e29SChris Mason { 960784b4e29SChris Mason /* 961784b4e29SChris Mason * we can't safely write a btree page from here, 962784b4e29SChris Mason * we haven't done the locking hook 963784b4e29SChris Mason */ 964784b4e29SChris Mason if (PageDirty(page)) 965784b4e29SChris Mason return -EAGAIN; 966784b4e29SChris Mason /* 967784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 968784b4e29SChris Mason * We must have no buffers or drop them. 969784b4e29SChris Mason */ 970784b4e29SChris Mason if (page_has_private(page) && 971784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 972784b4e29SChris Mason return -EAGAIN; 973a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 974784b4e29SChris Mason } 9753dd1462eSJan Beulich #endif 976784b4e29SChris Mason 9770da5468fSChris Mason 9780da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9790da5468fSChris Mason struct writeback_control *wbc) 9800da5468fSChris Mason { 981e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 982e2d84521SMiao Xie int ret; 983e2d84521SMiao Xie 984d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 985448d640bSChris Mason 986448d640bSChris Mason if (wbc->for_kupdate) 987448d640bSChris Mason return 0; 988448d640bSChris Mason 989e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 990b9473439SChris Mason /* this is a bit racy, but that's ok */ 991d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 992d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 993d814a491SEthan Lien fs_info->dirty_metadata_batch); 994e2d84521SMiao Xie if (ret < 0) 995793955bcSChris Mason return 0; 996793955bcSChris Mason } 9970b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9980da5468fSChris Mason } 9990da5468fSChris Mason 1000f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 10015f39d397SChris Mason { 1002f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 1003f913cff3SMatthew Wilcox (Oracle) return false; 10040c4e538bSDavid Sterba 1005f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 1006d98237b3SChris Mason } 1007d98237b3SChris Mason 1008895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 1009895586ebSMatthew Wilcox (Oracle) size_t length) 1010d98237b3SChris Mason { 1011d1310b2eSChris Mason struct extent_io_tree *tree; 1012895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 1013895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 1014f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 1015895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 1016895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 1017895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 1018895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 1019895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 10209ad6b7bcSChris Mason } 1021d98237b3SChris Mason } 1022d98237b3SChris Mason 1023bb146eb2SJosef Bacik #ifdef DEBUG 10240079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 10250079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 10260079c3b1SMatthew Wilcox (Oracle) { 10270079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 1028139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 10290b32f4bbSJosef Bacik struct extent_buffer *eb; 1030139e8cd3SQu Wenruo int cur_bit = 0; 10310079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 10320b32f4bbSJosef Bacik 1033139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 10340079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 10350b32f4bbSJosef Bacik BUG_ON(!eb); 10360b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10370b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 103849d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 10390079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 1040139e8cd3SQu Wenruo } 10410079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 1042139e8cd3SQu Wenruo 1043139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 1044139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 1045139e8cd3SQu Wenruo unsigned long flags; 1046139e8cd3SQu Wenruo u64 cur; 1047139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 1048139e8cd3SQu Wenruo 1049139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 1050139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 1051139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 1052139e8cd3SQu Wenruo cur_bit++; 1053139e8cd3SQu Wenruo continue; 1054139e8cd3SQu Wenruo } 1055139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 1056139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 1057139e8cd3SQu Wenruo 1058139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 1059139e8cd3SQu Wenruo ASSERT(eb); 1060139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 1061139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 106249d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 1063139e8cd3SQu Wenruo free_extent_buffer(eb); 1064139e8cd3SQu Wenruo 1065139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 1066139e8cd3SQu Wenruo } 10670079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 10680b32f4bbSJosef Bacik } 10690079c3b1SMatthew Wilcox (Oracle) #else 10700079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 10710079c3b1SMatthew Wilcox (Oracle) #endif 10720b32f4bbSJosef Bacik 10737f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 10740da5468fSChris Mason .writepages = btree_writepages, 1075f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 1076895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 10775a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1078784b4e29SChris Mason .migratepage = btree_migratepage, 10795a92bc88SChris Mason #endif 10800079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 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 11136a2e9dc4SFilipe Manana ret = btrfs_read_extent_buffer(buf, parent_transid, level, first_key); 11140f0fe8f7SFilipe David Borba Manana if (ret) { 1115537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 111664c043deSLiu Bo return ERR_PTR(ret); 11170f0fe8f7SFilipe David Borba Manana } 111888c602abSQu Wenruo if (btrfs_check_eb_owner(buf, owner_root)) { 111988c602abSQu Wenruo free_extent_buffer_stale(buf); 112088c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 112188c602abSQu Wenruo } 11225f39d397SChris Mason return buf; 1123ce9adaa5SChris Mason 1124eb60ceacSChris Mason } 1125eb60ceacSChris Mason 11266a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 1127ed2ff2cbSChris Mason { 11286a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 112955c69072SChris Mason if (btrfs_header_generation(buf) == 1130e2d84521SMiao Xie fs_info->running_transaction->transid) { 113149d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 1132b4ce94deSChris Mason 1133b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1134104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1135e2d84521SMiao Xie -buf->len, 1136e2d84521SMiao Xie fs_info->dirty_metadata_batch); 11370b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1138925baeddSChris Mason } 11395f39d397SChris Mason } 1140ed7b63ebSJosef Bacik } 11415f39d397SChris Mason 1142da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1143e20d96d6SChris Mason u64 objectid) 1144d97e63b6SChris Mason { 11457c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 11462e608bd1SJosef Bacik 11472e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 11482e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 11492e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 115096dfcb46SJosef Bacik root->fs_info = fs_info; 11512e608bd1SJosef Bacik root->root_key.objectid = objectid; 1152cfaa7295SChris Mason root->node = NULL; 1153a28ec197SChris Mason root->commit_root = NULL; 115427cdeb70SMiao Xie root->state = 0; 1155abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 11560b86a832SChris Mason 11570f7d52f4SChris Mason root->last_trans = 0; 11586b8fad57SNikolay Borisov root->free_objectid = 0; 1159eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1160199c2a9cSMiao Xie root->nr_ordered_extents = 0; 11616bef4d31SEric Paris root->inode_tree = RB_ROOT; 1162088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 11632e608bd1SJosef Bacik 11642e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 11650b86a832SChris Mason 11660b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 11675d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1168eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1169eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1170199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1171199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1172d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 11732ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 11742ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 11755d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1176eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1177199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1178f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 11792ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 11802ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 11818287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1182a2135011SChris Mason mutex_init(&root->objectid_mutex); 1183e02119d5SChris Mason mutex_init(&root->log_mutex); 118431f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1185573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1186c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 11877237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 11887237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 11897237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 11908b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 11918b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 11927237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 11937237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 11947237f183SYan Zheng atomic_set(&root->log_writers, 0); 11952ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11960700cea7SElena Reshetova refcount_set(&root->refs, 1); 11978ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1198eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 11997237f183SYan Zheng root->log_transid = 0; 1200d1433debSMiao Xie root->log_transid_committed = -1; 1201257c62e1SChris Mason root->last_log_commit = 0; 12022e608bd1SJosef Bacik root->anon_dev = 0; 1203e289f03eSFilipe Manana if (!dummy) { 120443eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 120543eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1206e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1207e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1208e289f03eSFilipe Manana } 1209017e5369SChris Mason 12105f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1211370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1212bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1213bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1214fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1215bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1216fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1217bd647ce3SJosef Bacik #endif 12183768f368SChris Mason } 12193768f368SChris Mason 122074e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 122196dfcb46SJosef Bacik u64 objectid, gfp_t flags) 12226f07e42eSAl Viro { 122374e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 12246f07e42eSAl Viro if (root) 122596dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 12266f07e42eSAl Viro return root; 12276f07e42eSAl Viro } 12286f07e42eSAl Viro 122906ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 123006ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1231da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 123206ea65a3SJosef Bacik { 123306ea65a3SJosef Bacik struct btrfs_root *root; 123406ea65a3SJosef Bacik 12357c0260eeSJeff Mahoney if (!fs_info) 12367c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 12377c0260eeSJeff Mahoney 123896dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 123906ea65a3SJosef Bacik if (!root) 124006ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1241da17066cSJeff Mahoney 1242b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1243faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 124406ea65a3SJosef Bacik 124506ea65a3SJosef Bacik return root; 124606ea65a3SJosef Bacik } 124706ea65a3SJosef Bacik #endif 124806ea65a3SJosef Bacik 1249abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 1250abed4aaaSJosef Bacik { 1251abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 1252abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 1253abed4aaaSJosef Bacik 1254abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 1255abed4aaaSJosef Bacik } 1256abed4aaaSJosef Bacik 1257abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 1258abed4aaaSJosef Bacik { 1259abed4aaaSJosef Bacik const struct btrfs_key *key = k; 1260abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 1261abed4aaaSJosef Bacik 1262abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 1263abed4aaaSJosef Bacik } 1264abed4aaaSJosef Bacik 1265abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 1266abed4aaaSJosef Bacik { 1267abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1268abed4aaaSJosef Bacik struct rb_node *tmp; 1269abed4aaaSJosef Bacik 1270abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1271abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 1272abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1273abed4aaaSJosef Bacik ASSERT(!tmp); 1274abed4aaaSJosef Bacik 1275abed4aaaSJosef Bacik return tmp ? -EEXIST : 0; 1276abed4aaaSJosef Bacik } 1277abed4aaaSJosef Bacik 1278abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 1279abed4aaaSJosef Bacik { 1280abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1281abed4aaaSJosef Bacik 1282abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1283abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1284abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1285abed4aaaSJosef Bacik } 1286abed4aaaSJosef Bacik 1287abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 1288abed4aaaSJosef Bacik struct btrfs_key *key) 1289abed4aaaSJosef Bacik { 1290abed4aaaSJosef Bacik struct rb_node *node; 1291abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 1292abed4aaaSJosef Bacik 1293abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 1294abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 1295abed4aaaSJosef Bacik if (node) 1296abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 1297abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 1298abed4aaaSJosef Bacik 1299abed4aaaSJosef Bacik return root; 1300abed4aaaSJosef Bacik } 1301abed4aaaSJosef Bacik 1302f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 1303f7238e50SJosef Bacik { 1304f7238e50SJosef Bacik struct btrfs_block_group *block_group; 1305f7238e50SJosef Bacik u64 ret; 1306f7238e50SJosef Bacik 1307f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 1308f7238e50SJosef Bacik return 0; 1309f7238e50SJosef Bacik 1310f7238e50SJosef Bacik if (bytenr) 1311f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 1312f7238e50SJosef Bacik else 1313f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 1314f7238e50SJosef Bacik ASSERT(block_group); 1315f7238e50SJosef Bacik if (!block_group) 1316f7238e50SJosef Bacik return 0; 1317f7238e50SJosef Bacik ret = block_group->global_root_id; 1318f7238e50SJosef Bacik btrfs_put_block_group(block_group); 1319f7238e50SJosef Bacik 1320f7238e50SJosef Bacik return ret; 1321f7238e50SJosef Bacik } 1322f7238e50SJosef Bacik 1323abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1324abed4aaaSJosef Bacik { 1325abed4aaaSJosef Bacik struct btrfs_key key = { 1326abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 1327abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1328f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1329abed4aaaSJosef Bacik }; 1330abed4aaaSJosef Bacik 1331abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1332abed4aaaSJosef Bacik } 1333abed4aaaSJosef Bacik 1334abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1335abed4aaaSJosef Bacik { 1336abed4aaaSJosef Bacik struct btrfs_key key = { 1337abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 1338abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1339f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1340abed4aaaSJosef Bacik }; 1341abed4aaaSJosef Bacik 1342abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1343abed4aaaSJosef Bacik } 1344abed4aaaSJosef Bacik 134520897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 134620897f5cSArne Jansen u64 objectid) 134720897f5cSArne Jansen { 13489b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 134920897f5cSArne Jansen struct extent_buffer *leaf; 135020897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 135120897f5cSArne Jansen struct btrfs_root *root; 135220897f5cSArne Jansen struct btrfs_key key; 1353b89f6d1fSFilipe Manana unsigned int nofs_flag; 135420897f5cSArne Jansen int ret = 0; 135520897f5cSArne Jansen 1356b89f6d1fSFilipe Manana /* 1357b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1358b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1359b89f6d1fSFilipe Manana */ 1360b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 136196dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1362b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 136320897f5cSArne Jansen if (!root) 136420897f5cSArne Jansen return ERR_PTR(-ENOMEM); 136520897f5cSArne Jansen 136620897f5cSArne Jansen root->root_key.objectid = objectid; 136720897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 136820897f5cSArne Jansen root->root_key.offset = 0; 136920897f5cSArne Jansen 13709631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 13719631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 137220897f5cSArne Jansen if (IS_ERR(leaf)) { 137320897f5cSArne Jansen ret = PTR_ERR(leaf); 13741dd05682STsutomu Itoh leaf = NULL; 13758a6a87cdSBoris Burkov goto fail_unlock; 137620897f5cSArne Jansen } 137720897f5cSArne Jansen 137820897f5cSArne Jansen root->node = leaf; 137920897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 138020897f5cSArne Jansen 138120897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 138227cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 138320897f5cSArne Jansen 1384f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1385f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 138620897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 138720897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 138820897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 138920897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 139020897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 139120897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 139220897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 139333d85fdaSQu Wenruo if (is_fstree(objectid)) 1394807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1395807fc790SAndy Shevchenko else 1396807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1397c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 139820897f5cSArne Jansen 13998a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 14008a6a87cdSBoris Burkov 140120897f5cSArne Jansen key.objectid = objectid; 140220897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 140320897f5cSArne Jansen key.offset = 0; 140420897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 140520897f5cSArne Jansen if (ret) 140620897f5cSArne Jansen goto fail; 140720897f5cSArne Jansen 140820897f5cSArne Jansen return root; 14091dd05682STsutomu Itoh 14108a6a87cdSBoris Burkov fail_unlock: 14118c38938cSJosef Bacik if (leaf) 14121dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 14138a6a87cdSBoris Burkov fail: 141400246528SJosef Bacik btrfs_put_root(root); 14151dd05682STsutomu Itoh 14161dd05682STsutomu Itoh return ERR_PTR(ret); 141720897f5cSArne Jansen } 141820897f5cSArne Jansen 14197237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1420e02119d5SChris Mason struct btrfs_fs_info *fs_info) 14210f7d52f4SChris Mason { 14220f7d52f4SChris Mason struct btrfs_root *root; 1423e02119d5SChris Mason 142496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1425e02119d5SChris Mason if (!root) 14267237f183SYan Zheng return ERR_PTR(-ENOMEM); 1427e02119d5SChris Mason 1428e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1429e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1430e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 143127cdeb70SMiao Xie 14326ab6ebb7SNaohiro Aota return root; 14336ab6ebb7SNaohiro Aota } 14346ab6ebb7SNaohiro Aota 14356ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 14366ab6ebb7SNaohiro Aota struct btrfs_root *root) 14376ab6ebb7SNaohiro Aota { 14386ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 14396ab6ebb7SNaohiro Aota 14407237f183SYan Zheng /* 144192a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 144227cdeb70SMiao Xie * 144392a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 144492a7cc42SQu Wenruo * and they go away before a real commit is actually done. 144592a7cc42SQu Wenruo * 144692a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 144792a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 14487237f183SYan Zheng */ 1449e02119d5SChris Mason 14504d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 14519631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 14526ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 14536ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 1454e02119d5SChris Mason 14557237f183SYan Zheng root->node = leaf; 1456e02119d5SChris Mason 1457e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1458e02119d5SChris Mason btrfs_tree_unlock(root->node); 14596ab6ebb7SNaohiro Aota 14606ab6ebb7SNaohiro Aota return 0; 14617237f183SYan Zheng } 14627237f183SYan Zheng 14637237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 14647237f183SYan Zheng struct btrfs_fs_info *fs_info) 14657237f183SYan Zheng { 14667237f183SYan Zheng struct btrfs_root *log_root; 14677237f183SYan Zheng 14687237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 14697237f183SYan Zheng if (IS_ERR(log_root)) 14707237f183SYan Zheng return PTR_ERR(log_root); 14716ab6ebb7SNaohiro Aota 14723ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 14733ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 14743ddebf27SNaohiro Aota 14756ab6ebb7SNaohiro Aota if (ret) { 14766ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 14776ab6ebb7SNaohiro Aota return ret; 14786ab6ebb7SNaohiro Aota } 14793ddebf27SNaohiro Aota } 14806ab6ebb7SNaohiro Aota 14817237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 14827237f183SYan Zheng fs_info->log_root_tree = log_root; 14837237f183SYan Zheng return 0; 14847237f183SYan Zheng } 14857237f183SYan Zheng 14867237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 14877237f183SYan Zheng struct btrfs_root *root) 14887237f183SYan Zheng { 14890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 14907237f183SYan Zheng struct btrfs_root *log_root; 14917237f183SYan Zheng struct btrfs_inode_item *inode_item; 14926ab6ebb7SNaohiro Aota int ret; 14937237f183SYan Zheng 14940b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 14957237f183SYan Zheng if (IS_ERR(log_root)) 14967237f183SYan Zheng return PTR_ERR(log_root); 14977237f183SYan Zheng 14986ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 14996ab6ebb7SNaohiro Aota if (ret) { 15006ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 15016ab6ebb7SNaohiro Aota return ret; 15026ab6ebb7SNaohiro Aota } 15036ab6ebb7SNaohiro Aota 15047237f183SYan Zheng log_root->last_trans = trans->transid; 15057237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 15067237f183SYan Zheng 15077237f183SYan Zheng inode_item = &log_root->root_item.inode; 15083cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 15093cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 15103cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1511da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 15120b246afaSJeff Mahoney fs_info->nodesize); 15133cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 15147237f183SYan Zheng 15155d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 15167237f183SYan Zheng 15177237f183SYan Zheng WARN_ON(root->log_root); 15187237f183SYan Zheng root->log_root = log_root; 15197237f183SYan Zheng root->log_transid = 0; 1520d1433debSMiao Xie root->log_transid_committed = -1; 1521257c62e1SChris Mason root->last_log_commit = 0; 1522e02119d5SChris Mason return 0; 1523e02119d5SChris Mason } 1524e02119d5SChris Mason 152549d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 152649d11beaSJosef Bacik struct btrfs_path *path, 1527cb517eabSMiao Xie struct btrfs_key *key) 1528e02119d5SChris Mason { 1529e02119d5SChris Mason struct btrfs_root *root; 1530e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 153184234f3aSYan Zheng u64 generation; 1532cb517eabSMiao Xie int ret; 1533581c1760SQu Wenruo int level; 1534cb517eabSMiao Xie 153596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 153649d11beaSJosef Bacik if (!root) 153749d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 15380f7d52f4SChris Mason 1539cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1540cb517eabSMiao Xie &root->root_item, &root->root_key); 15410f7d52f4SChris Mason if (ret) { 154213a8a7c8SYan, Zheng if (ret > 0) 154313a8a7c8SYan, Zheng ret = -ENOENT; 154449d11beaSJosef Bacik goto fail; 15450f7d52f4SChris Mason } 154613a8a7c8SYan, Zheng 154784234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1548581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 15492ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 15502ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 15511b7ec85eSJosef Bacik key->objectid, generation, level, NULL); 155264c043deSLiu Bo if (IS_ERR(root->node)) { 155364c043deSLiu Bo ret = PTR_ERR(root->node); 15548c38938cSJosef Bacik root->node = NULL; 155549d11beaSJosef Bacik goto fail; 15564eb150d6SQu Wenruo } 15574eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1558cb517eabSMiao Xie ret = -EIO; 155949d11beaSJosef Bacik goto fail; 1560416bc658SJosef Bacik } 156188c602abSQu Wenruo 156288c602abSQu Wenruo /* 156388c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 156488c602abSQu Wenruo * match its root node owner 156588c602abSQu Wenruo */ 156688c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 156788c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 156888c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 156988c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 157088c602abSQu Wenruo btrfs_crit(fs_info, 157188c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 157288c602abSQu Wenruo root->root_key.objectid, root->node->start, 157388c602abSQu Wenruo btrfs_header_owner(root->node), 157488c602abSQu Wenruo root->root_key.objectid); 157588c602abSQu Wenruo ret = -EUCLEAN; 157688c602abSQu Wenruo goto fail; 157788c602abSQu Wenruo } 15785d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1579cb517eabSMiao Xie return root; 158049d11beaSJosef Bacik fail: 158100246528SJosef Bacik btrfs_put_root(root); 158249d11beaSJosef Bacik return ERR_PTR(ret); 158349d11beaSJosef Bacik } 158449d11beaSJosef Bacik 158549d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 158649d11beaSJosef Bacik struct btrfs_key *key) 158749d11beaSJosef Bacik { 158849d11beaSJosef Bacik struct btrfs_root *root; 158949d11beaSJosef Bacik struct btrfs_path *path; 159049d11beaSJosef Bacik 159149d11beaSJosef Bacik path = btrfs_alloc_path(); 159249d11beaSJosef Bacik if (!path) 159349d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 159449d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 159549d11beaSJosef Bacik btrfs_free_path(path); 159649d11beaSJosef Bacik 159749d11beaSJosef Bacik return root; 1598cb517eabSMiao Xie } 1599cb517eabSMiao Xie 16002dfb1e43SQu Wenruo /* 16012dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 16022dfb1e43SQu Wenruo * 16032dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 16042dfb1e43SQu Wenruo */ 16052dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1606cb517eabSMiao Xie { 1607cb517eabSMiao Xie int ret; 1608dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1609cb517eabSMiao Xie 1610dcc3eb96SNikolay Borisov /* 1611dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1612dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1613dcc3eb96SNikolay Borisov */ 1614dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1615dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1616dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1617dcc3eb96SNikolay Borisov if (ret) 16188257b2dcSMiao Xie goto fail; 16198257b2dcSMiao Xie 1620aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 162137f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 162292a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1623f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1624f39e4571SJosef Bacik } 1625f39e4571SJosef Bacik 1626851fd730SQu Wenruo /* 1627851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1628851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1629851fd730SQu Wenruo */ 1630851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1631851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 16322dfb1e43SQu Wenruo if (!anon_dev) { 1633cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1634cb517eabSMiao Xie if (ret) 1635876d2cf1SLiu Bo goto fail; 16362dfb1e43SQu Wenruo } else { 16372dfb1e43SQu Wenruo root->anon_dev = anon_dev; 16382dfb1e43SQu Wenruo } 1639851fd730SQu Wenruo } 1640f32e48e9SChandan Rajendra 1641f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1642453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1643f32e48e9SChandan Rajendra if (ret) { 1644f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1645876d2cf1SLiu Bo goto fail; 1646f32e48e9SChandan Rajendra } 1647f32e48e9SChandan Rajendra 16486b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1649f32e48e9SChandan Rajendra 1650f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1651f32e48e9SChandan Rajendra 1652cb517eabSMiao Xie return 0; 1653cb517eabSMiao Xie fail: 165484db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1655cb517eabSMiao Xie return ret; 1656cb517eabSMiao Xie } 1657cb517eabSMiao Xie 1658a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1659cb517eabSMiao Xie u64 root_id) 1660cb517eabSMiao Xie { 1661cb517eabSMiao Xie struct btrfs_root *root; 1662cb517eabSMiao Xie 1663fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1664fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1665fc7cbcd4SDavid Sterba (unsigned long)root_id); 1666bc44d7c4SJosef Bacik if (root) 166700246528SJosef Bacik root = btrfs_grab_root(root); 1668fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1669cb517eabSMiao Xie return root; 1670cb517eabSMiao Xie } 1671cb517eabSMiao Xie 167249d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 167349d11beaSJosef Bacik u64 objectid) 167449d11beaSJosef Bacik { 1675abed4aaaSJosef Bacik struct btrfs_key key = { 1676abed4aaaSJosef Bacik .objectid = objectid, 1677abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1678abed4aaaSJosef Bacik .offset = 0, 1679abed4aaaSJosef Bacik }; 1680abed4aaaSJosef Bacik 168149d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 168249d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 168349d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 1684abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 168549d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 168649d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 168749d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 168849d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 168949d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 1690abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 169149d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 169249d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 169349d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 169449d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 169549d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 169649d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 1697abed4aaaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { 1698abed4aaaSJosef Bacik struct btrfs_root *root = btrfs_global_root(fs_info, &key); 1699abed4aaaSJosef Bacik 1700abed4aaaSJosef Bacik return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); 1701abed4aaaSJosef Bacik } 170249d11beaSJosef Bacik return NULL; 170349d11beaSJosef Bacik } 170449d11beaSJosef Bacik 1705cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1706cb517eabSMiao Xie struct btrfs_root *root) 1707cb517eabSMiao Xie { 1708cb517eabSMiao Xie int ret; 1709cb517eabSMiao Xie 1710fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1711fc7cbcd4SDavid Sterba if (ret) 1712fc7cbcd4SDavid Sterba return ret; 1713fc7cbcd4SDavid Sterba 1714fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1715fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1716fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1717fc7cbcd4SDavid Sterba root); 1718af01d2e5SJosef Bacik if (ret == 0) { 171900246528SJosef Bacik btrfs_grab_root(root); 1720fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1721af01d2e5SJosef Bacik } 1722fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1723fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1724cb517eabSMiao Xie 1725cb517eabSMiao Xie return ret; 1726cb517eabSMiao Xie } 1727cb517eabSMiao Xie 1728bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1729bd647ce3SJosef Bacik { 1730bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1731bd647ce3SJosef Bacik struct btrfs_root *root; 1732bd647ce3SJosef Bacik 1733bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1734457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1735457f1864SJosef Bacik 1736bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1737bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1738457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 173971008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1740bd647ce3SJosef Bacik refcount_read(&root->refs)); 1741bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 174200246528SJosef Bacik btrfs_put_root(root); 174300246528SJosef Bacik btrfs_put_root(root); 1744bd647ce3SJosef Bacik } 1745bd647ce3SJosef Bacik #endif 1746bd647ce3SJosef Bacik } 1747bd647ce3SJosef Bacik 1748abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1749abed4aaaSJosef Bacik { 1750abed4aaaSJosef Bacik struct btrfs_root *root; 1751abed4aaaSJosef Bacik struct rb_node *node; 1752abed4aaaSJosef Bacik 1753abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1754abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1755abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1756abed4aaaSJosef Bacik btrfs_put_root(root); 1757abed4aaaSJosef Bacik } 1758abed4aaaSJosef Bacik } 1759abed4aaaSJosef Bacik 17600d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 17610d4b0463SJosef Bacik { 1762141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1763141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 17645deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1765141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1766141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1767141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1768141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 17690d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 17700d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1771abed4aaaSJosef Bacik free_global_roots(fs_info); 177200246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 177300246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 177400246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 177500246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 177600246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 177700246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1778aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 17799c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1780bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 17813fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 17820d4b0463SJosef Bacik kfree(fs_info->super_copy); 17830d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 17848481dd80SQu Wenruo kfree(fs_info->subpage_info); 17850d4b0463SJosef Bacik kvfree(fs_info); 17860d4b0463SJosef Bacik } 17870d4b0463SJosef Bacik 17880d4b0463SJosef Bacik 17892dfb1e43SQu Wenruo /* 17902dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 17912dfb1e43SQu Wenruo * 17922dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 17932dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 17942dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 17952dfb1e43SQu Wenruo * 17962dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 17972dfb1e43SQu Wenruo * 17982dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 17992dfb1e43SQu Wenruo * same root objectid. 18002dfb1e43SQu Wenruo * 18012dfb1e43SQu Wenruo * @objectid: root id 18022dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 18032dfb1e43SQu Wenruo * pass 0 for new allocation. 18042dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 18052dfb1e43SQu Wenruo * for orphan roots 18062dfb1e43SQu Wenruo */ 18072dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 18082dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 18092dfb1e43SQu Wenruo bool check_ref) 18105eda7b5eSChris Mason { 18115eda7b5eSChris Mason struct btrfs_root *root; 1812381cf658SDavid Sterba struct btrfs_path *path; 18131d4c08e0SDavid Sterba struct btrfs_key key; 18145eda7b5eSChris Mason int ret; 18155eda7b5eSChris Mason 181649d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 181749d11beaSJosef Bacik if (root) 181849d11beaSJosef Bacik return root; 18194df27c4dSYan, Zheng again: 182056e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 182148475471SStefan Behrens if (root) { 18222dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 18232dfb1e43SQu Wenruo ASSERT(!anon_dev); 1824bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 182500246528SJosef Bacik btrfs_put_root(root); 182648475471SStefan Behrens return ERR_PTR(-ENOENT); 1827bc44d7c4SJosef Bacik } 18285eda7b5eSChris Mason return root; 182948475471SStefan Behrens } 18305eda7b5eSChris Mason 183156e9357aSDavid Sterba key.objectid = objectid; 183256e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 183356e9357aSDavid Sterba key.offset = (u64)-1; 183456e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 18355eda7b5eSChris Mason if (IS_ERR(root)) 18365eda7b5eSChris Mason return root; 18373394e160SChris Mason 1838c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1839d68fc57bSYan, Zheng ret = -ENOENT; 1840d68fc57bSYan, Zheng goto fail; 1841d68fc57bSYan, Zheng } 1842d68fc57bSYan, Zheng 18432dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1844cb517eabSMiao Xie if (ret) 1845cb517eabSMiao Xie goto fail; 1846cb517eabSMiao Xie 1847381cf658SDavid Sterba path = btrfs_alloc_path(); 1848381cf658SDavid Sterba if (!path) { 1849381cf658SDavid Sterba ret = -ENOMEM; 1850381cf658SDavid Sterba goto fail; 1851381cf658SDavid Sterba } 18521d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 18531d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 185456e9357aSDavid Sterba key.offset = objectid; 18551d4c08e0SDavid Sterba 18561d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1857381cf658SDavid Sterba btrfs_free_path(path); 1858d68fc57bSYan, Zheng if (ret < 0) 1859d68fc57bSYan, Zheng goto fail; 1860d68fc57bSYan, Zheng if (ret == 0) 186127cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1862d68fc57bSYan, Zheng 1863cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 18640f7d52f4SChris Mason if (ret) { 1865168a2f77SJia-Ju Bai if (ret == -EEXIST) { 186600246528SJosef Bacik btrfs_put_root(root); 18674df27c4dSYan, Zheng goto again; 1868168a2f77SJia-Ju Bai } 18694df27c4dSYan, Zheng goto fail; 18704df27c4dSYan, Zheng } 1871edbd8d4eSChris Mason return root; 18724df27c4dSYan, Zheng fail: 187333fab972SFilipe Manana /* 187433fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1875*143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 187633fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 187733fab972SFilipe Manana * and once again by our caller. 187833fab972SFilipe Manana */ 187933fab972SFilipe Manana if (anon_dev) 188033fab972SFilipe Manana root->anon_dev = 0; 18818c38938cSJosef Bacik btrfs_put_root(root); 18824df27c4dSYan, Zheng return ERR_PTR(ret); 1883edbd8d4eSChris Mason } 1884edbd8d4eSChris Mason 18852dfb1e43SQu Wenruo /* 18862dfb1e43SQu Wenruo * Get in-memory reference of a root structure 18872dfb1e43SQu Wenruo * 18882dfb1e43SQu Wenruo * @objectid: tree objectid 18892dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 18902dfb1e43SQu Wenruo * one reference 18912dfb1e43SQu Wenruo */ 18922dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 18932dfb1e43SQu Wenruo u64 objectid, bool check_ref) 18942dfb1e43SQu Wenruo { 18952dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 18962dfb1e43SQu Wenruo } 18972dfb1e43SQu Wenruo 18982dfb1e43SQu Wenruo /* 18992dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 19002dfb1e43SQu Wenruo * the anonymous block device id 19012dfb1e43SQu Wenruo * 19022dfb1e43SQu Wenruo * @objectid: tree objectid 19032dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 19042dfb1e43SQu Wenruo * parameter value 19052dfb1e43SQu Wenruo */ 19062dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 19072dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 19082dfb1e43SQu Wenruo { 19092dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 19102dfb1e43SQu Wenruo } 19112dfb1e43SQu Wenruo 19128b712842SChris Mason /* 191349d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 191449d11beaSJosef Bacik * @fs_info: the fs_info 191549d11beaSJosef Bacik * @objectid: the objectid we need to lookup 191649d11beaSJosef Bacik * 191749d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 191849d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 191949d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 192049d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 192149d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 192249d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 192349d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 192449d11beaSJosef Bacik * backref code is finished using the root. 192549d11beaSJosef Bacik */ 192649d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 192749d11beaSJosef Bacik struct btrfs_path *path, 192849d11beaSJosef Bacik u64 objectid) 192949d11beaSJosef Bacik { 193049d11beaSJosef Bacik struct btrfs_root *root; 193149d11beaSJosef Bacik struct btrfs_key key; 193249d11beaSJosef Bacik 193349d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 193449d11beaSJosef Bacik 193549d11beaSJosef Bacik /* 193649d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 193749d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 193849d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 193949d11beaSJosef Bacik * != NULL just return it. 194049d11beaSJosef Bacik */ 194149d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 194249d11beaSJosef Bacik if (root) 194349d11beaSJosef Bacik return root; 194449d11beaSJosef Bacik 194549d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 194649d11beaSJosef Bacik if (root) 194749d11beaSJosef Bacik return root; 194849d11beaSJosef Bacik 194949d11beaSJosef Bacik key.objectid = objectid; 195049d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 195149d11beaSJosef Bacik key.offset = (u64)-1; 195249d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 195349d11beaSJosef Bacik btrfs_release_path(path); 195449d11beaSJosef Bacik 195549d11beaSJosef Bacik return root; 195649d11beaSJosef Bacik } 195749d11beaSJosef Bacik 195849d11beaSJosef Bacik /* 19598b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 19608b712842SChris Mason * functions. This is where read checksum verification actually happens 19618b712842SChris Mason */ 19628b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1963ce9adaa5SChris Mason { 1964ce9adaa5SChris Mason struct bio *bio; 196597eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1966ce9adaa5SChris Mason 196797eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1968ce9adaa5SChris Mason bio = end_io_wq->bio; 19698b712842SChris Mason 19704e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1971ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1972ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 19734246a0b6SChristoph Hellwig bio_endio(bio); 19749be490f1SOmar Sandoval kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 1975ce9adaa5SChris Mason } 197644b8bd7eSChris Mason 1977a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1978a74a4b97SChris Mason { 19790d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1980d0278245SMiao Xie int again; 1981a74a4b97SChris Mason 1982d6fd0ae2SOmar Sandoval while (1) { 1983d0278245SMiao Xie again = 0; 19849d1a2a3aSDavid Sterba 1985fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1986fd340d0fSJosef Bacik 1987d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 19882ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1989d0278245SMiao Xie goto sleep; 1990d0278245SMiao Xie 199190c711abSZygo Blaxell /* 199290c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 199390c711abSZygo Blaxell * before we have finished mounting the filesystem. 199490c711abSZygo Blaxell */ 19950b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 199690c711abSZygo Blaxell goto sleep; 199790c711abSZygo Blaxell 19980b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1999d0278245SMiao Xie goto sleep; 2000d0278245SMiao Xie 2001dc7f370cSMiao Xie /* 2002dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 2003dc7f370cSMiao Xie * during the above check and trylock. 2004dc7f370cSMiao Xie */ 20052ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 20060b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 2007dc7f370cSMiao Xie goto sleep; 2008dc7f370cSMiao Xie } 2009dc7f370cSMiao Xie 20102ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 2011c2d6cb16SFilipe Manana 201233c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 20130b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 2014a74a4b97SChris Mason 2015d0278245SMiao Xie /* 201605323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 201705323cd1SMiao Xie * needn't do anything special here. 2018d0278245SMiao Xie */ 20190b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 202067c5e7d4SFilipe Manana 202167c5e7d4SFilipe Manana /* 2022f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 202367c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 202467c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 2025f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 202667c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 202767c5e7d4SFilipe Manana * unused block groups. 202867c5e7d4SFilipe Manana */ 20290b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 203018bb8bbfSJohannes Thumshirn 203118bb8bbfSJohannes Thumshirn /* 203218bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 203318bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 203418bb8bbfSJohannes Thumshirn * space. 203518bb8bbfSJohannes Thumshirn */ 203618bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 2037d0278245SMiao Xie sleep: 2038a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 2039d6fd0ae2SOmar Sandoval if (kthread_should_park()) 2040d6fd0ae2SOmar Sandoval kthread_parkme(); 2041d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 2042d6fd0ae2SOmar Sandoval return 0; 2043838fe188SJiri Kosina if (!again) { 2044a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 2045a74a4b97SChris Mason schedule(); 2046a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 2047a74a4b97SChris Mason } 2048da288d28SFilipe Manana } 2049a74a4b97SChris Mason } 2050a74a4b97SChris Mason 2051a74a4b97SChris Mason static int transaction_kthread(void *arg) 2052a74a4b97SChris Mason { 2053a74a4b97SChris Mason struct btrfs_root *root = arg; 20540b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 2055a74a4b97SChris Mason struct btrfs_trans_handle *trans; 2056a74a4b97SChris Mason struct btrfs_transaction *cur; 20578929ecfaSYan, Zheng u64 transid; 2058643900beSNikolay Borisov time64_t delta; 2059a74a4b97SChris Mason unsigned long delay; 2060914b2007SJan Kara bool cannot_commit; 2061a74a4b97SChris Mason 2062a74a4b97SChris Mason do { 2063914b2007SJan Kara cannot_commit = false; 2064ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 20650b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 2066a74a4b97SChris Mason 20670b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 20680b246afaSJeff Mahoney cur = fs_info->running_transaction; 2069a74a4b97SChris Mason if (!cur) { 20700b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 2071a74a4b97SChris Mason goto sleep; 2072a74a4b97SChris Mason } 207331153d81SYan Zheng 2074643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 2075fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 2076fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 2077643900beSNikolay Borisov delta < fs_info->commit_interval) { 20780b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 2079fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 2080fb8a7e94SNikolay Borisov delay = min(delay, 2081fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 2082a74a4b97SChris Mason goto sleep; 2083a74a4b97SChris Mason } 20848929ecfaSYan, Zheng transid = cur->transid; 20850b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 208656bec294SChris Mason 208779787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 2088354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 2089914b2007SJan Kara if (IS_ERR(trans)) { 2090354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 2091914b2007SJan Kara cannot_commit = true; 209279787eaaSJeff Mahoney goto sleep; 2093914b2007SJan Kara } 20948929ecfaSYan, Zheng if (transid == trans->transid) { 20953a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 20968929ecfaSYan, Zheng } else { 20973a45bb20SJeff Mahoney btrfs_end_transaction(trans); 20988929ecfaSYan, Zheng } 2099a74a4b97SChris Mason sleep: 21000b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 21010b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 2102a74a4b97SChris Mason 210384961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 21042ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 21058929ecfaSYan, Zheng if (!kthread_should_stop() && 21060b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 2107914b2007SJan Kara cannot_commit)) 2108bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 2109a74a4b97SChris Mason } while (!kthread_should_stop()); 2110a74a4b97SChris Mason return 0; 2111a74a4b97SChris Mason } 2112a74a4b97SChris Mason 2113af31f5e5SChris Mason /* 211401f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 211501f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 211601f0f9daSNikolay Borisov * anything. 2117af31f5e5SChris Mason * 2118af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 2119af31f5e5SChris Mason * generation of the latest root in the array with the generation 2120af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 2121af31f5e5SChris Mason */ 212201f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 2123af31f5e5SChris Mason { 212401f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 2125af31f5e5SChris Mason u64 cur; 2126af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2127af31f5e5SChris Mason int i; 2128af31f5e5SChris Mason 2129af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2130af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 2131af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 2132af31f5e5SChris Mason if (cur == newest_gen) 213301f0f9daSNikolay Borisov return i; 2134af31f5e5SChris Mason } 2135af31f5e5SChris Mason 213601f0f9daSNikolay Borisov return -EINVAL; 2137af31f5e5SChris Mason } 2138af31f5e5SChris Mason 2139af31f5e5SChris Mason /* 2140af31f5e5SChris Mason * copy all the root pointers into the super backup array. 2141af31f5e5SChris Mason * this will bump the backup pointer by one when it is 2142af31f5e5SChris Mason * done 2143af31f5e5SChris Mason */ 2144af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 2145af31f5e5SChris Mason { 21466ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 2147af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2148af31f5e5SChris Mason 2149af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 2150af31f5e5SChris Mason 2151af31f5e5SChris Mason /* 2152af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 2153af31f5e5SChris Mason * regardless of which ones we use today 2154af31f5e5SChris Mason */ 2155af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 2156af31f5e5SChris Mason 2157af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 2158af31f5e5SChris Mason 2159af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 2160af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 2161af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 2162af31f5e5SChris Mason 2163af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 2164af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 2165af31f5e5SChris Mason 2166af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 2167af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 2168af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 2169af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 2170af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 2171af31f5e5SChris Mason 21729c54e80dSJosef Bacik if (btrfs_fs_incompat(info, EXTENT_TREE_V2)) { 21739c54e80dSJosef Bacik btrfs_set_backup_block_group_root(root_backup, 21749c54e80dSJosef Bacik info->block_group_root->node->start); 21759c54e80dSJosef Bacik btrfs_set_backup_block_group_root_gen(root_backup, 21769c54e80dSJosef Bacik btrfs_header_generation(info->block_group_root->node)); 21779c54e80dSJosef Bacik btrfs_set_backup_block_group_root_level(root_backup, 21789c54e80dSJosef Bacik btrfs_header_level(info->block_group_root->node)); 21799c54e80dSJosef Bacik } else { 21809c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 2181f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 21829c54e80dSJosef Bacik 21839c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 21849c54e80dSJosef Bacik extent_root->node->start); 2185af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 218629cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 2187af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 218829cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 2189af31f5e5SChris Mason 2190f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 2191f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 2192f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 2193f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 2194f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 21959c54e80dSJosef Bacik } 2196af31f5e5SChris Mason 21977c7e82a7SChris Mason /* 21987c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 21997c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 22007c7e82a7SChris Mason */ 22017c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 22027c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 22037c7e82a7SChris Mason info->fs_root->node->start); 2204af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2205af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2206af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2207af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 22087c7e82a7SChris Mason } 2209af31f5e5SChris Mason 2210af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2211af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2212af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2213af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2214af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2215af31f5e5SChris Mason 2216af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2217af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2218af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2219af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2220af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2221af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2222af31f5e5SChris Mason 2223af31f5e5SChris Mason /* 2224af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2225af31f5e5SChris Mason * for the next commit 2226af31f5e5SChris Mason */ 2227af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2228af31f5e5SChris Mason &info->super_for_commit->super_roots, 2229af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2230af31f5e5SChris Mason } 2231af31f5e5SChris Mason 2232af31f5e5SChris Mason /* 2233bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 2234bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 2235bd2336b2SNikolay Borisov * 2236bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 2237bd2336b2SNikolay Borisov * priority - priority of backup root required 2238bd2336b2SNikolay Borisov * 2239bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 2240bd2336b2SNikolay Borisov */ 2241bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 2242bd2336b2SNikolay Borisov { 2243bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2244bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 2245bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 2246bd2336b2SNikolay Borisov 2247bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 2248bd2336b2SNikolay Borisov if (priority == 0) 2249bd2336b2SNikolay Borisov return backup_index; 2250bd2336b2SNikolay Borisov 2251bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 2252bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 2253bd2336b2SNikolay Borisov } else { 2254bd2336b2SNikolay Borisov return -EINVAL; 2255bd2336b2SNikolay Borisov } 2256bd2336b2SNikolay Borisov 2257bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 2258bd2336b2SNikolay Borisov 2259bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 2260bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 2261bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2262bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 2263bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 2264bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2265bd2336b2SNikolay Borisov 2266bd2336b2SNikolay Borisov /* 2267bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 2268bd2336b2SNikolay Borisov * need a fsck 2269bd2336b2SNikolay Borisov */ 2270bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2271bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2272bd2336b2SNikolay Borisov 2273bd2336b2SNikolay Borisov return backup_index; 2274bd2336b2SNikolay Borisov } 2275bd2336b2SNikolay Borisov 22767abadb64SLiu Bo /* helper to cleanup workers */ 22777abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 22787abadb64SLiu Bo { 2279dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2280afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 2281a31b4a43SChristoph Hellwig btrfs_destroy_workqueue(fs_info->hipri_workers); 22825cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2283fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2284fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 2285385de0efSChristoph Hellwig if (fs_info->rmw_workers) 2286385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 2287fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2288fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 22895b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2290e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2291a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2292fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2293b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 2294b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 2295a9b9477dSFilipe Manana /* 2296a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2297a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2298a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2299a9b9477dSFilipe Manana */ 2300a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2301a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 23027abadb64SLiu Bo } 23037abadb64SLiu Bo 23042e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 23052e9f5954SRashika { 23062e9f5954SRashika if (root) { 23072e9f5954SRashika free_extent_buffer(root->node); 23082e9f5954SRashika free_extent_buffer(root->commit_root); 23092e9f5954SRashika root->node = NULL; 23102e9f5954SRashika root->commit_root = NULL; 23112e9f5954SRashika } 23122e9f5954SRashika } 23132e9f5954SRashika 2314abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 2315abed4aaaSJosef Bacik { 2316abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 2317abed4aaaSJosef Bacik 2318abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 2319abed4aaaSJosef Bacik &fs_info->global_root_tree, 2320abed4aaaSJosef Bacik rb_node) 2321abed4aaaSJosef Bacik free_root_extent_buffers(root); 2322abed4aaaSJosef Bacik } 2323abed4aaaSJosef Bacik 2324af31f5e5SChris Mason /* helper to cleanup tree roots */ 23254273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2326af31f5e5SChris Mason { 23272e9f5954SRashika free_root_extent_buffers(info->tree_root); 2328655b09feSJosef Bacik 2329abed4aaaSJosef Bacik free_global_root_pointers(info); 23302e9f5954SRashika free_root_extent_buffers(info->dev_root); 23312e9f5954SRashika free_root_extent_buffers(info->quota_root); 23322e9f5954SRashika free_root_extent_buffers(info->uuid_root); 23338c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2334aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 23359c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 23364273eaffSAnand Jain if (free_chunk_root) 23372e9f5954SRashika free_root_extent_buffers(info->chunk_root); 2338af31f5e5SChris Mason } 2339af31f5e5SChris Mason 23408c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 23418c38938cSJosef Bacik { 23428c38938cSJosef Bacik if (!root) 23438c38938cSJosef Bacik return; 23448c38938cSJosef Bacik 23458c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 23468c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 23471dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 23488c38938cSJosef Bacik if (root->anon_dev) 23498c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 23508c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2351923eb523SJohannes Thumshirn free_root_extent_buffers(root); 23528c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2353fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 23548c38938cSJosef Bacik list_del_init(&root->leak_list); 2355fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 23568c38938cSJosef Bacik #endif 23578c38938cSJosef Bacik kfree(root); 23588c38938cSJosef Bacik } 23598c38938cSJosef Bacik } 23608c38938cSJosef Bacik 2361faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2362171f6537SJosef Bacik { 2363fc7cbcd4SDavid Sterba int ret; 2364fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 2365fc7cbcd4SDavid Sterba int i; 2366171f6537SJosef Bacik 2367171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2368fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 2369171f6537SJosef Bacik struct btrfs_root, root_list); 2370fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 2371171f6537SJosef Bacik 2372fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2373fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2374fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 2375171f6537SJosef Bacik } 2376171f6537SJosef Bacik 2377fc7cbcd4SDavid Sterba while (1) { 2378fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2379fc7cbcd4SDavid Sterba (void **)gang, 0, 2380fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 2381fc7cbcd4SDavid Sterba if (!ret) 2382fc7cbcd4SDavid Sterba break; 2383fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 2384fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2385171f6537SJosef Bacik } 2386171f6537SJosef Bacik } 2387af31f5e5SChris Mason 2388638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2389638aa7edSEric Sandeen { 2390638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2391638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2392638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2393638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2394638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2395638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2396ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2397638aa7edSEric Sandeen } 2398638aa7edSEric Sandeen 2399779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2400779a65a4SEric Sandeen { 2401779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2402779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2403779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2404779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2405779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2406779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2407907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 2408779a65a4SEric Sandeen } 2409779a65a4SEric Sandeen 24106bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2411f37938e0SEric Sandeen { 24122ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 24132ff7e61eSJeff Mahoney 24142ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 24152ff7e61eSJeff Mahoney set_nlink(inode, 1); 2416f37938e0SEric Sandeen /* 2417f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2418f37938e0SEric Sandeen * the real end of the address space is determined by all of 2419f37938e0SEric Sandeen * the devices in the system 2420f37938e0SEric Sandeen */ 24212ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 24222ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2423f37938e0SEric Sandeen 24242ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 242543eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 24262c53a14dSQu Wenruo IO_TREE_BTREE_INODE_IO, inode); 24277b439738SDavid Sterba BTRFS_I(inode)->io_tree.track_uptodate = false; 24282ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2429f37938e0SEric Sandeen 24305c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 24312ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 24322ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 24332ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2434f37938e0SEric Sandeen } 2435f37938e0SEric Sandeen 2436ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2437ad618368SEric Sandeen { 2438ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2439129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 24407f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2441ad618368SEric Sandeen } 2442ad618368SEric Sandeen 2443f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2444f9e92e40SEric Sandeen { 2445f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2446f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2447f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2448f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2449f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2450f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2451d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2452f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2453f9e92e40SEric Sandeen } 2454f9e92e40SEric Sandeen 2455d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 24562a458198SEric Sandeen { 2457f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 24586f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 24592a458198SEric Sandeen 24602a458198SEric Sandeen fs_info->workers = 2461a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 2462a31b4a43SChristoph Hellwig fs_info->hipri_workers = 2463a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker-high", 2464cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 24652a458198SEric Sandeen 24662a458198SEric Sandeen fs_info->delalloc_workers = 2467cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2468cb001095SJeff Mahoney flags, max_active, 2); 24692a458198SEric Sandeen 24702a458198SEric Sandeen fs_info->flush_workers = 2471cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2472cb001095SJeff Mahoney flags, max_active, 0); 24732a458198SEric Sandeen 24742a458198SEric Sandeen fs_info->caching_workers = 2475cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 24762a458198SEric Sandeen 24772a458198SEric Sandeen fs_info->fixup_workers = 2478cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 24792a458198SEric Sandeen 24802a458198SEric Sandeen /* 24812a458198SEric Sandeen * endios are largely parallel and should have a very 24822a458198SEric Sandeen * low idle thresh 24832a458198SEric Sandeen */ 24842a458198SEric Sandeen fs_info->endio_workers = 2485cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 24862a458198SEric Sandeen fs_info->endio_meta_workers = 2487cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2488cb001095SJeff Mahoney max_active, 4); 24892a458198SEric Sandeen fs_info->endio_meta_write_workers = 2490cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2491cb001095SJeff Mahoney max_active, 2); 24922a458198SEric Sandeen fs_info->endio_raid56_workers = 2493cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2494cb001095SJeff Mahoney max_active, 4); 2495385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 24962a458198SEric Sandeen fs_info->endio_write_workers = 2497cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2498cb001095SJeff Mahoney max_active, 2); 24992a458198SEric Sandeen fs_info->endio_freespace_worker = 2500cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2501cb001095SJeff Mahoney max_active, 0); 25022a458198SEric Sandeen fs_info->delayed_workers = 2503cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2504cb001095SJeff Mahoney max_active, 0); 25052a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2506cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2507b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2508b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 25092a458198SEric Sandeen 2510a31b4a43SChristoph Hellwig if (!(fs_info->workers && fs_info->hipri_workers && 2511a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 25122a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 25132a458198SEric Sandeen fs_info->endio_meta_write_workers && 25142a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 25152a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 2516f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 2517f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 2518b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 25192a458198SEric Sandeen return -ENOMEM; 25202a458198SEric Sandeen } 25212a458198SEric Sandeen 25222a458198SEric Sandeen return 0; 25232a458198SEric Sandeen } 25242a458198SEric Sandeen 25256d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 25266d97c6e3SJohannes Thumshirn { 25276d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2528b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 25296d97c6e3SJohannes Thumshirn 2530b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 25316d97c6e3SJohannes Thumshirn 25326d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 25336d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2534b4e967beSDavid Sterba csum_driver); 25356d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 25366d97c6e3SJohannes Thumshirn } 25376d97c6e3SJohannes Thumshirn 25386d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 25396d97c6e3SJohannes Thumshirn 25406d97c6e3SJohannes Thumshirn return 0; 25416d97c6e3SJohannes Thumshirn } 25426d97c6e3SJohannes Thumshirn 254363443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 254463443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 254563443bf5SEric Sandeen { 254663443bf5SEric Sandeen int ret; 254763443bf5SEric Sandeen struct btrfs_root *log_tree_root; 254863443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 254963443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2550581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 255163443bf5SEric Sandeen 255263443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2553f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 255463443bf5SEric Sandeen return -EIO; 255563443bf5SEric Sandeen } 255663443bf5SEric Sandeen 255796dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 255896dfcb46SJosef Bacik GFP_KERNEL); 255963443bf5SEric Sandeen if (!log_tree_root) 256063443bf5SEric Sandeen return -ENOMEM; 256163443bf5SEric Sandeen 25622ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 25631b7ec85eSJosef Bacik BTRFS_TREE_LOG_OBJECTID, 25641b7ec85eSJosef Bacik fs_info->generation + 1, level, 25651b7ec85eSJosef Bacik NULL); 256664c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2567f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 25680eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 25698c38938cSJosef Bacik log_tree_root->node = NULL; 257000246528SJosef Bacik btrfs_put_root(log_tree_root); 25710eeff236SLiu Bo return ret; 25724eb150d6SQu Wenruo } 25734eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2574f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 257500246528SJosef Bacik btrfs_put_root(log_tree_root); 257663443bf5SEric Sandeen return -EIO; 257763443bf5SEric Sandeen } 25784eb150d6SQu Wenruo 257963443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 258063443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 258163443bf5SEric Sandeen if (ret) { 25820b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 258363443bf5SEric Sandeen "Failed to recover log tree"); 258400246528SJosef Bacik btrfs_put_root(log_tree_root); 258563443bf5SEric Sandeen return ret; 258663443bf5SEric Sandeen } 258763443bf5SEric Sandeen 2588bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 25896bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 259063443bf5SEric Sandeen if (ret) 259163443bf5SEric Sandeen return ret; 259263443bf5SEric Sandeen } 259363443bf5SEric Sandeen 259463443bf5SEric Sandeen return 0; 259563443bf5SEric Sandeen } 259663443bf5SEric Sandeen 2597abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2598abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2599abed4aaaSJosef Bacik const char *name) 2600abed4aaaSJosef Bacik { 2601abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2602abed4aaaSJosef Bacik struct btrfs_root *root; 2603f7238e50SJosef Bacik u64 max_global_id = 0; 2604abed4aaaSJosef Bacik int ret; 2605abed4aaaSJosef Bacik struct btrfs_key key = { 2606abed4aaaSJosef Bacik .objectid = objectid, 2607abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2608abed4aaaSJosef Bacik .offset = 0, 2609abed4aaaSJosef Bacik }; 2610abed4aaaSJosef Bacik bool found = false; 2611abed4aaaSJosef Bacik 2612abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2613abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2614abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2615abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2616abed4aaaSJosef Bacik return 0; 2617abed4aaaSJosef Bacik } 2618abed4aaaSJosef Bacik 2619abed4aaaSJosef Bacik while (1) { 2620abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2621abed4aaaSJosef Bacik if (ret < 0) 2622abed4aaaSJosef Bacik break; 2623abed4aaaSJosef Bacik 2624abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2625abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2626abed4aaaSJosef Bacik if (ret) { 2627abed4aaaSJosef Bacik if (ret > 0) 2628abed4aaaSJosef Bacik ret = 0; 2629abed4aaaSJosef Bacik break; 2630abed4aaaSJosef Bacik } 2631abed4aaaSJosef Bacik } 2632abed4aaaSJosef Bacik ret = 0; 2633abed4aaaSJosef Bacik 2634abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2635abed4aaaSJosef Bacik if (key.objectid != objectid) 2636abed4aaaSJosef Bacik break; 2637abed4aaaSJosef Bacik btrfs_release_path(path); 2638abed4aaaSJosef Bacik 2639f7238e50SJosef Bacik /* 2640f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2641f7238e50SJosef Bacik * everybody. 2642f7238e50SJosef Bacik */ 2643f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2644f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2645f7238e50SJosef Bacik 2646abed4aaaSJosef Bacik found = true; 2647abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2648abed4aaaSJosef Bacik if (IS_ERR(root)) { 2649abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2650abed4aaaSJosef Bacik ret = PTR_ERR(root); 2651abed4aaaSJosef Bacik break; 2652abed4aaaSJosef Bacik } 2653abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2654abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2655abed4aaaSJosef Bacik if (ret) { 2656abed4aaaSJosef Bacik btrfs_put_root(root); 2657abed4aaaSJosef Bacik break; 2658abed4aaaSJosef Bacik } 2659abed4aaaSJosef Bacik key.offset++; 2660abed4aaaSJosef Bacik } 2661abed4aaaSJosef Bacik btrfs_release_path(path); 2662abed4aaaSJosef Bacik 2663f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2664f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2665f7238e50SJosef Bacik 2666abed4aaaSJosef Bacik if (!found || ret) { 2667abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2668abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2669abed4aaaSJosef Bacik 2670abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2671abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2672abed4aaaSJosef Bacik else 2673abed4aaaSJosef Bacik ret = 0; 2674abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2675abed4aaaSJosef Bacik } 2676abed4aaaSJosef Bacik return ret; 2677abed4aaaSJosef Bacik } 2678abed4aaaSJosef Bacik 2679abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2680abed4aaaSJosef Bacik { 2681abed4aaaSJosef Bacik struct btrfs_path *path; 2682abed4aaaSJosef Bacik int ret = 0; 2683abed4aaaSJosef Bacik 2684abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2685abed4aaaSJosef Bacik if (!path) 2686abed4aaaSJosef Bacik return -ENOMEM; 2687abed4aaaSJosef Bacik 2688abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2689abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2690abed4aaaSJosef Bacik if (ret) 2691abed4aaaSJosef Bacik goto out; 2692abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2693abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2694abed4aaaSJosef Bacik if (ret) 2695abed4aaaSJosef Bacik goto out; 2696abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2697abed4aaaSJosef Bacik goto out; 2698abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2699abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2700abed4aaaSJosef Bacik "free space"); 2701abed4aaaSJosef Bacik out: 2702abed4aaaSJosef Bacik btrfs_free_path(path); 2703abed4aaaSJosef Bacik return ret; 2704abed4aaaSJosef Bacik } 2705abed4aaaSJosef Bacik 27066bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 27074bbcaa64SEric Sandeen { 27086bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2709a4f3d2c4SDavid Sterba struct btrfs_root *root; 27104bbcaa64SEric Sandeen struct btrfs_key location; 27114bbcaa64SEric Sandeen int ret; 27124bbcaa64SEric Sandeen 27136bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 27146bccf3abSJeff Mahoney 2715abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2716abed4aaaSJosef Bacik if (ret) 2717abed4aaaSJosef Bacik return ret; 2718abed4aaaSJosef Bacik 2719abed4aaaSJosef Bacik location.objectid = BTRFS_DEV_TREE_OBJECTID; 27204bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 27214bbcaa64SEric Sandeen location.offset = 0; 27224bbcaa64SEric Sandeen 2723a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2724f50f4353SLiu Bo if (IS_ERR(root)) { 272542437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2726f50f4353SLiu Bo ret = PTR_ERR(root); 2727f50f4353SLiu Bo goto out; 2728f50f4353SLiu Bo } 272942437a63SJosef Bacik } else { 2730a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2731a4f3d2c4SDavid Sterba fs_info->dev_root = root; 273242437a63SJosef Bacik } 2733820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2734820a49daSJosef Bacik btrfs_init_devices_late(fs_info); 27354bbcaa64SEric Sandeen 2736aeb935a4SQu Wenruo /* 2737aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2738aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2739aeb935a4SQu Wenruo */ 274056e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 274156e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2742aeb935a4SQu Wenruo if (IS_ERR(root)) { 274342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2744aeb935a4SQu Wenruo ret = PTR_ERR(root); 2745aeb935a4SQu Wenruo goto out; 2746aeb935a4SQu Wenruo } 274742437a63SJosef Bacik } else { 2748aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2749aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 275042437a63SJosef Bacik } 2751aeb935a4SQu Wenruo 27524bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2753a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2754a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2755a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2756afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2757a4f3d2c4SDavid Sterba fs_info->quota_root = root; 27584bbcaa64SEric Sandeen } 27594bbcaa64SEric Sandeen 27604bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2761a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2762a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 276342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2764a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 27654bbcaa64SEric Sandeen if (ret != -ENOENT) 2766f50f4353SLiu Bo goto out; 276742437a63SJosef Bacik } 27684bbcaa64SEric Sandeen } else { 2769a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2770a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 27714bbcaa64SEric Sandeen } 27724bbcaa64SEric Sandeen 27734bbcaa64SEric Sandeen return 0; 2774f50f4353SLiu Bo out: 2775f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2776f50f4353SLiu Bo location.objectid, ret); 2777f50f4353SLiu Bo return ret; 27784bbcaa64SEric Sandeen } 27794bbcaa64SEric Sandeen 2780069ec957SQu Wenruo /* 2781069ec957SQu Wenruo * Real super block validation 2782069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2783069ec957SQu Wenruo * 2784069ec957SQu Wenruo * @sb: super block to check 2785069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2786069ec957SQu Wenruo * 0 the primary (1st) sb 2787069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2788069ec957SQu Wenruo * -1 skip bytenr check 2789069ec957SQu Wenruo */ 2790069ec957SQu Wenruo static int validate_super(struct btrfs_fs_info *fs_info, 2791069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 279221a852b0SQu Wenruo { 279321a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 279421a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 279521a852b0SQu Wenruo int ret = 0; 279621a852b0SQu Wenruo 279721a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 279821a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 279921a852b0SQu Wenruo ret = -EINVAL; 280021a852b0SQu Wenruo } 280121a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 280221a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 280321a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 280421a852b0SQu Wenruo ret = -EINVAL; 280521a852b0SQu Wenruo } 280621a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 280721a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 280821a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 280921a852b0SQu Wenruo ret = -EINVAL; 281021a852b0SQu Wenruo } 281121a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 281221a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 281321a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 281421a852b0SQu Wenruo ret = -EINVAL; 281521a852b0SQu Wenruo } 281621a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 281721a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 281821a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 281921a852b0SQu Wenruo ret = -EINVAL; 282021a852b0SQu Wenruo } 282121a852b0SQu Wenruo 282221a852b0SQu Wenruo /* 282321a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 282421a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 282521a852b0SQu Wenruo */ 282621a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 282721a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 282821a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 282921a852b0SQu Wenruo ret = -EINVAL; 283021a852b0SQu Wenruo } 28310bb3eb3eSQu Wenruo 28320bb3eb3eSQu Wenruo /* 28331a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 28341a42daabSQu Wenruo * 28351a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 28361a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 28371a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 28381a42daabSQu Wenruo * beginning. 28390bb3eb3eSQu Wenruo */ 28401a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 284121a852b0SQu Wenruo btrfs_err(fs_info, 28420bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 284321a852b0SQu Wenruo sectorsize, PAGE_SIZE); 284421a852b0SQu Wenruo ret = -EINVAL; 284521a852b0SQu Wenruo } 28460bb3eb3eSQu Wenruo 284721a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 284821a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 284921a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 285021a852b0SQu Wenruo ret = -EINVAL; 285121a852b0SQu Wenruo } 285221a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 285321a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 285421a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 285521a852b0SQu Wenruo ret = -EINVAL; 285621a852b0SQu Wenruo } 285721a852b0SQu Wenruo 285821a852b0SQu Wenruo /* Root alignment check */ 285921a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 286021a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 286121a852b0SQu Wenruo btrfs_super_root(sb)); 286221a852b0SQu Wenruo ret = -EINVAL; 286321a852b0SQu Wenruo } 286421a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 286521a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 286621a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 286721a852b0SQu Wenruo ret = -EINVAL; 286821a852b0SQu Wenruo } 286921a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 287021a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 287121a852b0SQu Wenruo btrfs_super_log_root(sb)); 287221a852b0SQu Wenruo ret = -EINVAL; 287321a852b0SQu Wenruo } 287421a852b0SQu Wenruo 2875aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2876aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2877aefd7f70SNikolay Borisov btrfs_err(fs_info, 2878aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2879aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2880aefd7f70SNikolay Borisov ret = -EINVAL; 2881aefd7f70SNikolay Borisov } 2882aefd7f70SNikolay Borisov 2883aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2884aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2885aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2886aefd7f70SNikolay Borisov btrfs_err(fs_info, 2887aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2888aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2889aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2890aefd7f70SNikolay Borisov ret = -EINVAL; 2891aefd7f70SNikolay Borisov } 2892aefd7f70SNikolay Borisov 2893de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 28947239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 289521a852b0SQu Wenruo btrfs_err(fs_info, 28967239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2897de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 289821a852b0SQu Wenruo ret = -EINVAL; 289921a852b0SQu Wenruo } 290021a852b0SQu Wenruo 290121a852b0SQu Wenruo /* 290221a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 290321a852b0SQu Wenruo * done later 290421a852b0SQu Wenruo */ 290521a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 290621a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 290721a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 290821a852b0SQu Wenruo ret = -EINVAL; 290921a852b0SQu Wenruo } 291021a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 291121a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 291221a852b0SQu Wenruo btrfs_super_stripesize(sb)); 291321a852b0SQu Wenruo ret = -EINVAL; 291421a852b0SQu Wenruo } 291521a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 291621a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 291721a852b0SQu Wenruo btrfs_super_num_devices(sb)); 291821a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 291921a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 292021a852b0SQu Wenruo ret = -EINVAL; 292121a852b0SQu Wenruo } 292221a852b0SQu Wenruo 2923069ec957SQu Wenruo if (mirror_num >= 0 && 2924069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 292521a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 292621a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 292721a852b0SQu Wenruo ret = -EINVAL; 292821a852b0SQu Wenruo } 292921a852b0SQu Wenruo 293021a852b0SQu Wenruo /* 293121a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 293221a852b0SQu Wenruo * and one chunk 293321a852b0SQu Wenruo */ 293421a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 293521a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 293621a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 293721a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 293821a852b0SQu Wenruo ret = -EINVAL; 293921a852b0SQu Wenruo } 294021a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 294121a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 294221a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 294321a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 294421a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 294521a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 294621a852b0SQu Wenruo ret = -EINVAL; 294721a852b0SQu Wenruo } 294821a852b0SQu Wenruo 294921a852b0SQu Wenruo /* 295021a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 295121a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 295221a852b0SQu Wenruo */ 295321a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 295421a852b0SQu Wenruo btrfs_warn(fs_info, 295521a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 295621a852b0SQu Wenruo btrfs_super_generation(sb), 295721a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 295821a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 295921a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 296021a852b0SQu Wenruo btrfs_warn(fs_info, 296121a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 296221a852b0SQu Wenruo btrfs_super_generation(sb), 296321a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 296421a852b0SQu Wenruo 296521a852b0SQu Wenruo return ret; 296621a852b0SQu Wenruo } 296721a852b0SQu Wenruo 2968069ec957SQu Wenruo /* 2969069ec957SQu Wenruo * Validation of super block at mount time. 2970069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2971069ec957SQu Wenruo * flags will be skipped. 2972069ec957SQu Wenruo */ 2973069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2974069ec957SQu Wenruo { 2975069ec957SQu Wenruo return validate_super(fs_info, fs_info->super_copy, 0); 2976069ec957SQu Wenruo } 2977069ec957SQu Wenruo 297875cb857dSQu Wenruo /* 297975cb857dSQu Wenruo * Validation of super block at write time. 298075cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 298175cb857dSQu Wenruo * overwritten soon. 298275cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 298375cb857dSQu Wenruo */ 298475cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 298575cb857dSQu Wenruo struct btrfs_super_block *sb) 298675cb857dSQu Wenruo { 298775cb857dSQu Wenruo int ret; 298875cb857dSQu Wenruo 298975cb857dSQu Wenruo ret = validate_super(fs_info, sb, -1); 299075cb857dSQu Wenruo if (ret < 0) 299175cb857dSQu Wenruo goto out; 2992e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 299375cb857dSQu Wenruo ret = -EUCLEAN; 299475cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 299575cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 299675cb857dSQu Wenruo goto out; 299775cb857dSQu Wenruo } 299875cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 299975cb857dSQu Wenruo ret = -EUCLEAN; 300075cb857dSQu Wenruo btrfs_err(fs_info, 300175cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 300275cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 300375cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 300475cb857dSQu Wenruo goto out; 300575cb857dSQu Wenruo } 300675cb857dSQu Wenruo out: 300775cb857dSQu Wenruo if (ret < 0) 300875cb857dSQu Wenruo btrfs_err(fs_info, 300975cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 301075cb857dSQu Wenruo return ret; 301175cb857dSQu Wenruo } 301275cb857dSQu Wenruo 3013bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 3014bd676446SJosef Bacik { 3015bd676446SJosef Bacik int ret = 0; 3016bd676446SJosef Bacik 3017bd676446SJosef Bacik root->node = read_tree_block(root->fs_info, bytenr, 3018bd676446SJosef Bacik root->root_key.objectid, gen, level, NULL); 3019bd676446SJosef Bacik if (IS_ERR(root->node)) { 3020bd676446SJosef Bacik ret = PTR_ERR(root->node); 3021bd676446SJosef Bacik root->node = NULL; 30224eb150d6SQu Wenruo return ret; 30234eb150d6SQu Wenruo } 30244eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 3025bd676446SJosef Bacik free_extent_buffer(root->node); 3026bd676446SJosef Bacik root->node = NULL; 30274eb150d6SQu Wenruo return -EIO; 3028bd676446SJosef Bacik } 3029bd676446SJosef Bacik 3030bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 3031bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 3032bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 3033bd676446SJosef Bacik return ret; 3034bd676446SJosef Bacik } 3035bd676446SJosef Bacik 3036bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 3037bd676446SJosef Bacik { 3038bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 3039bd676446SJosef Bacik u64 gen, bytenr; 3040bd676446SJosef Bacik int level, ret; 3041bd676446SJosef Bacik 3042bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 3043bd676446SJosef Bacik gen = btrfs_super_generation(sb); 3044bd676446SJosef Bacik level = btrfs_super_root_level(sb); 3045bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 30469c54e80dSJosef Bacik if (ret) { 3047bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 3048bd676446SJosef Bacik return ret; 3049bd676446SJosef Bacik } 3050bd676446SJosef Bacik 30519c54e80dSJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 30529c54e80dSJosef Bacik return 0; 30539c54e80dSJosef Bacik 30549c54e80dSJosef Bacik bytenr = btrfs_super_block_group_root(sb); 30559c54e80dSJosef Bacik gen = btrfs_super_block_group_root_generation(sb); 30569c54e80dSJosef Bacik level = btrfs_super_block_group_root_level(sb); 30579c54e80dSJosef Bacik ret = load_super_root(fs_info->block_group_root, bytenr, gen, level); 30589c54e80dSJosef Bacik if (ret) 30599c54e80dSJosef Bacik btrfs_warn(fs_info, "couldn't read block group root"); 30609c54e80dSJosef Bacik return ret; 30619c54e80dSJosef Bacik } 30629c54e80dSJosef Bacik 30636ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 3064b8522a1eSNikolay Borisov { 30656ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 3066b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 3067b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 3068b8522a1eSNikolay Borisov bool handle_error = false; 3069b8522a1eSNikolay Borisov int ret = 0; 3070b8522a1eSNikolay Borisov int i; 3071b8522a1eSNikolay Borisov 30729c54e80dSJosef Bacik if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) { 30739c54e80dSJosef Bacik struct btrfs_root *root; 3074b8522a1eSNikolay Borisov 30759c54e80dSJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_BLOCK_GROUP_TREE_OBJECTID, 30769c54e80dSJosef Bacik GFP_KERNEL); 30779c54e80dSJosef Bacik if (!root) 30789c54e80dSJosef Bacik return -ENOMEM; 30799c54e80dSJosef Bacik fs_info->block_group_root = root; 30809c54e80dSJosef Bacik } 30819c54e80dSJosef Bacik 3082b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 3083b8522a1eSNikolay Borisov if (handle_error) { 3084b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 3085b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 3086b8522a1eSNikolay Borisov tree_root->node = NULL; 3087b8522a1eSNikolay Borisov 3088b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 3089b8522a1eSNikolay Borisov break; 3090b8522a1eSNikolay Borisov 3091b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 3092b8522a1eSNikolay Borisov 3093b8522a1eSNikolay Borisov /* 3094b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 3095b8522a1eSNikolay Borisov * valid 3096b8522a1eSNikolay Borisov */ 3097b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 3098b8522a1eSNikolay Borisov 3099b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 3100b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3101b8522a1eSNikolay Borisov 3102b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 31036ef108ddSNikolay Borisov backup_index = ret; 3104b8522a1eSNikolay Borisov if (ret < 0) 3105b8522a1eSNikolay Borisov return ret; 3106b8522a1eSNikolay Borisov } 3107b8522a1eSNikolay Borisov 3108bd676446SJosef Bacik ret = load_important_roots(fs_info); 3109bd676446SJosef Bacik if (ret) { 3110217f5004SNikolay Borisov handle_error = true; 3111b8522a1eSNikolay Borisov continue; 3112b8522a1eSNikolay Borisov } 3113b8522a1eSNikolay Borisov 3114336a0d8dSNikolay Borisov /* 3115336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 3116336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 3117336a0d8dSNikolay Borisov */ 3118453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 3119b8522a1eSNikolay Borisov if (ret < 0) { 3120b8522a1eSNikolay Borisov handle_error = true; 3121b8522a1eSNikolay Borisov continue; 3122b8522a1eSNikolay Borisov } 3123b8522a1eSNikolay Borisov 31246b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 3125b8522a1eSNikolay Borisov 3126b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 3127b8522a1eSNikolay Borisov if (ret < 0) { 3128b8522a1eSNikolay Borisov handle_error = true; 3129b8522a1eSNikolay Borisov continue; 3130b8522a1eSNikolay Borisov } 3131b8522a1eSNikolay Borisov 3132b8522a1eSNikolay Borisov /* All successful */ 3133bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 3134bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 3135d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 31366ef108ddSNikolay Borisov 31376ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 31386ef108ddSNikolay Borisov if (backup_index < 0) { 31396ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 31406ef108ddSNikolay Borisov } else { 31416ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 31426ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 31436ef108ddSNikolay Borisov } 3144b8522a1eSNikolay Borisov break; 3145b8522a1eSNikolay Borisov } 3146b8522a1eSNikolay Borisov 3147b8522a1eSNikolay Borisov return ret; 3148b8522a1eSNikolay Borisov } 3149b8522a1eSNikolay Borisov 31508260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 3151eb60ceacSChris Mason { 3152fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 315301cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 31548fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 3155facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 315624bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 3157eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 315811833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 3159eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 3160a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 3161fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 316224bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 31634cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 3164ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 3165f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 316647ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 316740ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 3168afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 3169c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 3170f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 3171abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 3172d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 3173f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 31747585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 3175573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 31760bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 31775f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 3178de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 317919c00ddcSChris Mason 31800b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 31816324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 3182f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 318347ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 318418bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 3185afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 3186bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 3187bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 31883fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 31893fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 3190bd647ce3SJosef Bacik #endif 3191c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 319266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 319366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 319466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 319566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 319666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 319766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 319866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 3199ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 3200ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 3201ba2c4d4eSJosef Bacik 3202771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 32034cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 3204034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 3205fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 3206abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 320795ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 32089ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 32094cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 3210a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 3211f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 32128b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 3213f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 3214fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 3215c8b97818SChris Mason 3216b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 3217b34b086cSChris Mason num_online_cpus() + 2, 8); 32180afbaf8cSChris Mason 3219199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 3220199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 322169fe2d75SJosef Bacik 3222638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 322321adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 322421adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 322521adbd5cSStefan Behrens #endif 3226779a65a4SEric Sandeen btrfs_init_balance(fs_info); 322757056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 3228a2de733cSArne Jansen 322916b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 323008dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 32310f9dd46cSJosef Bacik 3232fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 3233fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 323439279cc3SChris Mason 32355a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 3236e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 3237925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 3238a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 3239a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 32401bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 32419e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 3242c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 324376dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 3244803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 3245fa9c0d79SChris Mason 3246ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 3247f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 3248b0643e59SDennis Zhou btrfs_discard_init(fs_info); 3249416ac51dSArne Jansen 3250fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 3251fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 3252fa9c0d79SChris Mason 3253e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 3254f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 3255bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 32564854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 3257034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 32583768f368SChris Mason 3259da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 3260da17066cSJeff Mahoney fs_info->nodesize = 4096; 3261da17066cSJeff Mahoney fs_info->sectorsize = 4096; 3262ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 3263da17066cSJeff Mahoney fs_info->stripesize = 4096; 3264da17066cSJeff Mahoney 3265eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 3266eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 3267eede2bf3SOmar Sandoval 326818bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 326918bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 32708260edbaSJosef Bacik } 32718260edbaSJosef Bacik 32728260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 32738260edbaSJosef Bacik { 32748260edbaSJosef Bacik int ret; 32758260edbaSJosef Bacik 32768260edbaSJosef Bacik fs_info->sb = sb; 32778260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 32788260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 32799e967495SFilipe Manana 32805deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 3281ae18c37aSJosef Bacik if (ret) 3282ae18c37aSJosef Bacik return ret; 3283ae18c37aSJosef Bacik 3284ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 3285ae18c37aSJosef Bacik if (ret) 3286c75e8394SJosef Bacik return ret; 3287ae18c37aSJosef Bacik 3288ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 3289ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 3290ae18c37aSJosef Bacik 3291ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 3292ae18c37aSJosef Bacik if (ret) 3293c75e8394SJosef Bacik return ret; 3294ae18c37aSJosef Bacik 3295ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 3296ae18c37aSJosef Bacik GFP_KERNEL); 3297ae18c37aSJosef Bacik if (ret) 3298c75e8394SJosef Bacik return ret; 3299ae18c37aSJosef Bacik 3300ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 3301ae18c37aSJosef Bacik GFP_KERNEL); 3302c75e8394SJosef Bacik if (!fs_info->delayed_root) 3303c75e8394SJosef Bacik return -ENOMEM; 3304ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 3305ae18c37aSJosef Bacik 3306a0a1db70SFilipe Manana if (sb_rdonly(sb)) 3307a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 3308a0a1db70SFilipe Manana 3309c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 3310ae18c37aSJosef Bacik } 3311ae18c37aSJosef Bacik 331297f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 331397f4dd09SNikolay Borisov { 33140d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 331597f4dd09SNikolay Borisov int ret; 331697f4dd09SNikolay Borisov 331797f4dd09SNikolay Borisov /* 331897f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 331997f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 332097f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 332197f4dd09SNikolay Borisov */ 332297f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 332397f4dd09SNikolay Borisov if (ret < 0) { 3324c94bec2cSJosef Bacik if (ret != -EINTR) 3325c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 3326c94bec2cSJosef Bacik ret); 332797f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 332897f4dd09SNikolay Borisov return ret; 332997f4dd09SNikolay Borisov } 333097f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 333197f4dd09SNikolay Borisov } 333297f4dd09SNikolay Borisov 333397f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 333497f4dd09SNikolay Borisov { 333597f4dd09SNikolay Borisov struct task_struct *task; 333697f4dd09SNikolay Borisov 333797f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 333897f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 333997f4dd09SNikolay Borisov if (IS_ERR(task)) { 334097f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 334197f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 334297f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 334397f4dd09SNikolay Borisov return PTR_ERR(task); 334497f4dd09SNikolay Borisov } 334597f4dd09SNikolay Borisov 334697f4dd09SNikolay Borisov return 0; 334797f4dd09SNikolay Borisov } 334897f4dd09SNikolay Borisov 334944c0ca21SBoris Burkov /* 33508cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 33518cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 33528cd29088SBoris Burkov * code paths. 33538cd29088SBoris Burkov */ 33548cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 33558cd29088SBoris Burkov { 33568cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 33578b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 33588cd29088SBoris Burkov } 33598cd29088SBoris Burkov 33608cd29088SBoris Burkov /* 336144c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 336244c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 336344c0ca21SBoris Burkov */ 336444c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 336544c0ca21SBoris Burkov { 336644c0ca21SBoris Burkov int ret; 336794846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 33688b228324SBoris Burkov bool clear_free_space_tree = false; 33698b228324SBoris Burkov 33708b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 33718b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 33728b228324SBoris Burkov clear_free_space_tree = true; 33738b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 33748b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 33758b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 33768b228324SBoris Burkov clear_free_space_tree = true; 33778b228324SBoris Burkov } 33788b228324SBoris Burkov 33798b228324SBoris Burkov if (clear_free_space_tree) { 33808b228324SBoris Burkov btrfs_info(fs_info, "clearing free space tree"); 33818b228324SBoris Burkov ret = btrfs_clear_free_space_tree(fs_info); 33828b228324SBoris Burkov if (ret) { 33838b228324SBoris Burkov btrfs_warn(fs_info, 33848b228324SBoris Burkov "failed to clear free space tree: %d", ret); 33858b228324SBoris Burkov goto out; 33868b228324SBoris Burkov } 33878b228324SBoris Burkov } 338844c0ca21SBoris Burkov 33898d488a8cSFilipe Manana /* 33908d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 33918d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 3392fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 33938d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 33948d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 33958d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 33968d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 33978d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 33988d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 33998d488a8cSFilipe Manana */ 34008d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 34018d488a8cSFilipe Manana if (ret) 34028d488a8cSFilipe Manana goto out; 34038d488a8cSFilipe Manana 340444c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 340544c0ca21SBoris Burkov if (ret) 340644c0ca21SBoris Burkov goto out; 340744c0ca21SBoris Burkov 34088f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 34098f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 34108f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 34118f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 34128f1c21d7SBoris Burkov goto out; 34138f1c21d7SBoris Burkov } 34148f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 34158f1c21d7SBoris Burkov 341644c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 34177eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 341844c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 341944c0ca21SBoris Burkov if (ret < 0) { 342044c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 342144c0ca21SBoris Burkov goto out; 342244c0ca21SBoris Burkov } 342344c0ca21SBoris Burkov 34245011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 34255011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 34265011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 34275011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 34285011139aSBoris Burkov if (ret) { 34295011139aSBoris Burkov btrfs_warn(fs_info, 34305011139aSBoris Burkov "failed to create free space tree: %d", ret); 34315011139aSBoris Burkov goto out; 34325011139aSBoris Burkov } 34335011139aSBoris Burkov } 34345011139aSBoris Burkov 343594846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 343694846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 343794846229SBoris Burkov if (ret) 343894846229SBoris Burkov goto out; 343994846229SBoris Burkov } 344094846229SBoris Burkov 344144c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 344244c0ca21SBoris Burkov if (ret) 344344c0ca21SBoris Burkov goto out; 344444c0ca21SBoris Burkov 344544c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 344644c0ca21SBoris Burkov if (ret) { 344744c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 344844c0ca21SBoris Burkov goto out; 344944c0ca21SBoris Burkov } 345044c0ca21SBoris Burkov 345144c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 345244c0ca21SBoris Burkov 345344c0ca21SBoris Burkov if (!fs_info->uuid_root) { 345444c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 345544c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 345644c0ca21SBoris Burkov if (ret) { 345744c0ca21SBoris Burkov btrfs_warn(fs_info, 345844c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 345944c0ca21SBoris Burkov goto out; 346044c0ca21SBoris Burkov } 346144c0ca21SBoris Burkov } 346244c0ca21SBoris Burkov 346344c0ca21SBoris Burkov out: 346444c0ca21SBoris Burkov return ret; 346544c0ca21SBoris Burkov } 346644c0ca21SBoris Burkov 3467ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3468ae18c37aSJosef Bacik char *options) 3469ae18c37aSJosef Bacik { 3470ae18c37aSJosef Bacik u32 sectorsize; 3471ae18c37aSJosef Bacik u32 nodesize; 3472ae18c37aSJosef Bacik u32 stripesize; 3473ae18c37aSJosef Bacik u64 generation; 3474ae18c37aSJosef Bacik u64 features; 3475ae18c37aSJosef Bacik u16 csum_type; 3476ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3477ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3478ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3479ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3480ae18c37aSJosef Bacik int ret; 3481ae18c37aSJosef Bacik int err = -EINVAL; 3482ae18c37aSJosef Bacik int level; 3483ae18c37aSJosef Bacik 34848260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 348553b381b3SDavid Woodhouse if (ret) { 348683c8266aSDavid Sterba err = ret; 3487ae18c37aSJosef Bacik goto fail; 348853b381b3SDavid Woodhouse } 348953b381b3SDavid Woodhouse 3490ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3491ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3492ae18c37aSJosef Bacik GFP_KERNEL); 3493ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3494ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3495ae18c37aSJosef Bacik GFP_KERNEL); 3496ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3497ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 3498ae18c37aSJosef Bacik err = -ENOMEM; 3499c75e8394SJosef Bacik goto fail; 3500ae18c37aSJosef Bacik } 3501ae18c37aSJosef Bacik 3502ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 3503ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 3504ae18c37aSJosef Bacik err = -ENOMEM; 3505c75e8394SJosef Bacik goto fail; 3506ae18c37aSJosef Bacik } 3507ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 3508ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 3509ae18c37aSJosef Bacik 3510d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 35111104a885SDavid Sterba 35121104a885SDavid Sterba /* 35131104a885SDavid Sterba * Read super block and check the signature bytes only 35141104a885SDavid Sterba */ 3515d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 35168f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 35178f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 351816cdcec7SMiao Xie goto fail_alloc; 351920b45077SDave Young } 352039279cc3SChris Mason 35211104a885SDavid Sterba /* 3522260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 35238dc3f22cSJohannes Thumshirn * corrupted, we'll find out 35248dc3f22cSJohannes Thumshirn */ 35258f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 352651bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 35278dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 352851bce6c9SJohannes Thumshirn csum_type); 35298dc3f22cSJohannes Thumshirn err = -EINVAL; 35308f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35318dc3f22cSJohannes Thumshirn goto fail_alloc; 35328dc3f22cSJohannes Thumshirn } 35338dc3f22cSJohannes Thumshirn 353483c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 353583c68bbcSSu Yue 35366d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 35376d97c6e3SJohannes Thumshirn if (ret) { 35386d97c6e3SJohannes Thumshirn err = ret; 35398f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35406d97c6e3SJohannes Thumshirn goto fail_alloc; 35416d97c6e3SJohannes Thumshirn } 35426d97c6e3SJohannes Thumshirn 35438dc3f22cSJohannes Thumshirn /* 35441104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 35451104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 35461104a885SDavid Sterba */ 35478f32380dSJohannes Thumshirn if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { 354805135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 35491104a885SDavid Sterba err = -EINVAL; 35508f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3551141386e1SJosef Bacik goto fail_alloc; 35521104a885SDavid Sterba } 35531104a885SDavid Sterba 35541104a885SDavid Sterba /* 35551104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 35561104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 35571104a885SDavid Sterba * the whole block of INFO_SIZE 35581104a885SDavid Sterba */ 35598f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 35608f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35615f39d397SChris Mason 3562fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3563fbc6feaeSNikolay Borisov 35640b86a832SChris Mason 3565fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3566fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3567fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3568fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3569fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3570fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3571fbc6feaeSNikolay Borisov } 3572fbc6feaeSNikolay Borisov 3573fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3574fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3575de37aa51SNikolay Borisov 3576069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 35771104a885SDavid Sterba if (ret) { 357805135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 35791104a885SDavid Sterba err = -EINVAL; 3580141386e1SJosef Bacik goto fail_alloc; 35811104a885SDavid Sterba } 35821104a885SDavid Sterba 35830f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3584141386e1SJosef Bacik goto fail_alloc; 35850f7d52f4SChris Mason 3586acce952bSliubo /* check FS state, whether FS is broken. */ 358787533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 358887533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3589acce952bSliubo 359075e7cb7fSLiu Bo /* 359175e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 359275e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 359375e7cb7fSLiu Bo */ 359475e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 359575e7cb7fSLiu Bo 35966f93e834SAnand Jain /* 35976f93e834SAnand Jain * Flag our filesystem as having big metadata blocks if they are bigger 35986f93e834SAnand Jain * than the page size. 35996f93e834SAnand Jain */ 36006f93e834SAnand Jain if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 36016f93e834SAnand Jain if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 36026f93e834SAnand Jain btrfs_info(fs_info, 36036f93e834SAnand Jain "flagging fs with big metadata feature"); 36046f93e834SAnand Jain features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 36056f93e834SAnand Jain } 36066f93e834SAnand Jain 36076f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 36086f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 36096f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 36106f93e834SAnand Jain stripesize = sectorsize; 36116f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 36126f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 36136f93e834SAnand Jain 36146f93e834SAnand Jain fs_info->nodesize = nodesize; 36156f93e834SAnand Jain fs_info->sectorsize = sectorsize; 36166f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 36176f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 36186f93e834SAnand Jain fs_info->stripesize = stripesize; 36196f93e834SAnand Jain 36202ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 36212b82032cSYan Zheng if (ret) { 36222b82032cSYan Zheng err = ret; 3623141386e1SJosef Bacik goto fail_alloc; 36242b82032cSYan Zheng } 3625dfe25020SChris Mason 3626f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 3627f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 3628f2b636e8SJosef Bacik if (features) { 362905135f59SDavid Sterba btrfs_err(fs_info, 3630d5321a0fSQu Wenruo "cannot mount because of unsupported optional features (0x%llx)", 3631c1c9ff7cSGeert Uytterhoeven features); 3632f2b636e8SJosef Bacik err = -EINVAL; 3633141386e1SJosef Bacik goto fail_alloc; 3634f2b636e8SJosef Bacik } 3635f2b636e8SJosef Bacik 36365d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 36375d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 36380b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3639a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 36405c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 36415c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3642727011e0SChris Mason 36433173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 364405135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 36453173a18fSJosef Bacik 3646727011e0SChris Mason /* 3647bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 3648bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 3649bc3f116fSChris Mason */ 3650bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3651707e8a07SDavid Sterba (sectorsize != nodesize)) { 365205135f59SDavid Sterba btrfs_err(fs_info, 365305135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 365405135f59SDavid Sterba nodesize, sectorsize); 3655141386e1SJosef Bacik goto fail_alloc; 3656bc3f116fSChris Mason } 3657bc3f116fSChris Mason 3658ceda0864SMiao Xie /* 3659ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 3660ceda0864SMiao Xie * update the flag. 3661ceda0864SMiao Xie */ 36625d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 36635d4f98a2SYan Zheng 3664f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 3665f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 3666bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 366705135f59SDavid Sterba btrfs_err(fs_info, 3668d5321a0fSQu Wenruo "cannot mount read-write because of unsupported optional features (0x%llx)", 3669c1c9ff7cSGeert Uytterhoeven features); 3670f2b636e8SJosef Bacik err = -EINVAL; 3671141386e1SJosef Bacik goto fail_alloc; 3672f2b636e8SJosef Bacik } 367361d92c32SChris Mason 36748481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 36758481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 36768481dd80SQu Wenruo 36779f73f1aeSQu Wenruo /* 36789f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 36799f73f1aeSQu Wenruo * going to be deprecated. 36809f73f1aeSQu Wenruo * 36819f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 36829f73f1aeSQu Wenruo */ 36839f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 36849f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 36859f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 36869f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 36879f73f1aeSQu Wenruo 368895ea0486SQu Wenruo btrfs_warn(fs_info, 368995ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 36900bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 36918481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 36928481dd80SQu Wenruo if (!subpage_info) 36938481dd80SQu Wenruo goto fail_alloc; 36948481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 36958481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3696c8050b3bSQu Wenruo } 36970bb3eb3eSQu Wenruo 3698d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 36992a458198SEric Sandeen if (ret) { 37002a458198SEric Sandeen err = ret; 37010dc3b84aSJosef Bacik goto fail_sb_buffer; 37020dc3b84aSJosef Bacik } 37034543df7eSChris Mason 37049e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 37059e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 37064575c9ccSChris Mason 3707a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3708a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3709de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3710db94535dSChris Mason 3711925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 37126bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3713925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 371484eed90fSChris Mason if (ret) { 371505135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 37165d4f98a2SYan Zheng goto fail_sb_buffer; 371784eed90fSChris Mason } 37180b86a832SChris Mason 371984234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3720581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3721bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3722bd676446SJosef Bacik generation, level); 3723bd676446SJosef Bacik if (ret) { 372405135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3725af31f5e5SChris Mason goto fail_tree_roots; 372683121942SDavid Woodhouse } 37270b86a832SChris Mason 3728e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3729c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3730c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3731e17cade2SChris Mason 37325b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 37332b82032cSYan Zheng if (ret) { 373405135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3735af31f5e5SChris Mason goto fail_tree_roots; 37362b82032cSYan Zheng } 37370b86a832SChris Mason 37388dabb742SStefan Behrens /* 3739bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3740bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3741bacce86aSAnand Jain * target is required. So free devices that are not part of this 37421a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3743bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 37448dabb742SStefan Behrens */ 3745bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3746d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 374705135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3748a6b0d5c8SChris Mason goto fail_tree_roots; 3749a6b0d5c8SChris Mason } 3750a6b0d5c8SChris Mason 3751b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 37524bbcaa64SEric Sandeen if (ret) 3753b8522a1eSNikolay Borisov goto fail_tree_roots; 37548929ecfaSYan, Zheng 375575ec1db8SJosef Bacik /* 375673651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 375773651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 375873651042SNaohiro Aota * zoned incompat feature flag set. 375973651042SNaohiro Aota */ 376073651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 376173651042SNaohiro Aota if (ret) { 376273651042SNaohiro Aota btrfs_err(fs_info, 376373651042SNaohiro Aota "zoned: failed to read device zone info: %d", 376473651042SNaohiro Aota ret); 376573651042SNaohiro Aota goto fail_block_groups; 376673651042SNaohiro Aota } 376773651042SNaohiro Aota 376873651042SNaohiro Aota /* 376975ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 377075ec1db8SJosef Bacik * check the generation here so we can set the 377175ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 377275ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 377375ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 377475ec1db8SJosef Bacik * even though it was perfectly fine. 377575ec1db8SJosef Bacik */ 377675ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 377775ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 377875ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 377975ec1db8SJosef Bacik 3780cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3781cf90d884SQu Wenruo if (ret) { 3782cf90d884SQu Wenruo btrfs_err(fs_info, 3783cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3784cf90d884SQu Wenruo ret); 3785cf90d884SQu Wenruo goto fail_block_groups; 3786cf90d884SQu Wenruo } 378768310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 378868310a5eSIlya Dryomov if (ret) { 378905135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 379068310a5eSIlya Dryomov goto fail_block_groups; 379168310a5eSIlya Dryomov } 379268310a5eSIlya Dryomov 3793733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3794733f4fbbSStefan Behrens if (ret) { 379505135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3796733f4fbbSStefan Behrens goto fail_block_groups; 3797733f4fbbSStefan Behrens } 3798733f4fbbSStefan Behrens 37998dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 38008dabb742SStefan Behrens if (ret) { 380105135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 38028dabb742SStefan Behrens goto fail_block_groups; 38038dabb742SStefan Behrens } 38048dabb742SStefan Behrens 3805b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3806b70f5097SNaohiro Aota if (ret) { 3807b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3808b70f5097SNaohiro Aota ret); 3809b70f5097SNaohiro Aota goto fail_block_groups; 3810b70f5097SNaohiro Aota } 3811b70f5097SNaohiro Aota 3812c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3813b7c35e81SAnand Jain if (ret) { 381405135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 381505135f59SDavid Sterba ret); 3816b7c35e81SAnand Jain goto fail_block_groups; 3817b7c35e81SAnand Jain } 3818b7c35e81SAnand Jain 381996f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 38205ac1d209SJeff Mahoney if (ret) { 382105135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3822b7c35e81SAnand Jain goto fail_fsdev_sysfs; 38235ac1d209SJeff Mahoney } 38245ac1d209SJeff Mahoney 3825c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3826c59021f8Sliubo if (ret) { 382705135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 38282365dd3cSAnand Jain goto fail_sysfs; 3829c59021f8Sliubo } 3830c59021f8Sliubo 38315b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 38321b1d1f66SJosef Bacik if (ret) { 383305135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 38342365dd3cSAnand Jain goto fail_sysfs; 38351b1d1f66SJosef Bacik } 38364330e183SQu Wenruo 383716beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 383816beac87SNaohiro Aota 38395c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 38405c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 384105135f59SDavid Sterba btrfs_warn(fs_info, 384252042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 38432365dd3cSAnand Jain goto fail_sysfs; 3844292fd7fcSStefan Behrens } 38459078a3e1SChris Mason 384633c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3847a74a4b97SChris Mason "btrfs-cleaner"); 384857506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 38492365dd3cSAnand Jain goto fail_sysfs; 3850a74a4b97SChris Mason 3851a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3852a74a4b97SChris Mason tree_root, 3853a74a4b97SChris Mason "btrfs-transaction"); 385457506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 38553f157a2fSChris Mason goto fail_cleaner; 3856a74a4b97SChris Mason 3857583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3858c289811cSChris Mason !fs_info->fs_devices->rotating) { 3859583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3860c289811cSChris Mason } 3861c289811cSChris Mason 3862572d9ab7SDavid Sterba /* 386301327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 3864572d9ab7SDavid Sterba * not have to wait for transaction commit 3865572d9ab7SDavid Sterba */ 3866572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 38673818aea2SQu Wenruo 386821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 38690b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 38702ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 38710b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3872cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 387321adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 387421adbd5cSStefan Behrens if (ret) 387505135f59SDavid Sterba btrfs_warn(fs_info, 387605135f59SDavid Sterba "failed to initialize integrity check module: %d", 387705135f59SDavid Sterba ret); 387821adbd5cSStefan Behrens } 387921adbd5cSStefan Behrens #endif 3880bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3881bcef60f2SArne Jansen if (ret) 3882bcef60f2SArne Jansen goto fail_trans_kthread; 388321adbd5cSStefan Behrens 3884fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3885fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3886fd708b81SJosef Bacik 388796da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 388896da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 38890b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3890e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 389163443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 389279787eaaSJeff Mahoney if (ret) { 389363443bf5SEric Sandeen err = ret; 389428c16cbbSWang Shilong goto fail_qgroup; 3895e556ce2cSYan Zheng } 3896e02119d5SChris Mason } 38971a40e23bSZheng Yan 389856e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 38993140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 39003140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3901f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3902315bf8efSJosef Bacik fs_info->fs_root = NULL; 3903bcef60f2SArne Jansen goto fail_qgroup; 39043140c9a3SDan Carpenter } 3905c289811cSChris Mason 3906bc98a42cSDavid Howells if (sb_rdonly(sb)) 39078cd29088SBoris Burkov goto clear_oneshot; 39082b6ba629SIlya Dryomov 390944c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 39102b6ba629SIlya Dryomov if (ret) { 39116bccf3abSJeff Mahoney close_ctree(fs_info); 39122b6ba629SIlya Dryomov return ret; 3913e3acc2a6SJosef Bacik } 3914b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3915b382a324SJan Schmidt 391644c0ca21SBoris Burkov if (fs_info->uuid_root && 391744c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 391844c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 391905135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 392070f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 392170f80175SStefan Behrens if (ret) { 392205135f59SDavid Sterba btrfs_warn(fs_info, 392305135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 39246bccf3abSJeff Mahoney close_ctree(fs_info); 392570f80175SStefan Behrens return ret; 392670f80175SStefan Behrens } 3927f7a81ea4SStefan Behrens } 392894846229SBoris Burkov 3929afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 393047ab2a6cSJosef Bacik 3931b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3932b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3933b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3934b4be6aefSJosef Bacik 39358cd29088SBoris Burkov clear_oneshot: 39368cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3937ad2b2c80SAl Viro return 0; 393839279cc3SChris Mason 3939bcef60f2SArne Jansen fail_qgroup: 3940bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 39417c2ca468SChris Mason fail_trans_kthread: 39427c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 39432ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3944faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 39453f157a2fSChris Mason fail_cleaner: 3946a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 39477c2ca468SChris Mason 39487c2ca468SChris Mason /* 39497c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 39507c2ca468SChris Mason * kthreads 39517c2ca468SChris Mason */ 39527c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 39537c2ca468SChris Mason 39542365dd3cSAnand Jain fail_sysfs: 39556618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 39562365dd3cSAnand Jain 3957b7c35e81SAnand Jain fail_fsdev_sysfs: 3958b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3959b7c35e81SAnand Jain 39601b1d1f66SJosef Bacik fail_block_groups: 396154067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3962af31f5e5SChris Mason 3963af31f5e5SChris Mason fail_tree_roots: 39649e3aa805SJosef Bacik if (fs_info->data_reloc_root) 39659e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 39664273eaffSAnand Jain free_root_pointers(fs_info, true); 39672b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3968af31f5e5SChris Mason 396939279cc3SChris Mason fail_sb_buffer: 39707abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 39715cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 397216cdcec7SMiao Xie fail_alloc: 3973586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3974586e46e2SIlya Dryomov 39754543df7eSChris Mason iput(fs_info->btree_inode); 39767e662854SQinghuang Feng fail: 3977586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3978ad2b2c80SAl Viro return err; 3979eb60ceacSChris Mason } 3980663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3981eb60ceacSChris Mason 3982314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3983f2984462SChris Mason { 3984314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3985314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3986314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3987314b6dd0SJohannes Thumshirn struct page *page; 3988442a4f63SStefan Behrens 3989314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3990314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3991314b6dd0SJohannes Thumshirn 3992314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3993fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3994314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3995314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3996314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3997314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3998314b6dd0SJohannes Thumshirn SetPageError(page); 3999314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 4000314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 4001314b6dd0SJohannes Thumshirn } else { 4002314b6dd0SJohannes Thumshirn SetPageUptodate(page); 4003f2984462SChris Mason } 4004314b6dd0SJohannes Thumshirn 4005314b6dd0SJohannes Thumshirn put_page(page); 4006314b6dd0SJohannes Thumshirn unlock_page(page); 4007314b6dd0SJohannes Thumshirn } 4008314b6dd0SJohannes Thumshirn 4009314b6dd0SJohannes Thumshirn bio_put(bio); 4010f2984462SChris Mason } 4011f2984462SChris Mason 40128f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 40138f32380dSJohannes Thumshirn int copy_num) 401429c36d72SAnand Jain { 401529c36d72SAnand Jain struct btrfs_super_block *super; 40168f32380dSJohannes Thumshirn struct page *page; 401712659251SNaohiro Aota u64 bytenr, bytenr_orig; 40188f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 401912659251SNaohiro Aota int ret; 402029c36d72SAnand Jain 402112659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 402212659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 402312659251SNaohiro Aota if (ret == -ENOENT) 402412659251SNaohiro Aota return ERR_PTR(-EINVAL); 402512659251SNaohiro Aota else if (ret) 402612659251SNaohiro Aota return ERR_PTR(ret); 402712659251SNaohiro Aota 4028cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 40298f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 403029c36d72SAnand Jain 40318f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 40328f32380dSJohannes Thumshirn if (IS_ERR(page)) 40338f32380dSJohannes Thumshirn return ERR_CAST(page); 403429c36d72SAnand Jain 40358f32380dSJohannes Thumshirn super = page_address(page); 403696c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 403796c2e067SAnand Jain btrfs_release_disk_super(super); 403896c2e067SAnand Jain return ERR_PTR(-ENODATA); 403996c2e067SAnand Jain } 404096c2e067SAnand Jain 404112659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 40428f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 40438f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 404429c36d72SAnand Jain } 404529c36d72SAnand Jain 40468f32380dSJohannes Thumshirn return super; 404729c36d72SAnand Jain } 404829c36d72SAnand Jain 404929c36d72SAnand Jain 40508f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 4051a512bbf8SYan Zheng { 40528f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 4053a512bbf8SYan Zheng int i; 4054a512bbf8SYan Zheng u64 transid = 0; 4055a512bbf8SYan Zheng 4056a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 4057a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 4058a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 4059a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 4060a512bbf8SYan Zheng */ 4061a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 40628f32380dSJohannes Thumshirn super = btrfs_read_dev_one_super(bdev, i); 40638f32380dSJohannes Thumshirn if (IS_ERR(super)) 4064a512bbf8SYan Zheng continue; 4065a512bbf8SYan Zheng 4066a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 40678f32380dSJohannes Thumshirn if (latest) 40688f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 40698f32380dSJohannes Thumshirn 40708f32380dSJohannes Thumshirn latest = super; 4071a512bbf8SYan Zheng transid = btrfs_super_generation(super); 4072a512bbf8SYan Zheng } 4073a512bbf8SYan Zheng } 407492fc03fbSAnand Jain 40758f32380dSJohannes Thumshirn return super; 4076a512bbf8SYan Zheng } 4077a512bbf8SYan Zheng 40784eedeb75SHisashi Hifumi /* 4079abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 4080314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 40814eedeb75SHisashi Hifumi * 4082abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 4083abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 4084abbb3b8eSDavid Sterba * same for write and wait phases. 40854eedeb75SHisashi Hifumi * 4086314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 40874eedeb75SHisashi Hifumi */ 4088a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 4089abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 4090a512bbf8SYan Zheng { 4091d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 4092314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 4093d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 4094a512bbf8SYan Zheng int i; 4095a512bbf8SYan Zheng int errors = 0; 409612659251SNaohiro Aota int ret; 409712659251SNaohiro Aota u64 bytenr, bytenr_orig; 4098a512bbf8SYan Zheng 4099a512bbf8SYan Zheng if (max_mirrors == 0) 4100a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4101a512bbf8SYan Zheng 4102d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 4103d5178578SJohannes Thumshirn 4104a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 4105314b6dd0SJohannes Thumshirn struct page *page; 4106314b6dd0SJohannes Thumshirn struct bio *bio; 4107314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 4108314b6dd0SJohannes Thumshirn 410912659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 411012659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 411112659251SNaohiro Aota if (ret == -ENOENT) { 411212659251SNaohiro Aota continue; 411312659251SNaohiro Aota } else if (ret < 0) { 411412659251SNaohiro Aota btrfs_err(device->fs_info, 411512659251SNaohiro Aota "couldn't get super block location for mirror %d", 411612659251SNaohiro Aota i); 411712659251SNaohiro Aota errors++; 411812659251SNaohiro Aota continue; 411912659251SNaohiro Aota } 4120935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4121935e5cc9SMiao Xie device->commit_total_bytes) 4122a512bbf8SYan Zheng break; 4123a512bbf8SYan Zheng 412412659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 4125a512bbf8SYan Zheng 4126fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 4127fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 4128fd08001fSEric Biggers sb->csum); 4129a512bbf8SYan Zheng 4130314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 4131314b6dd0SJohannes Thumshirn GFP_NOFS); 4132314b6dd0SJohannes Thumshirn if (!page) { 4133fb456252SJeff Mahoney btrfs_err(device->fs_info, 4134314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 4135f14d104dSDavid Sterba bytenr); 4136634554dcSJosef Bacik errors++; 4137634554dcSJosef Bacik continue; 4138634554dcSJosef Bacik } 4139634554dcSJosef Bacik 4140314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 4141314b6dd0SJohannes Thumshirn get_page(page); 4142a512bbf8SYan Zheng 4143314b6dd0SJohannes Thumshirn disk_super = page_address(page); 4144314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 4145a512bbf8SYan Zheng 4146387125fcSChris Mason /* 4147314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 4148314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 4149314b6dd0SJohannes Thumshirn * checking. 4150387125fcSChris Mason */ 415107888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 415207888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 415307888c66SChristoph Hellwig GFP_NOFS); 4154314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 4155314b6dd0SJohannes Thumshirn bio->bi_private = device; 4156314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 4157314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 4158314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 4159314b6dd0SJohannes Thumshirn 4160314b6dd0SJohannes Thumshirn /* 4161314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 4162314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 4163314b6dd0SJohannes Thumshirn * copies might still contain the older version. 4164314b6dd0SJohannes Thumshirn */ 41651b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 4166314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 4167314b6dd0SJohannes Thumshirn 416858ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 416958ff51f1SChristoph Hellwig submit_bio(bio); 41708376d9e1SNaohiro Aota 41718376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 41728376d9e1SNaohiro Aota errors++; 4173a512bbf8SYan Zheng } 4174a512bbf8SYan Zheng return errors < i ? 0 : -1; 4175a512bbf8SYan Zheng } 4176a512bbf8SYan Zheng 4177387125fcSChris Mason /* 4178abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 4179abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 4180abbb3b8eSDavid Sterba * 4181314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 4182abbb3b8eSDavid Sterba * date. 4183abbb3b8eSDavid Sterba */ 4184abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 4185abbb3b8eSDavid Sterba { 4186abbb3b8eSDavid Sterba int i; 4187abbb3b8eSDavid Sterba int errors = 0; 4188b6a535faSHoward McLauchlan bool primary_failed = false; 418912659251SNaohiro Aota int ret; 4190abbb3b8eSDavid Sterba u64 bytenr; 4191abbb3b8eSDavid Sterba 4192abbb3b8eSDavid Sterba if (max_mirrors == 0) 4193abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4194abbb3b8eSDavid Sterba 4195abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 4196314b6dd0SJohannes Thumshirn struct page *page; 4197314b6dd0SJohannes Thumshirn 419812659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 419912659251SNaohiro Aota if (ret == -ENOENT) { 420012659251SNaohiro Aota break; 420112659251SNaohiro Aota } else if (ret < 0) { 420212659251SNaohiro Aota errors++; 420312659251SNaohiro Aota if (i == 0) 420412659251SNaohiro Aota primary_failed = true; 420512659251SNaohiro Aota continue; 420612659251SNaohiro Aota } 4207abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4208abbb3b8eSDavid Sterba device->commit_total_bytes) 4209abbb3b8eSDavid Sterba break; 4210abbb3b8eSDavid Sterba 4211314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 4212314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 4213314b6dd0SJohannes Thumshirn if (!page) { 4214abbb3b8eSDavid Sterba errors++; 4215b6a535faSHoward McLauchlan if (i == 0) 4216b6a535faSHoward McLauchlan primary_failed = true; 4217abbb3b8eSDavid Sterba continue; 4218abbb3b8eSDavid Sterba } 4219314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 4220314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 4221314b6dd0SJohannes Thumshirn if (PageError(page)) { 4222abbb3b8eSDavid Sterba errors++; 4223b6a535faSHoward McLauchlan if (i == 0) 4224b6a535faSHoward McLauchlan primary_failed = true; 4225b6a535faSHoward McLauchlan } 4226abbb3b8eSDavid Sterba 4227314b6dd0SJohannes Thumshirn /* Drop our reference */ 4228314b6dd0SJohannes Thumshirn put_page(page); 4229abbb3b8eSDavid Sterba 4230314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 4231314b6dd0SJohannes Thumshirn put_page(page); 4232abbb3b8eSDavid Sterba } 4233abbb3b8eSDavid Sterba 4234b6a535faSHoward McLauchlan /* log error, force error return */ 4235b6a535faSHoward McLauchlan if (primary_failed) { 4236b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 4237b6a535faSHoward McLauchlan device->devid); 4238b6a535faSHoward McLauchlan return -1; 4239b6a535faSHoward McLauchlan } 4240b6a535faSHoward McLauchlan 4241abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 4242abbb3b8eSDavid Sterba } 4243abbb3b8eSDavid Sterba 4244abbb3b8eSDavid Sterba /* 4245387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 4246387125fcSChris Mason * for the barrier when it is done 4247387125fcSChris Mason */ 42484246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 4249387125fcSChris Mason { 4250f9e69aa9SChristoph Hellwig bio_uninit(bio); 4251387125fcSChris Mason complete(bio->bi_private); 4252387125fcSChris Mason } 4253387125fcSChris Mason 4254387125fcSChris Mason /* 42554fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 42564fc6441aSAnand Jain * done in the waiting counterpart. 4257387125fcSChris Mason */ 42584fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 4259387125fcSChris Mason { 4260f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 4261387125fcSChris Mason 4262a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 4263a91cf0ffSWang Yugui /* 4264a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 4265a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 4266a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 4267a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 4268a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 4269a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 4270a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 4271a91cf0ffSWang Yugui * non-debug builds. 4272a91cf0ffSWang Yugui */ 427308e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 42744fc6441aSAnand Jain return; 4275a91cf0ffSWang Yugui #endif 4276387125fcSChris Mason 4277f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 4278f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 4279387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 4280387125fcSChris Mason init_completion(&device->flush_wait); 4281387125fcSChris Mason bio->bi_private = &device->flush_wait; 4282387125fcSChris Mason 428358ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 428458ff51f1SChristoph Hellwig submit_bio(bio); 42851c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 42864fc6441aSAnand Jain } 4287387125fcSChris Mason 42884fc6441aSAnand Jain /* 42894fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 42904fc6441aSAnand Jain */ 42918c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 42924fc6441aSAnand Jain { 4293f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 42944fc6441aSAnand Jain 42951c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 429658efbc9fSOmar Sandoval return BLK_STS_OK; 42974fc6441aSAnand Jain 42981c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 42992980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 43004fc6441aSAnand Jain 43018c27cb35SLinus Torvalds return bio->bi_status; 4302387125fcSChris Mason } 4303387125fcSChris Mason 4304d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 4305401b41e5SAnand Jain { 43066528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 4307401b41e5SAnand Jain return -EIO; 4308387125fcSChris Mason return 0; 4309387125fcSChris Mason } 4310387125fcSChris Mason 4311387125fcSChris Mason /* 4312387125fcSChris Mason * send an empty flush down to each device in parallel, 4313387125fcSChris Mason * then wait for them 4314387125fcSChris Mason */ 4315387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 4316387125fcSChris Mason { 4317387125fcSChris Mason struct list_head *head; 4318387125fcSChris Mason struct btrfs_device *dev; 43195af3e8ccSStefan Behrens int errors_wait = 0; 43204e4cbee9SChristoph Hellwig blk_status_t ret; 4321387125fcSChris Mason 43221538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 4323387125fcSChris Mason /* send down all the barriers */ 4324387125fcSChris Mason head = &info->fs_devices->devices; 43251538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4326e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4327f88ba6a2SHidetoshi Seto continue; 4328cea7c8bfSAnand Jain if (!dev->bdev) 4329387125fcSChris Mason continue; 4330e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4331ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4332387125fcSChris Mason continue; 4333387125fcSChris Mason 43344fc6441aSAnand Jain write_dev_flush(dev); 433558efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 4336387125fcSChris Mason } 4337387125fcSChris Mason 4338387125fcSChris Mason /* wait for all the barriers */ 43391538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4340e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4341f88ba6a2SHidetoshi Seto continue; 4342387125fcSChris Mason if (!dev->bdev) { 43435af3e8ccSStefan Behrens errors_wait++; 4344387125fcSChris Mason continue; 4345387125fcSChris Mason } 4346e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4347ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4348387125fcSChris Mason continue; 4349387125fcSChris Mason 43504fc6441aSAnand Jain ret = wait_dev_flush(dev); 4351401b41e5SAnand Jain if (ret) { 4352401b41e5SAnand Jain dev->last_flush_error = ret; 435366b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 435466b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 43555af3e8ccSStefan Behrens errors_wait++; 4356387125fcSChris Mason } 4357401b41e5SAnand Jain } 4358401b41e5SAnand Jain 4359cea7c8bfSAnand Jain if (errors_wait) { 4360401b41e5SAnand Jain /* 4361401b41e5SAnand Jain * At some point we need the status of all disks 4362401b41e5SAnand Jain * to arrive at the volume status. So error checking 4363401b41e5SAnand Jain * is being pushed to a separate loop. 4364401b41e5SAnand Jain */ 4365d10b82feSQu Wenruo return check_barrier_error(info); 4366401b41e5SAnand Jain } 4367387125fcSChris Mason return 0; 4368387125fcSChris Mason } 4369387125fcSChris Mason 4370943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4371943c6e99SZhao Lei { 43728789f4feSZhao Lei int raid_type; 43738789f4feSZhao Lei int min_tolerated = INT_MAX; 4374943c6e99SZhao Lei 43758789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 43768789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 43778c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 43788789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 43798789f4feSZhao Lei tolerated_failures); 4380943c6e99SZhao Lei 43818789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 43828789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 43838789f4feSZhao Lei continue; 438441a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 43858789f4feSZhao Lei continue; 43868c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 43878789f4feSZhao Lei btrfs_raid_array[raid_type]. 43888789f4feSZhao Lei tolerated_failures); 43898789f4feSZhao Lei } 4390943c6e99SZhao Lei 43918789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4392ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 43938789f4feSZhao Lei min_tolerated = 0; 43948789f4feSZhao Lei } 43958789f4feSZhao Lei 43968789f4feSZhao Lei return min_tolerated; 4397943c6e99SZhao Lei } 4398943c6e99SZhao Lei 4399eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4400f2984462SChris Mason { 4401e5e9a520SChris Mason struct list_head *head; 4402f2984462SChris Mason struct btrfs_device *dev; 4403a061fc8dSChris Mason struct btrfs_super_block *sb; 4404f2984462SChris Mason struct btrfs_dev_item *dev_item; 4405f2984462SChris Mason int ret; 4406f2984462SChris Mason int do_barriers; 4407a236aed1SChris Mason int max_errors; 4408a236aed1SChris Mason int total_errors = 0; 4409a061fc8dSChris Mason u64 flags; 4410f2984462SChris Mason 44110b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4412fed3b381SLiu Bo 4413fed3b381SLiu Bo /* 4414fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4415fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4416fed3b381SLiu Bo * been consistent on disk. 4417fed3b381SLiu Bo */ 4418fed3b381SLiu Bo if (max_mirrors == 0) 44190b246afaSJeff Mahoney backup_super_roots(fs_info); 4420f2984462SChris Mason 44210b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4422a061fc8dSChris Mason dev_item = &sb->dev_item; 4423e5e9a520SChris Mason 44240b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 44250b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 44260b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4427387125fcSChris Mason 44285af3e8ccSStefan Behrens if (do_barriers) { 44290b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 44305af3e8ccSStefan Behrens if (ret) { 44315af3e8ccSStefan Behrens mutex_unlock( 44320b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 44330b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 44345af3e8ccSStefan Behrens "errors while submitting device barriers."); 44355af3e8ccSStefan Behrens return ret; 44365af3e8ccSStefan Behrens } 44375af3e8ccSStefan Behrens } 4438387125fcSChris Mason 44391538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4440dfe25020SChris Mason if (!dev->bdev) { 4441dfe25020SChris Mason total_errors++; 4442dfe25020SChris Mason continue; 4443dfe25020SChris Mason } 4444e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4445ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4446dfe25020SChris Mason continue; 4447dfe25020SChris Mason 44482b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4449a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4450a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 44517df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4452935e5cc9SMiao Xie dev->commit_total_bytes); 4453ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4454ce7213c7SMiao Xie dev->commit_bytes_used); 4455a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4456a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4457a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4458a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 44597239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 44607239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4461a512bbf8SYan Zheng 4462a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4463a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4464f2984462SChris Mason 446575cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 446675cb857dSQu Wenruo if (ret < 0) { 446775cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 446875cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 446975cb857dSQu Wenruo "unexpected superblock corruption detected"); 447075cb857dSQu Wenruo return -EUCLEAN; 447175cb857dSQu Wenruo } 447275cb857dSQu Wenruo 4473abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4474a236aed1SChris Mason if (ret) 4475a236aed1SChris Mason total_errors++; 4476f2984462SChris Mason } 4477a236aed1SChris Mason if (total_errors > max_errors) { 44780b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4479d397712bSChris Mason total_errors); 44800b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 448179787eaaSJeff Mahoney 44829d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 44830b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 44840b246afaSJeff Mahoney "%d errors while writing supers", 44850b246afaSJeff Mahoney total_errors); 44869d565ba4SStefan Behrens return -EIO; 4487a236aed1SChris Mason } 4488f2984462SChris Mason 4489a512bbf8SYan Zheng total_errors = 0; 44901538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4491dfe25020SChris Mason if (!dev->bdev) 4492dfe25020SChris Mason continue; 4493e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4494ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4495dfe25020SChris Mason continue; 4496dfe25020SChris Mason 4497abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4498a512bbf8SYan Zheng if (ret) 44991259ab75SChris Mason total_errors++; 4500f2984462SChris Mason } 45010b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4502a236aed1SChris Mason if (total_errors > max_errors) { 45030b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 45040b246afaSJeff Mahoney "%d errors while writing supers", 45050b246afaSJeff Mahoney total_errors); 450679787eaaSJeff Mahoney return -EIO; 4507a236aed1SChris Mason } 4508f2984462SChris Mason return 0; 4509f2984462SChris Mason } 4510f2984462SChris Mason 4511cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4512cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4513cb517eabSMiao Xie struct btrfs_root *root) 45142619ba1fSChris Mason { 45154785e24fSJosef Bacik bool drop_ref = false; 45164785e24fSJosef Bacik 4517fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4518fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4519fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4520fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 45214785e24fSJosef Bacik drop_ref = true; 4522fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 452376dda93cSYan, Zheng 452484961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4525ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 45261c1ea4f7SLiu Bo if (root->reloc_root) { 452700246528SJosef Bacik btrfs_put_root(root->reloc_root); 45281c1ea4f7SLiu Bo root->reloc_root = NULL; 45291c1ea4f7SLiu Bo } 45301c1ea4f7SLiu Bo } 45313321719eSLiu Bo 45324785e24fSJosef Bacik if (drop_ref) 453300246528SJosef Bacik btrfs_put_root(root); 45342619ba1fSChris Mason } 45352619ba1fSChris Mason 4536c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4537c146afadSYan Zheng { 4538fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4539fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4540fc7cbcd4SDavid Sterba int i = 0; 454165d33fd7SQu Wenruo int err = 0; 4542fc7cbcd4SDavid Sterba unsigned int ret = 0; 4543c146afadSYan Zheng 4544c146afadSYan Zheng while (1) { 4545fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4546fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4547fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4548fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4549fc7cbcd4SDavid Sterba if (!ret) { 4550fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4551c146afadSYan Zheng break; 455265d33fd7SQu Wenruo } 4553fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 455466b4ffd1SJosef Bacik 4555fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4556fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4557fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4558fc7cbcd4SDavid Sterba gang[i] = NULL; 455965d33fd7SQu Wenruo continue; 4560c146afadSYan Zheng } 4561fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4562fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4563fc7cbcd4SDavid Sterba } 4564fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4565fc7cbcd4SDavid Sterba 4566fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4567fc7cbcd4SDavid Sterba if (!gang[i]) 4568fc7cbcd4SDavid Sterba continue; 4569fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4570fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4571fc7cbcd4SDavid Sterba if (err) 4572fc7cbcd4SDavid Sterba break; 4573fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4574fc7cbcd4SDavid Sterba } 4575fc7cbcd4SDavid Sterba root_objectid++; 4576c146afadSYan Zheng } 457765d33fd7SQu Wenruo 4578fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4579fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4580fc7cbcd4SDavid Sterba if (gang[i]) 4581fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 458265d33fd7SQu Wenruo } 458365d33fd7SQu Wenruo return err; 4584c146afadSYan Zheng } 4585c146afadSYan Zheng 45866bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4587c146afadSYan Zheng { 45886bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4589c146afadSYan Zheng struct btrfs_trans_handle *trans; 4590c146afadSYan Zheng 45910b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45922ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45930b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 45940b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4595c71bf099SYan, Zheng 4596c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 45970b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45980b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4599c71bf099SYan, Zheng 46007a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 46013612b495STsutomu Itoh if (IS_ERR(trans)) 46023612b495STsutomu Itoh return PTR_ERR(trans); 46033a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4604c146afadSYan Zheng } 4605c146afadSYan Zheng 460636c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 460736c86a9eSQu Wenruo { 460836c86a9eSQu Wenruo struct btrfs_transaction *trans; 460936c86a9eSQu Wenruo struct btrfs_transaction *tmp; 461036c86a9eSQu Wenruo bool found = false; 461136c86a9eSQu Wenruo 461236c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 461336c86a9eSQu Wenruo return; 461436c86a9eSQu Wenruo 461536c86a9eSQu Wenruo /* 461636c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 461736c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 461836c86a9eSQu Wenruo */ 461936c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 462036c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 462136c86a9eSQu Wenruo struct extent_state *cached = NULL; 462236c86a9eSQu Wenruo u64 dirty_bytes = 0; 462336c86a9eSQu Wenruo u64 cur = 0; 462436c86a9eSQu Wenruo u64 found_start; 462536c86a9eSQu Wenruo u64 found_end; 462636c86a9eSQu Wenruo 462736c86a9eSQu Wenruo found = true; 462836c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 462936c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 463036c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 463136c86a9eSQu Wenruo cur = found_end + 1; 463236c86a9eSQu Wenruo } 463336c86a9eSQu Wenruo btrfs_warn(fs_info, 463436c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 463536c86a9eSQu Wenruo trans->transid, dirty_bytes); 463636c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 463736c86a9eSQu Wenruo 463836c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 463936c86a9eSQu Wenruo fs_info->running_transaction = NULL; 464036c86a9eSQu Wenruo list_del_init(&trans->list); 464136c86a9eSQu Wenruo 464236c86a9eSQu Wenruo btrfs_put_transaction(trans); 464336c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 464436c86a9eSQu Wenruo } 464536c86a9eSQu Wenruo ASSERT(!found); 464636c86a9eSQu Wenruo } 464736c86a9eSQu Wenruo 4648b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4649eb60ceacSChris Mason { 4650c146afadSYan Zheng int ret; 4651e089f05cSChris Mason 4652afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 465331e70e52SFilipe Manana 465431e70e52SFilipe Manana /* 465531e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 465631e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 465731e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 465831e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 465931e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 466031e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 466131e70e52SFilipe Manana * trying to stop the async reclaim task. 466231e70e52SFilipe Manana */ 466331e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4664d6fd0ae2SOmar Sandoval /* 4665d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4666d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4667d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4668d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4669d6fd0ae2SOmar Sandoval */ 4670d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4671a2135011SChris Mason 4672b4be6aefSJosef Bacik /* 4673b4be6aefSJosef Bacik * If we had UNFINISHED_DROPS we could still be processing them, so 4674b4be6aefSJosef Bacik * clear that bit and wake up relocation so it can stop. 4675b4be6aefSJosef Bacik */ 4676b4be6aefSJosef Bacik btrfs_wake_unfinished_drop(fs_info); 4677b4be6aefSJosef Bacik 46787343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4679d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 46807343dd61SJustin Maggard 4681803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4682803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4683803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4684803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4685803b2f54SStefan Behrens 4686837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4687aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4688837d5b6eSIlya Dryomov 46898dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 46908dabb742SStefan Behrens 4691aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 46924cb5300bSChris Mason 46934cb5300bSChris Mason /* wait for any defraggers to finish */ 46944cb5300bSChris Mason wait_event(fs_info->transaction_wait, 46954cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 46964cb5300bSChris Mason 46974cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 469826176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 46994cb5300bSChris Mason 470021c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 470157056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4702576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 470321c7e756SMiao Xie 4704b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4705b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4706b0643e59SDennis Zhou 4707bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4708e44163e1SJeff Mahoney /* 4709d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4710d6fd0ae2SOmar Sandoval * unused block groups. 4711e44163e1SJeff Mahoney */ 47120b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4713e44163e1SJeff Mahoney 4714f0cc2cd7SFilipe Manana /* 4715f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4716f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4717f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4718f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4719f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4720f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4721f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4722f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4723f0cc2cd7SFilipe Manana * This is a very rare case. 4724f0cc2cd7SFilipe Manana */ 4725f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4726f0cc2cd7SFilipe Manana 47276bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4728d397712bSChris Mason if (ret) 472904892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4730c146afadSYan Zheng } 4731ed2ff2cbSChris Mason 473284961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 47332ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4734acce952bSliubo 4735e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4736e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 47378929ecfaSYan, Zheng 4738e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4739afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4740f25784b3SYan Zheng 47415958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 47425958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 47435958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 47445958253cSQu Wenruo } 47455958253cSQu Wenruo 474604892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4747fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4748bcef60f2SArne Jansen 4749963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 475004892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4751963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4752b0c68f8bSChris Mason } 475331153d81SYan Zheng 47545deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 47554297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 47565deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 47574297ff84SJosef Bacik 47586618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4759b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 47605ac1d209SJeff Mahoney 47611a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 47621a4319ccSLiu Bo 4763de348ee0SWang Shilong /* 4764de348ee0SWang Shilong * we must make sure there is not any read request to 4765de348ee0SWang Shilong * submit after we stopping all workers. 4766de348ee0SWang Shilong */ 4767de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 476896192499SJosef Bacik btrfs_stop_all_workers(fs_info); 476996192499SJosef Bacik 47700a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 477136c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 47720a31daa4SFilipe Manana 4773afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 47744273eaffSAnand Jain free_root_pointers(fs_info, true); 47758c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 47769ad6b7bcSChris Mason 47774e19443dSJosef Bacik /* 47784e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 47794e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 47804e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 47814e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 47824e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 47834e19443dSJosef Bacik */ 47844e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 47854e19443dSJosef Bacik 478613e6c37bSJosef Bacik iput(fs_info->btree_inode); 4787d6bfde87SChris Mason 478821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 47890b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 47902ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 479121adbd5cSStefan Behrens #endif 479221adbd5cSStefan Behrens 47930b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 479468c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4795eb60ceacSChris Mason } 4796eb60ceacSChris Mason 4797b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4798b9fab919SChris Mason int atomic) 4799ccd467d6SChris Mason { 48001259ab75SChris Mason int ret; 4801727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 48021259ab75SChris Mason 48030b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 48041259ab75SChris Mason if (!ret) 48051259ab75SChris Mason return ret; 48061259ab75SChris Mason 48071259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4808b9fab919SChris Mason parent_transid, atomic); 4809b9fab919SChris Mason if (ret == -EAGAIN) 4810b9fab919SChris Mason return ret; 48111259ab75SChris Mason return !ret; 48125f39d397SChris Mason } 48136702ed49SChris Mason 48145f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 48155f39d397SChris Mason { 48162f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 48175f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4818b9473439SChris Mason int was_dirty; 4819b4ce94deSChris Mason 482006ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 482106ea65a3SJosef Bacik /* 482206ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 482352042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 482406ea65a3SJosef Bacik * outside of the sanity tests. 482506ea65a3SJosef Bacik */ 4826b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 482706ea65a3SJosef Bacik return; 482806ea65a3SJosef Bacik #endif 482949d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 48300b246afaSJeff Mahoney if (transid != fs_info->generation) 48315d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 48320b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 48330b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4834e2d84521SMiao Xie if (!was_dirty) 4835104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4836e2d84521SMiao Xie buf->len, 48370b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 48381f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 483969fc6cbbSQu Wenruo /* 484069fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 484169fc6cbbSQu Wenruo * but item data not updated. 484269fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 484369fc6cbbSQu Wenruo */ 484469fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4845cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4846a4f78750SDavid Sterba btrfs_print_leaf(buf); 48471f21ef0aSFilipe Manana ASSERT(0); 48481f21ef0aSFilipe Manana } 48491f21ef0aSFilipe Manana #endif 4850eb60ceacSChris Mason } 4851eb60ceacSChris Mason 48522ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4853b53d3f5dSLiu Bo int flush_delayed) 485435b7e476SChris Mason { 4855188de649SChris Mason /* 4856188de649SChris Mason * looks as though older kernels can get into trouble with 4857188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4858188de649SChris Mason */ 4859e2d84521SMiao Xie int ret; 4860d6bfde87SChris Mason 48616933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4862d6bfde87SChris Mason return; 4863d6bfde87SChris Mason 4864b53d3f5dSLiu Bo if (flush_delayed) 48652ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 486616cdcec7SMiao Xie 4867d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4868d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4869d814a491SEthan Lien fs_info->dirty_metadata_batch); 4870e2d84521SMiao Xie if (ret > 0) { 48710b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 487216cdcec7SMiao Xie } 487316cdcec7SMiao Xie } 487416cdcec7SMiao Xie 48752ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 487616cdcec7SMiao Xie { 48772ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 487835b7e476SChris Mason } 4879b53d3f5dSLiu Bo 48802ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4881b53d3f5dSLiu Bo { 48822ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4883d6bfde87SChris Mason } 48846b80053dSChris Mason 48852ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4886acce952bSliubo { 4887fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4888fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4889fe816d0fSNikolay Borisov 48900b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 48912ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 48920b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4893acce952bSliubo 48940b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 48950b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4896acce952bSliubo } 4897acce952bSliubo 4898ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4899ef67963dSJosef Bacik { 4900fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4901fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4902fc7cbcd4SDavid Sterba int ret; 4903ef67963dSJosef Bacik 4904fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4905fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4906fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4907fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4908fc7cbcd4SDavid Sterba int i; 4909ef67963dSJosef Bacik 4910fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4911fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4912fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4913fc7cbcd4SDavid Sterba 4914fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4915fc7cbcd4SDavid Sterba if (!gang[i]) 4916ef67963dSJosef Bacik continue; 4917fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4918fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4919fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4920ef67963dSJosef Bacik } 4921fc7cbcd4SDavid Sterba root_objectid++; 4922fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4923ef67963dSJosef Bacik } 4924fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4925ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4926ef67963dSJosef Bacik } 4927ef67963dSJosef Bacik 4928143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4929acce952bSliubo { 4930acce952bSliubo struct btrfs_ordered_extent *ordered; 4931acce952bSliubo 4932199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4933779880efSJosef Bacik /* 4934779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4935779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4936779880efSJosef Bacik */ 4937199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4938779880efSJosef Bacik root_extent_list) 4939779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4940199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4941199c2a9cSMiao Xie } 4942199c2a9cSMiao Xie 4943199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4944199c2a9cSMiao Xie { 4945199c2a9cSMiao Xie struct btrfs_root *root; 4946199c2a9cSMiao Xie struct list_head splice; 4947199c2a9cSMiao Xie 4948199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4949199c2a9cSMiao Xie 4950199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4951199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4952199c2a9cSMiao Xie while (!list_empty(&splice)) { 4953199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4954199c2a9cSMiao Xie ordered_root); 49551de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 49561de2cfdeSJosef Bacik &fs_info->ordered_roots); 4957199c2a9cSMiao Xie 49582a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4959199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4960199c2a9cSMiao Xie 49612a85d9caSLiu Bo cond_resched(); 49622a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4963199c2a9cSMiao Xie } 4964199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 496574d5d229SJosef Bacik 496674d5d229SJosef Bacik /* 496774d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 496874d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 496974d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 497074d5d229SJosef Bacik * actually get run and error out properly. 497174d5d229SJosef Bacik */ 497274d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4973acce952bSliubo } 4974acce952bSliubo 497535a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 49762ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4977acce952bSliubo { 4978acce952bSliubo struct rb_node *node; 4979acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4980acce952bSliubo struct btrfs_delayed_ref_node *ref; 4981acce952bSliubo int ret = 0; 4982acce952bSliubo 4983acce952bSliubo delayed_refs = &trans->delayed_refs; 4984acce952bSliubo 4985acce952bSliubo spin_lock(&delayed_refs->lock); 4986d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4987cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4988b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4989acce952bSliubo return ret; 4990acce952bSliubo } 4991acce952bSliubo 49925c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4993d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 49940e0adbcfSJosef Bacik struct rb_node *n; 4995e78417d1SJosef Bacik bool pin_bytes = false; 4996acce952bSliubo 4997d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4998d7df2c79SJosef Bacik href_node); 49993069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 5000b939d1abSJosef Bacik continue; 50013069bd26SJosef Bacik 5002d7df2c79SJosef Bacik spin_lock(&head->lock); 5003e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 50040e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 50050e0adbcfSJosef Bacik ref_node); 5006d7df2c79SJosef Bacik ref->in_tree = 0; 5007e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 50080e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 50091d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 50101d57ee94SWang Xiaoguang list_del(&ref->add_list); 5011d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 5012d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 5013d7df2c79SJosef Bacik } 501454067ae9SJosef Bacik if (head->must_insert_reserved) 5015e78417d1SJosef Bacik pin_bytes = true; 501678a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 5017fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 5018d7df2c79SJosef Bacik spin_unlock(&head->lock); 5019acce952bSliubo spin_unlock(&delayed_refs->lock); 5020e78417d1SJosef Bacik mutex_unlock(&head->mutex); 5021acce952bSliubo 5022f603bb94SNikolay Borisov if (pin_bytes) { 5023f603bb94SNikolay Borisov struct btrfs_block_group *cache; 5024f603bb94SNikolay Borisov 5025f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 5026f603bb94SNikolay Borisov BUG_ON(!cache); 5027f603bb94SNikolay Borisov 5028f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 5029f603bb94SNikolay Borisov spin_lock(&cache->lock); 5030f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 5031f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 5032f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 5033f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 5034f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 5035f603bb94SNikolay Borisov spin_unlock(&cache->lock); 5036f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 5037f603bb94SNikolay Borisov 5038f603bb94SNikolay Borisov btrfs_put_block_group(cache); 5039f603bb94SNikolay Borisov 5040f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 5041f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 5042f603bb94SNikolay Borisov } 504331890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 5044d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 5045acce952bSliubo cond_resched(); 5046acce952bSliubo spin_lock(&delayed_refs->lock); 5047acce952bSliubo } 504881f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 5049acce952bSliubo 5050acce952bSliubo spin_unlock(&delayed_refs->lock); 5051acce952bSliubo 5052acce952bSliubo return ret; 5053acce952bSliubo } 5054acce952bSliubo 5055143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 5056acce952bSliubo { 5057acce952bSliubo struct btrfs_inode *btrfs_inode; 5058acce952bSliubo struct list_head splice; 5059acce952bSliubo 5060acce952bSliubo INIT_LIST_HEAD(&splice); 5061acce952bSliubo 5062eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5063eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 5064acce952bSliubo 5065acce952bSliubo while (!list_empty(&splice)) { 5066fe816d0fSNikolay Borisov struct inode *inode = NULL; 5067eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 5068acce952bSliubo delalloc_inodes); 5069fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 5070eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5071acce952bSliubo 5072fe816d0fSNikolay Borisov /* 5073fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 5074fe816d0fSNikolay Borisov * meanwhile. 5075fe816d0fSNikolay Borisov */ 5076fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 5077fe816d0fSNikolay Borisov if (inode) { 5078fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 5079fe816d0fSNikolay Borisov iput(inode); 5080fe816d0fSNikolay Borisov } 5081eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5082acce952bSliubo } 5083eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5084eb73c1b7SMiao Xie } 5085eb73c1b7SMiao Xie 5086eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 5087eb73c1b7SMiao Xie { 5088eb73c1b7SMiao Xie struct btrfs_root *root; 5089eb73c1b7SMiao Xie struct list_head splice; 5090eb73c1b7SMiao Xie 5091eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 5092eb73c1b7SMiao Xie 5093eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5094eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 5095eb73c1b7SMiao Xie while (!list_empty(&splice)) { 5096eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 5097eb73c1b7SMiao Xie delalloc_root); 509800246528SJosef Bacik root = btrfs_grab_root(root); 5099eb73c1b7SMiao Xie BUG_ON(!root); 5100eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5101eb73c1b7SMiao Xie 5102eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 510300246528SJosef Bacik btrfs_put_root(root); 5104eb73c1b7SMiao Xie 5105eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5106eb73c1b7SMiao Xie } 5107eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5108acce952bSliubo } 5109acce952bSliubo 51102ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 5111acce952bSliubo struct extent_io_tree *dirty_pages, 5112acce952bSliubo int mark) 5113acce952bSliubo { 5114acce952bSliubo int ret; 5115acce952bSliubo struct extent_buffer *eb; 5116acce952bSliubo u64 start = 0; 5117acce952bSliubo u64 end; 5118acce952bSliubo 5119acce952bSliubo while (1) { 5120acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 5121e6138876SJosef Bacik mark, NULL); 5122acce952bSliubo if (ret) 5123acce952bSliubo break; 5124acce952bSliubo 512591166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 5126acce952bSliubo while (start <= end) { 51270b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 51280b246afaSJeff Mahoney start += fs_info->nodesize; 5129fd8b2b61SJosef Bacik if (!eb) 5130acce952bSliubo continue; 5131fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 5132acce952bSliubo 5133fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 5134fd8b2b61SJosef Bacik &eb->bflags)) 5135fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 5136fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 5137acce952bSliubo } 5138acce952bSliubo } 5139acce952bSliubo 5140acce952bSliubo return ret; 5141acce952bSliubo } 5142acce952bSliubo 51432ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 5144fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 5145acce952bSliubo { 5146acce952bSliubo u64 start; 5147acce952bSliubo u64 end; 5148acce952bSliubo int ret; 5149acce952bSliubo 5150acce952bSliubo while (1) { 51510e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 51520e6ec385SFilipe Manana 5153fcd5e742SLu Fengqi /* 5154fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 5155fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 5156fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 5157fcd5e742SLu Fengqi * the same extent range. 5158fcd5e742SLu Fengqi */ 5159fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 5160acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 51610e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 5162fcd5e742SLu Fengqi if (ret) { 5163fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5164acce952bSliubo break; 5165fcd5e742SLu Fengqi } 5166acce952bSliubo 51670e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 51680e6ec385SFilipe Manana free_extent_state(cached_state); 51692ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 5170fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5171acce952bSliubo cond_resched(); 5172acce952bSliubo } 5173acce952bSliubo 5174acce952bSliubo return 0; 5175acce952bSliubo } 5176acce952bSliubo 517732da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 5178c79a1751SLiu Bo { 5179c79a1751SLiu Bo struct inode *inode; 5180c79a1751SLiu Bo 5181c79a1751SLiu Bo inode = cache->io_ctl.inode; 5182c79a1751SLiu Bo if (inode) { 5183c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 5184c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 5185c79a1751SLiu Bo cache->io_ctl.inode = NULL; 5186c79a1751SLiu Bo iput(inode); 5187c79a1751SLiu Bo } 5188bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 5189c79a1751SLiu Bo btrfs_put_block_group(cache); 5190c79a1751SLiu Bo } 5191c79a1751SLiu Bo 5192c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 51932ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 5194c79a1751SLiu Bo { 519532da5386SDavid Sterba struct btrfs_block_group *cache; 5196c79a1751SLiu Bo 5197c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5198c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 5199c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 520032da5386SDavid Sterba struct btrfs_block_group, 5201c79a1751SLiu Bo dirty_list); 5202c79a1751SLiu Bo 5203c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 5204c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5205c79a1751SLiu Bo list_del_init(&cache->io_list); 5206c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5207c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5208c79a1751SLiu Bo } 5209c79a1751SLiu Bo 5210c79a1751SLiu Bo list_del_init(&cache->dirty_list); 5211c79a1751SLiu Bo spin_lock(&cache->lock); 5212c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5213c79a1751SLiu Bo spin_unlock(&cache->lock); 5214c79a1751SLiu Bo 5215c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5216c79a1751SLiu Bo btrfs_put_block_group(cache); 5217ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 5218c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5219c79a1751SLiu Bo } 5220c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5221c79a1751SLiu Bo 522245ae2c18SNikolay Borisov /* 522345ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 522445ae2c18SNikolay Borisov * to use it without any locking 522545ae2c18SNikolay Borisov */ 5226c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 5227c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 522832da5386SDavid Sterba struct btrfs_block_group, 5229c79a1751SLiu Bo io_list); 5230c79a1751SLiu Bo 5231c79a1751SLiu Bo list_del_init(&cache->io_list); 5232c79a1751SLiu Bo spin_lock(&cache->lock); 5233c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5234c79a1751SLiu Bo spin_unlock(&cache->lock); 5235c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5236c79a1751SLiu Bo } 5237c79a1751SLiu Bo } 5238c79a1751SLiu Bo 523949b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 52402ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 524149b25e05SJeff Mahoney { 5242bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 5243bbbf7243SNikolay Borisov 52442ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 5245c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 5246c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 5247c79a1751SLiu Bo 5248bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 5249bbbf7243SNikolay Borisov post_commit_list) { 5250bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 5251bbbf7243SNikolay Borisov } 5252bbbf7243SNikolay Borisov 52532ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 525449b25e05SJeff Mahoney 52554a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 52560b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 525749b25e05SJeff Mahoney 52584a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 52590b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 526049b25e05SJeff Mahoney 5261ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 526267cde344SMiao Xie 52632ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 526449b25e05SJeff Mahoney EXTENT_DIRTY); 5265fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 526649b25e05SJeff Mahoney 5267d3575156SNaohiro Aota btrfs_free_redirty_list(cur_trans); 5268d3575156SNaohiro Aota 52694a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 52704a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 527149b25e05SJeff Mahoney } 527249b25e05SJeff Mahoney 52732ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 5274acce952bSliubo { 5275acce952bSliubo struct btrfs_transaction *t; 5276acce952bSliubo 52770b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 5278acce952bSliubo 52790b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 52800b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 52810b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 5282724e2315SJosef Bacik struct btrfs_transaction, list); 5283724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 52849b64f57dSElena Reshetova refcount_inc(&t->use_count); 52850b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 52862ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 5287724e2315SJosef Bacik btrfs_put_transaction(t); 52880b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5289724e2315SJosef Bacik continue; 5290724e2315SJosef Bacik } 52910b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 5292724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 52930b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5294724e2315SJosef Bacik /* 5295724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 5296724e2315SJosef Bacik * handle open currently for this transaction. 5297724e2315SJosef Bacik */ 5298724e2315SJosef Bacik wait_event(t->writer_wait, 5299724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 5300724e2315SJosef Bacik } else { 53010b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5302724e2315SJosef Bacik } 53032ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 5304724e2315SJosef Bacik 53050b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 53060b246afaSJeff Mahoney if (t == fs_info->running_transaction) 53070b246afaSJeff Mahoney fs_info->running_transaction = NULL; 5308724e2315SJosef Bacik list_del_init(&t->list); 53090b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5310a4abeea4SJosef Bacik 5311724e2315SJosef Bacik btrfs_put_transaction(t); 53122e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 53130b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5314724e2315SJosef Bacik } 53150b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 53160b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 5317ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 5318ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 53190b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 5320ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 53210b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 5322acce952bSliubo 5323acce952bSliubo return 0; 5324acce952bSliubo } 5325ec7d6dfdSNikolay Borisov 5326453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 5327ec7d6dfdSNikolay Borisov { 5328ec7d6dfdSNikolay Borisov struct btrfs_path *path; 5329ec7d6dfdSNikolay Borisov int ret; 5330ec7d6dfdSNikolay Borisov struct extent_buffer *l; 5331ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 5332ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 5333ec7d6dfdSNikolay Borisov int slot; 5334ec7d6dfdSNikolay Borisov 5335ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 5336ec7d6dfdSNikolay Borisov if (!path) 5337ec7d6dfdSNikolay Borisov return -ENOMEM; 5338ec7d6dfdSNikolay Borisov 5339ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 5340ec7d6dfdSNikolay Borisov search_key.type = -1; 5341ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 5342ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 5343ec7d6dfdSNikolay Borisov if (ret < 0) 5344ec7d6dfdSNikolay Borisov goto error; 5345ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 5346ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 5347ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 5348ec7d6dfdSNikolay Borisov l = path->nodes[0]; 5349ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 535023125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 535123125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 5352ec7d6dfdSNikolay Borisov } else { 535323125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 5354ec7d6dfdSNikolay Borisov } 5355ec7d6dfdSNikolay Borisov ret = 0; 5356ec7d6dfdSNikolay Borisov error: 5357ec7d6dfdSNikolay Borisov btrfs_free_path(path); 5358ec7d6dfdSNikolay Borisov return ret; 5359ec7d6dfdSNikolay Borisov } 5360ec7d6dfdSNikolay Borisov 5361543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 5362ec7d6dfdSNikolay Borisov { 5363ec7d6dfdSNikolay Borisov int ret; 5364ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5365ec7d6dfdSNikolay Borisov 53666b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5367ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5368ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5369ec7d6dfdSNikolay Borisov root->root_key.objectid); 5370ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5371ec7d6dfdSNikolay Borisov goto out; 5372ec7d6dfdSNikolay Borisov } 5373ec7d6dfdSNikolay Borisov 537423125104SNikolay Borisov *objectid = root->free_objectid++; 5375ec7d6dfdSNikolay Borisov ret = 0; 5376ec7d6dfdSNikolay Borisov out: 5377ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5378ec7d6dfdSNikolay Borisov return ret; 5379ec7d6dfdSNikolay Borisov } 5380