1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 26cbd5570SChris Mason /* 36cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 46cbd5570SChris Mason */ 56cbd5570SChris Mason 6e20d96d6SChris Mason #include <linux/fs.h> 7d98237b3SChris Mason #include <linux/blkdev.h> 80f7d52f4SChris Mason #include <linux/radix-tree.h> 935b7e476SChris Mason #include <linux/writeback.h> 10ce9adaa5SChris Mason #include <linux/workqueue.h> 11a74a4b97SChris Mason #include <linux/kthread.h> 125a0e3ad6STejun Heo #include <linux/slab.h> 13784b4e29SChris Mason #include <linux/migrate.h> 147a36ddecSDavid Sterba #include <linux/ratelimit.h> 156463fe58SStefan Behrens #include <linux/uuid.h> 16803b2f54SStefan Behrens #include <linux/semaphore.h> 17540adea3SMasami Hiramatsu #include <linux/error-injection.h> 189678c543SNikolay Borisov #include <linux/crc32c.h> 19b89f6d1fSFilipe Manana #include <linux/sched/mm.h> 207e75bf3fSDavid Sterba #include <asm/unaligned.h> 216d97c6e3SJohannes Thumshirn #include <crypto/hash.h> 22eb60ceacSChris Mason #include "ctree.h" 23eb60ceacSChris Mason #include "disk-io.h" 24e089f05cSChris Mason #include "transaction.h" 250f7d52f4SChris Mason #include "btrfs_inode.h" 260b86a832SChris Mason #include "volumes.h" 27db94535dSChris Mason #include "print-tree.h" 28925baeddSChris Mason #include "locking.h" 29e02119d5SChris Mason #include "tree-log.h" 30fa9c0d79SChris Mason #include "free-space-cache.h" 3170f6d82eSOmar Sandoval #include "free-space-tree.h" 3221adbd5cSStefan Behrens #include "check-integrity.h" 33606686eeSJosef Bacik #include "rcu-string.h" 348dabb742SStefan Behrens #include "dev-replace.h" 3553b381b3SDavid Woodhouse #include "raid56.h" 365ac1d209SJeff Mahoney #include "sysfs.h" 37fcebe456SJosef Bacik #include "qgroup.h" 38ebb8765bSAnand Jain #include "compression.h" 39557ea5ddSQu Wenruo #include "tree-checker.h" 40fd708b81SJosef Bacik #include "ref-verify.h" 41aac0023cSJosef Bacik #include "block-group.h" 42b0643e59SDennis Zhou #include "discard.h" 43f603bb94SNikolay Borisov #include "space-info.h" 44b70f5097SNaohiro Aota #include "zoned.h" 45139e8cd3SQu Wenruo #include "subpage.h" 46eb60ceacSChris Mason 47319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 48319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 49319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 50319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 51e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 52e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 53319e4d06SQu Wenruo 548b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 55143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 56acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 572ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 58143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 592ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 60acce952bSliubo struct extent_io_tree *dirty_pages, 61acce952bSliubo int mark); 622ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 63acce952bSliubo struct extent_io_tree *pinned_extents); 642ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 652ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 66ce9adaa5SChris Mason 67d352ac68SChris Mason /* 6897eb6b69SDavid Sterba * btrfs_end_io_wq structs are used to do processing in task context when an IO 6997eb6b69SDavid Sterba * is complete. This is used during reads to verify checksums, and it is used 70d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 71d352ac68SChris Mason */ 7297eb6b69SDavid Sterba struct btrfs_end_io_wq { 73ce9adaa5SChris Mason struct bio *bio; 74ce9adaa5SChris Mason bio_end_io_t *end_io; 75ce9adaa5SChris Mason void *private; 76ce9adaa5SChris Mason struct btrfs_fs_info *info; 774e4cbee9SChristoph Hellwig blk_status_t status; 78bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata; 798b712842SChris Mason struct btrfs_work work; 80ce9adaa5SChris Mason }; 810da5468fSChris Mason 8297eb6b69SDavid Sterba static struct kmem_cache *btrfs_end_io_wq_cache; 8397eb6b69SDavid Sterba 8497eb6b69SDavid Sterba int __init btrfs_end_io_wq_init(void) 8597eb6b69SDavid Sterba { 8697eb6b69SDavid Sterba btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", 8797eb6b69SDavid Sterba sizeof(struct btrfs_end_io_wq), 8897eb6b69SDavid Sterba 0, 89fba4b697SNikolay Borisov SLAB_MEM_SPREAD, 9097eb6b69SDavid Sterba NULL); 9197eb6b69SDavid Sterba if (!btrfs_end_io_wq_cache) 9297eb6b69SDavid Sterba return -ENOMEM; 9397eb6b69SDavid Sterba return 0; 9497eb6b69SDavid Sterba } 9597eb6b69SDavid Sterba 96e67c718bSDavid Sterba void __cold btrfs_end_io_wq_exit(void) 9797eb6b69SDavid Sterba { 9897eb6b69SDavid Sterba kmem_cache_destroy(btrfs_end_io_wq_cache); 9997eb6b69SDavid Sterba } 10097eb6b69SDavid Sterba 101141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 102141386e1SJosef Bacik { 103141386e1SJosef Bacik if (fs_info->csum_shash) 104141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 105141386e1SJosef Bacik } 106141386e1SJosef Bacik 107d352ac68SChris Mason /* 108d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 109d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 110d352ac68SChris Mason * just before they are sent down the IO stack. 111d352ac68SChris Mason */ 11244b8bd7eSChris Mason struct async_submit_bio { 1138896a08dSQu Wenruo struct inode *inode; 11444b8bd7eSChris Mason struct bio *bio; 115a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 11644b8bd7eSChris Mason int mirror_num; 1171941b64bSQu Wenruo 1181941b64bSQu Wenruo /* Optional parameter for submit_bio_start used by direct io */ 1191941b64bSQu Wenruo u64 dio_file_offset; 1208b712842SChris Mason struct btrfs_work work; 1214e4cbee9SChristoph Hellwig blk_status_t status; 12244b8bd7eSChris Mason }; 12344b8bd7eSChris Mason 12485d4e461SChris Mason /* 12585d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 12685d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 12785d4e461SChris Mason * the level the eb occupies in the tree. 1284008c04aSChris Mason * 12985d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 13085d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 13185d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 1324fd786e6SMisono Tomohiro * by btrfs_root->root_key.objectid. This ensures that all special purpose 1334fd786e6SMisono Tomohiro * roots have separate keysets. 1344008c04aSChris Mason * 13585d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 13685d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 13785d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1384008c04aSChris Mason * 13985d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 14085d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 14185d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 14285d4e461SChris Mason * 14385d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 14485d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 14585d4e461SChris Mason * needs update as well. 1464008c04aSChris Mason */ 1474008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1484008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1494008c04aSChris Mason # error 1504008c04aSChris Mason # endif 15185d4e461SChris Mason 152ab1405aaSDavid Sterba #define DEFINE_LEVEL(stem, level) \ 153ab1405aaSDavid Sterba .names[level] = "btrfs-" stem "-0" #level, 154ab1405aaSDavid Sterba 155ab1405aaSDavid Sterba #define DEFINE_NAME(stem) \ 156ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 0) \ 157ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 1) \ 158ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 2) \ 159ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 3) \ 160ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 4) \ 161ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 5) \ 162ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 6) \ 163ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 7) 164ab1405aaSDavid Sterba 16585d4e461SChris Mason static struct btrfs_lockdep_keyset { 16685d4e461SChris Mason u64 id; /* root objectid */ 167ab1405aaSDavid Sterba /* Longest entry: btrfs-free-space-00 */ 168387824afSDavid Sterba char names[BTRFS_MAX_LEVEL][20]; 169387824afSDavid Sterba struct lock_class_key keys[BTRFS_MAX_LEVEL]; 17085d4e461SChris Mason } btrfs_lockdep_keysets[] = { 171ab1405aaSDavid Sterba { .id = BTRFS_ROOT_TREE_OBJECTID, DEFINE_NAME("root") }, 172ab1405aaSDavid Sterba { .id = BTRFS_EXTENT_TREE_OBJECTID, DEFINE_NAME("extent") }, 173ab1405aaSDavid Sterba { .id = BTRFS_CHUNK_TREE_OBJECTID, DEFINE_NAME("chunk") }, 174ab1405aaSDavid Sterba { .id = BTRFS_DEV_TREE_OBJECTID, DEFINE_NAME("dev") }, 175ab1405aaSDavid Sterba { .id = BTRFS_CSUM_TREE_OBJECTID, DEFINE_NAME("csum") }, 176ab1405aaSDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, DEFINE_NAME("quota") }, 177ab1405aaSDavid Sterba { .id = BTRFS_TREE_LOG_OBJECTID, DEFINE_NAME("log") }, 178ab1405aaSDavid Sterba { .id = BTRFS_TREE_RELOC_OBJECTID, DEFINE_NAME("treloc") }, 179ab1405aaSDavid Sterba { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, DEFINE_NAME("dreloc") }, 180ab1405aaSDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, DEFINE_NAME("uuid") }, 181ab1405aaSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, DEFINE_NAME("free-space") }, 182ab1405aaSDavid Sterba { .id = 0, DEFINE_NAME("tree") }, 1834008c04aSChris Mason }; 18485d4e461SChris Mason 185ab1405aaSDavid Sterba #undef DEFINE_LEVEL 186ab1405aaSDavid Sterba #undef DEFINE_NAME 18785d4e461SChris Mason 18885d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 18985d4e461SChris Mason int level) 19085d4e461SChris Mason { 19185d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 19285d4e461SChris Mason 19385d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 19485d4e461SChris Mason 19585d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 19685d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 19785d4e461SChris Mason if (ks->id == objectid) 19885d4e461SChris Mason break; 19985d4e461SChris Mason 20085d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 20185d4e461SChris Mason &ks->keys[level], ks->names[level]); 20285d4e461SChris Mason } 20385d4e461SChris Mason 2044008c04aSChris Mason #endif 2054008c04aSChris Mason 206d352ac68SChris Mason /* 2072996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 208d352ac68SChris Mason */ 209c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 21019c00ddcSChris Mason { 211d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 2127280305eSDavid Sterba const int num_pages = num_extent_pages(buf); 213a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 214d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 21519c00ddcSChris Mason char *kaddr; 216e9be5a30SDavid Sterba int i; 217d5178578SJohannes Thumshirn 218d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 219d5178578SJohannes Thumshirn crypto_shash_init(shash); 220a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 221e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 222a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 22319c00ddcSChris Mason 224e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 225e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 226e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 22719c00ddcSChris Mason } 22871a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 229d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 23019c00ddcSChris Mason } 23119c00ddcSChris Mason 232d352ac68SChris Mason /* 233d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 234d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 235d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 236d352ac68SChris Mason * in the wrong place. 237d352ac68SChris Mason */ 2381259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 239b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 240b9fab919SChris Mason int atomic) 2411259ab75SChris Mason { 2422ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 2431259ab75SChris Mason int ret; 2441259ab75SChris Mason 2451259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 2461259ab75SChris Mason return 0; 2471259ab75SChris Mason 248b9fab919SChris Mason if (atomic) 249b9fab919SChris Mason return -EAGAIN; 250b9fab919SChris Mason 2512ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 252ff13db41SDavid Sterba &cached_state); 2530b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 2541259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 2551259ab75SChris Mason ret = 0; 2561259ab75SChris Mason goto out; 2571259ab75SChris Mason } 25894647322SDavid Sterba btrfs_err_rl(eb->fs_info, 25994647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 26094647322SDavid Sterba eb->start, 26129549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 2621259ab75SChris Mason ret = 1; 2630b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 26433958dc6SChris Mason out: 2652ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 266e43bbe5eSDavid Sterba &cached_state); 2671259ab75SChris Mason return ret; 2681259ab75SChris Mason } 2691259ab75SChris Mason 270e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 271e7e16f48SJohannes Thumshirn { 272e7e16f48SJohannes Thumshirn switch (csum_type) { 273e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 2743951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 2753831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 276352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 277e7e16f48SJohannes Thumshirn return true; 278e7e16f48SJohannes Thumshirn default: 279e7e16f48SJohannes Thumshirn return false; 280e7e16f48SJohannes Thumshirn } 281e7e16f48SJohannes Thumshirn } 282e7e16f48SJohannes Thumshirn 283d352ac68SChris Mason /* 2841104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 2851104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 2861104a885SDavid Sterba */ 287ab8d0fc4SJeff Mahoney static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 288ab8d0fc4SJeff Mahoney char *raw_disk_sb) 2891104a885SDavid Sterba { 2901104a885SDavid Sterba struct btrfs_super_block *disk_sb = 2911104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 29251bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 293d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 294d5178578SJohannes Thumshirn 295d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 2961104a885SDavid Sterba 2971104a885SDavid Sterba /* 2981104a885SDavid Sterba * The super_block structure does not span the whole 29951bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 30051bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 3011104a885SDavid Sterba */ 302fd08001fSEric Biggers crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, 303fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 3041104a885SDavid Sterba 30555fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 306e7e16f48SJohannes Thumshirn return 1; 3071104a885SDavid Sterba 308e7e16f48SJohannes Thumshirn return 0; 3091104a885SDavid Sterba } 3101104a885SDavid Sterba 311e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 312ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 313581c1760SQu Wenruo { 314e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 315581c1760SQu Wenruo int found_level; 316581c1760SQu Wenruo struct btrfs_key found_key; 317581c1760SQu Wenruo int ret; 318581c1760SQu Wenruo 319581c1760SQu Wenruo found_level = btrfs_header_level(eb); 320581c1760SQu Wenruo if (found_level != level) { 32163489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 32263489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 323581c1760SQu Wenruo btrfs_err(fs_info, 324581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 325581c1760SQu Wenruo eb->start, level, found_level); 326581c1760SQu Wenruo return -EIO; 327581c1760SQu Wenruo } 328581c1760SQu Wenruo 329581c1760SQu Wenruo if (!first_key) 330581c1760SQu Wenruo return 0; 331581c1760SQu Wenruo 3325d41be6fSQu Wenruo /* 3335d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 3345d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 3355d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 3365d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 3375d41be6fSQu Wenruo */ 3385d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 3395d41be6fSQu Wenruo return 0; 34062fdaa52SQu Wenruo 34162fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 34262fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 34362fdaa52SQu Wenruo btrfs_err(fs_info, 34462fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 34562fdaa52SQu Wenruo eb->start); 34662fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 34762fdaa52SQu Wenruo return -EUCLEAN; 34862fdaa52SQu Wenruo } 34962fdaa52SQu Wenruo 350581c1760SQu Wenruo if (found_level) 351581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 352581c1760SQu Wenruo else 353581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 354581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 355581c1760SQu Wenruo 356581c1760SQu Wenruo if (ret) { 35763489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 35863489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 359581c1760SQu Wenruo btrfs_err(fs_info, 360ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 361ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 362ff76a864SLiu Bo first_key->type, first_key->offset, 363ff76a864SLiu Bo found_key.objectid, found_key.type, 364ff76a864SLiu Bo found_key.offset); 365581c1760SQu Wenruo } 366581c1760SQu Wenruo return ret; 367581c1760SQu Wenruo } 368581c1760SQu Wenruo 3691104a885SDavid Sterba /* 370d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 371d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 372581c1760SQu Wenruo * 373581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 374581c1760SQu Wenruo * @level: expected level, mandatory check 375581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 376d352ac68SChris Mason */ 3775ab12d1fSDavid Sterba static int btree_read_extent_buffer_pages(struct extent_buffer *eb, 378581c1760SQu Wenruo u64 parent_transid, int level, 379581c1760SQu Wenruo struct btrfs_key *first_key) 380f188591eSChris Mason { 3815ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 382f188591eSChris Mason struct extent_io_tree *io_tree; 383ea466794SJosef Bacik int failed = 0; 384f188591eSChris Mason int ret; 385f188591eSChris Mason int num_copies = 0; 386f188591eSChris Mason int mirror_num = 0; 387ea466794SJosef Bacik int failed_mirror = 0; 388f188591eSChris Mason 3890b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 390f188591eSChris Mason while (1) { 391f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 392c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); 393256dd1bbSStefan Behrens if (!ret) { 394581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 395b9fab919SChris Mason parent_transid, 0)) 396256dd1bbSStefan Behrens ret = -EIO; 397e064d5e9SDavid Sterba else if (btrfs_verify_level_key(eb, level, 398ff76a864SLiu Bo first_key, parent_transid)) 399581c1760SQu Wenruo ret = -EUCLEAN; 400581c1760SQu Wenruo else 401581c1760SQu Wenruo break; 402256dd1bbSStefan Behrens } 403d397712bSChris Mason 4040b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 405f188591eSChris Mason eb->start, eb->len); 4064235298eSChris Mason if (num_copies == 1) 407ea466794SJosef Bacik break; 4084235298eSChris Mason 4095cf1ab56SJosef Bacik if (!failed_mirror) { 4105cf1ab56SJosef Bacik failed = 1; 4115cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4125cf1ab56SJosef Bacik } 4135cf1ab56SJosef Bacik 414f188591eSChris Mason mirror_num++; 415ea466794SJosef Bacik if (mirror_num == failed_mirror) 416ea466794SJosef Bacik mirror_num++; 417ea466794SJosef Bacik 4184235298eSChris Mason if (mirror_num > num_copies) 419ea466794SJosef Bacik break; 420f188591eSChris Mason } 421ea466794SJosef Bacik 422c0901581SStefan Behrens if (failed && !ret && failed_mirror) 42320a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 424ea466794SJosef Bacik 425ea466794SJosef Bacik return ret; 426f188591eSChris Mason } 42719c00ddcSChris Mason 428eca0f6f6SQu Wenruo static int csum_one_extent_buffer(struct extent_buffer *eb) 429eca0f6f6SQu Wenruo { 430eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 431eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 432eca0f6f6SQu Wenruo int ret; 433eca0f6f6SQu Wenruo 434eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 435eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 436eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 437eca0f6f6SQu Wenruo csum_tree_block(eb, result); 438eca0f6f6SQu Wenruo 439eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 440eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 441eca0f6f6SQu Wenruo else 442eca0f6f6SQu Wenruo ret = btrfs_check_leaf_full(eb); 443eca0f6f6SQu Wenruo 444eca0f6f6SQu Wenruo if (ret < 0) { 445eca0f6f6SQu Wenruo btrfs_print_tree(eb, 0); 446eca0f6f6SQu Wenruo btrfs_err(fs_info, 447eca0f6f6SQu Wenruo "block=%llu write time tree block corruption detected", 448eca0f6f6SQu Wenruo eb->start); 449eca0f6f6SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 450eca0f6f6SQu Wenruo return ret; 451eca0f6f6SQu Wenruo } 452eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 453eca0f6f6SQu Wenruo 454eca0f6f6SQu Wenruo return 0; 455eca0f6f6SQu Wenruo } 456eca0f6f6SQu Wenruo 457eca0f6f6SQu Wenruo /* Checksum all dirty extent buffers in one bio_vec */ 458eca0f6f6SQu Wenruo static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, 459eca0f6f6SQu Wenruo struct bio_vec *bvec) 460eca0f6f6SQu Wenruo { 461eca0f6f6SQu Wenruo struct page *page = bvec->bv_page; 462eca0f6f6SQu Wenruo u64 bvec_start = page_offset(page) + bvec->bv_offset; 463eca0f6f6SQu Wenruo u64 cur; 464eca0f6f6SQu Wenruo int ret = 0; 465eca0f6f6SQu Wenruo 466eca0f6f6SQu Wenruo for (cur = bvec_start; cur < bvec_start + bvec->bv_len; 467eca0f6f6SQu Wenruo cur += fs_info->nodesize) { 468eca0f6f6SQu Wenruo struct extent_buffer *eb; 469eca0f6f6SQu Wenruo bool uptodate; 470eca0f6f6SQu Wenruo 471eca0f6f6SQu Wenruo eb = find_extent_buffer(fs_info, cur); 472eca0f6f6SQu Wenruo uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, 473eca0f6f6SQu Wenruo fs_info->nodesize); 474eca0f6f6SQu Wenruo 475eca0f6f6SQu Wenruo /* A dirty eb shouldn't disappear from buffer_radix */ 476eca0f6f6SQu Wenruo if (WARN_ON(!eb)) 477eca0f6f6SQu Wenruo return -EUCLEAN; 478eca0f6f6SQu Wenruo 479eca0f6f6SQu Wenruo if (WARN_ON(cur != btrfs_header_bytenr(eb))) { 480eca0f6f6SQu Wenruo free_extent_buffer(eb); 481eca0f6f6SQu Wenruo return -EUCLEAN; 482eca0f6f6SQu Wenruo } 483eca0f6f6SQu Wenruo if (WARN_ON(!uptodate)) { 484eca0f6f6SQu Wenruo free_extent_buffer(eb); 485eca0f6f6SQu Wenruo return -EUCLEAN; 486eca0f6f6SQu Wenruo } 487eca0f6f6SQu Wenruo 488eca0f6f6SQu Wenruo ret = csum_one_extent_buffer(eb); 489eca0f6f6SQu Wenruo free_extent_buffer(eb); 490eca0f6f6SQu Wenruo if (ret < 0) 491eca0f6f6SQu Wenruo return ret; 492eca0f6f6SQu Wenruo } 493eca0f6f6SQu Wenruo return ret; 494eca0f6f6SQu Wenruo } 495eca0f6f6SQu Wenruo 496d352ac68SChris Mason /* 497ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 498ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 499ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 500d352ac68SChris Mason */ 501ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 50219c00ddcSChris Mason { 503ac303b69SQu Wenruo struct page *page = bvec->bv_page; 5044eee4fa4SMiao Xie u64 start = page_offset(page); 50519c00ddcSChris Mason u64 found_start; 50619c00ddcSChris Mason struct extent_buffer *eb; 507eca0f6f6SQu Wenruo 508eca0f6f6SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE) 509eca0f6f6SQu Wenruo return csum_dirty_subpage_buffers(fs_info, bvec); 510f188591eSChris Mason 5114f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 5124f2de97aSJosef Bacik if (page != eb->pages[0]) 5134f2de97aSJosef Bacik return 0; 5140f805531SAlex Lyakas 51519c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 516d3575156SNaohiro Aota 517d3575156SNaohiro Aota if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 518d3575156SNaohiro Aota WARN_ON(found_start != 0); 519d3575156SNaohiro Aota return 0; 520d3575156SNaohiro Aota } 521d3575156SNaohiro Aota 5220f805531SAlex Lyakas /* 5230f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 5240f805531SAlex Lyakas * It is useful to know what went wrong. 5250f805531SAlex Lyakas */ 5260f805531SAlex Lyakas if (WARN_ON(found_start != start)) 5270f805531SAlex Lyakas return -EUCLEAN; 5280f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 5290f805531SAlex Lyakas return -EUCLEAN; 5300f805531SAlex Lyakas 531eca0f6f6SQu Wenruo return csum_one_extent_buffer(eb); 53219c00ddcSChris Mason } 53319c00ddcSChris Mason 534b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 5352b82032cSYan Zheng { 536b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 537944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 53844880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 539944d3f9fSNikolay Borisov u8 *metadata_uuid; 5402b82032cSYan Zheng 5419a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 5429a8658e3SDavid Sterba BTRFS_FSID_SIZE); 5437239ff4bSNikolay Borisov /* 544944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 545944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 546944d3f9fSNikolay Borisov * ->fsid in this case is fine 5477239ff4bSNikolay Borisov */ 548944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 5497239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 5507239ff4bSNikolay Borisov else 5517239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 5527239ff4bSNikolay Borisov 553944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 554944d3f9fSNikolay Borisov return 0; 555944d3f9fSNikolay Borisov 556944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 557944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 558944d3f9fSNikolay Borisov return 0; 559944d3f9fSNikolay Borisov 560944d3f9fSNikolay Borisov return 1; 5612b82032cSYan Zheng } 5622b82032cSYan Zheng 56377bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 56477bf40a2SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb) 565ce9adaa5SChris Mason { 56677bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 567ce9adaa5SChris Mason u64 found_start; 56877bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 56977bf40a2SQu Wenruo u8 found_level; 5702996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 571dfd29eedSDavid Sterba const u8 *header_csum; 57277bf40a2SQu Wenruo int ret = 0; 573ea466794SJosef Bacik 574ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 575727011e0SChris Mason if (found_start != eb->start) { 576893bf4b1SSu Yue btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", 577893bf4b1SSu Yue eb->start, found_start); 578f188591eSChris Mason ret = -EIO; 57977bf40a2SQu Wenruo goto out; 580ce9adaa5SChris Mason } 581b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 58202873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 58394647322SDavid Sterba eb->start); 5841259ab75SChris Mason ret = -EIO; 58577bf40a2SQu Wenruo goto out; 5861259ab75SChris Mason } 587ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 5881c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 589893bf4b1SSu Yue btrfs_err(fs_info, "bad tree block level %d on %llu", 590893bf4b1SSu Yue (int)btrfs_header_level(eb), eb->start); 5911c24c3ceSJosef Bacik ret = -EIO; 59277bf40a2SQu Wenruo goto out; 5931c24c3ceSJosef Bacik } 594ce9adaa5SChris Mason 595c67b3892SDavid Sterba csum_tree_block(eb, result); 596dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 597dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 598a826d6dcSJosef Bacik 599dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 6002996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 601ff14aa79SDavid Sterba "checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", 602ff14aa79SDavid Sterba eb->start, 603dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 60435be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 60535be8851SJohannes Thumshirn btrfs_header_level(eb)); 6062996e1f8SJohannes Thumshirn ret = -EUCLEAN; 60777bf40a2SQu Wenruo goto out; 6082996e1f8SJohannes Thumshirn } 6092996e1f8SJohannes Thumshirn 610a826d6dcSJosef Bacik /* 611a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 612a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 613a826d6dcSJosef Bacik * return -EIO. 614a826d6dcSJosef Bacik */ 6151c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 616a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 617a826d6dcSJosef Bacik ret = -EIO; 618a826d6dcSJosef Bacik } 619ce9adaa5SChris Mason 620813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 621053ab70fSLiu Bo ret = -EIO; 622053ab70fSLiu Bo 6230b32f4bbSJosef Bacik if (!ret) 6240b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 62575391f0dSQu Wenruo else 62675391f0dSQu Wenruo btrfs_err(fs_info, 62775391f0dSQu Wenruo "block=%llu read time tree block corruption detected", 62875391f0dSQu Wenruo eb->start); 62977bf40a2SQu Wenruo out: 63077bf40a2SQu Wenruo return ret; 63177bf40a2SQu Wenruo } 63277bf40a2SQu Wenruo 633371cdc07SQu Wenruo static int validate_subpage_buffer(struct page *page, u64 start, u64 end, 634371cdc07SQu Wenruo int mirror) 635371cdc07SQu Wenruo { 636371cdc07SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 637371cdc07SQu Wenruo struct extent_buffer *eb; 638371cdc07SQu Wenruo bool reads_done; 639371cdc07SQu Wenruo int ret = 0; 640371cdc07SQu Wenruo 641371cdc07SQu Wenruo /* 642371cdc07SQu Wenruo * We don't allow bio merge for subpage metadata read, so we should 643371cdc07SQu Wenruo * only get one eb for each endio hook. 644371cdc07SQu Wenruo */ 645371cdc07SQu Wenruo ASSERT(end == start + fs_info->nodesize - 1); 646371cdc07SQu Wenruo ASSERT(PagePrivate(page)); 647371cdc07SQu Wenruo 648371cdc07SQu Wenruo eb = find_extent_buffer(fs_info, start); 649371cdc07SQu Wenruo /* 650371cdc07SQu Wenruo * When we are reading one tree block, eb must have been inserted into 651371cdc07SQu Wenruo * the radix tree. If not, something is wrong. 652371cdc07SQu Wenruo */ 653371cdc07SQu Wenruo ASSERT(eb); 654371cdc07SQu Wenruo 655371cdc07SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 656371cdc07SQu Wenruo /* Subpage read must finish in page read */ 657371cdc07SQu Wenruo ASSERT(reads_done); 658371cdc07SQu Wenruo 659371cdc07SQu Wenruo eb->read_mirror = mirror; 660371cdc07SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 661371cdc07SQu Wenruo ret = -EIO; 662371cdc07SQu Wenruo goto err; 663371cdc07SQu Wenruo } 664371cdc07SQu Wenruo ret = validate_extent_buffer(eb); 665371cdc07SQu Wenruo if (ret < 0) 666371cdc07SQu Wenruo goto err; 667371cdc07SQu Wenruo 668371cdc07SQu Wenruo if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 669371cdc07SQu Wenruo btree_readahead_hook(eb, ret); 670371cdc07SQu Wenruo 671371cdc07SQu Wenruo set_extent_buffer_uptodate(eb); 672371cdc07SQu Wenruo 673371cdc07SQu Wenruo free_extent_buffer(eb); 674371cdc07SQu Wenruo return ret; 675371cdc07SQu Wenruo err: 676371cdc07SQu Wenruo /* 677371cdc07SQu Wenruo * end_bio_extent_readpage decrements io_pages in case of error, 678371cdc07SQu Wenruo * make sure it has something to decrement. 679371cdc07SQu Wenruo */ 680371cdc07SQu Wenruo atomic_inc(&eb->io_pages); 681371cdc07SQu Wenruo clear_extent_buffer_uptodate(eb); 682371cdc07SQu Wenruo free_extent_buffer(eb); 683371cdc07SQu Wenruo return ret; 684371cdc07SQu Wenruo } 685371cdc07SQu Wenruo 686c3a3b19bSQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, 68777bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 68877bf40a2SQu Wenruo int mirror) 68977bf40a2SQu Wenruo { 69077bf40a2SQu Wenruo struct extent_buffer *eb; 69177bf40a2SQu Wenruo int ret = 0; 69277bf40a2SQu Wenruo int reads_done; 69377bf40a2SQu Wenruo 69477bf40a2SQu Wenruo ASSERT(page->private); 695371cdc07SQu Wenruo 696371cdc07SQu Wenruo if (btrfs_sb(page->mapping->host->i_sb)->sectorsize < PAGE_SIZE) 697371cdc07SQu Wenruo return validate_subpage_buffer(page, start, end, mirror); 698371cdc07SQu Wenruo 69977bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 70077bf40a2SQu Wenruo 70177bf40a2SQu Wenruo /* 70277bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 70377bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 70477bf40a2SQu Wenruo */ 70577bf40a2SQu Wenruo atomic_inc(&eb->refs); 70677bf40a2SQu Wenruo 70777bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 70877bf40a2SQu Wenruo if (!reads_done) 70977bf40a2SQu Wenruo goto err; 71077bf40a2SQu Wenruo 71177bf40a2SQu Wenruo eb->read_mirror = mirror; 71277bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 71377bf40a2SQu Wenruo ret = -EIO; 71477bf40a2SQu Wenruo goto err; 71577bf40a2SQu Wenruo } 71677bf40a2SQu Wenruo ret = validate_extent_buffer(eb); 717ce9adaa5SChris Mason err: 71879fb65a1SJosef Bacik if (reads_done && 71979fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 720d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 7214bb31e92SArne Jansen 72253b381b3SDavid Woodhouse if (ret) { 72353b381b3SDavid Woodhouse /* 72453b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 72553b381b3SDavid Woodhouse * again, we have to make sure it has something 72653b381b3SDavid Woodhouse * to decrement 72753b381b3SDavid Woodhouse */ 72853b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 7290b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 73053b381b3SDavid Woodhouse } 7310b32f4bbSJosef Bacik free_extent_buffer(eb); 73277bf40a2SQu Wenruo 733f188591eSChris Mason return ret; 734ce9adaa5SChris Mason } 735ce9adaa5SChris Mason 7364246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 737ce9adaa5SChris Mason { 73897eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 739ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 7409e0af237SLiu Bo struct btrfs_workqueue *wq; 741ce9adaa5SChris Mason 742ce9adaa5SChris Mason fs_info = end_io_wq->info; 7434e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 744d20f7043SChris Mason 745cfe94440SNaohiro Aota if (btrfs_op(bio) == BTRFS_MAP_WRITE) { 746a0cac0ecSOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) 7479e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 748a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) 7499e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 750a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 7519e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 752a0cac0ecSOmar Sandoval else 7539e0af237SLiu Bo wq = fs_info->endio_write_workers; 7549e0af237SLiu Bo } else { 7555c047a69SOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 7569e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 757a0cac0ecSOmar Sandoval else if (end_io_wq->metadata) 7589e0af237SLiu Bo wq = fs_info->endio_meta_workers; 759a0cac0ecSOmar Sandoval else 7609e0af237SLiu Bo wq = fs_info->endio_workers; 7619e0af237SLiu Bo } 7629e0af237SLiu Bo 763a0cac0ecSOmar Sandoval btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); 7649e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 765ce9adaa5SChris Mason } 766ce9adaa5SChris Mason 7674e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 768bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 7690b86a832SChris Mason { 77097eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 7718b110e39SMiao Xie 77297eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 773ce9adaa5SChris Mason if (!end_io_wq) 7744e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 775ce9adaa5SChris Mason 776ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 777ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 77822c59948SChris Mason end_io_wq->info = info; 7794e4cbee9SChristoph Hellwig end_io_wq->status = 0; 780ce9adaa5SChris Mason end_io_wq->bio = bio; 78122c59948SChris Mason end_io_wq->metadata = metadata; 782ce9adaa5SChris Mason 783ce9adaa5SChris Mason bio->bi_private = end_io_wq; 784ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 78522c59948SChris Mason return 0; 78622c59948SChris Mason } 78722c59948SChris Mason 7884a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7894a69a410SChris Mason { 7904a69a410SChris Mason struct async_submit_bio *async; 7914e4cbee9SChristoph Hellwig blk_status_t ret; 7924a69a410SChris Mason 7934a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7941941b64bSQu Wenruo ret = async->submit_bio_start(async->inode, async->bio, 7951941b64bSQu Wenruo async->dio_file_offset); 79679787eaaSJeff Mahoney if (ret) 7974e4cbee9SChristoph Hellwig async->status = ret; 7984a69a410SChris Mason } 7994a69a410SChris Mason 80006ea01b1SDavid Sterba /* 80106ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 80206ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 80306ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 80406ea01b1SDavid Sterba * 80506ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 80606ea01b1SDavid Sterba * inserted into the tree. 80706ea01b1SDavid Sterba */ 8084a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 8098b712842SChris Mason { 8108b712842SChris Mason struct async_submit_bio *async; 81106ea01b1SDavid Sterba struct inode *inode; 81206ea01b1SDavid Sterba blk_status_t ret; 8138b712842SChris Mason 8148b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 8158896a08dSQu Wenruo inode = async->inode; 8164854ddd0SChris Mason 817bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 8184e4cbee9SChristoph Hellwig if (async->status) { 8194e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 8204246a0b6SChristoph Hellwig bio_endio(async->bio); 82179787eaaSJeff Mahoney return; 82279787eaaSJeff Mahoney } 82379787eaaSJeff Mahoney 824ec39f769SChris Mason /* 825ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 826ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 827ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 828ec39f769SChris Mason */ 829ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 83008635baeSChris Mason ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 83106ea01b1SDavid Sterba if (ret) { 83206ea01b1SDavid Sterba async->bio->bi_status = ret; 83306ea01b1SDavid Sterba bio_endio(async->bio); 83406ea01b1SDavid Sterba } 8354a69a410SChris Mason } 8364a69a410SChris Mason 8374a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 8384a69a410SChris Mason { 8394a69a410SChris Mason struct async_submit_bio *async; 8404a69a410SChris Mason 8414a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 8428b712842SChris Mason kfree(async); 8438b712842SChris Mason } 8448b712842SChris Mason 8458896a08dSQu Wenruo blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, 846c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 8471941b64bSQu Wenruo u64 dio_file_offset, 848e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 84944b8bd7eSChris Mason { 8508896a08dSQu Wenruo struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 85144b8bd7eSChris Mason struct async_submit_bio *async; 85244b8bd7eSChris Mason 85344b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 85444b8bd7eSChris Mason if (!async) 8554e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 85644b8bd7eSChris Mason 8578896a08dSQu Wenruo async->inode = inode; 85844b8bd7eSChris Mason async->bio = bio; 85944b8bd7eSChris Mason async->mirror_num = mirror_num; 8604a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8614a69a410SChris Mason 862a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 863a0cac0ecSOmar Sandoval run_one_async_free); 8644a69a410SChris Mason 8651941b64bSQu Wenruo async->dio_file_offset = dio_file_offset; 8668c8bee1dSChris Mason 8674e4cbee9SChristoph Hellwig async->status = 0; 86879787eaaSJeff Mahoney 86967f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 8705cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 871d313d7a3SChris Mason 8725cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 87344b8bd7eSChris Mason return 0; 87444b8bd7eSChris Mason } 87544b8bd7eSChris Mason 8764e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 877ce3ed71aSChris Mason { 8782c30c71bSKent Overstreet struct bio_vec *bvec; 879ce3ed71aSChris Mason struct btrfs_root *root; 8802b070cfeSChristoph Hellwig int ret = 0; 8816dc4f100SMing Lei struct bvec_iter_all iter_all; 882ce3ed71aSChris Mason 883c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 8842b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 885ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 886ac303b69SQu Wenruo ret = csum_dirty_buffer(root->fs_info, bvec); 88779787eaaSJeff Mahoney if (ret) 88879787eaaSJeff Mahoney break; 889ce3ed71aSChris Mason } 8902c30c71bSKent Overstreet 8914e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 892ce3ed71aSChris Mason } 893ce3ed71aSChris Mason 8948896a08dSQu Wenruo static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, 8951941b64bSQu Wenruo u64 dio_file_offset) 89622c59948SChris Mason { 8978b712842SChris Mason /* 8988b712842SChris Mason * when we're called for a write, we're already in the async 8995443be45SChris Mason * submission context. Just jump into btrfs_map_bio 9008b712842SChris Mason */ 90179787eaaSJeff Mahoney return btree_csum_one_bio(bio); 90222c59948SChris Mason } 90322c59948SChris Mason 904f4dcfb30SJohannes Thumshirn static bool should_async_write(struct btrfs_fs_info *fs_info, 9059b4e675aSDavid Sterba struct btrfs_inode *bi) 906de0022b9SJosef Bacik { 9074eef29efSNaohiro Aota if (btrfs_is_zoned(fs_info)) 908f4dcfb30SJohannes Thumshirn return false; 9096300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 910f4dcfb30SJohannes Thumshirn return false; 9119b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 912f4dcfb30SJohannes Thumshirn return false; 913f4dcfb30SJohannes Thumshirn return true; 914de0022b9SJosef Bacik } 915de0022b9SJosef Bacik 9161b36294aSNikolay Borisov blk_status_t btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, 9171b36294aSNikolay Borisov int mirror_num, unsigned long bio_flags) 91844b8bd7eSChris Mason { 9190b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 9204e4cbee9SChristoph Hellwig blk_status_t ret; 921cad321adSChris Mason 922cfe94440SNaohiro Aota if (btrfs_op(bio) != BTRFS_MAP_WRITE) { 923cad321adSChris Mason /* 924cad321adSChris Mason * called for a read, do the setup so that checksum validation 925cad321adSChris Mason * can happen in the async kernel threads 926cad321adSChris Mason */ 9270b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 9280b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 9291d4284bdSChris Mason if (ret) 93061891923SStefan Behrens goto out_w_error; 93108635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 932f4dcfb30SJohannes Thumshirn } else if (!should_async_write(fs_info, BTRFS_I(inode))) { 933de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 934de0022b9SJosef Bacik if (ret) 93561891923SStefan Behrens goto out_w_error; 93608635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 93761891923SStefan Behrens } else { 938cad321adSChris Mason /* 93961891923SStefan Behrens * kthread helpers are used to submit writes so that 94061891923SStefan Behrens * checksumming can happen in parallel across all CPUs 941cad321adSChris Mason */ 9428896a08dSQu Wenruo ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, 9438896a08dSQu Wenruo 0, btree_submit_bio_start); 94444b8bd7eSChris Mason } 94544b8bd7eSChris Mason 9464246a0b6SChristoph Hellwig if (ret) 9474246a0b6SChristoph Hellwig goto out_w_error; 9484246a0b6SChristoph Hellwig return 0; 9494246a0b6SChristoph Hellwig 95061891923SStefan Behrens out_w_error: 9514e4cbee9SChristoph Hellwig bio->bi_status = ret; 9524246a0b6SChristoph Hellwig bio_endio(bio); 95361891923SStefan Behrens return ret; 95461891923SStefan Behrens } 95561891923SStefan Behrens 9563dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 957784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 958a6bc32b8SMel Gorman struct page *newpage, struct page *page, 959a6bc32b8SMel Gorman enum migrate_mode mode) 960784b4e29SChris Mason { 961784b4e29SChris Mason /* 962784b4e29SChris Mason * we can't safely write a btree page from here, 963784b4e29SChris Mason * we haven't done the locking hook 964784b4e29SChris Mason */ 965784b4e29SChris Mason if (PageDirty(page)) 966784b4e29SChris Mason return -EAGAIN; 967784b4e29SChris Mason /* 968784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 969784b4e29SChris Mason * We must have no buffers or drop them. 970784b4e29SChris Mason */ 971784b4e29SChris Mason if (page_has_private(page) && 972784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 973784b4e29SChris Mason return -EAGAIN; 974a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 975784b4e29SChris Mason } 9763dd1462eSJan Beulich #endif 977784b4e29SChris Mason 9780da5468fSChris Mason 9790da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9800da5468fSChris Mason struct writeback_control *wbc) 9810da5468fSChris Mason { 982e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 983e2d84521SMiao Xie int ret; 984e2d84521SMiao Xie 985d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 986448d640bSChris Mason 987448d640bSChris Mason if (wbc->for_kupdate) 988448d640bSChris Mason return 0; 989448d640bSChris Mason 990e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 991b9473439SChris Mason /* this is a bit racy, but that's ok */ 992d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 993d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 994d814a491SEthan Lien fs_info->dirty_metadata_batch); 995e2d84521SMiao Xie if (ret < 0) 996793955bcSChris Mason return 0; 997793955bcSChris Mason } 9980b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9990da5468fSChris Mason } 10000da5468fSChris Mason 100170dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 10025f39d397SChris Mason { 100398509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 100498509cfcSChris Mason return 0; 10050c4e538bSDavid Sterba 1006f7a52a40SDavid Sterba return try_release_extent_buffer(page); 1007d98237b3SChris Mason } 1008d98237b3SChris Mason 1009d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 1010d47992f8SLukas Czerner unsigned int length) 1011d98237b3SChris Mason { 1012d1310b2eSChris Mason struct extent_io_tree *tree; 1013d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10145f39d397SChris Mason extent_invalidatepage(tree, page, offset); 10155f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 10169ad6b7bcSChris Mason if (PagePrivate(page)) { 1017efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 1018efe120a0SFrank Holton "page private not zero on page %llu", 1019efe120a0SFrank Holton (unsigned long long)page_offset(page)); 1020d1b89bc0SGuoqing Jiang detach_page_private(page); 10219ad6b7bcSChris Mason } 1022d98237b3SChris Mason } 1023d98237b3SChris Mason 10240b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 10250b32f4bbSJosef Bacik { 1026bb146eb2SJosef Bacik #ifdef DEBUG 1027139e8cd3SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 1028139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 10290b32f4bbSJosef Bacik struct extent_buffer *eb; 1030139e8cd3SQu Wenruo int cur_bit = 0; 1031139e8cd3SQu Wenruo u64 page_start = page_offset(page); 10320b32f4bbSJosef Bacik 1033139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 10340b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 10350b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 10360b32f4bbSJosef Bacik BUG_ON(!eb); 10370b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10380b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 103949d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 1040139e8cd3SQu Wenruo return __set_page_dirty_nobuffers(page); 1041139e8cd3SQu Wenruo } 1042139e8cd3SQu Wenruo ASSERT(PagePrivate(page) && page->private); 1043139e8cd3SQu Wenruo subpage = (struct btrfs_subpage *)page->private; 1044139e8cd3SQu Wenruo 1045139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 1046139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 1047139e8cd3SQu Wenruo unsigned long flags; 1048139e8cd3SQu Wenruo u64 cur; 1049139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 1050139e8cd3SQu Wenruo 1051139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 1052139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 1053139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 1054139e8cd3SQu Wenruo cur_bit++; 1055139e8cd3SQu Wenruo continue; 1056139e8cd3SQu Wenruo } 1057139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 1058139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 1059139e8cd3SQu Wenruo 1060139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 1061139e8cd3SQu Wenruo ASSERT(eb); 1062139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 1063139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 106449d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 1065139e8cd3SQu Wenruo free_extent_buffer(eb); 1066139e8cd3SQu Wenruo 1067139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 1068139e8cd3SQu Wenruo } 1069bb146eb2SJosef Bacik #endif 10700b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10710b32f4bbSJosef Bacik } 10720b32f4bbSJosef Bacik 10737f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 10740da5468fSChris Mason .writepages = btree_writepages, 10755f39d397SChris Mason .releasepage = btree_releasepage, 10765f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10775a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1078784b4e29SChris Mason .migratepage = btree_migratepage, 10795a92bc88SChris Mason #endif 10800b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1081d98237b3SChris Mason }; 1082123abc88SChris Mason 10832ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 10842ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 10853fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 10863fbaf258SJosef Bacik int level) 10870999df54SChris Mason { 10880b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 10890b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 10903fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 10910999df54SChris Mason } 10920999df54SChris Mason 1093581c1760SQu Wenruo /* 1094581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 1095581c1760SQu Wenruo * verification. 1096581c1760SQu Wenruo * 10971b7ec85eSJosef Bacik * @owner_root: the objectid of the root owner for this block. 1098581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 1099581c1760SQu Wenruo * @level: expected level, mandatory check 1100581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 1101581c1760SQu Wenruo */ 11022ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 11031b7ec85eSJosef Bacik u64 owner_root, u64 parent_transid, 11041b7ec85eSJosef Bacik int level, struct btrfs_key *first_key) 1105e20d96d6SChris Mason { 11065f39d397SChris Mason struct extent_buffer *buf = NULL; 110719c00ddcSChris Mason int ret; 110819c00ddcSChris Mason 11093fbaf258SJosef Bacik buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); 1110c871b0f2SLiu Bo if (IS_ERR(buf)) 1111c871b0f2SLiu Bo return buf; 1112e4204dedSChris Mason 11135ab12d1fSDavid Sterba ret = btree_read_extent_buffer_pages(buf, parent_transid, 1114581c1760SQu Wenruo level, first_key); 11150f0fe8f7SFilipe David Borba Manana if (ret) { 1116537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 111764c043deSLiu Bo return ERR_PTR(ret); 11180f0fe8f7SFilipe David Borba Manana } 11195f39d397SChris Mason return buf; 1120ce9adaa5SChris Mason 1121eb60ceacSChris Mason } 1122eb60ceacSChris Mason 11236a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 1124ed2ff2cbSChris Mason { 11256a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 112655c69072SChris Mason if (btrfs_header_generation(buf) == 1127e2d84521SMiao Xie fs_info->running_transaction->transid) { 112849d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 1129b4ce94deSChris Mason 1130b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1131104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1132e2d84521SMiao Xie -buf->len, 1133e2d84521SMiao Xie fs_info->dirty_metadata_batch); 11340b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1135925baeddSChris Mason } 11365f39d397SChris Mason } 1137ed7b63ebSJosef Bacik } 11385f39d397SChris Mason 1139da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1140e20d96d6SChris Mason u64 objectid) 1141d97e63b6SChris Mason { 11427c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 114396dfcb46SJosef Bacik root->fs_info = fs_info; 1144cfaa7295SChris Mason root->node = NULL; 1145a28ec197SChris Mason root->commit_root = NULL; 114627cdeb70SMiao Xie root->state = 0; 1147d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 11480b86a832SChris Mason 11490f7d52f4SChris Mason root->last_trans = 0; 11506b8fad57SNikolay Borisov root->free_objectid = 0; 1151eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1152199c2a9cSMiao Xie root->nr_ordered_extents = 0; 11536bef4d31SEric Paris root->inode_tree = RB_ROOT; 115416cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1155f0486c68SYan, Zheng root->block_rsv = NULL; 11560b86a832SChris Mason 11570b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 11585d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1159eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1160eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1161199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1162199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1163d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 11642ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 11652ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 11665d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1167eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1168199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1169f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 11702ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 11712ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 11728287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1173a2135011SChris Mason mutex_init(&root->objectid_mutex); 1174e02119d5SChris Mason mutex_init(&root->log_mutex); 117531f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1176573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1177c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 11787237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 11797237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 11807237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 11818b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 11828b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 11837237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 11847237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 11857237f183SYan Zheng atomic_set(&root->log_writers, 0); 11862ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11870700cea7SElena Reshetova refcount_set(&root->refs, 1); 11888ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1189eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 11907237f183SYan Zheng root->log_transid = 0; 1191d1433debSMiao Xie root->log_transid_committed = -1; 1192257c62e1SChris Mason root->last_log_commit = 0; 1193e289f03eSFilipe Manana if (!dummy) { 119443eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 119543eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1196e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1197e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1198e289f03eSFilipe Manana } 1199017e5369SChris Mason 12003768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 12013768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 12026702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 12034d775673SChris Mason root->root_key.objectid = objectid; 12040ee5dc67SAl Viro root->anon_dev = 0; 12058ea05e3aSAlexander Block 12065f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1207370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1208bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1209bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1210bd647ce3SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 1211bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1212bd647ce3SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 1213bd647ce3SJosef Bacik #endif 12143768f368SChris Mason } 12153768f368SChris Mason 121674e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 121796dfcb46SJosef Bacik u64 objectid, gfp_t flags) 12186f07e42eSAl Viro { 121974e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 12206f07e42eSAl Viro if (root) 122196dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 12226f07e42eSAl Viro return root; 12236f07e42eSAl Viro } 12246f07e42eSAl Viro 122506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 122606ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1227da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 122806ea65a3SJosef Bacik { 122906ea65a3SJosef Bacik struct btrfs_root *root; 123006ea65a3SJosef Bacik 12317c0260eeSJeff Mahoney if (!fs_info) 12327c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 12337c0260eeSJeff Mahoney 123496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 123506ea65a3SJosef Bacik if (!root) 123606ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1237da17066cSJeff Mahoney 1238b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1239faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 124006ea65a3SJosef Bacik 124106ea65a3SJosef Bacik return root; 124206ea65a3SJosef Bacik } 124306ea65a3SJosef Bacik #endif 124406ea65a3SJosef Bacik 124520897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 124620897f5cSArne Jansen u64 objectid) 124720897f5cSArne Jansen { 12489b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 124920897f5cSArne Jansen struct extent_buffer *leaf; 125020897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 125120897f5cSArne Jansen struct btrfs_root *root; 125220897f5cSArne Jansen struct btrfs_key key; 1253b89f6d1fSFilipe Manana unsigned int nofs_flag; 125420897f5cSArne Jansen int ret = 0; 125520897f5cSArne Jansen 1256b89f6d1fSFilipe Manana /* 1257b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1258b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1259b89f6d1fSFilipe Manana */ 1260b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 126196dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1262b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 126320897f5cSArne Jansen if (!root) 126420897f5cSArne Jansen return ERR_PTR(-ENOMEM); 126520897f5cSArne Jansen 126620897f5cSArne Jansen root->root_key.objectid = objectid; 126720897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 126820897f5cSArne Jansen root->root_key.offset = 0; 126920897f5cSArne Jansen 12709631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 12719631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 127220897f5cSArne Jansen if (IS_ERR(leaf)) { 127320897f5cSArne Jansen ret = PTR_ERR(leaf); 12741dd05682STsutomu Itoh leaf = NULL; 12758a6a87cdSBoris Burkov goto fail_unlock; 127620897f5cSArne Jansen } 127720897f5cSArne Jansen 127820897f5cSArne Jansen root->node = leaf; 127920897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 128020897f5cSArne Jansen 128120897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 128227cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 128320897f5cSArne Jansen 1284f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1285f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 128620897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 128720897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 128820897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 128920897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 129020897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 129120897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 129220897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 129333d85fdaSQu Wenruo if (is_fstree(objectid)) 1294807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1295807fc790SAndy Shevchenko else 1296807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1297c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 129820897f5cSArne Jansen 12998a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 13008a6a87cdSBoris Burkov 130120897f5cSArne Jansen key.objectid = objectid; 130220897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 130320897f5cSArne Jansen key.offset = 0; 130420897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 130520897f5cSArne Jansen if (ret) 130620897f5cSArne Jansen goto fail; 130720897f5cSArne Jansen 130820897f5cSArne Jansen return root; 13091dd05682STsutomu Itoh 13108a6a87cdSBoris Burkov fail_unlock: 13118c38938cSJosef Bacik if (leaf) 13121dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 13138a6a87cdSBoris Burkov fail: 131400246528SJosef Bacik btrfs_put_root(root); 13151dd05682STsutomu Itoh 13161dd05682STsutomu Itoh return ERR_PTR(ret); 131720897f5cSArne Jansen } 131820897f5cSArne Jansen 13197237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1320e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13210f7d52f4SChris Mason { 13220f7d52f4SChris Mason struct btrfs_root *root; 1323e02119d5SChris Mason 132496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1325e02119d5SChris Mason if (!root) 13267237f183SYan Zheng return ERR_PTR(-ENOMEM); 1327e02119d5SChris Mason 1328e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1329e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1330e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 133127cdeb70SMiao Xie 13326ab6ebb7SNaohiro Aota return root; 13336ab6ebb7SNaohiro Aota } 13346ab6ebb7SNaohiro Aota 13356ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 13366ab6ebb7SNaohiro Aota struct btrfs_root *root) 13376ab6ebb7SNaohiro Aota { 13386ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 13396ab6ebb7SNaohiro Aota 13407237f183SYan Zheng /* 134192a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 134227cdeb70SMiao Xie * 134392a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 134492a7cc42SQu Wenruo * and they go away before a real commit is actually done. 134592a7cc42SQu Wenruo * 134692a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 134792a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 13487237f183SYan Zheng */ 1349e02119d5SChris Mason 13504d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 13519631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 13526ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 13536ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 1354e02119d5SChris Mason 13557237f183SYan Zheng root->node = leaf; 1356e02119d5SChris Mason 1357e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1358e02119d5SChris Mason btrfs_tree_unlock(root->node); 13596ab6ebb7SNaohiro Aota 13606ab6ebb7SNaohiro Aota return 0; 13617237f183SYan Zheng } 13627237f183SYan Zheng 13637237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13647237f183SYan Zheng struct btrfs_fs_info *fs_info) 13657237f183SYan Zheng { 13667237f183SYan Zheng struct btrfs_root *log_root; 13677237f183SYan Zheng 13687237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13697237f183SYan Zheng if (IS_ERR(log_root)) 13707237f183SYan Zheng return PTR_ERR(log_root); 13716ab6ebb7SNaohiro Aota 13723ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 13733ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 13743ddebf27SNaohiro Aota 13756ab6ebb7SNaohiro Aota if (ret) { 13766ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13776ab6ebb7SNaohiro Aota return ret; 13786ab6ebb7SNaohiro Aota } 13793ddebf27SNaohiro Aota } 13806ab6ebb7SNaohiro Aota 13817237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13827237f183SYan Zheng fs_info->log_root_tree = log_root; 13837237f183SYan Zheng return 0; 13847237f183SYan Zheng } 13857237f183SYan Zheng 13867237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13877237f183SYan Zheng struct btrfs_root *root) 13887237f183SYan Zheng { 13890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13907237f183SYan Zheng struct btrfs_root *log_root; 13917237f183SYan Zheng struct btrfs_inode_item *inode_item; 13926ab6ebb7SNaohiro Aota int ret; 13937237f183SYan Zheng 13940b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13957237f183SYan Zheng if (IS_ERR(log_root)) 13967237f183SYan Zheng return PTR_ERR(log_root); 13977237f183SYan Zheng 13986ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 13996ab6ebb7SNaohiro Aota if (ret) { 14006ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 14016ab6ebb7SNaohiro Aota return ret; 14026ab6ebb7SNaohiro Aota } 14036ab6ebb7SNaohiro Aota 14047237f183SYan Zheng log_root->last_trans = trans->transid; 14057237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14067237f183SYan Zheng 14077237f183SYan Zheng inode_item = &log_root->root_item.inode; 14083cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14093cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14103cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1411da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 14120b246afaSJeff Mahoney fs_info->nodesize); 14133cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14147237f183SYan Zheng 14155d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14167237f183SYan Zheng 14177237f183SYan Zheng WARN_ON(root->log_root); 14187237f183SYan Zheng root->log_root = log_root; 14197237f183SYan Zheng root->log_transid = 0; 1420d1433debSMiao Xie root->log_transid_committed = -1; 1421257c62e1SChris Mason root->last_log_commit = 0; 1422e02119d5SChris Mason return 0; 1423e02119d5SChris Mason } 1424e02119d5SChris Mason 142549d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 142649d11beaSJosef Bacik struct btrfs_path *path, 1427cb517eabSMiao Xie struct btrfs_key *key) 1428e02119d5SChris Mason { 1429e02119d5SChris Mason struct btrfs_root *root; 1430e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 143184234f3aSYan Zheng u64 generation; 1432cb517eabSMiao Xie int ret; 1433581c1760SQu Wenruo int level; 1434cb517eabSMiao Xie 143596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 143649d11beaSJosef Bacik if (!root) 143749d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 14380f7d52f4SChris Mason 1439cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1440cb517eabSMiao Xie &root->root_item, &root->root_key); 14410f7d52f4SChris Mason if (ret) { 144213a8a7c8SYan, Zheng if (ret > 0) 144313a8a7c8SYan, Zheng ret = -ENOENT; 144449d11beaSJosef Bacik goto fail; 14450f7d52f4SChris Mason } 144613a8a7c8SYan, Zheng 144784234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1448581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 14492ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 14502ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 14511b7ec85eSJosef Bacik key->objectid, generation, level, NULL); 145264c043deSLiu Bo if (IS_ERR(root->node)) { 145364c043deSLiu Bo ret = PTR_ERR(root->node); 14548c38938cSJosef Bacik root->node = NULL; 145549d11beaSJosef Bacik goto fail; 1456cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1457cb517eabSMiao Xie ret = -EIO; 145849d11beaSJosef Bacik goto fail; 1459416bc658SJosef Bacik } 14605d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1461cb517eabSMiao Xie return root; 146249d11beaSJosef Bacik fail: 146300246528SJosef Bacik btrfs_put_root(root); 146449d11beaSJosef Bacik return ERR_PTR(ret); 146549d11beaSJosef Bacik } 146649d11beaSJosef Bacik 146749d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 146849d11beaSJosef Bacik struct btrfs_key *key) 146949d11beaSJosef Bacik { 147049d11beaSJosef Bacik struct btrfs_root *root; 147149d11beaSJosef Bacik struct btrfs_path *path; 147249d11beaSJosef Bacik 147349d11beaSJosef Bacik path = btrfs_alloc_path(); 147449d11beaSJosef Bacik if (!path) 147549d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 147649d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 147749d11beaSJosef Bacik btrfs_free_path(path); 147849d11beaSJosef Bacik 147949d11beaSJosef Bacik return root; 1480cb517eabSMiao Xie } 1481cb517eabSMiao Xie 14822dfb1e43SQu Wenruo /* 14832dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 14842dfb1e43SQu Wenruo * 14852dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 14862dfb1e43SQu Wenruo */ 14872dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1488cb517eabSMiao Xie { 1489cb517eabSMiao Xie int ret; 1490dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1491cb517eabSMiao Xie 1492dcc3eb96SNikolay Borisov /* 1493dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1494dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1495dcc3eb96SNikolay Borisov */ 1496dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1497dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1498dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1499dcc3eb96SNikolay Borisov if (ret) 15008257b2dcSMiao Xie goto fail; 15018257b2dcSMiao Xie 1502aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 150337f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 150492a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1505f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1506f39e4571SJosef Bacik } 1507f39e4571SJosef Bacik 1508851fd730SQu Wenruo /* 1509851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1510851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1511851fd730SQu Wenruo */ 1512851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1513851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 15142dfb1e43SQu Wenruo if (!anon_dev) { 1515cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1516cb517eabSMiao Xie if (ret) 1517876d2cf1SLiu Bo goto fail; 15182dfb1e43SQu Wenruo } else { 15192dfb1e43SQu Wenruo root->anon_dev = anon_dev; 15202dfb1e43SQu Wenruo } 1521851fd730SQu Wenruo } 1522f32e48e9SChandan Rajendra 1523f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1524453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1525f32e48e9SChandan Rajendra if (ret) { 1526f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1527876d2cf1SLiu Bo goto fail; 1528f32e48e9SChandan Rajendra } 1529f32e48e9SChandan Rajendra 15306b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1531f32e48e9SChandan Rajendra 1532f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1533f32e48e9SChandan Rajendra 1534cb517eabSMiao Xie return 0; 1535cb517eabSMiao Xie fail: 153684db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1537cb517eabSMiao Xie return ret; 1538cb517eabSMiao Xie } 1539cb517eabSMiao Xie 1540a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1541cb517eabSMiao Xie u64 root_id) 1542cb517eabSMiao Xie { 1543cb517eabSMiao Xie struct btrfs_root *root; 1544cb517eabSMiao Xie 1545cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1546cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1547cb517eabSMiao Xie (unsigned long)root_id); 1548bc44d7c4SJosef Bacik if (root) 154900246528SJosef Bacik root = btrfs_grab_root(root); 1550cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1551cb517eabSMiao Xie return root; 1552cb517eabSMiao Xie } 1553cb517eabSMiao Xie 155449d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 155549d11beaSJosef Bacik u64 objectid) 155649d11beaSJosef Bacik { 155749d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 155849d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 155949d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 156049d11beaSJosef Bacik return btrfs_grab_root(fs_info->extent_root); 156149d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 156249d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 156349d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 156449d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 156549d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 156649d11beaSJosef Bacik return btrfs_grab_root(fs_info->csum_root); 156749d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 156849d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 156949d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 157049d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 157149d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 157249d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 157349d11beaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 157449d11beaSJosef Bacik return btrfs_grab_root(fs_info->free_space_root) ? 157549d11beaSJosef Bacik fs_info->free_space_root : ERR_PTR(-ENOENT); 157649d11beaSJosef Bacik return NULL; 157749d11beaSJosef Bacik } 157849d11beaSJosef Bacik 1579cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1580cb517eabSMiao Xie struct btrfs_root *root) 1581cb517eabSMiao Xie { 1582cb517eabSMiao Xie int ret; 1583cb517eabSMiao Xie 1584e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1585cb517eabSMiao Xie if (ret) 1586cb517eabSMiao Xie return ret; 1587cb517eabSMiao Xie 1588cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1589cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1590cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1591cb517eabSMiao Xie root); 1592af01d2e5SJosef Bacik if (ret == 0) { 159300246528SJosef Bacik btrfs_grab_root(root); 159427cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1595af01d2e5SJosef Bacik } 1596cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1597cb517eabSMiao Xie radix_tree_preload_end(); 1598cb517eabSMiao Xie 1599cb517eabSMiao Xie return ret; 1600cb517eabSMiao Xie } 1601cb517eabSMiao Xie 1602bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1603bd647ce3SJosef Bacik { 1604bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1605bd647ce3SJosef Bacik struct btrfs_root *root; 1606bd647ce3SJosef Bacik 1607bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1608457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1609457f1864SJosef Bacik 1610bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1611bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1612457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 161371008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1614bd647ce3SJosef Bacik refcount_read(&root->refs)); 1615bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 161600246528SJosef Bacik btrfs_put_root(root); 161700246528SJosef Bacik btrfs_put_root(root); 1618bd647ce3SJosef Bacik } 1619bd647ce3SJosef Bacik #endif 1620bd647ce3SJosef Bacik } 1621bd647ce3SJosef Bacik 16220d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 16230d4b0463SJosef Bacik { 1624141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1625141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 16265deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1627141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1628141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1629141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1630141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 16310d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 16320d4b0463SJosef Bacik kfree(fs_info->delayed_root); 163300246528SJosef Bacik btrfs_put_root(fs_info->extent_root); 163400246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 163500246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 163600246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 163700246528SJosef Bacik btrfs_put_root(fs_info->csum_root); 163800246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 163900246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 164000246528SJosef Bacik btrfs_put_root(fs_info->free_space_root); 164100246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1642aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 1643bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 16443fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 16450d4b0463SJosef Bacik kfree(fs_info->super_copy); 16460d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 16478481dd80SQu Wenruo kfree(fs_info->subpage_info); 16480d4b0463SJosef Bacik kvfree(fs_info); 16490d4b0463SJosef Bacik } 16500d4b0463SJosef Bacik 16510d4b0463SJosef Bacik 16522dfb1e43SQu Wenruo /* 16532dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 16542dfb1e43SQu Wenruo * 16552dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 16562dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 16572dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 16582dfb1e43SQu Wenruo * 16592dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 16602dfb1e43SQu Wenruo * 16612dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 16622dfb1e43SQu Wenruo * same root objectid. 16632dfb1e43SQu Wenruo * 16642dfb1e43SQu Wenruo * @objectid: root id 16652dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 16662dfb1e43SQu Wenruo * pass 0 for new allocation. 16672dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 16682dfb1e43SQu Wenruo * for orphan roots 16692dfb1e43SQu Wenruo */ 16702dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 16712dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 16722dfb1e43SQu Wenruo bool check_ref) 16735eda7b5eSChris Mason { 16745eda7b5eSChris Mason struct btrfs_root *root; 1675381cf658SDavid Sterba struct btrfs_path *path; 16761d4c08e0SDavid Sterba struct btrfs_key key; 16775eda7b5eSChris Mason int ret; 16785eda7b5eSChris Mason 167949d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 168049d11beaSJosef Bacik if (root) 168149d11beaSJosef Bacik return root; 16824df27c4dSYan, Zheng again: 168356e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 168448475471SStefan Behrens if (root) { 16852dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 16862dfb1e43SQu Wenruo ASSERT(!anon_dev); 1687bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 168800246528SJosef Bacik btrfs_put_root(root); 168948475471SStefan Behrens return ERR_PTR(-ENOENT); 1690bc44d7c4SJosef Bacik } 16915eda7b5eSChris Mason return root; 169248475471SStefan Behrens } 16935eda7b5eSChris Mason 169456e9357aSDavid Sterba key.objectid = objectid; 169556e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 169656e9357aSDavid Sterba key.offset = (u64)-1; 169756e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 16985eda7b5eSChris Mason if (IS_ERR(root)) 16995eda7b5eSChris Mason return root; 17003394e160SChris Mason 1701c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1702d68fc57bSYan, Zheng ret = -ENOENT; 1703d68fc57bSYan, Zheng goto fail; 1704d68fc57bSYan, Zheng } 1705d68fc57bSYan, Zheng 17062dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1707cb517eabSMiao Xie if (ret) 1708cb517eabSMiao Xie goto fail; 1709cb517eabSMiao Xie 1710381cf658SDavid Sterba path = btrfs_alloc_path(); 1711381cf658SDavid Sterba if (!path) { 1712381cf658SDavid Sterba ret = -ENOMEM; 1713381cf658SDavid Sterba goto fail; 1714381cf658SDavid Sterba } 17151d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 17161d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 171756e9357aSDavid Sterba key.offset = objectid; 17181d4c08e0SDavid Sterba 17191d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1720381cf658SDavid Sterba btrfs_free_path(path); 1721d68fc57bSYan, Zheng if (ret < 0) 1722d68fc57bSYan, Zheng goto fail; 1723d68fc57bSYan, Zheng if (ret == 0) 172427cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1725d68fc57bSYan, Zheng 1726cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 17270f7d52f4SChris Mason if (ret) { 172800246528SJosef Bacik btrfs_put_root(root); 17294785e24fSJosef Bacik if (ret == -EEXIST) 17304df27c4dSYan, Zheng goto again; 17314df27c4dSYan, Zheng goto fail; 17324df27c4dSYan, Zheng } 1733edbd8d4eSChris Mason return root; 17344df27c4dSYan, Zheng fail: 17358c38938cSJosef Bacik btrfs_put_root(root); 17364df27c4dSYan, Zheng return ERR_PTR(ret); 1737edbd8d4eSChris Mason } 1738edbd8d4eSChris Mason 17392dfb1e43SQu Wenruo /* 17402dfb1e43SQu Wenruo * Get in-memory reference of a root structure 17412dfb1e43SQu Wenruo * 17422dfb1e43SQu Wenruo * @objectid: tree objectid 17432dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 17442dfb1e43SQu Wenruo * one reference 17452dfb1e43SQu Wenruo */ 17462dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 17472dfb1e43SQu Wenruo u64 objectid, bool check_ref) 17482dfb1e43SQu Wenruo { 17492dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 17502dfb1e43SQu Wenruo } 17512dfb1e43SQu Wenruo 17522dfb1e43SQu Wenruo /* 17532dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 17542dfb1e43SQu Wenruo * the anonymous block device id 17552dfb1e43SQu Wenruo * 17562dfb1e43SQu Wenruo * @objectid: tree objectid 17572dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 17582dfb1e43SQu Wenruo * parameter value 17592dfb1e43SQu Wenruo */ 17602dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 17612dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 17622dfb1e43SQu Wenruo { 17632dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 17642dfb1e43SQu Wenruo } 17652dfb1e43SQu Wenruo 17668b712842SChris Mason /* 176749d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 176849d11beaSJosef Bacik * @fs_info: the fs_info 176949d11beaSJosef Bacik * @objectid: the objectid we need to lookup 177049d11beaSJosef Bacik * 177149d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 177249d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 177349d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 177449d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 177549d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 177649d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 177749d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 177849d11beaSJosef Bacik * backref code is finished using the root. 177949d11beaSJosef Bacik */ 178049d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 178149d11beaSJosef Bacik struct btrfs_path *path, 178249d11beaSJosef Bacik u64 objectid) 178349d11beaSJosef Bacik { 178449d11beaSJosef Bacik struct btrfs_root *root; 178549d11beaSJosef Bacik struct btrfs_key key; 178649d11beaSJosef Bacik 178749d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 178849d11beaSJosef Bacik 178949d11beaSJosef Bacik /* 179049d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 179149d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 179249d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 179349d11beaSJosef Bacik * != NULL just return it. 179449d11beaSJosef Bacik */ 179549d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 179649d11beaSJosef Bacik if (root) 179749d11beaSJosef Bacik return root; 179849d11beaSJosef Bacik 179949d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 180049d11beaSJosef Bacik if (root) 180149d11beaSJosef Bacik return root; 180249d11beaSJosef Bacik 180349d11beaSJosef Bacik key.objectid = objectid; 180449d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 180549d11beaSJosef Bacik key.offset = (u64)-1; 180649d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 180749d11beaSJosef Bacik btrfs_release_path(path); 180849d11beaSJosef Bacik 180949d11beaSJosef Bacik return root; 181049d11beaSJosef Bacik } 181149d11beaSJosef Bacik 181249d11beaSJosef Bacik /* 18138b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 18148b712842SChris Mason * functions. This is where read checksum verification actually happens 18158b712842SChris Mason */ 18168b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1817ce9adaa5SChris Mason { 1818ce9adaa5SChris Mason struct bio *bio; 181997eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1820ce9adaa5SChris Mason 182197eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1822ce9adaa5SChris Mason bio = end_io_wq->bio; 18238b712842SChris Mason 18244e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1825ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1826ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 18274246a0b6SChristoph Hellwig bio_endio(bio); 18289be490f1SOmar Sandoval kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 1829ce9adaa5SChris Mason } 183044b8bd7eSChris Mason 1831a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1832a74a4b97SChris Mason { 1833a74a4b97SChris Mason struct btrfs_root *root = arg; 18340b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1835d0278245SMiao Xie int again; 1836a74a4b97SChris Mason 1837d6fd0ae2SOmar Sandoval while (1) { 1838d0278245SMiao Xie again = 0; 18399d1a2a3aSDavid Sterba 1840fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1841fd340d0fSJosef Bacik 1842d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 18432ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1844d0278245SMiao Xie goto sleep; 1845d0278245SMiao Xie 184690c711abSZygo Blaxell /* 184790c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 184890c711abSZygo Blaxell * before we have finished mounting the filesystem. 184990c711abSZygo Blaxell */ 18500b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 185190c711abSZygo Blaxell goto sleep; 185290c711abSZygo Blaxell 18530b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1854d0278245SMiao Xie goto sleep; 1855d0278245SMiao Xie 1856dc7f370cSMiao Xie /* 1857dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1858dc7f370cSMiao Xie * during the above check and trylock. 1859dc7f370cSMiao Xie */ 18602ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 18610b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1862dc7f370cSMiao Xie goto sleep; 1863dc7f370cSMiao Xie } 1864dc7f370cSMiao Xie 18652ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1866c2d6cb16SFilipe Manana 18679d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 18680b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1869a74a4b97SChris Mason 1870d0278245SMiao Xie /* 187105323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 187205323cd1SMiao Xie * needn't do anything special here. 1873d0278245SMiao Xie */ 18740b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 187567c5e7d4SFilipe Manana 187667c5e7d4SFilipe Manana /* 1877f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 187867c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 187967c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1880f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 188167c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 188267c5e7d4SFilipe Manana * unused block groups. 188367c5e7d4SFilipe Manana */ 18840b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 188518bb8bbfSJohannes Thumshirn 188618bb8bbfSJohannes Thumshirn /* 188718bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 188818bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 188918bb8bbfSJohannes Thumshirn * space. 189018bb8bbfSJohannes Thumshirn */ 189118bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1892d0278245SMiao Xie sleep: 1893a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1894d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1895d6fd0ae2SOmar Sandoval kthread_parkme(); 1896d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1897d6fd0ae2SOmar Sandoval return 0; 1898838fe188SJiri Kosina if (!again) { 1899a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1900a74a4b97SChris Mason schedule(); 1901a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1902a74a4b97SChris Mason } 1903da288d28SFilipe Manana } 1904a74a4b97SChris Mason } 1905a74a4b97SChris Mason 1906a74a4b97SChris Mason static int transaction_kthread(void *arg) 1907a74a4b97SChris Mason { 1908a74a4b97SChris Mason struct btrfs_root *root = arg; 19090b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1910a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1911a74a4b97SChris Mason struct btrfs_transaction *cur; 19128929ecfaSYan, Zheng u64 transid; 1913643900beSNikolay Borisov time64_t delta; 1914a74a4b97SChris Mason unsigned long delay; 1915914b2007SJan Kara bool cannot_commit; 1916a74a4b97SChris Mason 1917a74a4b97SChris Mason do { 1918914b2007SJan Kara cannot_commit = false; 1919ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 19200b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1921a74a4b97SChris Mason 19220b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 19230b246afaSJeff Mahoney cur = fs_info->running_transaction; 1924a74a4b97SChris Mason if (!cur) { 19250b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1926a74a4b97SChris Mason goto sleep; 1927a74a4b97SChris Mason } 192831153d81SYan Zheng 1929643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 19303296bf56SQu Wenruo if (cur->state < TRANS_STATE_COMMIT_START && 1931643900beSNikolay Borisov delta < fs_info->commit_interval) { 19320b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1933fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1934fb8a7e94SNikolay Borisov delay = min(delay, 1935fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1936a74a4b97SChris Mason goto sleep; 1937a74a4b97SChris Mason } 19388929ecfaSYan, Zheng transid = cur->transid; 19390b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 194056bec294SChris Mason 194179787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1942354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1943914b2007SJan Kara if (IS_ERR(trans)) { 1944354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1945914b2007SJan Kara cannot_commit = true; 194679787eaaSJeff Mahoney goto sleep; 1947914b2007SJan Kara } 19488929ecfaSYan, Zheng if (transid == trans->transid) { 19493a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 19508929ecfaSYan, Zheng } else { 19513a45bb20SJeff Mahoney btrfs_end_transaction(trans); 19528929ecfaSYan, Zheng } 1953a74a4b97SChris Mason sleep: 19540b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 19550b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1956a74a4b97SChris Mason 195784961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 19582ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 19598929ecfaSYan, Zheng if (!kthread_should_stop() && 19600b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1961914b2007SJan Kara cannot_commit)) 1962bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1963a74a4b97SChris Mason } while (!kthread_should_stop()); 1964a74a4b97SChris Mason return 0; 1965a74a4b97SChris Mason } 1966a74a4b97SChris Mason 1967af31f5e5SChris Mason /* 196801f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 196901f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 197001f0f9daSNikolay Borisov * anything. 1971af31f5e5SChris Mason * 1972af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1973af31f5e5SChris Mason * generation of the latest root in the array with the generation 1974af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1975af31f5e5SChris Mason */ 197601f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1977af31f5e5SChris Mason { 197801f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1979af31f5e5SChris Mason u64 cur; 1980af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1981af31f5e5SChris Mason int i; 1982af31f5e5SChris Mason 1983af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1984af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1985af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1986af31f5e5SChris Mason if (cur == newest_gen) 198701f0f9daSNikolay Borisov return i; 1988af31f5e5SChris Mason } 1989af31f5e5SChris Mason 199001f0f9daSNikolay Borisov return -EINVAL; 1991af31f5e5SChris Mason } 1992af31f5e5SChris Mason 1993af31f5e5SChris Mason /* 1994af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1995af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1996af31f5e5SChris Mason * done 1997af31f5e5SChris Mason */ 1998af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1999af31f5e5SChris Mason { 20006ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 2001af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2002af31f5e5SChris Mason 2003af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 2004af31f5e5SChris Mason 2005af31f5e5SChris Mason /* 2006af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 2007af31f5e5SChris Mason * regardless of which ones we use today 2008af31f5e5SChris Mason */ 2009af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 2010af31f5e5SChris Mason 2011af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 2012af31f5e5SChris Mason 2013af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 2014af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 2015af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 2016af31f5e5SChris Mason 2017af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 2018af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 2019af31f5e5SChris Mason 2020af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 2021af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 2022af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 2023af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 2024af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 2025af31f5e5SChris Mason 2026af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 2027af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 2028af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 2029af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 2030af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 2031af31f5e5SChris Mason 20327c7e82a7SChris Mason /* 20337c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 20347c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 20357c7e82a7SChris Mason */ 20367c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 20377c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 20387c7e82a7SChris Mason info->fs_root->node->start); 2039af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2040af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2041af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2042af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 20437c7e82a7SChris Mason } 2044af31f5e5SChris Mason 2045af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2046af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2047af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2048af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2049af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2050af31f5e5SChris Mason 2051af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 2052af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 2053af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 2054af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 2055af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 2056af31f5e5SChris Mason 2057af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2058af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2059af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2060af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2061af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2062af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2063af31f5e5SChris Mason 2064af31f5e5SChris Mason /* 2065af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2066af31f5e5SChris Mason * for the next commit 2067af31f5e5SChris Mason */ 2068af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2069af31f5e5SChris Mason &info->super_for_commit->super_roots, 2070af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2071af31f5e5SChris Mason } 2072af31f5e5SChris Mason 2073af31f5e5SChris Mason /* 2074bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 2075bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 2076bd2336b2SNikolay Borisov * 2077bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 2078bd2336b2SNikolay Borisov * priority - priority of backup root required 2079bd2336b2SNikolay Borisov * 2080bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 2081bd2336b2SNikolay Borisov */ 2082bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 2083bd2336b2SNikolay Borisov { 2084bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2085bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 2086bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 2087bd2336b2SNikolay Borisov 2088bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 2089bd2336b2SNikolay Borisov if (priority == 0) 2090bd2336b2SNikolay Borisov return backup_index; 2091bd2336b2SNikolay Borisov 2092bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 2093bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 2094bd2336b2SNikolay Borisov } else { 2095bd2336b2SNikolay Borisov return -EINVAL; 2096bd2336b2SNikolay Borisov } 2097bd2336b2SNikolay Borisov 2098bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 2099bd2336b2SNikolay Borisov 2100bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 2101bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 2102bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2103bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 2104bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 2105bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2106bd2336b2SNikolay Borisov 2107bd2336b2SNikolay Borisov /* 2108bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 2109bd2336b2SNikolay Borisov * need a fsck 2110bd2336b2SNikolay Borisov */ 2111bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2112bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2113bd2336b2SNikolay Borisov 2114bd2336b2SNikolay Borisov return backup_index; 2115bd2336b2SNikolay Borisov } 2116bd2336b2SNikolay Borisov 21177abadb64SLiu Bo /* helper to cleanup workers */ 21187abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 21197abadb64SLiu Bo { 2120dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2121afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 21225cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2123fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2124fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 2125d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2126fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2127fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 21285b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2129e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2130736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2131a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2132fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2133b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 2134b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 2135a9b9477dSFilipe Manana /* 2136a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2137a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2138a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2139a9b9477dSFilipe Manana */ 2140a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2141a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 21427abadb64SLiu Bo } 21437abadb64SLiu Bo 21442e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 21452e9f5954SRashika { 21462e9f5954SRashika if (root) { 21472e9f5954SRashika free_extent_buffer(root->node); 21482e9f5954SRashika free_extent_buffer(root->commit_root); 21492e9f5954SRashika root->node = NULL; 21502e9f5954SRashika root->commit_root = NULL; 21512e9f5954SRashika } 21522e9f5954SRashika } 21532e9f5954SRashika 2154af31f5e5SChris Mason /* helper to cleanup tree roots */ 21554273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2156af31f5e5SChris Mason { 21572e9f5954SRashika free_root_extent_buffers(info->tree_root); 2158655b09feSJosef Bacik 21592e9f5954SRashika free_root_extent_buffers(info->dev_root); 21602e9f5954SRashika free_root_extent_buffers(info->extent_root); 21612e9f5954SRashika free_root_extent_buffers(info->csum_root); 21622e9f5954SRashika free_root_extent_buffers(info->quota_root); 21632e9f5954SRashika free_root_extent_buffers(info->uuid_root); 21648c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2165aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 21664273eaffSAnand Jain if (free_chunk_root) 21672e9f5954SRashika free_root_extent_buffers(info->chunk_root); 216870f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2169af31f5e5SChris Mason } 2170af31f5e5SChris Mason 21718c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 21728c38938cSJosef Bacik { 21738c38938cSJosef Bacik if (!root) 21748c38938cSJosef Bacik return; 21758c38938cSJosef Bacik 21768c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 21778c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 21781dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 21798c38938cSJosef Bacik if (root->anon_dev) 21808c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 21818c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2182923eb523SJohannes Thumshirn free_root_extent_buffers(root); 21838c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 21848c38938cSJosef Bacik spin_lock(&root->fs_info->fs_roots_radix_lock); 21858c38938cSJosef Bacik list_del_init(&root->leak_list); 21868c38938cSJosef Bacik spin_unlock(&root->fs_info->fs_roots_radix_lock); 21878c38938cSJosef Bacik #endif 21888c38938cSJosef Bacik kfree(root); 21898c38938cSJosef Bacik } 21908c38938cSJosef Bacik } 21918c38938cSJosef Bacik 2192faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2193171f6537SJosef Bacik { 2194171f6537SJosef Bacik int ret; 2195171f6537SJosef Bacik struct btrfs_root *gang[8]; 2196171f6537SJosef Bacik int i; 2197171f6537SJosef Bacik 2198171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2199171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2200171f6537SJosef Bacik struct btrfs_root, root_list); 2201171f6537SJosef Bacik list_del(&gang[0]->root_list); 2202171f6537SJosef Bacik 22038c38938cSJosef Bacik if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2204cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 220500246528SJosef Bacik btrfs_put_root(gang[0]); 2206171f6537SJosef Bacik } 2207171f6537SJosef Bacik 2208171f6537SJosef Bacik while (1) { 2209171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2210171f6537SJosef Bacik (void **)gang, 0, 2211171f6537SJosef Bacik ARRAY_SIZE(gang)); 2212171f6537SJosef Bacik if (!ret) 2213171f6537SJosef Bacik break; 2214171f6537SJosef Bacik for (i = 0; i < ret; i++) 2215cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2216171f6537SJosef Bacik } 2217171f6537SJosef Bacik } 2218af31f5e5SChris Mason 2219638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2220638aa7edSEric Sandeen { 2221638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2222638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2223638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2224638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2225638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2226638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2227ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2228638aa7edSEric Sandeen } 2229638aa7edSEric Sandeen 2230779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2231779a65a4SEric Sandeen { 2232779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2233779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2234779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2235779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2236779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2237779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2238907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 2239779a65a4SEric Sandeen } 2240779a65a4SEric Sandeen 22416bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2242f37938e0SEric Sandeen { 22432ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 22442ff7e61eSJeff Mahoney 22452ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 22462ff7e61eSJeff Mahoney set_nlink(inode, 1); 2247f37938e0SEric Sandeen /* 2248f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2249f37938e0SEric Sandeen * the real end of the address space is determined by all of 2250f37938e0SEric Sandeen * the devices in the system 2251f37938e0SEric Sandeen */ 22522ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 22532ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2254f37938e0SEric Sandeen 22552ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 225643eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 22572c53a14dSQu Wenruo IO_TREE_BTREE_INODE_IO, inode); 22587b439738SDavid Sterba BTRFS_I(inode)->io_tree.track_uptodate = false; 22592ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2260f37938e0SEric Sandeen 22615c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 22622ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 22632ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 22642ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2265f37938e0SEric Sandeen } 2266f37938e0SEric Sandeen 2267ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2268ad618368SEric Sandeen { 2269ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2270129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 22717f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2272ad618368SEric Sandeen } 2273ad618368SEric Sandeen 2274f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2275f9e92e40SEric Sandeen { 2276f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2277f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2278f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2279f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2280f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2281f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2282d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2283f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2284f9e92e40SEric Sandeen } 2285f9e92e40SEric Sandeen 22862a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 22872a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 22882a458198SEric Sandeen { 2289f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 22906f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 22912a458198SEric Sandeen 22922a458198SEric Sandeen fs_info->workers = 2293cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2294cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 22952a458198SEric Sandeen 22962a458198SEric Sandeen fs_info->delalloc_workers = 2297cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2298cb001095SJeff Mahoney flags, max_active, 2); 22992a458198SEric Sandeen 23002a458198SEric Sandeen fs_info->flush_workers = 2301cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2302cb001095SJeff Mahoney flags, max_active, 0); 23032a458198SEric Sandeen 23042a458198SEric Sandeen fs_info->caching_workers = 2305cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 23062a458198SEric Sandeen 23072a458198SEric Sandeen fs_info->fixup_workers = 2308cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 23092a458198SEric Sandeen 23102a458198SEric Sandeen /* 23112a458198SEric Sandeen * endios are largely parallel and should have a very 23122a458198SEric Sandeen * low idle thresh 23132a458198SEric Sandeen */ 23142a458198SEric Sandeen fs_info->endio_workers = 2315cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 23162a458198SEric Sandeen fs_info->endio_meta_workers = 2317cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2318cb001095SJeff Mahoney max_active, 4); 23192a458198SEric Sandeen fs_info->endio_meta_write_workers = 2320cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2321cb001095SJeff Mahoney max_active, 2); 23222a458198SEric Sandeen fs_info->endio_raid56_workers = 2323cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2324cb001095SJeff Mahoney max_active, 4); 23252a458198SEric Sandeen fs_info->rmw_workers = 2326cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 23272a458198SEric Sandeen fs_info->endio_write_workers = 2328cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2329cb001095SJeff Mahoney max_active, 2); 23302a458198SEric Sandeen fs_info->endio_freespace_worker = 2331cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2332cb001095SJeff Mahoney max_active, 0); 23332a458198SEric Sandeen fs_info->delayed_workers = 2334cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2335cb001095SJeff Mahoney max_active, 0); 23362a458198SEric Sandeen fs_info->readahead_workers = 2337cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2338cb001095SJeff Mahoney max_active, 2); 23392a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2340cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2341b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2342b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 23432a458198SEric Sandeen 23442a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 2345ba8a9d07SChris Mason fs_info->flush_workers && 23462a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 23472a458198SEric Sandeen fs_info->endio_meta_write_workers && 23482a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 23492a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 23502a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 23512a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 2352b0643e59SDennis Zhou fs_info->qgroup_rescan_workers && 2353b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 23542a458198SEric Sandeen return -ENOMEM; 23552a458198SEric Sandeen } 23562a458198SEric Sandeen 23572a458198SEric Sandeen return 0; 23582a458198SEric Sandeen } 23592a458198SEric Sandeen 23606d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 23616d97c6e3SJohannes Thumshirn { 23626d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2363b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 23646d97c6e3SJohannes Thumshirn 2365b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 23666d97c6e3SJohannes Thumshirn 23676d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 23686d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2369b4e967beSDavid Sterba csum_driver); 23706d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 23716d97c6e3SJohannes Thumshirn } 23726d97c6e3SJohannes Thumshirn 23736d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 23746d97c6e3SJohannes Thumshirn 23756d97c6e3SJohannes Thumshirn return 0; 23766d97c6e3SJohannes Thumshirn } 23776d97c6e3SJohannes Thumshirn 237863443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 237963443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 238063443bf5SEric Sandeen { 238163443bf5SEric Sandeen int ret; 238263443bf5SEric Sandeen struct btrfs_root *log_tree_root; 238363443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 238463443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2385581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 238663443bf5SEric Sandeen 238763443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2388f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 238963443bf5SEric Sandeen return -EIO; 239063443bf5SEric Sandeen } 239163443bf5SEric Sandeen 239296dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 239396dfcb46SJosef Bacik GFP_KERNEL); 239463443bf5SEric Sandeen if (!log_tree_root) 239563443bf5SEric Sandeen return -ENOMEM; 239663443bf5SEric Sandeen 23972ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 23981b7ec85eSJosef Bacik BTRFS_TREE_LOG_OBJECTID, 23991b7ec85eSJosef Bacik fs_info->generation + 1, level, 24001b7ec85eSJosef Bacik NULL); 240164c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2402f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 24030eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 24048c38938cSJosef Bacik log_tree_root->node = NULL; 240500246528SJosef Bacik btrfs_put_root(log_tree_root); 24060eeff236SLiu Bo return ret; 240764c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2408f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 240900246528SJosef Bacik btrfs_put_root(log_tree_root); 241063443bf5SEric Sandeen return -EIO; 241163443bf5SEric Sandeen } 241263443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 241363443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 241463443bf5SEric Sandeen if (ret) { 24150b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 241663443bf5SEric Sandeen "Failed to recover log tree"); 241700246528SJosef Bacik btrfs_put_root(log_tree_root); 241863443bf5SEric Sandeen return ret; 241963443bf5SEric Sandeen } 242063443bf5SEric Sandeen 2421bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 24226bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 242363443bf5SEric Sandeen if (ret) 242463443bf5SEric Sandeen return ret; 242563443bf5SEric Sandeen } 242663443bf5SEric Sandeen 242763443bf5SEric Sandeen return 0; 242863443bf5SEric Sandeen } 242963443bf5SEric Sandeen 24306bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 24314bbcaa64SEric Sandeen { 24326bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2433a4f3d2c4SDavid Sterba struct btrfs_root *root; 24344bbcaa64SEric Sandeen struct btrfs_key location; 24354bbcaa64SEric Sandeen int ret; 24364bbcaa64SEric Sandeen 24376bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 24386bccf3abSJeff Mahoney 24394bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 24404bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 24414bbcaa64SEric Sandeen location.offset = 0; 24424bbcaa64SEric Sandeen 2443a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2444f50f4353SLiu Bo if (IS_ERR(root)) { 244542437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2446f50f4353SLiu Bo ret = PTR_ERR(root); 2447f50f4353SLiu Bo goto out; 2448f50f4353SLiu Bo } 244942437a63SJosef Bacik } else { 2450a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2451a4f3d2c4SDavid Sterba fs_info->extent_root = root; 245242437a63SJosef Bacik } 24534bbcaa64SEric Sandeen 24544bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2455a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2456f50f4353SLiu Bo if (IS_ERR(root)) { 245742437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2458f50f4353SLiu Bo ret = PTR_ERR(root); 2459f50f4353SLiu Bo goto out; 2460f50f4353SLiu Bo } 246142437a63SJosef Bacik } else { 2462a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2463a4f3d2c4SDavid Sterba fs_info->dev_root = root; 246442437a63SJosef Bacik } 2465820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2466820a49daSJosef Bacik btrfs_init_devices_late(fs_info); 24674bbcaa64SEric Sandeen 2468882dbe0cSJosef Bacik /* If IGNOREDATACSUMS is set don't bother reading the csum root. */ 2469882dbe0cSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 24704bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2471a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2472f50f4353SLiu Bo if (IS_ERR(root)) { 247342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2474f50f4353SLiu Bo ret = PTR_ERR(root); 2475f50f4353SLiu Bo goto out; 2476f50f4353SLiu Bo } 247742437a63SJosef Bacik } else { 2478a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2479a4f3d2c4SDavid Sterba fs_info->csum_root = root; 248042437a63SJosef Bacik } 2481882dbe0cSJosef Bacik } 24824bbcaa64SEric Sandeen 2483aeb935a4SQu Wenruo /* 2484aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2485aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2486aeb935a4SQu Wenruo */ 248756e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 248856e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2489aeb935a4SQu Wenruo if (IS_ERR(root)) { 249042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2491aeb935a4SQu Wenruo ret = PTR_ERR(root); 2492aeb935a4SQu Wenruo goto out; 2493aeb935a4SQu Wenruo } 249442437a63SJosef Bacik } else { 2495aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2496aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 249742437a63SJosef Bacik } 2498aeb935a4SQu Wenruo 24994bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2500a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2501a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2502a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2503afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2504a4f3d2c4SDavid Sterba fs_info->quota_root = root; 25054bbcaa64SEric Sandeen } 25064bbcaa64SEric Sandeen 25074bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2508a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2509a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 251042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2511a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 25124bbcaa64SEric Sandeen if (ret != -ENOENT) 2513f50f4353SLiu Bo goto out; 251442437a63SJosef Bacik } 25154bbcaa64SEric Sandeen } else { 2516a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2517a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 25184bbcaa64SEric Sandeen } 25194bbcaa64SEric Sandeen 252070f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 252170f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 252270f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 2523f50f4353SLiu Bo if (IS_ERR(root)) { 252442437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2525f50f4353SLiu Bo ret = PTR_ERR(root); 2526f50f4353SLiu Bo goto out; 2527f50f4353SLiu Bo } 252842437a63SJosef Bacik } else { 252970f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 253070f6d82eSOmar Sandoval fs_info->free_space_root = root; 253170f6d82eSOmar Sandoval } 253242437a63SJosef Bacik } 253370f6d82eSOmar Sandoval 25344bbcaa64SEric Sandeen return 0; 2535f50f4353SLiu Bo out: 2536f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2537f50f4353SLiu Bo location.objectid, ret); 2538f50f4353SLiu Bo return ret; 25394bbcaa64SEric Sandeen } 25404bbcaa64SEric Sandeen 2541069ec957SQu Wenruo /* 2542069ec957SQu Wenruo * Real super block validation 2543069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2544069ec957SQu Wenruo * 2545069ec957SQu Wenruo * @sb: super block to check 2546069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2547069ec957SQu Wenruo * 0 the primary (1st) sb 2548069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2549069ec957SQu Wenruo * -1 skip bytenr check 2550069ec957SQu Wenruo */ 2551069ec957SQu Wenruo static int validate_super(struct btrfs_fs_info *fs_info, 2552069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 255321a852b0SQu Wenruo { 255421a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 255521a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 255621a852b0SQu Wenruo int ret = 0; 255721a852b0SQu Wenruo 255821a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 255921a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 256021a852b0SQu Wenruo ret = -EINVAL; 256121a852b0SQu Wenruo } 256221a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 256321a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 256421a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 256521a852b0SQu Wenruo ret = -EINVAL; 256621a852b0SQu Wenruo } 256721a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 256821a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 256921a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 257021a852b0SQu Wenruo ret = -EINVAL; 257121a852b0SQu Wenruo } 257221a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 257321a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 257421a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 257521a852b0SQu Wenruo ret = -EINVAL; 257621a852b0SQu Wenruo } 257721a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 257821a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 257921a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 258021a852b0SQu Wenruo ret = -EINVAL; 258121a852b0SQu Wenruo } 258221a852b0SQu Wenruo 258321a852b0SQu Wenruo /* 258421a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 258521a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 258621a852b0SQu Wenruo */ 258721a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 258821a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 258921a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 259021a852b0SQu Wenruo ret = -EINVAL; 259121a852b0SQu Wenruo } 25920bb3eb3eSQu Wenruo 25930bb3eb3eSQu Wenruo /* 25940bb3eb3eSQu Wenruo * For 4K page size, we only support 4K sector size. 259550780d9bSAnand Jain * For 64K page size, we support 64K and 4K sector sizes. 25960bb3eb3eSQu Wenruo */ 25970bb3eb3eSQu Wenruo if ((PAGE_SIZE == SZ_4K && sectorsize != PAGE_SIZE) || 25980bb3eb3eSQu Wenruo (PAGE_SIZE == SZ_64K && (sectorsize != SZ_4K && 25990bb3eb3eSQu Wenruo sectorsize != SZ_64K))) { 260021a852b0SQu Wenruo btrfs_err(fs_info, 26010bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 260221a852b0SQu Wenruo sectorsize, PAGE_SIZE); 260321a852b0SQu Wenruo ret = -EINVAL; 260421a852b0SQu Wenruo } 26050bb3eb3eSQu Wenruo 260621a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 260721a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 260821a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 260921a852b0SQu Wenruo ret = -EINVAL; 261021a852b0SQu Wenruo } 261121a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 261221a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 261321a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 261421a852b0SQu Wenruo ret = -EINVAL; 261521a852b0SQu Wenruo } 261621a852b0SQu Wenruo 261721a852b0SQu Wenruo /* Root alignment check */ 261821a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 261921a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 262021a852b0SQu Wenruo btrfs_super_root(sb)); 262121a852b0SQu Wenruo ret = -EINVAL; 262221a852b0SQu Wenruo } 262321a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 262421a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 262521a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 262621a852b0SQu Wenruo ret = -EINVAL; 262721a852b0SQu Wenruo } 262821a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 262921a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 263021a852b0SQu Wenruo btrfs_super_log_root(sb)); 263121a852b0SQu Wenruo ret = -EINVAL; 263221a852b0SQu Wenruo } 263321a852b0SQu Wenruo 2634aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2635aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2636aefd7f70SNikolay Borisov btrfs_err(fs_info, 2637aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2638aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2639aefd7f70SNikolay Borisov ret = -EINVAL; 2640aefd7f70SNikolay Borisov } 2641aefd7f70SNikolay Borisov 2642aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2643aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2644aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2645aefd7f70SNikolay Borisov btrfs_err(fs_info, 2646aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2647aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2648aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2649aefd7f70SNikolay Borisov ret = -EINVAL; 2650aefd7f70SNikolay Borisov } 2651aefd7f70SNikolay Borisov 2652de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 26537239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 265421a852b0SQu Wenruo btrfs_err(fs_info, 26557239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2656de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 265721a852b0SQu Wenruo ret = -EINVAL; 265821a852b0SQu Wenruo } 265921a852b0SQu Wenruo 266021a852b0SQu Wenruo /* 266121a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 266221a852b0SQu Wenruo * done later 266321a852b0SQu Wenruo */ 266421a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 266521a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 266621a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 266721a852b0SQu Wenruo ret = -EINVAL; 266821a852b0SQu Wenruo } 266921a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 267021a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 267121a852b0SQu Wenruo btrfs_super_stripesize(sb)); 267221a852b0SQu Wenruo ret = -EINVAL; 267321a852b0SQu Wenruo } 267421a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 267521a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 267621a852b0SQu Wenruo btrfs_super_num_devices(sb)); 267721a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 267821a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 267921a852b0SQu Wenruo ret = -EINVAL; 268021a852b0SQu Wenruo } 268121a852b0SQu Wenruo 2682069ec957SQu Wenruo if (mirror_num >= 0 && 2683069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 268421a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 268521a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 268621a852b0SQu Wenruo ret = -EINVAL; 268721a852b0SQu Wenruo } 268821a852b0SQu Wenruo 268921a852b0SQu Wenruo /* 269021a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 269121a852b0SQu Wenruo * and one chunk 269221a852b0SQu Wenruo */ 269321a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 269421a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 269521a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 269621a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 269721a852b0SQu Wenruo ret = -EINVAL; 269821a852b0SQu Wenruo } 269921a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 270021a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 270121a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 270221a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 270321a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 270421a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 270521a852b0SQu Wenruo ret = -EINVAL; 270621a852b0SQu Wenruo } 270721a852b0SQu Wenruo 270821a852b0SQu Wenruo /* 270921a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 271021a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 271121a852b0SQu Wenruo */ 271221a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 271321a852b0SQu Wenruo btrfs_warn(fs_info, 271421a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 271521a852b0SQu Wenruo btrfs_super_generation(sb), 271621a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 271721a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 271821a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 271921a852b0SQu Wenruo btrfs_warn(fs_info, 272021a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 272121a852b0SQu Wenruo btrfs_super_generation(sb), 272221a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 272321a852b0SQu Wenruo 272421a852b0SQu Wenruo return ret; 272521a852b0SQu Wenruo } 272621a852b0SQu Wenruo 2727069ec957SQu Wenruo /* 2728069ec957SQu Wenruo * Validation of super block at mount time. 2729069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2730069ec957SQu Wenruo * flags will be skipped. 2731069ec957SQu Wenruo */ 2732069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2733069ec957SQu Wenruo { 2734069ec957SQu Wenruo return validate_super(fs_info, fs_info->super_copy, 0); 2735069ec957SQu Wenruo } 2736069ec957SQu Wenruo 273775cb857dSQu Wenruo /* 273875cb857dSQu Wenruo * Validation of super block at write time. 273975cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 274075cb857dSQu Wenruo * overwritten soon. 274175cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 274275cb857dSQu Wenruo */ 274375cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 274475cb857dSQu Wenruo struct btrfs_super_block *sb) 274575cb857dSQu Wenruo { 274675cb857dSQu Wenruo int ret; 274775cb857dSQu Wenruo 274875cb857dSQu Wenruo ret = validate_super(fs_info, sb, -1); 274975cb857dSQu Wenruo if (ret < 0) 275075cb857dSQu Wenruo goto out; 2751e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 275275cb857dSQu Wenruo ret = -EUCLEAN; 275375cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 275475cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 275575cb857dSQu Wenruo goto out; 275675cb857dSQu Wenruo } 275775cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 275875cb857dSQu Wenruo ret = -EUCLEAN; 275975cb857dSQu Wenruo btrfs_err(fs_info, 276075cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 276175cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 276275cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 276375cb857dSQu Wenruo goto out; 276475cb857dSQu Wenruo } 276575cb857dSQu Wenruo out: 276675cb857dSQu Wenruo if (ret < 0) 276775cb857dSQu Wenruo btrfs_err(fs_info, 276875cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 276975cb857dSQu Wenruo return ret; 277075cb857dSQu Wenruo } 277175cb857dSQu Wenruo 27726ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2773b8522a1eSNikolay Borisov { 27746ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2775b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2776b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2777b8522a1eSNikolay Borisov bool handle_error = false; 2778b8522a1eSNikolay Borisov int ret = 0; 2779b8522a1eSNikolay Borisov int i; 2780b8522a1eSNikolay Borisov 2781b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2782b8522a1eSNikolay Borisov u64 generation; 2783b8522a1eSNikolay Borisov int level; 2784b8522a1eSNikolay Borisov 2785b8522a1eSNikolay Borisov if (handle_error) { 2786b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2787b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2788b8522a1eSNikolay Borisov tree_root->node = NULL; 2789b8522a1eSNikolay Borisov 2790b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2791b8522a1eSNikolay Borisov break; 2792b8522a1eSNikolay Borisov 2793b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2794b8522a1eSNikolay Borisov 2795b8522a1eSNikolay Borisov /* 2796b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2797b8522a1eSNikolay Borisov * valid 2798b8522a1eSNikolay Borisov */ 2799b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2800b8522a1eSNikolay Borisov 2801b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2802b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2803b8522a1eSNikolay Borisov 2804b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 28056ef108ddSNikolay Borisov backup_index = ret; 2806b8522a1eSNikolay Borisov if (ret < 0) 2807b8522a1eSNikolay Borisov return ret; 2808b8522a1eSNikolay Borisov } 2809b8522a1eSNikolay Borisov generation = btrfs_super_generation(sb); 2810b8522a1eSNikolay Borisov level = btrfs_super_root_level(sb); 2811b8522a1eSNikolay Borisov tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), 28121b7ec85eSJosef Bacik BTRFS_ROOT_TREE_OBJECTID, 2813b8522a1eSNikolay Borisov generation, level, NULL); 28140465337cSJosef Bacik if (IS_ERR(tree_root->node)) { 2815217f5004SNikolay Borisov handle_error = true; 2816b8522a1eSNikolay Borisov ret = PTR_ERR(tree_root->node); 28170465337cSJosef Bacik tree_root->node = NULL; 2818217f5004SNikolay Borisov btrfs_warn(fs_info, "couldn't read tree root"); 2819217f5004SNikolay Borisov continue; 2820b8522a1eSNikolay Borisov 2821217f5004SNikolay Borisov } else if (!extent_buffer_uptodate(tree_root->node)) { 2822217f5004SNikolay Borisov handle_error = true; 2823217f5004SNikolay Borisov ret = -EIO; 2824217f5004SNikolay Borisov btrfs_warn(fs_info, "error while reading tree root"); 2825b8522a1eSNikolay Borisov continue; 2826b8522a1eSNikolay Borisov } 2827b8522a1eSNikolay Borisov 2828b8522a1eSNikolay Borisov btrfs_set_root_node(&tree_root->root_item, tree_root->node); 2829b8522a1eSNikolay Borisov tree_root->commit_root = btrfs_root_node(tree_root); 2830b8522a1eSNikolay Borisov btrfs_set_root_refs(&tree_root->root_item, 1); 2831b8522a1eSNikolay Borisov 2832336a0d8dSNikolay Borisov /* 2833336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2834336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2835336a0d8dSNikolay Borisov */ 2836453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2837b8522a1eSNikolay Borisov if (ret < 0) { 2838b8522a1eSNikolay Borisov handle_error = true; 2839b8522a1eSNikolay Borisov continue; 2840b8522a1eSNikolay Borisov } 2841b8522a1eSNikolay Borisov 28426b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2843b8522a1eSNikolay Borisov 2844b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2845b8522a1eSNikolay Borisov if (ret < 0) { 2846b8522a1eSNikolay Borisov handle_error = true; 2847b8522a1eSNikolay Borisov continue; 2848b8522a1eSNikolay Borisov } 2849b8522a1eSNikolay Borisov 2850b8522a1eSNikolay Borisov /* All successful */ 2851b8522a1eSNikolay Borisov fs_info->generation = generation; 2852b8522a1eSNikolay Borisov fs_info->last_trans_committed = generation; 28536ef108ddSNikolay Borisov 28546ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 28556ef108ddSNikolay Borisov if (backup_index < 0) { 28566ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 28576ef108ddSNikolay Borisov } else { 28586ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 28596ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 28606ef108ddSNikolay Borisov } 2861b8522a1eSNikolay Borisov break; 2862b8522a1eSNikolay Borisov } 2863b8522a1eSNikolay Borisov 2864b8522a1eSNikolay Borisov return ret; 2865b8522a1eSNikolay Borisov } 2866b8522a1eSNikolay Borisov 28678260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2868eb60ceacSChris Mason { 286976dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2870f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 28718fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2872facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 287324bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2874eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 287511833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2876eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2877a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 287876dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 287924bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 28804cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2881ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2882f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 288347ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 288440ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2885afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2886c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2887f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2888d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2889f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 28907585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2891573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 28920bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 2893de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 289419c00ddcSChris Mason 28950b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 28966324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2897f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 289847ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 289918bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2900afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2901bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2902bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 29033fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 29043fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2905bd647ce3SJosef Bacik #endif 2906c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 290766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 290866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 290966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 291066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 291166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 291266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 291366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2914ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2915ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2916ba2c4d4eSJosef Bacik 2917771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 29184cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 29192fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2920034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2921fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 292295ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 29239ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 29244cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2925a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2926f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 29278b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2928f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 292990519d66SArne Jansen /* readahead state */ 2930d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 293190519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2932fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2933c8b97818SChris Mason 2934b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2935b34b086cSChris Mason num_online_cpus() + 2, 8); 29360afbaf8cSChris Mason 2937199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2938199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 293969fe2d75SJosef Bacik 2940638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 294121adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 294221adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 294321adbd5cSStefan Behrens #endif 2944779a65a4SEric Sandeen btrfs_init_balance(fs_info); 294557056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2946a2de733cSArne Jansen 29470f9dd46cSJosef Bacik spin_lock_init(&fs_info->block_group_cache_lock); 29486bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2949a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 29500f9dd46cSJosef Bacik 2951fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 2952fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 2953afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 295439279cc3SChris Mason 29555a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2956e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2957925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2958a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2959a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 29601bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 29619e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2962c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 296376dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2964803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2965fa9c0d79SChris Mason 2966ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2967f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2968b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2969416ac51dSArne Jansen 2970fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2971fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2972fa9c0d79SChris Mason 2973e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2974f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2975bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 29764854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2977034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 29783768f368SChris Mason 2979da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2980da17066cSJeff Mahoney fs_info->nodesize = 4096; 2981da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2982ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2983da17066cSJeff Mahoney fs_info->stripesize = 4096; 2984da17066cSJeff Mahoney 2985eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2986eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2987eede2bf3SOmar Sandoval 29881cea5cf0SFilipe Manana spin_lock_init(&fs_info->send_reloc_lock); 29899e967495SFilipe Manana fs_info->send_in_progress = 0; 299018bb8bbfSJohannes Thumshirn 299118bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 299218bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 29938260edbaSJosef Bacik } 29948260edbaSJosef Bacik 29958260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 29968260edbaSJosef Bacik { 29978260edbaSJosef Bacik int ret; 29988260edbaSJosef Bacik 29998260edbaSJosef Bacik fs_info->sb = sb; 30008260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 30018260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 30029e967495SFilipe Manana 30035deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 3004ae18c37aSJosef Bacik if (ret) 3005ae18c37aSJosef Bacik return ret; 3006ae18c37aSJosef Bacik 3007ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 3008ae18c37aSJosef Bacik if (ret) 3009c75e8394SJosef Bacik return ret; 3010ae18c37aSJosef Bacik 3011ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 3012ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 3013ae18c37aSJosef Bacik 3014ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 3015ae18c37aSJosef Bacik if (ret) 3016c75e8394SJosef Bacik return ret; 3017ae18c37aSJosef Bacik 3018ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 3019ae18c37aSJosef Bacik GFP_KERNEL); 3020ae18c37aSJosef Bacik if (ret) 3021c75e8394SJosef Bacik return ret; 3022ae18c37aSJosef Bacik 3023ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 3024ae18c37aSJosef Bacik GFP_KERNEL); 3025c75e8394SJosef Bacik if (!fs_info->delayed_root) 3026c75e8394SJosef Bacik return -ENOMEM; 3027ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 3028ae18c37aSJosef Bacik 3029a0a1db70SFilipe Manana if (sb_rdonly(sb)) 3030a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 3031a0a1db70SFilipe Manana 3032c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 3033ae18c37aSJosef Bacik } 3034ae18c37aSJosef Bacik 303597f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 303697f4dd09SNikolay Borisov { 303797f4dd09SNikolay Borisov struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; 303897f4dd09SNikolay Borisov int ret; 303997f4dd09SNikolay Borisov 304097f4dd09SNikolay Borisov /* 304197f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 304297f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 304397f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 304497f4dd09SNikolay Borisov */ 304597f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 304697f4dd09SNikolay Borisov if (ret < 0) { 3047c94bec2cSJosef Bacik if (ret != -EINTR) 3048c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 3049c94bec2cSJosef Bacik ret); 305097f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 305197f4dd09SNikolay Borisov return ret; 305297f4dd09SNikolay Borisov } 305397f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 305497f4dd09SNikolay Borisov } 305597f4dd09SNikolay Borisov 305697f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 305797f4dd09SNikolay Borisov { 305897f4dd09SNikolay Borisov struct task_struct *task; 305997f4dd09SNikolay Borisov 306097f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 306197f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 306297f4dd09SNikolay Borisov if (IS_ERR(task)) { 306397f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 306497f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 306597f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 306697f4dd09SNikolay Borisov return PTR_ERR(task); 306797f4dd09SNikolay Borisov } 306897f4dd09SNikolay Borisov 306997f4dd09SNikolay Borisov return 0; 307097f4dd09SNikolay Borisov } 307197f4dd09SNikolay Borisov 307244c0ca21SBoris Burkov /* 30738cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 30748cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 30758cd29088SBoris Burkov * code paths. 30768cd29088SBoris Burkov */ 30778cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 30788cd29088SBoris Burkov { 30798cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 30808b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 30818cd29088SBoris Burkov } 30828cd29088SBoris Burkov 30838cd29088SBoris Burkov /* 308444c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 308544c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 308644c0ca21SBoris Burkov */ 308744c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 308844c0ca21SBoris Burkov { 308944c0ca21SBoris Burkov int ret; 309094846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 30918b228324SBoris Burkov bool clear_free_space_tree = false; 30928b228324SBoris Burkov 30938b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 30948b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 30958b228324SBoris Burkov clear_free_space_tree = true; 30968b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 30978b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 30988b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 30998b228324SBoris Burkov clear_free_space_tree = true; 31008b228324SBoris Burkov } 31018b228324SBoris Burkov 31028b228324SBoris Burkov if (clear_free_space_tree) { 31038b228324SBoris Burkov btrfs_info(fs_info, "clearing free space tree"); 31048b228324SBoris Burkov ret = btrfs_clear_free_space_tree(fs_info); 31058b228324SBoris Burkov if (ret) { 31068b228324SBoris Burkov btrfs_warn(fs_info, 31078b228324SBoris Burkov "failed to clear free space tree: %d", ret); 31088b228324SBoris Burkov goto out; 31098b228324SBoris Burkov } 31108b228324SBoris Burkov } 311144c0ca21SBoris Burkov 31128d488a8cSFilipe Manana /* 31138d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 31148d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 31158d488a8cSFilipe Manana * them into the fs_info->fs_roots_radix tree. This must be done before 31168d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 31178d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 31188d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 31198d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 31208d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 31218d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 31228d488a8cSFilipe Manana */ 31238d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 31248d488a8cSFilipe Manana if (ret) 31258d488a8cSFilipe Manana goto out; 31268d488a8cSFilipe Manana 312744c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 312844c0ca21SBoris Burkov if (ret) 312944c0ca21SBoris Burkov goto out; 313044c0ca21SBoris Burkov 31318f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 31328f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 31338f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 31348f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 31358f1c21d7SBoris Burkov goto out; 31368f1c21d7SBoris Burkov } 31378f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 31388f1c21d7SBoris Burkov 313944c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 314044c0ca21SBoris Burkov ret = btrfs_recover_relocation(fs_info->tree_root); 314144c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 314244c0ca21SBoris Burkov if (ret < 0) { 314344c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 314444c0ca21SBoris Burkov goto out; 314544c0ca21SBoris Burkov } 314644c0ca21SBoris Burkov 31475011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 31485011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 31495011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 31505011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 31515011139aSBoris Burkov if (ret) { 31525011139aSBoris Burkov btrfs_warn(fs_info, 31535011139aSBoris Burkov "failed to create free space tree: %d", ret); 31545011139aSBoris Burkov goto out; 31555011139aSBoris Burkov } 31565011139aSBoris Burkov } 31575011139aSBoris Burkov 315894846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 315994846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 316094846229SBoris Burkov if (ret) 316194846229SBoris Burkov goto out; 316294846229SBoris Burkov } 316394846229SBoris Burkov 316444c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 316544c0ca21SBoris Burkov if (ret) 316644c0ca21SBoris Burkov goto out; 316744c0ca21SBoris Burkov 316844c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 316944c0ca21SBoris Burkov if (ret) { 317044c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 317144c0ca21SBoris Burkov goto out; 317244c0ca21SBoris Burkov } 317344c0ca21SBoris Burkov 317444c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 317544c0ca21SBoris Burkov 317644c0ca21SBoris Burkov if (!fs_info->uuid_root) { 317744c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 317844c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 317944c0ca21SBoris Burkov if (ret) { 318044c0ca21SBoris Burkov btrfs_warn(fs_info, 318144c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 318244c0ca21SBoris Burkov goto out; 318344c0ca21SBoris Burkov } 318444c0ca21SBoris Burkov } 318544c0ca21SBoris Burkov 318644c0ca21SBoris Burkov out: 318744c0ca21SBoris Burkov return ret; 318844c0ca21SBoris Burkov } 318944c0ca21SBoris Burkov 3190ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3191ae18c37aSJosef Bacik char *options) 3192ae18c37aSJosef Bacik { 3193ae18c37aSJosef Bacik u32 sectorsize; 3194ae18c37aSJosef Bacik u32 nodesize; 3195ae18c37aSJosef Bacik u32 stripesize; 3196ae18c37aSJosef Bacik u64 generation; 3197ae18c37aSJosef Bacik u64 features; 3198ae18c37aSJosef Bacik u16 csum_type; 3199ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3200ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3201ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3202ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3203ae18c37aSJosef Bacik int ret; 3204ae18c37aSJosef Bacik int err = -EINVAL; 3205ae18c37aSJosef Bacik int level; 3206ae18c37aSJosef Bacik 32078260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 320853b381b3SDavid Woodhouse if (ret) { 320983c8266aSDavid Sterba err = ret; 3210ae18c37aSJosef Bacik goto fail; 321153b381b3SDavid Woodhouse } 321253b381b3SDavid Woodhouse 3213ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3214ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3215ae18c37aSJosef Bacik GFP_KERNEL); 3216ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3217ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3218ae18c37aSJosef Bacik GFP_KERNEL); 3219ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3220ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 3221ae18c37aSJosef Bacik err = -ENOMEM; 3222c75e8394SJosef Bacik goto fail; 3223ae18c37aSJosef Bacik } 3224ae18c37aSJosef Bacik 3225ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 3226ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 3227ae18c37aSJosef Bacik err = -ENOMEM; 3228c75e8394SJosef Bacik goto fail; 3229ae18c37aSJosef Bacik } 3230ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 3231ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 3232ae18c37aSJosef Bacik 3233d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 32341104a885SDavid Sterba 32351104a885SDavid Sterba /* 32361104a885SDavid Sterba * Read super block and check the signature bytes only 32371104a885SDavid Sterba */ 3238d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 32398f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 32408f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 324116cdcec7SMiao Xie goto fail_alloc; 324220b45077SDave Young } 324339279cc3SChris Mason 32441104a885SDavid Sterba /* 3245260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 32468dc3f22cSJohannes Thumshirn * corrupted, we'll find out 32478dc3f22cSJohannes Thumshirn */ 32488f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 324951bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 32508dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 325151bce6c9SJohannes Thumshirn csum_type); 32528dc3f22cSJohannes Thumshirn err = -EINVAL; 32538f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32548dc3f22cSJohannes Thumshirn goto fail_alloc; 32558dc3f22cSJohannes Thumshirn } 32568dc3f22cSJohannes Thumshirn 325783c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 325883c68bbcSSu Yue 32596d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 32606d97c6e3SJohannes Thumshirn if (ret) { 32616d97c6e3SJohannes Thumshirn err = ret; 32628f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32636d97c6e3SJohannes Thumshirn goto fail_alloc; 32646d97c6e3SJohannes Thumshirn } 32656d97c6e3SJohannes Thumshirn 32668dc3f22cSJohannes Thumshirn /* 32671104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 32681104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 32691104a885SDavid Sterba */ 32708f32380dSJohannes Thumshirn if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { 327105135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 32721104a885SDavid Sterba err = -EINVAL; 32738f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3274141386e1SJosef Bacik goto fail_alloc; 32751104a885SDavid Sterba } 32761104a885SDavid Sterba 32771104a885SDavid Sterba /* 32781104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 32791104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 32801104a885SDavid Sterba * the whole block of INFO_SIZE 32811104a885SDavid Sterba */ 32828f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 32838f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32845f39d397SChris Mason 3285fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3286fbc6feaeSNikolay Borisov 32870b86a832SChris Mason 3288fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3289fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3290fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3291fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3292fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3293fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3294fbc6feaeSNikolay Borisov } 3295fbc6feaeSNikolay Borisov 3296fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3297fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3298de37aa51SNikolay Borisov 3299069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 33001104a885SDavid Sterba if (ret) { 330105135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 33021104a885SDavid Sterba err = -EINVAL; 3303141386e1SJosef Bacik goto fail_alloc; 33041104a885SDavid Sterba } 33051104a885SDavid Sterba 33060f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3307141386e1SJosef Bacik goto fail_alloc; 33080f7d52f4SChris Mason 3309acce952bSliubo /* check FS state, whether FS is broken. */ 331087533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 331187533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3312acce952bSliubo 331375e7cb7fSLiu Bo /* 331475e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 331575e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 331675e7cb7fSLiu Bo */ 331775e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 331875e7cb7fSLiu Bo 33196f93e834SAnand Jain /* 33206f93e834SAnand Jain * Flag our filesystem as having big metadata blocks if they are bigger 33216f93e834SAnand Jain * than the page size. 33226f93e834SAnand Jain */ 33236f93e834SAnand Jain if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 33246f93e834SAnand Jain if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 33256f93e834SAnand Jain btrfs_info(fs_info, 33266f93e834SAnand Jain "flagging fs with big metadata feature"); 33276f93e834SAnand Jain features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 33286f93e834SAnand Jain } 33296f93e834SAnand Jain 33306f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 33316f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 33326f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 33336f93e834SAnand Jain stripesize = sectorsize; 33346f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 33356f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 33366f93e834SAnand Jain 33376f93e834SAnand Jain fs_info->nodesize = nodesize; 33386f93e834SAnand Jain fs_info->sectorsize = sectorsize; 33396f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 33406f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 33416f93e834SAnand Jain fs_info->stripesize = stripesize; 33426f93e834SAnand Jain 33432ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 33442b82032cSYan Zheng if (ret) { 33452b82032cSYan Zheng err = ret; 3346141386e1SJosef Bacik goto fail_alloc; 33472b82032cSYan Zheng } 3348dfe25020SChris Mason 3349f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 3350f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 3351f2b636e8SJosef Bacik if (features) { 335205135f59SDavid Sterba btrfs_err(fs_info, 335305135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 3354c1c9ff7cSGeert Uytterhoeven features); 3355f2b636e8SJosef Bacik err = -EINVAL; 3356141386e1SJosef Bacik goto fail_alloc; 3357f2b636e8SJosef Bacik } 3358f2b636e8SJosef Bacik 33595d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 33605d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 33610b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3362a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 33635c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 33645c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3365727011e0SChris Mason 33663173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 336705135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 33683173a18fSJosef Bacik 3369727011e0SChris Mason /* 3370bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 3371bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 3372bc3f116fSChris Mason */ 3373bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3374707e8a07SDavid Sterba (sectorsize != nodesize)) { 337505135f59SDavid Sterba btrfs_err(fs_info, 337605135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 337705135f59SDavid Sterba nodesize, sectorsize); 3378141386e1SJosef Bacik goto fail_alloc; 3379bc3f116fSChris Mason } 3380bc3f116fSChris Mason 3381ceda0864SMiao Xie /* 3382ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 3383ceda0864SMiao Xie * update the flag. 3384ceda0864SMiao Xie */ 33855d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 33865d4f98a2SYan Zheng 3387f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 3388f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 3389bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 339005135f59SDavid Sterba btrfs_err(fs_info, 339105135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 3392c1c9ff7cSGeert Uytterhoeven features); 3393f2b636e8SJosef Bacik err = -EINVAL; 3394141386e1SJosef Bacik goto fail_alloc; 3395f2b636e8SJosef Bacik } 339661d92c32SChris Mason 33978481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 33988481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 33998481dd80SQu Wenruo 340095ea0486SQu Wenruo btrfs_warn(fs_info, 340195ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 34020bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 3403c8050b3bSQu Wenruo if (btrfs_super_incompat_flags(fs_info->super_copy) & 3404c8050b3bSQu Wenruo BTRFS_FEATURE_INCOMPAT_RAID56) { 3405c8050b3bSQu Wenruo btrfs_err(fs_info, 3406c8050b3bSQu Wenruo "RAID56 is not yet supported for sector size %u with page size %lu", 3407c8050b3bSQu Wenruo sectorsize, PAGE_SIZE); 3408c8050b3bSQu Wenruo err = -EINVAL; 3409c8050b3bSQu Wenruo goto fail_alloc; 3410c8050b3bSQu Wenruo } 34118481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 34128481dd80SQu Wenruo if (!subpage_info) 34138481dd80SQu Wenruo goto fail_alloc; 34148481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 34158481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3416c8050b3bSQu Wenruo } 34170bb3eb3eSQu Wenruo 34182a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 34192a458198SEric Sandeen if (ret) { 34202a458198SEric Sandeen err = ret; 34210dc3b84aSJosef Bacik goto fail_sb_buffer; 34220dc3b84aSJosef Bacik } 34234543df7eSChris Mason 34249e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 34259e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 34264575c9ccSChris Mason 3427a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3428a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3429de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3430db94535dSChris Mason 3431925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 34326bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3433925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 343484eed90fSChris Mason if (ret) { 343505135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 34365d4f98a2SYan Zheng goto fail_sb_buffer; 343784eed90fSChris Mason } 34380b86a832SChris Mason 343984234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3440581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 34410b86a832SChris Mason 34422ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 34430b86a832SChris Mason btrfs_super_chunk_root(disk_super), 34441b7ec85eSJosef Bacik BTRFS_CHUNK_TREE_OBJECTID, 3445581c1760SQu Wenruo generation, level, NULL); 344664c043deSLiu Bo if (IS_ERR(chunk_root->node) || 344764c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 344805135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3449e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 3450e5fffbacSchandan free_extent_buffer(chunk_root->node); 345195ab1f64SZhao Lei chunk_root->node = NULL; 3452af31f5e5SChris Mason goto fail_tree_roots; 345383121942SDavid Woodhouse } 34545d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 34555d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 34560b86a832SChris Mason 3457e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3458c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3459c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3460e17cade2SChris Mason 34615b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 34622b82032cSYan Zheng if (ret) { 346305135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3464af31f5e5SChris Mason goto fail_tree_roots; 34652b82032cSYan Zheng } 34660b86a832SChris Mason 34678dabb742SStefan Behrens /* 3468bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3469bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3470bacce86aSAnand Jain * target is required. So free devices that are not part of this 34711a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3472bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 34738dabb742SStefan Behrens */ 3474bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3475d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 347605135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3477a6b0d5c8SChris Mason goto fail_tree_roots; 3478a6b0d5c8SChris Mason } 3479a6b0d5c8SChris Mason 3480b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 34814bbcaa64SEric Sandeen if (ret) 3482b8522a1eSNikolay Borisov goto fail_tree_roots; 34838929ecfaSYan, Zheng 348475ec1db8SJosef Bacik /* 348573651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 348673651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 348773651042SNaohiro Aota * zoned incompat feature flag set. 348873651042SNaohiro Aota */ 348973651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 349073651042SNaohiro Aota if (ret) { 349173651042SNaohiro Aota btrfs_err(fs_info, 349273651042SNaohiro Aota "zoned: failed to read device zone info: %d", 349373651042SNaohiro Aota ret); 349473651042SNaohiro Aota goto fail_block_groups; 349573651042SNaohiro Aota } 349673651042SNaohiro Aota 349773651042SNaohiro Aota /* 349875ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 349975ec1db8SJosef Bacik * check the generation here so we can set the 350075ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 350175ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 350275ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 350375ec1db8SJosef Bacik * even though it was perfectly fine. 350475ec1db8SJosef Bacik */ 350575ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 350675ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 350775ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 350875ec1db8SJosef Bacik 3509cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3510cf90d884SQu Wenruo if (ret) { 3511cf90d884SQu Wenruo btrfs_err(fs_info, 3512cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3513cf90d884SQu Wenruo ret); 3514cf90d884SQu Wenruo goto fail_block_groups; 3515cf90d884SQu Wenruo } 351668310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 351768310a5eSIlya Dryomov if (ret) { 351805135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 351968310a5eSIlya Dryomov goto fail_block_groups; 352068310a5eSIlya Dryomov } 352168310a5eSIlya Dryomov 3522733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3523733f4fbbSStefan Behrens if (ret) { 352405135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3525733f4fbbSStefan Behrens goto fail_block_groups; 3526733f4fbbSStefan Behrens } 3527733f4fbbSStefan Behrens 35288dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 35298dabb742SStefan Behrens if (ret) { 353005135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 35318dabb742SStefan Behrens goto fail_block_groups; 35328dabb742SStefan Behrens } 35338dabb742SStefan Behrens 3534b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3535b70f5097SNaohiro Aota if (ret) { 3536b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3537b70f5097SNaohiro Aota ret); 3538b70f5097SNaohiro Aota goto fail_block_groups; 3539b70f5097SNaohiro Aota } 3540b70f5097SNaohiro Aota 3541c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3542b7c35e81SAnand Jain if (ret) { 354305135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 354405135f59SDavid Sterba ret); 3545b7c35e81SAnand Jain goto fail_block_groups; 3546b7c35e81SAnand Jain } 3547b7c35e81SAnand Jain 354896f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 35495ac1d209SJeff Mahoney if (ret) { 355005135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3551b7c35e81SAnand Jain goto fail_fsdev_sysfs; 35525ac1d209SJeff Mahoney } 35535ac1d209SJeff Mahoney 3554c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3555c59021f8Sliubo if (ret) { 355605135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 35572365dd3cSAnand Jain goto fail_sysfs; 3558c59021f8Sliubo } 3559c59021f8Sliubo 35605b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 35611b1d1f66SJosef Bacik if (ret) { 356205135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 35632365dd3cSAnand Jain goto fail_sysfs; 35641b1d1f66SJosef Bacik } 35654330e183SQu Wenruo 35665c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 35675c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 356805135f59SDavid Sterba btrfs_warn(fs_info, 356952042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 35702365dd3cSAnand Jain goto fail_sysfs; 3571292fd7fcSStefan Behrens } 35729078a3e1SChris Mason 3573a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 3574a74a4b97SChris Mason "btrfs-cleaner"); 357557506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 35762365dd3cSAnand Jain goto fail_sysfs; 3577a74a4b97SChris Mason 3578a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3579a74a4b97SChris Mason tree_root, 3580a74a4b97SChris Mason "btrfs-transaction"); 358157506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 35823f157a2fSChris Mason goto fail_cleaner; 3583a74a4b97SChris Mason 3584583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3585c289811cSChris Mason !fs_info->fs_devices->rotating) { 3586583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3587c289811cSChris Mason } 3588c289811cSChris Mason 3589572d9ab7SDavid Sterba /* 359001327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 3591572d9ab7SDavid Sterba * not have to wait for transaction commit 3592572d9ab7SDavid Sterba */ 3593572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 35943818aea2SQu Wenruo 359521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 35960b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 35972ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 35980b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3599cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 360021adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 360121adbd5cSStefan Behrens if (ret) 360205135f59SDavid Sterba btrfs_warn(fs_info, 360305135f59SDavid Sterba "failed to initialize integrity check module: %d", 360405135f59SDavid Sterba ret); 360521adbd5cSStefan Behrens } 360621adbd5cSStefan Behrens #endif 3607bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3608bcef60f2SArne Jansen if (ret) 3609bcef60f2SArne Jansen goto fail_trans_kthread; 361021adbd5cSStefan Behrens 3611fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3612fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3613fd708b81SJosef Bacik 361496da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 361596da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 36160b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3617e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 361863443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 361979787eaaSJeff Mahoney if (ret) { 362063443bf5SEric Sandeen err = ret; 362128c16cbbSWang Shilong goto fail_qgroup; 3622e556ce2cSYan Zheng } 3623e02119d5SChris Mason } 36241a40e23bSZheng Yan 362556e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 36263140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 36273140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3628f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3629315bf8efSJosef Bacik fs_info->fs_root = NULL; 3630bcef60f2SArne Jansen goto fail_qgroup; 36313140c9a3SDan Carpenter } 3632c289811cSChris Mason 3633bc98a42cSDavid Howells if (sb_rdonly(sb)) 36348cd29088SBoris Burkov goto clear_oneshot; 36352b6ba629SIlya Dryomov 363644c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 36372b6ba629SIlya Dryomov if (ret) { 36386bccf3abSJeff Mahoney close_ctree(fs_info); 36392b6ba629SIlya Dryomov return ret; 3640e3acc2a6SJosef Bacik } 3641b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3642b382a324SJan Schmidt 364344c0ca21SBoris Burkov if (fs_info->uuid_root && 364444c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 364544c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 364605135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 364770f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 364870f80175SStefan Behrens if (ret) { 364905135f59SDavid Sterba btrfs_warn(fs_info, 365005135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 36516bccf3abSJeff Mahoney close_ctree(fs_info); 365270f80175SStefan Behrens return ret; 365370f80175SStefan Behrens } 3654f7a81ea4SStefan Behrens } 365594846229SBoris Burkov 3656afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 365747ab2a6cSJosef Bacik 36588cd29088SBoris Burkov clear_oneshot: 36598cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3660ad2b2c80SAl Viro return 0; 366139279cc3SChris Mason 3662bcef60f2SArne Jansen fail_qgroup: 3663bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 36647c2ca468SChris Mason fail_trans_kthread: 36657c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 36662ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3667faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 36683f157a2fSChris Mason fail_cleaner: 3669a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 36707c2ca468SChris Mason 36717c2ca468SChris Mason /* 36727c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 36737c2ca468SChris Mason * kthreads 36747c2ca468SChris Mason */ 36757c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 36767c2ca468SChris Mason 36772365dd3cSAnand Jain fail_sysfs: 36786618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 36792365dd3cSAnand Jain 3680b7c35e81SAnand Jain fail_fsdev_sysfs: 3681b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3682b7c35e81SAnand Jain 36831b1d1f66SJosef Bacik fail_block_groups: 368454067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3685af31f5e5SChris Mason 3686af31f5e5SChris Mason fail_tree_roots: 36879e3aa805SJosef Bacik if (fs_info->data_reloc_root) 36889e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 36894273eaffSAnand Jain free_root_pointers(fs_info, true); 36902b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3691af31f5e5SChris Mason 369239279cc3SChris Mason fail_sb_buffer: 36937abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 36945cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 369516cdcec7SMiao Xie fail_alloc: 3696586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3697586e46e2SIlya Dryomov 36984543df7eSChris Mason iput(fs_info->btree_inode); 36997e662854SQinghuang Feng fail: 3700586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3701ad2b2c80SAl Viro return err; 3702eb60ceacSChris Mason } 3703663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3704eb60ceacSChris Mason 3705314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3706f2984462SChris Mason { 3707314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3708314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3709314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3710314b6dd0SJohannes Thumshirn struct page *page; 3711442a4f63SStefan Behrens 3712314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3713314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3714314b6dd0SJohannes Thumshirn 3715314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3716fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3717314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3718314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3719314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3720314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3721314b6dd0SJohannes Thumshirn SetPageError(page); 3722314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3723314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3724314b6dd0SJohannes Thumshirn } else { 3725314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3726f2984462SChris Mason } 3727314b6dd0SJohannes Thumshirn 3728314b6dd0SJohannes Thumshirn put_page(page); 3729314b6dd0SJohannes Thumshirn unlock_page(page); 3730314b6dd0SJohannes Thumshirn } 3731314b6dd0SJohannes Thumshirn 3732314b6dd0SJohannes Thumshirn bio_put(bio); 3733f2984462SChris Mason } 3734f2984462SChris Mason 37358f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 37368f32380dSJohannes Thumshirn int copy_num) 373729c36d72SAnand Jain { 373829c36d72SAnand Jain struct btrfs_super_block *super; 37398f32380dSJohannes Thumshirn struct page *page; 374012659251SNaohiro Aota u64 bytenr, bytenr_orig; 37418f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 374212659251SNaohiro Aota int ret; 374329c36d72SAnand Jain 374412659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 374512659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 374612659251SNaohiro Aota if (ret == -ENOENT) 374712659251SNaohiro Aota return ERR_PTR(-EINVAL); 374812659251SNaohiro Aota else if (ret) 374912659251SNaohiro Aota return ERR_PTR(ret); 375012659251SNaohiro Aota 375129c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 37528f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 375329c36d72SAnand Jain 37548f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 37558f32380dSJohannes Thumshirn if (IS_ERR(page)) 37568f32380dSJohannes Thumshirn return ERR_CAST(page); 375729c36d72SAnand Jain 37588f32380dSJohannes Thumshirn super = page_address(page); 375996c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 376096c2e067SAnand Jain btrfs_release_disk_super(super); 376196c2e067SAnand Jain return ERR_PTR(-ENODATA); 376296c2e067SAnand Jain } 376396c2e067SAnand Jain 376412659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 37658f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 37668f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 376729c36d72SAnand Jain } 376829c36d72SAnand Jain 37698f32380dSJohannes Thumshirn return super; 377029c36d72SAnand Jain } 377129c36d72SAnand Jain 377229c36d72SAnand Jain 37738f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3774a512bbf8SYan Zheng { 37758f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3776a512bbf8SYan Zheng int i; 3777a512bbf8SYan Zheng u64 transid = 0; 3778a512bbf8SYan Zheng 3779a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3780a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3781a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3782a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3783a512bbf8SYan Zheng */ 3784a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 37858f32380dSJohannes Thumshirn super = btrfs_read_dev_one_super(bdev, i); 37868f32380dSJohannes Thumshirn if (IS_ERR(super)) 3787a512bbf8SYan Zheng continue; 3788a512bbf8SYan Zheng 3789a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 37908f32380dSJohannes Thumshirn if (latest) 37918f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 37928f32380dSJohannes Thumshirn 37938f32380dSJohannes Thumshirn latest = super; 3794a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3795a512bbf8SYan Zheng } 3796a512bbf8SYan Zheng } 379792fc03fbSAnand Jain 37988f32380dSJohannes Thumshirn return super; 3799a512bbf8SYan Zheng } 3800a512bbf8SYan Zheng 38014eedeb75SHisashi Hifumi /* 3802abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3803314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 38044eedeb75SHisashi Hifumi * 3805abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3806abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3807abbb3b8eSDavid Sterba * same for write and wait phases. 38084eedeb75SHisashi Hifumi * 3809314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 38104eedeb75SHisashi Hifumi */ 3811a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3812abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3813a512bbf8SYan Zheng { 3814d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3815314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3816d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3817a512bbf8SYan Zheng int i; 3818a512bbf8SYan Zheng int errors = 0; 381912659251SNaohiro Aota int ret; 382012659251SNaohiro Aota u64 bytenr, bytenr_orig; 3821a512bbf8SYan Zheng 3822a512bbf8SYan Zheng if (max_mirrors == 0) 3823a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3824a512bbf8SYan Zheng 3825d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3826d5178578SJohannes Thumshirn 3827a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3828314b6dd0SJohannes Thumshirn struct page *page; 3829314b6dd0SJohannes Thumshirn struct bio *bio; 3830314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3831314b6dd0SJohannes Thumshirn 383212659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 383312659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 383412659251SNaohiro Aota if (ret == -ENOENT) { 383512659251SNaohiro Aota continue; 383612659251SNaohiro Aota } else if (ret < 0) { 383712659251SNaohiro Aota btrfs_err(device->fs_info, 383812659251SNaohiro Aota "couldn't get super block location for mirror %d", 383912659251SNaohiro Aota i); 384012659251SNaohiro Aota errors++; 384112659251SNaohiro Aota continue; 384212659251SNaohiro Aota } 3843935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3844935e5cc9SMiao Xie device->commit_total_bytes) 3845a512bbf8SYan Zheng break; 3846a512bbf8SYan Zheng 384712659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3848a512bbf8SYan Zheng 3849fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3850fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3851fd08001fSEric Biggers sb->csum); 3852a512bbf8SYan Zheng 3853314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3854314b6dd0SJohannes Thumshirn GFP_NOFS); 3855314b6dd0SJohannes Thumshirn if (!page) { 3856fb456252SJeff Mahoney btrfs_err(device->fs_info, 3857314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3858f14d104dSDavid Sterba bytenr); 3859634554dcSJosef Bacik errors++; 3860634554dcSJosef Bacik continue; 3861634554dcSJosef Bacik } 3862634554dcSJosef Bacik 3863314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3864314b6dd0SJohannes Thumshirn get_page(page); 3865a512bbf8SYan Zheng 3866314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3867314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3868a512bbf8SYan Zheng 3869387125fcSChris Mason /* 3870314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3871314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3872314b6dd0SJohannes Thumshirn * checking. 3873387125fcSChris Mason */ 3874314b6dd0SJohannes Thumshirn bio = bio_alloc(GFP_NOFS, 1); 3875314b6dd0SJohannes Thumshirn bio_set_dev(bio, device->bdev); 3876314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3877314b6dd0SJohannes Thumshirn bio->bi_private = device; 3878314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3879314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3880314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3881314b6dd0SJohannes Thumshirn 3882314b6dd0SJohannes Thumshirn /* 3883314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3884314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3885314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3886314b6dd0SJohannes Thumshirn */ 3887314b6dd0SJohannes Thumshirn bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; 38881b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3889314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3890314b6dd0SJohannes Thumshirn 3891314b6dd0SJohannes Thumshirn btrfsic_submit_bio(bio); 38928376d9e1SNaohiro Aota 38938376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 38948376d9e1SNaohiro Aota errors++; 3895a512bbf8SYan Zheng } 3896a512bbf8SYan Zheng return errors < i ? 0 : -1; 3897a512bbf8SYan Zheng } 3898a512bbf8SYan Zheng 3899387125fcSChris Mason /* 3900abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3901abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3902abbb3b8eSDavid Sterba * 3903314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3904abbb3b8eSDavid Sterba * date. 3905abbb3b8eSDavid Sterba */ 3906abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3907abbb3b8eSDavid Sterba { 3908abbb3b8eSDavid Sterba int i; 3909abbb3b8eSDavid Sterba int errors = 0; 3910b6a535faSHoward McLauchlan bool primary_failed = false; 391112659251SNaohiro Aota int ret; 3912abbb3b8eSDavid Sterba u64 bytenr; 3913abbb3b8eSDavid Sterba 3914abbb3b8eSDavid Sterba if (max_mirrors == 0) 3915abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3916abbb3b8eSDavid Sterba 3917abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3918314b6dd0SJohannes Thumshirn struct page *page; 3919314b6dd0SJohannes Thumshirn 392012659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 392112659251SNaohiro Aota if (ret == -ENOENT) { 392212659251SNaohiro Aota break; 392312659251SNaohiro Aota } else if (ret < 0) { 392412659251SNaohiro Aota errors++; 392512659251SNaohiro Aota if (i == 0) 392612659251SNaohiro Aota primary_failed = true; 392712659251SNaohiro Aota continue; 392812659251SNaohiro Aota } 3929abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3930abbb3b8eSDavid Sterba device->commit_total_bytes) 3931abbb3b8eSDavid Sterba break; 3932abbb3b8eSDavid Sterba 3933314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3934314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3935314b6dd0SJohannes Thumshirn if (!page) { 3936abbb3b8eSDavid Sterba errors++; 3937b6a535faSHoward McLauchlan if (i == 0) 3938b6a535faSHoward McLauchlan primary_failed = true; 3939abbb3b8eSDavid Sterba continue; 3940abbb3b8eSDavid Sterba } 3941314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3942314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3943314b6dd0SJohannes Thumshirn if (PageError(page)) { 3944abbb3b8eSDavid Sterba errors++; 3945b6a535faSHoward McLauchlan if (i == 0) 3946b6a535faSHoward McLauchlan primary_failed = true; 3947b6a535faSHoward McLauchlan } 3948abbb3b8eSDavid Sterba 3949314b6dd0SJohannes Thumshirn /* Drop our reference */ 3950314b6dd0SJohannes Thumshirn put_page(page); 3951abbb3b8eSDavid Sterba 3952314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3953314b6dd0SJohannes Thumshirn put_page(page); 3954abbb3b8eSDavid Sterba } 3955abbb3b8eSDavid Sterba 3956b6a535faSHoward McLauchlan /* log error, force error return */ 3957b6a535faSHoward McLauchlan if (primary_failed) { 3958b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3959b6a535faSHoward McLauchlan device->devid); 3960b6a535faSHoward McLauchlan return -1; 3961b6a535faSHoward McLauchlan } 3962b6a535faSHoward McLauchlan 3963abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3964abbb3b8eSDavid Sterba } 3965abbb3b8eSDavid Sterba 3966abbb3b8eSDavid Sterba /* 3967387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3968387125fcSChris Mason * for the barrier when it is done 3969387125fcSChris Mason */ 39704246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3971387125fcSChris Mason { 3972387125fcSChris Mason complete(bio->bi_private); 3973387125fcSChris Mason } 3974387125fcSChris Mason 3975387125fcSChris Mason /* 39764fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 39774fc6441aSAnand Jain * done in the waiting counterpart. 3978387125fcSChris Mason */ 39794fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3980387125fcSChris Mason { 3981e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3982387125fcSChris Mason 3983*a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3984*a91cf0ffSWang Yugui /* 3985*a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 3986*a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 3987*a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 3988*a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 3989*a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 3990*a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 3991*a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 3992*a91cf0ffSWang Yugui * non-debug builds. 3993*a91cf0ffSWang Yugui */ 3994*a91cf0ffSWang Yugui struct request_queue *q = bdev_get_queue(device->bdev); 3995c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 39964fc6441aSAnand Jain return; 3997*a91cf0ffSWang Yugui #endif 3998387125fcSChris Mason 3999e0ae9994SDavid Sterba bio_reset(bio); 4000387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 400174d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 40028d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 4003387125fcSChris Mason init_completion(&device->flush_wait); 4004387125fcSChris Mason bio->bi_private = &device->flush_wait; 4005387125fcSChris Mason 400643a01111SLu Fengqi btrfsic_submit_bio(bio); 40071c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 40084fc6441aSAnand Jain } 4009387125fcSChris Mason 40104fc6441aSAnand Jain /* 40114fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 40124fc6441aSAnand Jain */ 40138c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 40144fc6441aSAnand Jain { 40154fc6441aSAnand Jain struct bio *bio = device->flush_bio; 40164fc6441aSAnand Jain 40171c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 401858efbc9fSOmar Sandoval return BLK_STS_OK; 40194fc6441aSAnand Jain 40201c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 40212980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 40224fc6441aSAnand Jain 40238c27cb35SLinus Torvalds return bio->bi_status; 4024387125fcSChris Mason } 4025387125fcSChris Mason 4026d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 4027401b41e5SAnand Jain { 40286528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 4029401b41e5SAnand Jain return -EIO; 4030387125fcSChris Mason return 0; 4031387125fcSChris Mason } 4032387125fcSChris Mason 4033387125fcSChris Mason /* 4034387125fcSChris Mason * send an empty flush down to each device in parallel, 4035387125fcSChris Mason * then wait for them 4036387125fcSChris Mason */ 4037387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 4038387125fcSChris Mason { 4039387125fcSChris Mason struct list_head *head; 4040387125fcSChris Mason struct btrfs_device *dev; 40415af3e8ccSStefan Behrens int errors_wait = 0; 40424e4cbee9SChristoph Hellwig blk_status_t ret; 4043387125fcSChris Mason 40441538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 4045387125fcSChris Mason /* send down all the barriers */ 4046387125fcSChris Mason head = &info->fs_devices->devices; 40471538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4048e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4049f88ba6a2SHidetoshi Seto continue; 4050cea7c8bfSAnand Jain if (!dev->bdev) 4051387125fcSChris Mason continue; 4052e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4053ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4054387125fcSChris Mason continue; 4055387125fcSChris Mason 40564fc6441aSAnand Jain write_dev_flush(dev); 405758efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 4058387125fcSChris Mason } 4059387125fcSChris Mason 4060387125fcSChris Mason /* wait for all the barriers */ 40611538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4062e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4063f88ba6a2SHidetoshi Seto continue; 4064387125fcSChris Mason if (!dev->bdev) { 40655af3e8ccSStefan Behrens errors_wait++; 4066387125fcSChris Mason continue; 4067387125fcSChris Mason } 4068e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4069ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4070387125fcSChris Mason continue; 4071387125fcSChris Mason 40724fc6441aSAnand Jain ret = wait_dev_flush(dev); 4073401b41e5SAnand Jain if (ret) { 4074401b41e5SAnand Jain dev->last_flush_error = ret; 407566b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 407666b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 40775af3e8ccSStefan Behrens errors_wait++; 4078387125fcSChris Mason } 4079401b41e5SAnand Jain } 4080401b41e5SAnand Jain 4081cea7c8bfSAnand Jain if (errors_wait) { 4082401b41e5SAnand Jain /* 4083401b41e5SAnand Jain * At some point we need the status of all disks 4084401b41e5SAnand Jain * to arrive at the volume status. So error checking 4085401b41e5SAnand Jain * is being pushed to a separate loop. 4086401b41e5SAnand Jain */ 4087d10b82feSQu Wenruo return check_barrier_error(info); 4088401b41e5SAnand Jain } 4089387125fcSChris Mason return 0; 4090387125fcSChris Mason } 4091387125fcSChris Mason 4092943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4093943c6e99SZhao Lei { 40948789f4feSZhao Lei int raid_type; 40958789f4feSZhao Lei int min_tolerated = INT_MAX; 4096943c6e99SZhao Lei 40978789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 40988789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 40998c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 41008789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 41018789f4feSZhao Lei tolerated_failures); 4102943c6e99SZhao Lei 41038789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 41048789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 41058789f4feSZhao Lei continue; 410641a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 41078789f4feSZhao Lei continue; 41088c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 41098789f4feSZhao Lei btrfs_raid_array[raid_type]. 41108789f4feSZhao Lei tolerated_failures); 41118789f4feSZhao Lei } 4112943c6e99SZhao Lei 41138789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4114ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 41158789f4feSZhao Lei min_tolerated = 0; 41168789f4feSZhao Lei } 41178789f4feSZhao Lei 41188789f4feSZhao Lei return min_tolerated; 4119943c6e99SZhao Lei } 4120943c6e99SZhao Lei 4121eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4122f2984462SChris Mason { 4123e5e9a520SChris Mason struct list_head *head; 4124f2984462SChris Mason struct btrfs_device *dev; 4125a061fc8dSChris Mason struct btrfs_super_block *sb; 4126f2984462SChris Mason struct btrfs_dev_item *dev_item; 4127f2984462SChris Mason int ret; 4128f2984462SChris Mason int do_barriers; 4129a236aed1SChris Mason int max_errors; 4130a236aed1SChris Mason int total_errors = 0; 4131a061fc8dSChris Mason u64 flags; 4132f2984462SChris Mason 41330b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4134fed3b381SLiu Bo 4135fed3b381SLiu Bo /* 4136fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4137fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4138fed3b381SLiu Bo * been consistent on disk. 4139fed3b381SLiu Bo */ 4140fed3b381SLiu Bo if (max_mirrors == 0) 41410b246afaSJeff Mahoney backup_super_roots(fs_info); 4142f2984462SChris Mason 41430b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4144a061fc8dSChris Mason dev_item = &sb->dev_item; 4145e5e9a520SChris Mason 41460b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 41470b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 41480b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4149387125fcSChris Mason 41505af3e8ccSStefan Behrens if (do_barriers) { 41510b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 41525af3e8ccSStefan Behrens if (ret) { 41535af3e8ccSStefan Behrens mutex_unlock( 41540b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 41550b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 41565af3e8ccSStefan Behrens "errors while submitting device barriers."); 41575af3e8ccSStefan Behrens return ret; 41585af3e8ccSStefan Behrens } 41595af3e8ccSStefan Behrens } 4160387125fcSChris Mason 41611538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4162dfe25020SChris Mason if (!dev->bdev) { 4163dfe25020SChris Mason total_errors++; 4164dfe25020SChris Mason continue; 4165dfe25020SChris Mason } 4166e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4167ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4168dfe25020SChris Mason continue; 4169dfe25020SChris Mason 41702b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4171a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4172a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 41737df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4174935e5cc9SMiao Xie dev->commit_total_bytes); 4175ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4176ce7213c7SMiao Xie dev->commit_bytes_used); 4177a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4178a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4179a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4180a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 41817239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 41827239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4183a512bbf8SYan Zheng 4184a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4185a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4186f2984462SChris Mason 418775cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 418875cb857dSQu Wenruo if (ret < 0) { 418975cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 419075cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 419175cb857dSQu Wenruo "unexpected superblock corruption detected"); 419275cb857dSQu Wenruo return -EUCLEAN; 419375cb857dSQu Wenruo } 419475cb857dSQu Wenruo 4195abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4196a236aed1SChris Mason if (ret) 4197a236aed1SChris Mason total_errors++; 4198f2984462SChris Mason } 4199a236aed1SChris Mason if (total_errors > max_errors) { 42000b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4201d397712bSChris Mason total_errors); 42020b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 420379787eaaSJeff Mahoney 42049d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 42050b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 42060b246afaSJeff Mahoney "%d errors while writing supers", 42070b246afaSJeff Mahoney total_errors); 42089d565ba4SStefan Behrens return -EIO; 4209a236aed1SChris Mason } 4210f2984462SChris Mason 4211a512bbf8SYan Zheng total_errors = 0; 42121538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4213dfe25020SChris Mason if (!dev->bdev) 4214dfe25020SChris Mason continue; 4215e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4216ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4217dfe25020SChris Mason continue; 4218dfe25020SChris Mason 4219abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4220a512bbf8SYan Zheng if (ret) 42211259ab75SChris Mason total_errors++; 4222f2984462SChris Mason } 42230b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4224a236aed1SChris Mason if (total_errors > max_errors) { 42250b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 42260b246afaSJeff Mahoney "%d errors while writing supers", 42270b246afaSJeff Mahoney total_errors); 422879787eaaSJeff Mahoney return -EIO; 4229a236aed1SChris Mason } 4230f2984462SChris Mason return 0; 4231f2984462SChris Mason } 4232f2984462SChris Mason 4233cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4234cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4235cb517eabSMiao Xie struct btrfs_root *root) 42362619ba1fSChris Mason { 42374785e24fSJosef Bacik bool drop_ref = false; 42384785e24fSJosef Bacik 42394df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 42402619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 42412619ba1fSChris Mason (unsigned long)root->root_key.objectid); 4242af01d2e5SJosef Bacik if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 42434785e24fSJosef Bacik drop_ref = true; 42444df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 424576dda93cSYan, Zheng 424684961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4247ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 42481c1ea4f7SLiu Bo if (root->reloc_root) { 424900246528SJosef Bacik btrfs_put_root(root->reloc_root); 42501c1ea4f7SLiu Bo root->reloc_root = NULL; 42511c1ea4f7SLiu Bo } 42521c1ea4f7SLiu Bo } 42533321719eSLiu Bo 42544785e24fSJosef Bacik if (drop_ref) 425500246528SJosef Bacik btrfs_put_root(root); 42562619ba1fSChris Mason } 42572619ba1fSChris Mason 4258c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4259c146afadSYan Zheng { 4260c146afadSYan Zheng u64 root_objectid = 0; 4261c146afadSYan Zheng struct btrfs_root *gang[8]; 426265d33fd7SQu Wenruo int i = 0; 426365d33fd7SQu Wenruo int err = 0; 426465d33fd7SQu Wenruo unsigned int ret = 0; 4265c146afadSYan Zheng 4266c146afadSYan Zheng while (1) { 4267efc34534SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4268c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4269c146afadSYan Zheng (void **)gang, root_objectid, 4270c146afadSYan Zheng ARRAY_SIZE(gang)); 427165d33fd7SQu Wenruo if (!ret) { 4272efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4273c146afadSYan Zheng break; 427465d33fd7SQu Wenruo } 42755d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 427666b4ffd1SJosef Bacik 427765d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 427865d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 427965d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 428065d33fd7SQu Wenruo gang[i] = NULL; 428165d33fd7SQu Wenruo continue; 428265d33fd7SQu Wenruo } 428365d33fd7SQu Wenruo /* grab all the search result for later use */ 428400246528SJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 428565d33fd7SQu Wenruo } 4286efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 428765d33fd7SQu Wenruo 428865d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 428965d33fd7SQu Wenruo if (!gang[i]) 429065d33fd7SQu Wenruo continue; 4291c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 429266b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 429366b4ffd1SJosef Bacik if (err) 429465d33fd7SQu Wenruo break; 429500246528SJosef Bacik btrfs_put_root(gang[i]); 4296c146afadSYan Zheng } 4297c146afadSYan Zheng root_objectid++; 4298c146afadSYan Zheng } 429965d33fd7SQu Wenruo 430065d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 430165d33fd7SQu Wenruo for (; i < ret; i++) { 430265d33fd7SQu Wenruo if (gang[i]) 430300246528SJosef Bacik btrfs_put_root(gang[i]); 430465d33fd7SQu Wenruo } 430565d33fd7SQu Wenruo return err; 4306c146afadSYan Zheng } 4307c146afadSYan Zheng 43086bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4309c146afadSYan Zheng { 43106bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4311c146afadSYan Zheng struct btrfs_trans_handle *trans; 4312c146afadSYan Zheng 43130b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 43142ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 43150b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 43160b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4317c71bf099SYan, Zheng 4318c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 43190b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 43200b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4321c71bf099SYan, Zheng 43227a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 43233612b495STsutomu Itoh if (IS_ERR(trans)) 43243612b495STsutomu Itoh return PTR_ERR(trans); 43253a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4326c146afadSYan Zheng } 4327c146afadSYan Zheng 4328b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4329eb60ceacSChris Mason { 4330c146afadSYan Zheng int ret; 4331e089f05cSChris Mason 4332afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 4333d6fd0ae2SOmar Sandoval /* 4334d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4335d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4336d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4337d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4338d6fd0ae2SOmar Sandoval */ 4339d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4340a2135011SChris Mason 43417343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4342d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 43437343dd61SJustin Maggard 4344803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4345803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4346803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4347803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4348803b2f54SStefan Behrens 4349837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4350aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4351837d5b6eSIlya Dryomov 43528dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 43538dabb742SStefan Behrens 4354aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 43554cb5300bSChris Mason 43564cb5300bSChris Mason /* wait for any defraggers to finish */ 43574cb5300bSChris Mason wait_event(fs_info->transaction_wait, 43584cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 43594cb5300bSChris Mason 43604cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 436126176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 43624cb5300bSChris Mason 436321c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 436457056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4365576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 436621c7e756SMiao Xie 436718bb8bbfSJohannes Thumshirn cancel_work_sync(&fs_info->reclaim_bgs_work); 436818bb8bbfSJohannes Thumshirn 4369b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4370b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4371b0643e59SDennis Zhou 4372bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4373e44163e1SJeff Mahoney /* 4374d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4375d6fd0ae2SOmar Sandoval * unused block groups. 4376e44163e1SJeff Mahoney */ 43770b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4378e44163e1SJeff Mahoney 4379f0cc2cd7SFilipe Manana /* 4380f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4381f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4382f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4383f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4384f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4385f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4386f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4387f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4388f0cc2cd7SFilipe Manana * This is a very rare case. 4389f0cc2cd7SFilipe Manana */ 4390f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4391f0cc2cd7SFilipe Manana 43926bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4393d397712bSChris Mason if (ret) 439404892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4395c146afadSYan Zheng } 4396ed2ff2cbSChris Mason 439784961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 43982ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4399acce952bSliubo 4400e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4401e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 44028929ecfaSYan, Zheng 4403e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4404afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4405f25784b3SYan Zheng 44065958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 44075958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 44085958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 44095958253cSQu Wenruo } 44105958253cSQu Wenruo 441104892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4412fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4413bcef60f2SArne Jansen 4414963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 441504892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4416963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4417b0c68f8bSChris Mason } 441831153d81SYan Zheng 44195deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 44204297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 44215deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 44224297ff84SJosef Bacik 44236618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4424b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 44255ac1d209SJeff Mahoney 44261a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 44271a4319ccSLiu Bo 4428de348ee0SWang Shilong /* 4429de348ee0SWang Shilong * we must make sure there is not any read request to 4430de348ee0SWang Shilong * submit after we stopping all workers. 4431de348ee0SWang Shilong */ 4432de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 443396192499SJosef Bacik btrfs_stop_all_workers(fs_info); 443496192499SJosef Bacik 44350a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 44360a31daa4SFilipe Manana ASSERT(list_empty(&fs_info->trans_list)); 44370a31daa4SFilipe Manana 4438afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 44394273eaffSAnand Jain free_root_pointers(fs_info, true); 44408c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 44419ad6b7bcSChris Mason 44424e19443dSJosef Bacik /* 44434e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 44444e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 44454e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 44464e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 44474e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 44484e19443dSJosef Bacik */ 44494e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 44504e19443dSJosef Bacik 445113e6c37bSJosef Bacik iput(fs_info->btree_inode); 4452d6bfde87SChris Mason 445321adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 44540b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 44552ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 445621adbd5cSStefan Behrens #endif 445721adbd5cSStefan Behrens 44580b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 445968c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4460eb60ceacSChris Mason } 4461eb60ceacSChris Mason 4462b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4463b9fab919SChris Mason int atomic) 4464ccd467d6SChris Mason { 44651259ab75SChris Mason int ret; 4466727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 44671259ab75SChris Mason 44680b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 44691259ab75SChris Mason if (!ret) 44701259ab75SChris Mason return ret; 44711259ab75SChris Mason 44721259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4473b9fab919SChris Mason parent_transid, atomic); 4474b9fab919SChris Mason if (ret == -EAGAIN) 4475b9fab919SChris Mason return ret; 44761259ab75SChris Mason return !ret; 44775f39d397SChris Mason } 44786702ed49SChris Mason 44795f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 44805f39d397SChris Mason { 44812f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 44825f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4483b9473439SChris Mason int was_dirty; 4484b4ce94deSChris Mason 448506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 448606ea65a3SJosef Bacik /* 448706ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 448852042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 448906ea65a3SJosef Bacik * outside of the sanity tests. 449006ea65a3SJosef Bacik */ 4491b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 449206ea65a3SJosef Bacik return; 449306ea65a3SJosef Bacik #endif 449449d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 44950b246afaSJeff Mahoney if (transid != fs_info->generation) 44965d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 44970b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 44980b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4499e2d84521SMiao Xie if (!was_dirty) 4500104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4501e2d84521SMiao Xie buf->len, 45020b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 45031f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 450469fc6cbbSQu Wenruo /* 450569fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 450669fc6cbbSQu Wenruo * but item data not updated. 450769fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 450869fc6cbbSQu Wenruo */ 450969fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4510cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4511a4f78750SDavid Sterba btrfs_print_leaf(buf); 45121f21ef0aSFilipe Manana ASSERT(0); 45131f21ef0aSFilipe Manana } 45141f21ef0aSFilipe Manana #endif 4515eb60ceacSChris Mason } 4516eb60ceacSChris Mason 45172ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4518b53d3f5dSLiu Bo int flush_delayed) 451935b7e476SChris Mason { 4520188de649SChris Mason /* 4521188de649SChris Mason * looks as though older kernels can get into trouble with 4522188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4523188de649SChris Mason */ 4524e2d84521SMiao Xie int ret; 4525d6bfde87SChris Mason 45266933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4527d6bfde87SChris Mason return; 4528d6bfde87SChris Mason 4529b53d3f5dSLiu Bo if (flush_delayed) 45302ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 453116cdcec7SMiao Xie 4532d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4533d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4534d814a491SEthan Lien fs_info->dirty_metadata_batch); 4535e2d84521SMiao Xie if (ret > 0) { 45360b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 453716cdcec7SMiao Xie } 453816cdcec7SMiao Xie } 453916cdcec7SMiao Xie 45402ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 454116cdcec7SMiao Xie { 45422ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 454335b7e476SChris Mason } 4544b53d3f5dSLiu Bo 45452ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4546b53d3f5dSLiu Bo { 45472ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4548d6bfde87SChris Mason } 45496b80053dSChris Mason 4550581c1760SQu Wenruo int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, 4551581c1760SQu Wenruo struct btrfs_key *first_key) 45526b80053dSChris Mason { 45535ab12d1fSDavid Sterba return btree_read_extent_buffer_pages(buf, parent_transid, 4554581c1760SQu Wenruo level, first_key); 45556b80053dSChris Mason } 45560da5468fSChris Mason 45572ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4558acce952bSliubo { 4559fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4560fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4561fe816d0fSNikolay Borisov 45620b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45632ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45640b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4565acce952bSliubo 45660b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45670b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4568acce952bSliubo } 4569acce952bSliubo 4570ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4571ef67963dSJosef Bacik { 4572ef67963dSJosef Bacik struct btrfs_root *gang[8]; 4573ef67963dSJosef Bacik u64 root_objectid = 0; 4574ef67963dSJosef Bacik int ret; 4575ef67963dSJosef Bacik 4576ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4577ef67963dSJosef Bacik while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4578ef67963dSJosef Bacik (void **)gang, root_objectid, 4579ef67963dSJosef Bacik ARRAY_SIZE(gang))) != 0) { 4580ef67963dSJosef Bacik int i; 4581ef67963dSJosef Bacik 4582ef67963dSJosef Bacik for (i = 0; i < ret; i++) 4583ef67963dSJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 4584ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4585ef67963dSJosef Bacik 4586ef67963dSJosef Bacik for (i = 0; i < ret; i++) { 4587ef67963dSJosef Bacik if (!gang[i]) 4588ef67963dSJosef Bacik continue; 4589ef67963dSJosef Bacik root_objectid = gang[i]->root_key.objectid; 4590ef67963dSJosef Bacik btrfs_free_log(NULL, gang[i]); 4591ef67963dSJosef Bacik btrfs_put_root(gang[i]); 4592ef67963dSJosef Bacik } 4593ef67963dSJosef Bacik root_objectid++; 4594ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4595ef67963dSJosef Bacik } 4596ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4597ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4598ef67963dSJosef Bacik } 4599ef67963dSJosef Bacik 4600143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4601acce952bSliubo { 4602acce952bSliubo struct btrfs_ordered_extent *ordered; 4603acce952bSliubo 4604199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4605779880efSJosef Bacik /* 4606779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4607779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4608779880efSJosef Bacik */ 4609199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4610779880efSJosef Bacik root_extent_list) 4611779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4612199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4613199c2a9cSMiao Xie } 4614199c2a9cSMiao Xie 4615199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4616199c2a9cSMiao Xie { 4617199c2a9cSMiao Xie struct btrfs_root *root; 4618199c2a9cSMiao Xie struct list_head splice; 4619199c2a9cSMiao Xie 4620199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4621199c2a9cSMiao Xie 4622199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4623199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4624199c2a9cSMiao Xie while (!list_empty(&splice)) { 4625199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4626199c2a9cSMiao Xie ordered_root); 46271de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 46281de2cfdeSJosef Bacik &fs_info->ordered_roots); 4629199c2a9cSMiao Xie 46302a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4631199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4632199c2a9cSMiao Xie 46332a85d9caSLiu Bo cond_resched(); 46342a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4635199c2a9cSMiao Xie } 4636199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 463774d5d229SJosef Bacik 463874d5d229SJosef Bacik /* 463974d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 464074d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 464174d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 464274d5d229SJosef Bacik * actually get run and error out properly. 464374d5d229SJosef Bacik */ 464474d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4645acce952bSliubo } 4646acce952bSliubo 464735a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 46482ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4649acce952bSliubo { 4650acce952bSliubo struct rb_node *node; 4651acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4652acce952bSliubo struct btrfs_delayed_ref_node *ref; 4653acce952bSliubo int ret = 0; 4654acce952bSliubo 4655acce952bSliubo delayed_refs = &trans->delayed_refs; 4656acce952bSliubo 4657acce952bSliubo spin_lock(&delayed_refs->lock); 4658d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4659cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4660b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4661acce952bSliubo return ret; 4662acce952bSliubo } 4663acce952bSliubo 46645c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4665d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 46660e0adbcfSJosef Bacik struct rb_node *n; 4667e78417d1SJosef Bacik bool pin_bytes = false; 4668acce952bSliubo 4669d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4670d7df2c79SJosef Bacik href_node); 46713069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4672b939d1abSJosef Bacik continue; 46733069bd26SJosef Bacik 4674d7df2c79SJosef Bacik spin_lock(&head->lock); 4675e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 46760e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 46770e0adbcfSJosef Bacik ref_node); 4678d7df2c79SJosef Bacik ref->in_tree = 0; 4679e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 46800e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 46811d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 46821d57ee94SWang Xiaoguang list_del(&ref->add_list); 4683d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4684d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4685d7df2c79SJosef Bacik } 468654067ae9SJosef Bacik if (head->must_insert_reserved) 4687e78417d1SJosef Bacik pin_bytes = true; 468878a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4689fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4690d7df2c79SJosef Bacik spin_unlock(&head->lock); 4691acce952bSliubo spin_unlock(&delayed_refs->lock); 4692e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4693acce952bSliubo 4694f603bb94SNikolay Borisov if (pin_bytes) { 4695f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4696f603bb94SNikolay Borisov 4697f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4698f603bb94SNikolay Borisov BUG_ON(!cache); 4699f603bb94SNikolay Borisov 4700f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4701f603bb94SNikolay Borisov spin_lock(&cache->lock); 4702f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4703f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4704f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4705f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4706f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4707f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4708f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4709f603bb94SNikolay Borisov 4710f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4711f603bb94SNikolay Borisov 4712f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4713f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4714f603bb94SNikolay Borisov } 471531890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4716d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4717acce952bSliubo cond_resched(); 4718acce952bSliubo spin_lock(&delayed_refs->lock); 4719acce952bSliubo } 472081f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4721acce952bSliubo 4722acce952bSliubo spin_unlock(&delayed_refs->lock); 4723acce952bSliubo 4724acce952bSliubo return ret; 4725acce952bSliubo } 4726acce952bSliubo 4727143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4728acce952bSliubo { 4729acce952bSliubo struct btrfs_inode *btrfs_inode; 4730acce952bSliubo struct list_head splice; 4731acce952bSliubo 4732acce952bSliubo INIT_LIST_HEAD(&splice); 4733acce952bSliubo 4734eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4735eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4736acce952bSliubo 4737acce952bSliubo while (!list_empty(&splice)) { 4738fe816d0fSNikolay Borisov struct inode *inode = NULL; 4739eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4740acce952bSliubo delalloc_inodes); 4741fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4742eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4743acce952bSliubo 4744fe816d0fSNikolay Borisov /* 4745fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4746fe816d0fSNikolay Borisov * meanwhile. 4747fe816d0fSNikolay Borisov */ 4748fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4749fe816d0fSNikolay Borisov if (inode) { 4750fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4751fe816d0fSNikolay Borisov iput(inode); 4752fe816d0fSNikolay Borisov } 4753eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4754acce952bSliubo } 4755eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4756eb73c1b7SMiao Xie } 4757eb73c1b7SMiao Xie 4758eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4759eb73c1b7SMiao Xie { 4760eb73c1b7SMiao Xie struct btrfs_root *root; 4761eb73c1b7SMiao Xie struct list_head splice; 4762eb73c1b7SMiao Xie 4763eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4764eb73c1b7SMiao Xie 4765eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4766eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4767eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4768eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4769eb73c1b7SMiao Xie delalloc_root); 477000246528SJosef Bacik root = btrfs_grab_root(root); 4771eb73c1b7SMiao Xie BUG_ON(!root); 4772eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4773eb73c1b7SMiao Xie 4774eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 477500246528SJosef Bacik btrfs_put_root(root); 4776eb73c1b7SMiao Xie 4777eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4778eb73c1b7SMiao Xie } 4779eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4780acce952bSliubo } 4781acce952bSliubo 47822ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4783acce952bSliubo struct extent_io_tree *dirty_pages, 4784acce952bSliubo int mark) 4785acce952bSliubo { 4786acce952bSliubo int ret; 4787acce952bSliubo struct extent_buffer *eb; 4788acce952bSliubo u64 start = 0; 4789acce952bSliubo u64 end; 4790acce952bSliubo 4791acce952bSliubo while (1) { 4792acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4793e6138876SJosef Bacik mark, NULL); 4794acce952bSliubo if (ret) 4795acce952bSliubo break; 4796acce952bSliubo 479791166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4798acce952bSliubo while (start <= end) { 47990b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 48000b246afaSJeff Mahoney start += fs_info->nodesize; 4801fd8b2b61SJosef Bacik if (!eb) 4802acce952bSliubo continue; 4803fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4804acce952bSliubo 4805fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4806fd8b2b61SJosef Bacik &eb->bflags)) 4807fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4808fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4809acce952bSliubo } 4810acce952bSliubo } 4811acce952bSliubo 4812acce952bSliubo return ret; 4813acce952bSliubo } 4814acce952bSliubo 48152ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4816fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4817acce952bSliubo { 4818acce952bSliubo u64 start; 4819acce952bSliubo u64 end; 4820acce952bSliubo int ret; 4821acce952bSliubo 4822acce952bSliubo while (1) { 48230e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 48240e6ec385SFilipe Manana 4825fcd5e742SLu Fengqi /* 4826fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4827fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4828fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4829fcd5e742SLu Fengqi * the same extent range. 4830fcd5e742SLu Fengqi */ 4831fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4832acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 48330e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4834fcd5e742SLu Fengqi if (ret) { 4835fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4836acce952bSliubo break; 4837fcd5e742SLu Fengqi } 4838acce952bSliubo 48390e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 48400e6ec385SFilipe Manana free_extent_state(cached_state); 48412ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4842fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4843acce952bSliubo cond_resched(); 4844acce952bSliubo } 4845acce952bSliubo 4846acce952bSliubo return 0; 4847acce952bSliubo } 4848acce952bSliubo 484932da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4850c79a1751SLiu Bo { 4851c79a1751SLiu Bo struct inode *inode; 4852c79a1751SLiu Bo 4853c79a1751SLiu Bo inode = cache->io_ctl.inode; 4854c79a1751SLiu Bo if (inode) { 4855c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4856c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4857c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4858c79a1751SLiu Bo iput(inode); 4859c79a1751SLiu Bo } 4860bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4861c79a1751SLiu Bo btrfs_put_block_group(cache); 4862c79a1751SLiu Bo } 4863c79a1751SLiu Bo 4864c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 48652ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4866c79a1751SLiu Bo { 486732da5386SDavid Sterba struct btrfs_block_group *cache; 4868c79a1751SLiu Bo 4869c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4870c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4871c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 487232da5386SDavid Sterba struct btrfs_block_group, 4873c79a1751SLiu Bo dirty_list); 4874c79a1751SLiu Bo 4875c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4876c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4877c79a1751SLiu Bo list_del_init(&cache->io_list); 4878c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4879c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4880c79a1751SLiu Bo } 4881c79a1751SLiu Bo 4882c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4883c79a1751SLiu Bo spin_lock(&cache->lock); 4884c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4885c79a1751SLiu Bo spin_unlock(&cache->lock); 4886c79a1751SLiu Bo 4887c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4888c79a1751SLiu Bo btrfs_put_block_group(cache); 4889ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4890c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4891c79a1751SLiu Bo } 4892c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4893c79a1751SLiu Bo 489445ae2c18SNikolay Borisov /* 489545ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 489645ae2c18SNikolay Borisov * to use it without any locking 489745ae2c18SNikolay Borisov */ 4898c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4899c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 490032da5386SDavid Sterba struct btrfs_block_group, 4901c79a1751SLiu Bo io_list); 4902c79a1751SLiu Bo 4903c79a1751SLiu Bo list_del_init(&cache->io_list); 4904c79a1751SLiu Bo spin_lock(&cache->lock); 4905c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4906c79a1751SLiu Bo spin_unlock(&cache->lock); 4907c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4908c79a1751SLiu Bo } 4909c79a1751SLiu Bo } 4910c79a1751SLiu Bo 491149b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 49122ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 491349b25e05SJeff Mahoney { 4914bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4915bbbf7243SNikolay Borisov 49162ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4917c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4918c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4919c79a1751SLiu Bo 4920bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4921bbbf7243SNikolay Borisov post_commit_list) { 4922bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4923bbbf7243SNikolay Borisov } 4924bbbf7243SNikolay Borisov 49252ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 492649b25e05SJeff Mahoney 49274a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 49280b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 492949b25e05SJeff Mahoney 49304a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 49310b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 493249b25e05SJeff Mahoney 4933ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 493467cde344SMiao Xie 49352ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 493649b25e05SJeff Mahoney EXTENT_DIRTY); 4937fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 493849b25e05SJeff Mahoney 4939d3575156SNaohiro Aota btrfs_free_redirty_list(cur_trans); 4940d3575156SNaohiro Aota 49414a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 49424a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 494349b25e05SJeff Mahoney } 494449b25e05SJeff Mahoney 49452ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4946acce952bSliubo { 4947acce952bSliubo struct btrfs_transaction *t; 4948acce952bSliubo 49490b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4950acce952bSliubo 49510b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49520b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 49530b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4954724e2315SJosef Bacik struct btrfs_transaction, list); 4955724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 49569b64f57dSElena Reshetova refcount_inc(&t->use_count); 49570b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49582ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4959724e2315SJosef Bacik btrfs_put_transaction(t); 49600b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4961724e2315SJosef Bacik continue; 4962724e2315SJosef Bacik } 49630b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4964724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 49650b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4966724e2315SJosef Bacik /* 4967724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4968724e2315SJosef Bacik * handle open currently for this transaction. 4969724e2315SJosef Bacik */ 4970724e2315SJosef Bacik wait_event(t->writer_wait, 4971724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4972724e2315SJosef Bacik } else { 49730b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4974724e2315SJosef Bacik } 49752ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4976724e2315SJosef Bacik 49770b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49780b246afaSJeff Mahoney if (t == fs_info->running_transaction) 49790b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4980724e2315SJosef Bacik list_del_init(&t->list); 49810b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4982a4abeea4SJosef Bacik 4983724e2315SJosef Bacik btrfs_put_transaction(t); 49842ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 49850b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4986724e2315SJosef Bacik } 49870b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49880b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4989ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4990ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 49910b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4992ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 49930b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4994acce952bSliubo 4995acce952bSliubo return 0; 4996acce952bSliubo } 4997ec7d6dfdSNikolay Borisov 4998453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 4999ec7d6dfdSNikolay Borisov { 5000ec7d6dfdSNikolay Borisov struct btrfs_path *path; 5001ec7d6dfdSNikolay Borisov int ret; 5002ec7d6dfdSNikolay Borisov struct extent_buffer *l; 5003ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 5004ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 5005ec7d6dfdSNikolay Borisov int slot; 5006ec7d6dfdSNikolay Borisov 5007ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 5008ec7d6dfdSNikolay Borisov if (!path) 5009ec7d6dfdSNikolay Borisov return -ENOMEM; 5010ec7d6dfdSNikolay Borisov 5011ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 5012ec7d6dfdSNikolay Borisov search_key.type = -1; 5013ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 5014ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 5015ec7d6dfdSNikolay Borisov if (ret < 0) 5016ec7d6dfdSNikolay Borisov goto error; 5017ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 5018ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 5019ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 5020ec7d6dfdSNikolay Borisov l = path->nodes[0]; 5021ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 502223125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 502323125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 5024ec7d6dfdSNikolay Borisov } else { 502523125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 5026ec7d6dfdSNikolay Borisov } 5027ec7d6dfdSNikolay Borisov ret = 0; 5028ec7d6dfdSNikolay Borisov error: 5029ec7d6dfdSNikolay Borisov btrfs_free_path(path); 5030ec7d6dfdSNikolay Borisov return ret; 5031ec7d6dfdSNikolay Borisov } 5032ec7d6dfdSNikolay Borisov 5033543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 5034ec7d6dfdSNikolay Borisov { 5035ec7d6dfdSNikolay Borisov int ret; 5036ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5037ec7d6dfdSNikolay Borisov 50386b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5039ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5040ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5041ec7d6dfdSNikolay Borisov root->root_key.objectid); 5042ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5043ec7d6dfdSNikolay Borisov goto out; 5044ec7d6dfdSNikolay Borisov } 5045ec7d6dfdSNikolay Borisov 504623125104SNikolay Borisov *objectid = root->free_objectid++; 5047ec7d6dfdSNikolay Borisov ret = 0; 5048ec7d6dfdSNikolay Borisov out: 5049ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5050ec7d6dfdSNikolay Borisov return ret; 5051ec7d6dfdSNikolay Borisov } 5052