1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 26cbd5570SChris Mason /* 36cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 46cbd5570SChris Mason */ 56cbd5570SChris Mason 6e20d96d6SChris Mason #include <linux/fs.h> 7d98237b3SChris Mason #include <linux/blkdev.h> 8fc7cbcd4SDavid Sterba #include <linux/radix-tree.h> 935b7e476SChris Mason #include <linux/writeback.h> 10ce9adaa5SChris Mason #include <linux/workqueue.h> 11a74a4b97SChris Mason #include <linux/kthread.h> 125a0e3ad6STejun Heo #include <linux/slab.h> 13784b4e29SChris Mason #include <linux/migrate.h> 147a36ddecSDavid Sterba #include <linux/ratelimit.h> 156463fe58SStefan Behrens #include <linux/uuid.h> 16803b2f54SStefan Behrens #include <linux/semaphore.h> 17540adea3SMasami Hiramatsu #include <linux/error-injection.h> 189678c543SNikolay Borisov #include <linux/crc32c.h> 19b89f6d1fSFilipe Manana #include <linux/sched/mm.h> 207e75bf3fSDavid Sterba #include <asm/unaligned.h> 216d97c6e3SJohannes Thumshirn #include <crypto/hash.h> 22eb60ceacSChris Mason #include "ctree.h" 23eb60ceacSChris Mason #include "disk-io.h" 24e089f05cSChris Mason #include "transaction.h" 250f7d52f4SChris Mason #include "btrfs_inode.h" 26103c1972SChristoph Hellwig #include "bio.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" 46c7f13d42SJosef Bacik #include "fs.h" 4707e81dc9SJosef Bacik #include "accessors.h" 48a0231804SJosef Bacik #include "extent-tree.h" 4945c40c8fSJosef Bacik #include "root-tree.h" 5059b818e0SJosef Bacik #include "defrag.h" 51c7a03b52SJosef Bacik #include "uuid-tree.h" 5267707479SJosef Bacik #include "relocation.h" 532fc6822cSJosef Bacik #include "scrub.h" 54c03b2207SJosef Bacik #include "super.h" 55eb60ceacSChris Mason 56319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 57319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 58319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 59319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 60e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 61e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 62319e4d06SQu Wenruo 632ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 642ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 65ce9adaa5SChris Mason 66141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 67141386e1SJosef Bacik { 68141386e1SJosef Bacik if (fs_info->csum_shash) 69141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 70141386e1SJosef Bacik } 71141386e1SJosef Bacik 72d352ac68SChris Mason /* 732996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 74d352ac68SChris Mason */ 75c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 7619c00ddcSChris Mason { 77d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 787280305eSDavid Sterba const int num_pages = num_extent_pages(buf); 79a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 80d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 8119c00ddcSChris Mason char *kaddr; 82e9be5a30SDavid Sterba int i; 83d5178578SJohannes Thumshirn 84d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 85d5178578SJohannes Thumshirn crypto_shash_init(shash); 86a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 87e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 88a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 8919c00ddcSChris Mason 905ad9b471Spengfuyuan for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) { 91e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 92e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 9319c00ddcSChris Mason } 9471a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 95d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 9619c00ddcSChris Mason } 9719c00ddcSChris Mason 98d352ac68SChris Mason /* 99d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 100d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 101d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 102d352ac68SChris Mason * in the wrong place. 103d352ac68SChris Mason */ 104d87e6575SChristoph Hellwig int btrfs_buffer_uptodate(struct extent_buffer *eb, u64 parent_transid, int atomic) 1051259ab75SChris Mason { 106d87e6575SChristoph Hellwig if (!extent_buffer_uptodate(eb)) 107d87e6575SChristoph Hellwig return 0; 1081259ab75SChris Mason 1091259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 110d87e6575SChristoph Hellwig return 1; 1111259ab75SChris Mason 112b9fab919SChris Mason if (atomic) 113b9fab919SChris Mason return -EAGAIN; 114b9fab919SChris Mason 115d87e6575SChristoph Hellwig if (!extent_buffer_uptodate(eb) || 116d87e6575SChristoph Hellwig btrfs_header_generation(eb) != parent_transid) { 11794647322SDavid Sterba btrfs_err_rl(eb->fs_info, 1188f0ed7d4SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 1198f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 12029549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 1210b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 1229e2aff90SChristoph Hellwig return 0; 123d87e6575SChristoph Hellwig } 1249e2aff90SChristoph Hellwig return 1; 1251259ab75SChris Mason } 1261259ab75SChris Mason 127e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 128e7e16f48SJohannes Thumshirn { 129e7e16f48SJohannes Thumshirn switch (csum_type) { 130e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 1313951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 1323831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 133352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 134e7e16f48SJohannes Thumshirn return true; 135e7e16f48SJohannes Thumshirn default: 136e7e16f48SJohannes Thumshirn return false; 137e7e16f48SJohannes Thumshirn } 138e7e16f48SJohannes Thumshirn } 139e7e16f48SJohannes Thumshirn 140d352ac68SChris Mason /* 1411104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 1421104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 1431104a885SDavid Sterba */ 1443d17adeaSQu Wenruo int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 1453d17adeaSQu Wenruo const struct btrfs_super_block *disk_sb) 1461104a885SDavid Sterba { 14751bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 148d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 149d5178578SJohannes Thumshirn 150d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 1511104a885SDavid Sterba 1521104a885SDavid Sterba /* 1531104a885SDavid Sterba * The super_block structure does not span the whole 15451bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 15551bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 1561104a885SDavid Sterba */ 1573d17adeaSQu Wenruo crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, 158fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 1591104a885SDavid Sterba 16055fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 161e7e16f48SJohannes Thumshirn return 1; 1621104a885SDavid Sterba 163e7e16f48SJohannes Thumshirn return 0; 1641104a885SDavid Sterba } 1651104a885SDavid Sterba 166bacf60e5SChristoph Hellwig static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, 167bacf60e5SChristoph Hellwig int mirror_num) 168bacf60e5SChristoph Hellwig { 169bacf60e5SChristoph Hellwig struct btrfs_fs_info *fs_info = eb->fs_info; 170bacf60e5SChristoph Hellwig int i, num_pages = num_extent_pages(eb); 171bacf60e5SChristoph Hellwig int ret = 0; 172bacf60e5SChristoph Hellwig 173bacf60e5SChristoph Hellwig if (sb_rdonly(fs_info->sb)) 174bacf60e5SChristoph Hellwig return -EROFS; 175bacf60e5SChristoph Hellwig 176bacf60e5SChristoph Hellwig for (i = 0; i < num_pages; i++) { 177bacf60e5SChristoph Hellwig struct page *p = eb->pages[i]; 178917ac778SQu Wenruo u64 start = max_t(u64, eb->start, page_offset(p)); 179917ac778SQu Wenruo u64 end = min_t(u64, eb->start + eb->len, page_offset(p) + PAGE_SIZE); 180917ac778SQu Wenruo u32 len = end - start; 181bacf60e5SChristoph Hellwig 182917ac778SQu Wenruo ret = btrfs_repair_io_failure(fs_info, 0, start, len, 183917ac778SQu Wenruo start, p, offset_in_page(start), mirror_num); 184bacf60e5SChristoph Hellwig if (ret) 185bacf60e5SChristoph Hellwig break; 186bacf60e5SChristoph Hellwig } 187bacf60e5SChristoph Hellwig 188bacf60e5SChristoph Hellwig return ret; 189bacf60e5SChristoph Hellwig } 190bacf60e5SChristoph Hellwig 1911104a885SDavid Sterba /* 192d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 193d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 194581c1760SQu Wenruo * 195789d6a3aSQu Wenruo * @check: expected tree parentness check, see the comments of the 196789d6a3aSQu Wenruo * structure for details. 197d352ac68SChris Mason */ 1986a2e9dc4SFilipe Manana int btrfs_read_extent_buffer(struct extent_buffer *eb, 199789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 200f188591eSChris Mason { 2015ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 202ea466794SJosef Bacik int failed = 0; 203f188591eSChris Mason int ret; 204f188591eSChris Mason int num_copies = 0; 205f188591eSChris Mason int mirror_num = 0; 206ea466794SJosef Bacik int failed_mirror = 0; 207f188591eSChris Mason 208789d6a3aSQu Wenruo ASSERT(check); 209789d6a3aSQu Wenruo 210f188591eSChris Mason while (1) { 211f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 212947a6299SQu Wenruo ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num, check); 213947a6299SQu Wenruo if (!ret) 214581c1760SQu Wenruo break; 215d397712bSChris Mason 2160b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 217f188591eSChris Mason eb->start, eb->len); 2184235298eSChris Mason if (num_copies == 1) 219ea466794SJosef Bacik break; 2204235298eSChris Mason 2215cf1ab56SJosef Bacik if (!failed_mirror) { 2225cf1ab56SJosef Bacik failed = 1; 2235cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 2245cf1ab56SJosef Bacik } 2255cf1ab56SJosef Bacik 226f188591eSChris Mason mirror_num++; 227ea466794SJosef Bacik if (mirror_num == failed_mirror) 228ea466794SJosef Bacik mirror_num++; 229ea466794SJosef Bacik 2304235298eSChris Mason if (mirror_num > num_copies) 231ea466794SJosef Bacik break; 232f188591eSChris Mason } 233ea466794SJosef Bacik 234c0901581SStefan Behrens if (failed && !ret && failed_mirror) 23520a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 236ea466794SJosef Bacik 237ea466794SJosef Bacik return ret; 238f188591eSChris Mason } 23919c00ddcSChris Mason 24031d89399SChristoph Hellwig /* 24131d89399SChristoph Hellwig * Checksum a dirty tree block before IO. 24231d89399SChristoph Hellwig */ 24331d89399SChristoph Hellwig blk_status_t btree_csum_one_bio(struct btrfs_bio *bbio) 244eca0f6f6SQu Wenruo { 24531d89399SChristoph Hellwig struct extent_buffer *eb = bbio->private; 246eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 24731d89399SChristoph Hellwig u64 found_start = btrfs_header_bytenr(eb); 248eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 249eca0f6f6SQu Wenruo int ret; 250eca0f6f6SQu Wenruo 25131d89399SChristoph Hellwig /* Btree blocks are always contiguous on disk. */ 25231d89399SChristoph Hellwig if (WARN_ON_ONCE(bbio->file_offset != eb->start)) 25331d89399SChristoph Hellwig return BLK_STS_IOERR; 25431d89399SChristoph Hellwig if (WARN_ON_ONCE(bbio->bio.bi_iter.bi_size != eb->len)) 25531d89399SChristoph Hellwig return BLK_STS_IOERR; 25631d89399SChristoph Hellwig 25731d89399SChristoph Hellwig if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 25831d89399SChristoph Hellwig WARN_ON_ONCE(found_start != 0); 25931d89399SChristoph Hellwig return BLK_STS_OK; 26031d89399SChristoph Hellwig } 26131d89399SChristoph Hellwig 26231d89399SChristoph Hellwig if (WARN_ON_ONCE(found_start != eb->start)) 26331d89399SChristoph Hellwig return BLK_STS_IOERR; 26431d89399SChristoph Hellwig if (WARN_ON(!btrfs_page_test_uptodate(fs_info, eb->pages[0], eb->start, 26531d89399SChristoph Hellwig eb->len))) 26631d89399SChristoph Hellwig return BLK_STS_IOERR; 26731d89399SChristoph Hellwig 268eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 269eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 270eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 271eca0f6f6SQu Wenruo csum_tree_block(eb, result); 272eca0f6f6SQu Wenruo 273eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 274eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 275eca0f6f6SQu Wenruo else 27685d8a826SJosef Bacik ret = btrfs_check_leaf(eb); 277eca0f6f6SQu Wenruo 2783777369fSQu Wenruo if (ret < 0) 2793777369fSQu Wenruo goto error; 2803777369fSQu Wenruo 2813777369fSQu Wenruo /* 2823777369fSQu Wenruo * Also check the generation, the eb reached here must be newer than 2833777369fSQu Wenruo * last committed. Or something seriously wrong happened. 2843777369fSQu Wenruo */ 2853777369fSQu Wenruo if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { 2863777369fSQu Wenruo ret = -EUCLEAN; 287eca0f6f6SQu Wenruo btrfs_err(fs_info, 2883777369fSQu Wenruo "block=%llu bad generation, have %llu expect > %llu", 2893777369fSQu Wenruo eb->start, btrfs_header_generation(eb), 2903777369fSQu Wenruo fs_info->last_trans_committed); 2913777369fSQu Wenruo goto error; 292eca0f6f6SQu Wenruo } 293eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 29431d89399SChristoph Hellwig return BLK_STS_OK; 2953777369fSQu Wenruo 2963777369fSQu Wenruo error: 2973777369fSQu Wenruo btrfs_print_tree(eb, 0); 2983777369fSQu Wenruo btrfs_err(fs_info, "block=%llu write time tree block corruption detected", 2993777369fSQu Wenruo eb->start); 30016199ad9SFilipe Manana /* 30116199ad9SFilipe Manana * Be noisy if this is an extent buffer from a log tree. We don't abort 30216199ad9SFilipe Manana * a transaction in case there's a bad log tree extent buffer, we just 30316199ad9SFilipe Manana * fallback to a transaction commit. Still we want to know when there is 30416199ad9SFilipe Manana * a bad log tree extent buffer, as that may signal a bug somewhere. 30516199ad9SFilipe Manana */ 30616199ad9SFilipe Manana WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG) || 30716199ad9SFilipe Manana btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID); 308deb6216fSChristoph Hellwig return errno_to_blk_status(ret); 309deb6216fSChristoph Hellwig } 310deb6216fSChristoph Hellwig 311413fb1bcSAnand Jain static bool check_tree_block_fsid(struct extent_buffer *eb) 3122b82032cSYan Zheng { 313b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 314944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 31544880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 3162b82032cSYan Zheng 3179a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 3189a8658e3SDavid Sterba BTRFS_FSID_SIZE); 3197239ff4bSNikolay Borisov 32067bc5ad0SAnand Jain /* 32167bc5ad0SAnand Jain * alloc_fs_devices() copies the fsid into metadata_uuid if the 32267bc5ad0SAnand Jain * metadata_uuid is unset in the superblock, including for a seed device. 32367bc5ad0SAnand Jain * So, we can use fs_devices->metadata_uuid. 32467bc5ad0SAnand Jain */ 32567bc5ad0SAnand Jain if (memcmp(fsid, fs_info->fs_devices->metadata_uuid, BTRFS_FSID_SIZE) == 0) 326413fb1bcSAnand Jain return false; 327944d3f9fSNikolay Borisov 328944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 329944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 330413fb1bcSAnand Jain return false; 331944d3f9fSNikolay Borisov 332413fb1bcSAnand Jain return true; 3332b82032cSYan Zheng } 3342b82032cSYan Zheng 33577bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 336046b562bSChristoph Hellwig int btrfs_validate_extent_buffer(struct extent_buffer *eb, 337947a6299SQu Wenruo struct btrfs_tree_parent_check *check) 338ce9adaa5SChris Mason { 33977bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 340ce9adaa5SChris Mason u64 found_start; 34177bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 34277bf40a2SQu Wenruo u8 found_level; 3432996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 344dfd29eedSDavid Sterba const u8 *header_csum; 34577bf40a2SQu Wenruo int ret = 0; 346ea466794SJosef Bacik 347947a6299SQu Wenruo ASSERT(check); 348947a6299SQu Wenruo 349ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 350727011e0SChris Mason if (found_start != eb->start) { 3518f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, 3528f0ed7d4SQu Wenruo "bad tree block start, mirror %u want %llu have %llu", 3538f0ed7d4SQu Wenruo eb->read_mirror, eb->start, found_start); 354f188591eSChris Mason ret = -EIO; 35577bf40a2SQu Wenruo goto out; 356ce9adaa5SChris Mason } 357b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 3588f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", 3598f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 3601259ab75SChris Mason ret = -EIO; 36177bf40a2SQu Wenruo goto out; 3621259ab75SChris Mason } 363ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 3641c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 3658f0ed7d4SQu Wenruo btrfs_err(fs_info, 3668f0ed7d4SQu Wenruo "bad tree block level, mirror %u level %d on logical %llu", 3678f0ed7d4SQu Wenruo eb->read_mirror, btrfs_header_level(eb), eb->start); 3681c24c3ceSJosef Bacik ret = -EIO; 36977bf40a2SQu Wenruo goto out; 3701c24c3ceSJosef Bacik } 371ce9adaa5SChris Mason 372c67b3892SDavid Sterba csum_tree_block(eb, result); 373dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 374dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 375a826d6dcSJosef Bacik 376dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 3772996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 3788f0ed7d4SQu Wenruo "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", 3798f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 380dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 38135be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 38235be8851SJohannes Thumshirn btrfs_header_level(eb)); 3832996e1f8SJohannes Thumshirn ret = -EUCLEAN; 38477bf40a2SQu Wenruo goto out; 3852996e1f8SJohannes Thumshirn } 3862996e1f8SJohannes Thumshirn 387947a6299SQu Wenruo if (found_level != check->level) { 38877177ed1SQu Wenruo btrfs_err(fs_info, 38977177ed1SQu Wenruo "level verify failed on logical %llu mirror %u wanted %u found %u", 39077177ed1SQu Wenruo eb->start, eb->read_mirror, check->level, found_level); 391947a6299SQu Wenruo ret = -EIO; 392947a6299SQu Wenruo goto out; 393947a6299SQu Wenruo } 394947a6299SQu Wenruo if (unlikely(check->transid && 395947a6299SQu Wenruo btrfs_header_generation(eb) != check->transid)) { 396947a6299SQu Wenruo btrfs_err_rl(eb->fs_info, 397947a6299SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 398947a6299SQu Wenruo eb->start, eb->read_mirror, check->transid, 399947a6299SQu Wenruo btrfs_header_generation(eb)); 400947a6299SQu Wenruo ret = -EIO; 401947a6299SQu Wenruo goto out; 402947a6299SQu Wenruo } 403947a6299SQu Wenruo if (check->has_first_key) { 404947a6299SQu Wenruo struct btrfs_key *expect_key = &check->first_key; 405947a6299SQu Wenruo struct btrfs_key found_key; 406947a6299SQu Wenruo 407947a6299SQu Wenruo if (found_level) 408947a6299SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 409947a6299SQu Wenruo else 410947a6299SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 411947a6299SQu Wenruo if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { 412947a6299SQu Wenruo btrfs_err(fs_info, 413947a6299SQu Wenruo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 414947a6299SQu Wenruo eb->start, check->transid, 415947a6299SQu Wenruo expect_key->objectid, 416947a6299SQu Wenruo expect_key->type, expect_key->offset, 417947a6299SQu Wenruo found_key.objectid, found_key.type, 418947a6299SQu Wenruo found_key.offset); 419947a6299SQu Wenruo ret = -EUCLEAN; 420947a6299SQu Wenruo goto out; 421947a6299SQu Wenruo } 422947a6299SQu Wenruo } 423947a6299SQu Wenruo if (check->owner_root) { 424947a6299SQu Wenruo ret = btrfs_check_eb_owner(eb, check->owner_root); 425947a6299SQu Wenruo if (ret < 0) 426947a6299SQu Wenruo goto out; 427947a6299SQu Wenruo } 428947a6299SQu Wenruo 429a826d6dcSJosef Bacik /* 430a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 431a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 432a826d6dcSJosef Bacik * return -EIO. 433a826d6dcSJosef Bacik */ 43485d8a826SJosef Bacik if (found_level == 0 && btrfs_check_leaf(eb)) { 435a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 436a826d6dcSJosef Bacik ret = -EIO; 437a826d6dcSJosef Bacik } 438ce9adaa5SChris Mason 439813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 440053ab70fSLiu Bo ret = -EIO; 441053ab70fSLiu Bo 442aebcc159SChristoph Hellwig if (ret) 44375391f0dSQu Wenruo btrfs_err(fs_info, 4448f0ed7d4SQu Wenruo "read time tree block corruption detected on logical %llu mirror %u", 4458f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 44677bf40a2SQu Wenruo out: 44777bf40a2SQu Wenruo return ret; 44877bf40a2SQu Wenruo } 44977bf40a2SQu Wenruo 4503dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 4518958b551SMatthew Wilcox (Oracle) static int btree_migrate_folio(struct address_space *mapping, 4528958b551SMatthew Wilcox (Oracle) struct folio *dst, struct folio *src, enum migrate_mode mode) 453784b4e29SChris Mason { 454784b4e29SChris Mason /* 455784b4e29SChris Mason * we can't safely write a btree page from here, 456784b4e29SChris Mason * we haven't done the locking hook 457784b4e29SChris Mason */ 4588958b551SMatthew Wilcox (Oracle) if (folio_test_dirty(src)) 459784b4e29SChris Mason return -EAGAIN; 460784b4e29SChris Mason /* 461784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 462784b4e29SChris Mason * We must have no buffers or drop them. 463784b4e29SChris Mason */ 4648958b551SMatthew Wilcox (Oracle) if (folio_get_private(src) && 4658958b551SMatthew Wilcox (Oracle) !filemap_release_folio(src, GFP_KERNEL)) 466784b4e29SChris Mason return -EAGAIN; 46754184650SMatthew Wilcox (Oracle) return migrate_folio(mapping, dst, src, mode); 468784b4e29SChris Mason } 4698958b551SMatthew Wilcox (Oracle) #else 4708958b551SMatthew Wilcox (Oracle) #define btree_migrate_folio NULL 4713dd1462eSJan Beulich #endif 472784b4e29SChris Mason 4730da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 4740da5468fSChris Mason struct writeback_control *wbc) 4750da5468fSChris Mason { 476e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 477e2d84521SMiao Xie int ret; 478e2d84521SMiao Xie 479d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 480448d640bSChris Mason 481448d640bSChris Mason if (wbc->for_kupdate) 482448d640bSChris Mason return 0; 483448d640bSChris Mason 484e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 485b9473439SChris Mason /* this is a bit racy, but that's ok */ 486d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 487d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 488d814a491SEthan Lien fs_info->dirty_metadata_batch); 489e2d84521SMiao Xie if (ret < 0) 490793955bcSChris Mason return 0; 491793955bcSChris Mason } 4920b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 4930da5468fSChris Mason } 4940da5468fSChris Mason 495f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 4965f39d397SChris Mason { 497f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 498f913cff3SMatthew Wilcox (Oracle) return false; 4990c4e538bSDavid Sterba 500f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 501d98237b3SChris Mason } 502d98237b3SChris Mason 503895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 504895586ebSMatthew Wilcox (Oracle) size_t length) 505d98237b3SChris Mason { 506d1310b2eSChris Mason struct extent_io_tree *tree; 507895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 508895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 509f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 510895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 511895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 512895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 513895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 514895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 5159ad6b7bcSChris Mason } 516d98237b3SChris Mason } 517d98237b3SChris Mason 518bb146eb2SJosef Bacik #ifdef DEBUG 5190079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 5200079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 5210079c3b1SMatthew Wilcox (Oracle) { 5220079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 5235e0e8799SQu Wenruo struct btrfs_subpage_info *spi = fs_info->subpage_info; 524139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 5250b32f4bbSJosef Bacik struct extent_buffer *eb; 526139e8cd3SQu Wenruo int cur_bit = 0; 5270079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 5280b32f4bbSJosef Bacik 529139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 5300079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 5310b32f4bbSJosef Bacik BUG_ON(!eb); 5320b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 5330b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 53449d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 5350079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 536139e8cd3SQu Wenruo } 5375e0e8799SQu Wenruo 5385e0e8799SQu Wenruo ASSERT(spi); 5390079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 540139e8cd3SQu Wenruo 5415e0e8799SQu Wenruo for (cur_bit = spi->dirty_offset; 5425e0e8799SQu Wenruo cur_bit < spi->dirty_offset + spi->bitmap_nr_bits; 5435e0e8799SQu Wenruo cur_bit++) { 544139e8cd3SQu Wenruo unsigned long flags; 545139e8cd3SQu Wenruo u64 cur; 546139e8cd3SQu Wenruo 547139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 5485e0e8799SQu Wenruo if (!test_bit(cur_bit, subpage->bitmaps)) { 549139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 550139e8cd3SQu Wenruo continue; 551139e8cd3SQu Wenruo } 552139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 553139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 554139e8cd3SQu Wenruo 555139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 556139e8cd3SQu Wenruo ASSERT(eb); 557139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 558139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 55949d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 560139e8cd3SQu Wenruo free_extent_buffer(eb); 561139e8cd3SQu Wenruo 5625e0e8799SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits) - 1; 563139e8cd3SQu Wenruo } 5640079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 5650b32f4bbSJosef Bacik } 5660079c3b1SMatthew Wilcox (Oracle) #else 5670079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 5680079c3b1SMatthew Wilcox (Oracle) #endif 5690b32f4bbSJosef Bacik 5707f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 5710da5468fSChris Mason .writepages = btree_writepages, 572f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 573895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 5748958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 5750079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 576d98237b3SChris Mason }; 577123abc88SChris Mason 5782ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 5792ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 5803fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 5813fbaf258SJosef Bacik int level) 5820999df54SChris Mason { 5830b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 5840b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 5853fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 5860999df54SChris Mason } 5870999df54SChris Mason 588581c1760SQu Wenruo /* 589581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 590581c1760SQu Wenruo * verification. 591581c1760SQu Wenruo * 592789d6a3aSQu Wenruo * @check: expected tree parentness check, see comments of the 593789d6a3aSQu Wenruo * structure for details. 594581c1760SQu Wenruo */ 5952ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 596789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 597e20d96d6SChris Mason { 5985f39d397SChris Mason struct extent_buffer *buf = NULL; 59919c00ddcSChris Mason int ret; 60019c00ddcSChris Mason 601789d6a3aSQu Wenruo ASSERT(check); 602789d6a3aSQu Wenruo 603789d6a3aSQu Wenruo buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, 604789d6a3aSQu Wenruo check->level); 605c871b0f2SLiu Bo if (IS_ERR(buf)) 606c871b0f2SLiu Bo return buf; 607e4204dedSChris Mason 608789d6a3aSQu Wenruo ret = btrfs_read_extent_buffer(buf, check); 6090f0fe8f7SFilipe David Borba Manana if (ret) { 610537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 61164c043deSLiu Bo return ERR_PTR(ret); 6120f0fe8f7SFilipe David Borba Manana } 613789d6a3aSQu Wenruo if (btrfs_check_eb_owner(buf, check->owner_root)) { 61488c602abSQu Wenruo free_extent_buffer_stale(buf); 61588c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 61688c602abSQu Wenruo } 6175f39d397SChris Mason return buf; 618ce9adaa5SChris Mason 619eb60ceacSChris Mason } 620eb60ceacSChris Mason 621da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 622e20d96d6SChris Mason u64 objectid) 623d97e63b6SChris Mason { 6247c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 6252e608bd1SJosef Bacik 6262e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 6272e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 6282e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 62996dfcb46SJosef Bacik root->fs_info = fs_info; 6302e608bd1SJosef Bacik root->root_key.objectid = objectid; 631cfaa7295SChris Mason root->node = NULL; 632a28ec197SChris Mason root->commit_root = NULL; 63327cdeb70SMiao Xie root->state = 0; 634abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 6350b86a832SChris Mason 6360f7d52f4SChris Mason root->last_trans = 0; 6376b8fad57SNikolay Borisov root->free_objectid = 0; 638eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 639199c2a9cSMiao Xie root->nr_ordered_extents = 0; 6406bef4d31SEric Paris root->inode_tree = RB_ROOT; 641088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 6422e608bd1SJosef Bacik 6432e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 6440b86a832SChris Mason 6450b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 6465d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 647eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 648eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 649199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 650199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 651d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 6522ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 6532ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 6545d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 655eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 656199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 657f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 6582ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 6592ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 6608287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 661a2135011SChris Mason mutex_init(&root->objectid_mutex); 662e02119d5SChris Mason mutex_init(&root->log_mutex); 66331f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 664573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 665c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 6667237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 6677237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 6687237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 6698b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 6708b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 6717237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 6727237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 6737237f183SYan Zheng atomic_set(&root->log_writers, 0); 6742ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 6750700cea7SElena Reshetova refcount_set(&root->refs, 1); 6768ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 677eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 6787237f183SYan Zheng root->log_transid = 0; 679d1433debSMiao Xie root->log_transid_committed = -1; 680257c62e1SChris Mason root->last_log_commit = 0; 6812e608bd1SJosef Bacik root->anon_dev = 0; 682e289f03eSFilipe Manana if (!dummy) { 68343eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 68435da5a7eSDavid Sterba IO_TREE_ROOT_DIRTY_LOG_PAGES); 685e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 68635da5a7eSDavid Sterba IO_TREE_LOG_CSUM_RANGE); 687e289f03eSFilipe Manana } 688017e5369SChris Mason 6895f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 690370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 691bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 692bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 693fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 694bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 695fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 696bd647ce3SJosef Bacik #endif 6973768f368SChris Mason } 6983768f368SChris Mason 69974e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 70096dfcb46SJosef Bacik u64 objectid, gfp_t flags) 7016f07e42eSAl Viro { 70274e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 7036f07e42eSAl Viro if (root) 70496dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 7056f07e42eSAl Viro return root; 7066f07e42eSAl Viro } 7076f07e42eSAl Viro 70806ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 70906ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 710da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 71106ea65a3SJosef Bacik { 71206ea65a3SJosef Bacik struct btrfs_root *root; 71306ea65a3SJosef Bacik 7147c0260eeSJeff Mahoney if (!fs_info) 7157c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 7167c0260eeSJeff Mahoney 71796dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 71806ea65a3SJosef Bacik if (!root) 71906ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 720da17066cSJeff Mahoney 721b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 722faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 72306ea65a3SJosef Bacik 72406ea65a3SJosef Bacik return root; 72506ea65a3SJosef Bacik } 72606ea65a3SJosef Bacik #endif 72706ea65a3SJosef Bacik 728abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 729abed4aaaSJosef Bacik { 730abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 731abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 732abed4aaaSJosef Bacik 733abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 734abed4aaaSJosef Bacik } 735abed4aaaSJosef Bacik 736abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 737abed4aaaSJosef Bacik { 738abed4aaaSJosef Bacik const struct btrfs_key *key = k; 739abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 740abed4aaaSJosef Bacik 741abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 742abed4aaaSJosef Bacik } 743abed4aaaSJosef Bacik 744abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 745abed4aaaSJosef Bacik { 746abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 747abed4aaaSJosef Bacik struct rb_node *tmp; 748745806fbSQu Wenruo int ret = 0; 749abed4aaaSJosef Bacik 750abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 751abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 752abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 753abed4aaaSJosef Bacik 754745806fbSQu Wenruo if (tmp) { 755745806fbSQu Wenruo ret = -EEXIST; 756745806fbSQu Wenruo btrfs_warn(fs_info, "global root %llu %llu already exists", 757745806fbSQu Wenruo root->root_key.objectid, root->root_key.offset); 758745806fbSQu Wenruo } 759745806fbSQu Wenruo return ret; 760abed4aaaSJosef Bacik } 761abed4aaaSJosef Bacik 762abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 763abed4aaaSJosef Bacik { 764abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 765abed4aaaSJosef Bacik 766abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 767abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 768abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 769abed4aaaSJosef Bacik } 770abed4aaaSJosef Bacik 771abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 772abed4aaaSJosef Bacik struct btrfs_key *key) 773abed4aaaSJosef Bacik { 774abed4aaaSJosef Bacik struct rb_node *node; 775abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 776abed4aaaSJosef Bacik 777abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 778abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 779abed4aaaSJosef Bacik if (node) 780abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 781abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 782abed4aaaSJosef Bacik 783abed4aaaSJosef Bacik return root; 784abed4aaaSJosef Bacik } 785abed4aaaSJosef Bacik 786f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 787f7238e50SJosef Bacik { 788f7238e50SJosef Bacik struct btrfs_block_group *block_group; 789f7238e50SJosef Bacik u64 ret; 790f7238e50SJosef Bacik 791f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 792f7238e50SJosef Bacik return 0; 793f7238e50SJosef Bacik 794f7238e50SJosef Bacik if (bytenr) 795f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 796f7238e50SJosef Bacik else 797f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 798f7238e50SJosef Bacik ASSERT(block_group); 799f7238e50SJosef Bacik if (!block_group) 800f7238e50SJosef Bacik return 0; 801f7238e50SJosef Bacik ret = block_group->global_root_id; 802f7238e50SJosef Bacik btrfs_put_block_group(block_group); 803f7238e50SJosef Bacik 804f7238e50SJosef Bacik return ret; 805f7238e50SJosef Bacik } 806f7238e50SJosef Bacik 807abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 808abed4aaaSJosef Bacik { 809abed4aaaSJosef Bacik struct btrfs_key key = { 810abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 811abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 812f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 813abed4aaaSJosef Bacik }; 814abed4aaaSJosef Bacik 815abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 816abed4aaaSJosef Bacik } 817abed4aaaSJosef Bacik 818abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 819abed4aaaSJosef Bacik { 820abed4aaaSJosef Bacik struct btrfs_key key = { 821abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 822abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 823f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 824abed4aaaSJosef Bacik }; 825abed4aaaSJosef Bacik 826abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 827abed4aaaSJosef Bacik } 828abed4aaaSJosef Bacik 82951129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 83051129b33SJosef Bacik { 83151129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 83251129b33SJosef Bacik return fs_info->block_group_root; 83351129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 83451129b33SJosef Bacik } 83551129b33SJosef Bacik 83620897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 83720897f5cSArne Jansen u64 objectid) 83820897f5cSArne Jansen { 8399b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 84020897f5cSArne Jansen struct extent_buffer *leaf; 84120897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 84220897f5cSArne Jansen struct btrfs_root *root; 84320897f5cSArne Jansen struct btrfs_key key; 844b89f6d1fSFilipe Manana unsigned int nofs_flag; 84520897f5cSArne Jansen int ret = 0; 84620897f5cSArne Jansen 847b89f6d1fSFilipe Manana /* 848b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 849b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 850b89f6d1fSFilipe Manana */ 851b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 85296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 853b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 85420897f5cSArne Jansen if (!root) 85520897f5cSArne Jansen return ERR_PTR(-ENOMEM); 85620897f5cSArne Jansen 85720897f5cSArne Jansen root->root_key.objectid = objectid; 85820897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 85920897f5cSArne Jansen root->root_key.offset = 0; 86020897f5cSArne Jansen 8619631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 8629631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 86320897f5cSArne Jansen if (IS_ERR(leaf)) { 86420897f5cSArne Jansen ret = PTR_ERR(leaf); 8651dd05682STsutomu Itoh leaf = NULL; 866c1b07854SPeng Hao goto fail; 86720897f5cSArne Jansen } 86820897f5cSArne Jansen 86920897f5cSArne Jansen root->node = leaf; 870d5e09e38SFilipe Manana btrfs_mark_buffer_dirty(trans, leaf); 87120897f5cSArne Jansen 87220897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 87327cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 87420897f5cSArne Jansen 875f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 876f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 87720897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 87820897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 87920897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 88020897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 88120897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 88220897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 88320897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 88433d85fdaSQu Wenruo if (is_fstree(objectid)) 885807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 886807fc790SAndy Shevchenko else 887807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 888c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 88920897f5cSArne Jansen 8908a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 8918a6a87cdSBoris Burkov 89220897f5cSArne Jansen key.objectid = objectid; 89320897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 89420897f5cSArne Jansen key.offset = 0; 89520897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 89620897f5cSArne Jansen if (ret) 89720897f5cSArne Jansen goto fail; 89820897f5cSArne Jansen 89920897f5cSArne Jansen return root; 9001dd05682STsutomu Itoh 9018a6a87cdSBoris Burkov fail: 90200246528SJosef Bacik btrfs_put_root(root); 9031dd05682STsutomu Itoh 9041dd05682STsutomu Itoh return ERR_PTR(ret); 90520897f5cSArne Jansen } 90620897f5cSArne Jansen 9077237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 908e02119d5SChris Mason struct btrfs_fs_info *fs_info) 9090f7d52f4SChris Mason { 9100f7d52f4SChris Mason struct btrfs_root *root; 911e02119d5SChris Mason 91296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 913e02119d5SChris Mason if (!root) 9147237f183SYan Zheng return ERR_PTR(-ENOMEM); 915e02119d5SChris Mason 916e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 917e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 918e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 91927cdeb70SMiao Xie 9206ab6ebb7SNaohiro Aota return root; 9216ab6ebb7SNaohiro Aota } 9226ab6ebb7SNaohiro Aota 9236ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 9246ab6ebb7SNaohiro Aota struct btrfs_root *root) 9256ab6ebb7SNaohiro Aota { 9266ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 9276ab6ebb7SNaohiro Aota 9287237f183SYan Zheng /* 92992a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 93027cdeb70SMiao Xie * 93192a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 93292a7cc42SQu Wenruo * and they go away before a real commit is actually done. 93392a7cc42SQu Wenruo * 93492a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 93592a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 9367237f183SYan Zheng */ 937e02119d5SChris Mason 9384d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 9399631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 9406ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 9416ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 942e02119d5SChris Mason 9437237f183SYan Zheng root->node = leaf; 944e02119d5SChris Mason 945d5e09e38SFilipe Manana btrfs_mark_buffer_dirty(trans, root->node); 946e02119d5SChris Mason btrfs_tree_unlock(root->node); 9476ab6ebb7SNaohiro Aota 9486ab6ebb7SNaohiro Aota return 0; 9497237f183SYan Zheng } 9507237f183SYan Zheng 9517237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 9527237f183SYan Zheng struct btrfs_fs_info *fs_info) 9537237f183SYan Zheng { 9547237f183SYan Zheng struct btrfs_root *log_root; 9557237f183SYan Zheng 9567237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 9577237f183SYan Zheng if (IS_ERR(log_root)) 9587237f183SYan Zheng return PTR_ERR(log_root); 9596ab6ebb7SNaohiro Aota 9603ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 9613ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 9623ddebf27SNaohiro Aota 9636ab6ebb7SNaohiro Aota if (ret) { 9646ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 9656ab6ebb7SNaohiro Aota return ret; 9666ab6ebb7SNaohiro Aota } 9673ddebf27SNaohiro Aota } 9686ab6ebb7SNaohiro Aota 9697237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 9707237f183SYan Zheng fs_info->log_root_tree = log_root; 9717237f183SYan Zheng return 0; 9727237f183SYan Zheng } 9737237f183SYan Zheng 9747237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 9757237f183SYan Zheng struct btrfs_root *root) 9767237f183SYan Zheng { 9770b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 9787237f183SYan Zheng struct btrfs_root *log_root; 9797237f183SYan Zheng struct btrfs_inode_item *inode_item; 9806ab6ebb7SNaohiro Aota int ret; 9817237f183SYan Zheng 9820b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 9837237f183SYan Zheng if (IS_ERR(log_root)) 9847237f183SYan Zheng return PTR_ERR(log_root); 9857237f183SYan Zheng 9866ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 9876ab6ebb7SNaohiro Aota if (ret) { 9886ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 9896ab6ebb7SNaohiro Aota return ret; 9906ab6ebb7SNaohiro Aota } 9916ab6ebb7SNaohiro Aota 9927237f183SYan Zheng log_root->last_trans = trans->transid; 9937237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 9947237f183SYan Zheng 9957237f183SYan Zheng inode_item = &log_root->root_item.inode; 9963cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 9973cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 9983cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 999da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 10000b246afaSJeff Mahoney fs_info->nodesize); 10013cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 10027237f183SYan Zheng 10035d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 10047237f183SYan Zheng 10057237f183SYan Zheng WARN_ON(root->log_root); 10067237f183SYan Zheng root->log_root = log_root; 10077237f183SYan Zheng root->log_transid = 0; 1008d1433debSMiao Xie root->log_transid_committed = -1; 1009257c62e1SChris Mason root->last_log_commit = 0; 1010e02119d5SChris Mason return 0; 1011e02119d5SChris Mason } 1012e02119d5SChris Mason 101349d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 101449d11beaSJosef Bacik struct btrfs_path *path, 1015cb517eabSMiao Xie struct btrfs_key *key) 1016e02119d5SChris Mason { 1017e02119d5SChris Mason struct btrfs_root *root; 1018789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 1019e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 102084234f3aSYan Zheng u64 generation; 1021cb517eabSMiao Xie int ret; 1022581c1760SQu Wenruo int level; 1023cb517eabSMiao Xie 102496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 102549d11beaSJosef Bacik if (!root) 102649d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 10270f7d52f4SChris Mason 1028cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1029cb517eabSMiao Xie &root->root_item, &root->root_key); 10300f7d52f4SChris Mason if (ret) { 103113a8a7c8SYan, Zheng if (ret > 0) 103213a8a7c8SYan, Zheng ret = -ENOENT; 103349d11beaSJosef Bacik goto fail; 10340f7d52f4SChris Mason } 103513a8a7c8SYan, Zheng 103684234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1037581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 1038789d6a3aSQu Wenruo check.level = level; 1039789d6a3aSQu Wenruo check.transid = generation; 1040789d6a3aSQu Wenruo check.owner_root = key->objectid; 1041789d6a3aSQu Wenruo root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), 1042789d6a3aSQu Wenruo &check); 104364c043deSLiu Bo if (IS_ERR(root->node)) { 104464c043deSLiu Bo ret = PTR_ERR(root->node); 10458c38938cSJosef Bacik root->node = NULL; 104649d11beaSJosef Bacik goto fail; 10474eb150d6SQu Wenruo } 10484eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1049cb517eabSMiao Xie ret = -EIO; 105049d11beaSJosef Bacik goto fail; 1051416bc658SJosef Bacik } 105288c602abSQu Wenruo 105388c602abSQu Wenruo /* 105488c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 105588c602abSQu Wenruo * match its root node owner 105688c602abSQu Wenruo */ 105788c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 105888c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 105988c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 106088c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 106188c602abSQu Wenruo btrfs_crit(fs_info, 106288c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 106388c602abSQu Wenruo root->root_key.objectid, root->node->start, 106488c602abSQu Wenruo btrfs_header_owner(root->node), 106588c602abSQu Wenruo root->root_key.objectid); 106688c602abSQu Wenruo ret = -EUCLEAN; 106788c602abSQu Wenruo goto fail; 106888c602abSQu Wenruo } 10695d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1070cb517eabSMiao Xie return root; 107149d11beaSJosef Bacik fail: 107200246528SJosef Bacik btrfs_put_root(root); 107349d11beaSJosef Bacik return ERR_PTR(ret); 107449d11beaSJosef Bacik } 107549d11beaSJosef Bacik 107649d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 107749d11beaSJosef Bacik struct btrfs_key *key) 107849d11beaSJosef Bacik { 107949d11beaSJosef Bacik struct btrfs_root *root; 108049d11beaSJosef Bacik struct btrfs_path *path; 108149d11beaSJosef Bacik 108249d11beaSJosef Bacik path = btrfs_alloc_path(); 108349d11beaSJosef Bacik if (!path) 108449d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 108549d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 108649d11beaSJosef Bacik btrfs_free_path(path); 108749d11beaSJosef Bacik 108849d11beaSJosef Bacik return root; 1089cb517eabSMiao Xie } 1090cb517eabSMiao Xie 10912dfb1e43SQu Wenruo /* 10922dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 10932dfb1e43SQu Wenruo * 10942dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 10952dfb1e43SQu Wenruo */ 10962dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1097cb517eabSMiao Xie { 1098cb517eabSMiao Xie int ret; 1099cb517eabSMiao Xie 11000b548539SDavid Sterba btrfs_drew_lock_init(&root->snapshot_lock); 11018257b2dcSMiao Xie 1102aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 11036ebcd021SQu Wenruo !btrfs_is_data_reloc_root(root) && 11046ebcd021SQu Wenruo is_fstree(root->root_key.objectid)) { 110592a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1106f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1107f39e4571SJosef Bacik } 1108f39e4571SJosef Bacik 1109851fd730SQu Wenruo /* 1110851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1111851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1112851fd730SQu Wenruo */ 1113851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1114851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 11152dfb1e43SQu Wenruo if (!anon_dev) { 1116cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1117cb517eabSMiao Xie if (ret) 1118876d2cf1SLiu Bo goto fail; 11192dfb1e43SQu Wenruo } else { 11202dfb1e43SQu Wenruo root->anon_dev = anon_dev; 11212dfb1e43SQu Wenruo } 1122851fd730SQu Wenruo } 1123f32e48e9SChandan Rajendra 1124f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1125453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1126f32e48e9SChandan Rajendra if (ret) { 1127f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1128876d2cf1SLiu Bo goto fail; 1129f32e48e9SChandan Rajendra } 1130f32e48e9SChandan Rajendra 11316b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1132f32e48e9SChandan Rajendra 1133f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1134f32e48e9SChandan Rajendra 1135cb517eabSMiao Xie return 0; 1136cb517eabSMiao Xie fail: 113784db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1138cb517eabSMiao Xie return ret; 1139cb517eabSMiao Xie } 1140cb517eabSMiao Xie 1141a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1142cb517eabSMiao Xie u64 root_id) 1143cb517eabSMiao Xie { 1144cb517eabSMiao Xie struct btrfs_root *root; 1145cb517eabSMiao Xie 1146fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1147fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1148fc7cbcd4SDavid Sterba (unsigned long)root_id); 114900246528SJosef Bacik root = btrfs_grab_root(root); 1150fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1151cb517eabSMiao Xie return root; 1152cb517eabSMiao Xie } 1153cb517eabSMiao Xie 115449d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 115549d11beaSJosef Bacik u64 objectid) 115649d11beaSJosef Bacik { 1157abed4aaaSJosef Bacik struct btrfs_key key = { 1158abed4aaaSJosef Bacik .objectid = objectid, 1159abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1160abed4aaaSJosef Bacik .offset = 0, 1161abed4aaaSJosef Bacik }; 1162abed4aaaSJosef Bacik 1163e91909aaSChristoph Hellwig switch (objectid) { 1164e91909aaSChristoph Hellwig case BTRFS_ROOT_TREE_OBJECTID: 116549d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 1166e91909aaSChristoph Hellwig case BTRFS_EXTENT_TREE_OBJECTID: 1167abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1168e91909aaSChristoph Hellwig case BTRFS_CHUNK_TREE_OBJECTID: 116949d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 1170e91909aaSChristoph Hellwig case BTRFS_DEV_TREE_OBJECTID: 117149d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 1172e91909aaSChristoph Hellwig case BTRFS_CSUM_TREE_OBJECTID: 1173abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1174e91909aaSChristoph Hellwig case BTRFS_QUOTA_TREE_OBJECTID: 117585724171SChristoph Hellwig return btrfs_grab_root(fs_info->quota_root); 1176e91909aaSChristoph Hellwig case BTRFS_UUID_TREE_OBJECTID: 117785724171SChristoph Hellwig return btrfs_grab_root(fs_info->uuid_root); 1178e91909aaSChristoph Hellwig case BTRFS_BLOCK_GROUP_TREE_OBJECTID: 117985724171SChristoph Hellwig return btrfs_grab_root(fs_info->block_group_root); 1180e91909aaSChristoph Hellwig case BTRFS_FREE_SPACE_TREE_OBJECTID: 118185724171SChristoph Hellwig return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1182e91909aaSChristoph Hellwig default: 118349d11beaSJosef Bacik return NULL; 118449d11beaSJosef Bacik } 1185e91909aaSChristoph Hellwig } 118649d11beaSJosef Bacik 1187cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1188cb517eabSMiao Xie struct btrfs_root *root) 1189cb517eabSMiao Xie { 1190cb517eabSMiao Xie int ret; 1191cb517eabSMiao Xie 1192fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1193fc7cbcd4SDavid Sterba if (ret) 1194fc7cbcd4SDavid Sterba return ret; 1195fc7cbcd4SDavid Sterba 1196fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1197fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1198fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1199fc7cbcd4SDavid Sterba root); 1200af01d2e5SJosef Bacik if (ret == 0) { 120100246528SJosef Bacik btrfs_grab_root(root); 1202fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1203af01d2e5SJosef Bacik } 1204fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1205fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1206cb517eabSMiao Xie 1207cb517eabSMiao Xie return ret; 1208cb517eabSMiao Xie } 1209cb517eabSMiao Xie 1210bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1211bd647ce3SJosef Bacik { 1212bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1213bd647ce3SJosef Bacik struct btrfs_root *root; 1214bd647ce3SJosef Bacik 1215bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1216457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1217457f1864SJosef Bacik 1218bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1219bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1220457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 122171008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1222bd647ce3SJosef Bacik refcount_read(&root->refs)); 1223bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 122400246528SJosef Bacik btrfs_put_root(root); 122500246528SJosef Bacik btrfs_put_root(root); 1226bd647ce3SJosef Bacik } 1227bd647ce3SJosef Bacik #endif 1228bd647ce3SJosef Bacik } 1229bd647ce3SJosef Bacik 1230abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1231abed4aaaSJosef Bacik { 1232abed4aaaSJosef Bacik struct btrfs_root *root; 1233abed4aaaSJosef Bacik struct rb_node *node; 1234abed4aaaSJosef Bacik 1235abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1236abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1237abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1238abed4aaaSJosef Bacik btrfs_put_root(root); 1239abed4aaaSJosef Bacik } 1240abed4aaaSJosef Bacik } 1241abed4aaaSJosef Bacik 12420d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 12430d4b0463SJosef Bacik { 1244141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1245141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 12465deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1247141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1248141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1249141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1250141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 12510d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 12520d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1253abed4aaaSJosef Bacik free_global_roots(fs_info); 125400246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 125500246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 125600246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 125700246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 125800246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 125900246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1260aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 12619c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1262bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 12633fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 12640d4b0463SJosef Bacik kfree(fs_info->super_copy); 12650d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 12668481dd80SQu Wenruo kfree(fs_info->subpage_info); 12670d4b0463SJosef Bacik kvfree(fs_info); 12680d4b0463SJosef Bacik } 12690d4b0463SJosef Bacik 12700d4b0463SJosef Bacik 12712dfb1e43SQu Wenruo /* 12722dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 12732dfb1e43SQu Wenruo * 12742dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 12752dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 12762dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 12772dfb1e43SQu Wenruo * 12782dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 12792dfb1e43SQu Wenruo * 12802dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 12812dfb1e43SQu Wenruo * same root objectid. 12822dfb1e43SQu Wenruo * 12832dfb1e43SQu Wenruo * @objectid: root id 12842dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 12852dfb1e43SQu Wenruo * pass 0 for new allocation. 12862dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 12872dfb1e43SQu Wenruo * for orphan roots 12882dfb1e43SQu Wenruo */ 12892dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 12902dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 12912dfb1e43SQu Wenruo bool check_ref) 12925eda7b5eSChris Mason { 12935eda7b5eSChris Mason struct btrfs_root *root; 1294381cf658SDavid Sterba struct btrfs_path *path; 12951d4c08e0SDavid Sterba struct btrfs_key key; 12965eda7b5eSChris Mason int ret; 12975eda7b5eSChris Mason 129849d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 129949d11beaSJosef Bacik if (root) 130049d11beaSJosef Bacik return root; 1301773e722aSQu Wenruo 1302773e722aSQu Wenruo /* 1303773e722aSQu Wenruo * If we're called for non-subvolume trees, and above function didn't 1304773e722aSQu Wenruo * find one, do not try to read it from disk. 1305773e722aSQu Wenruo * 1306773e722aSQu Wenruo * This is namely for free-space-tree and quota tree, which can change 1307773e722aSQu Wenruo * at runtime and should only be grabbed from fs_info. 1308773e722aSQu Wenruo */ 1309773e722aSQu Wenruo if (!is_fstree(objectid) && objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) 1310773e722aSQu Wenruo return ERR_PTR(-ENOENT); 13114df27c4dSYan, Zheng again: 131256e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 131348475471SStefan Behrens if (root) { 1314*83377565SQu Wenruo /* 1315*83377565SQu Wenruo * Some other caller may have read out the newly inserted 1316*83377565SQu Wenruo * subvolume already (for things like backref walk etc). Not 1317*83377565SQu Wenruo * that common but still possible. In that case, we just need 1318*83377565SQu Wenruo * to free the anon_dev. 1319*83377565SQu Wenruo */ 1320*83377565SQu Wenruo if (unlikely(anon_dev)) { 1321*83377565SQu Wenruo free_anon_bdev(anon_dev); 1322*83377565SQu Wenruo anon_dev = 0; 1323*83377565SQu Wenruo } 1324*83377565SQu Wenruo 1325bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 132600246528SJosef Bacik btrfs_put_root(root); 132748475471SStefan Behrens return ERR_PTR(-ENOENT); 1328bc44d7c4SJosef Bacik } 13295eda7b5eSChris Mason return root; 133048475471SStefan Behrens } 13315eda7b5eSChris Mason 133256e9357aSDavid Sterba key.objectid = objectid; 133356e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 133456e9357aSDavid Sterba key.offset = (u64)-1; 133556e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 13365eda7b5eSChris Mason if (IS_ERR(root)) 13375eda7b5eSChris Mason return root; 13383394e160SChris Mason 1339c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1340d68fc57bSYan, Zheng ret = -ENOENT; 1341d68fc57bSYan, Zheng goto fail; 1342d68fc57bSYan, Zheng } 1343d68fc57bSYan, Zheng 13442dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1345cb517eabSMiao Xie if (ret) 1346cb517eabSMiao Xie goto fail; 1347cb517eabSMiao Xie 1348381cf658SDavid Sterba path = btrfs_alloc_path(); 1349381cf658SDavid Sterba if (!path) { 1350381cf658SDavid Sterba ret = -ENOMEM; 1351381cf658SDavid Sterba goto fail; 1352381cf658SDavid Sterba } 13531d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 13541d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 135556e9357aSDavid Sterba key.offset = objectid; 13561d4c08e0SDavid Sterba 13571d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1358381cf658SDavid Sterba btrfs_free_path(path); 1359d68fc57bSYan, Zheng if (ret < 0) 1360d68fc57bSYan, Zheng goto fail; 1361d68fc57bSYan, Zheng if (ret == 0) 136227cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1363d68fc57bSYan, Zheng 1364cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 13650f7d52f4SChris Mason if (ret) { 1366168a2f77SJia-Ju Bai if (ret == -EEXIST) { 136700246528SJosef Bacik btrfs_put_root(root); 13684df27c4dSYan, Zheng goto again; 1369168a2f77SJia-Ju Bai } 13704df27c4dSYan, Zheng goto fail; 13714df27c4dSYan, Zheng } 1372edbd8d4eSChris Mason return root; 13734df27c4dSYan, Zheng fail: 137433fab972SFilipe Manana /* 137533fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1376143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 137733fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 137833fab972SFilipe Manana * and once again by our caller. 137933fab972SFilipe Manana */ 138033fab972SFilipe Manana if (anon_dev) 138133fab972SFilipe Manana root->anon_dev = 0; 13828c38938cSJosef Bacik btrfs_put_root(root); 13834df27c4dSYan, Zheng return ERR_PTR(ret); 1384edbd8d4eSChris Mason } 1385edbd8d4eSChris Mason 13862dfb1e43SQu Wenruo /* 13872dfb1e43SQu Wenruo * Get in-memory reference of a root structure 13882dfb1e43SQu Wenruo * 13892dfb1e43SQu Wenruo * @objectid: tree objectid 13902dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 13912dfb1e43SQu Wenruo * one reference 13922dfb1e43SQu Wenruo */ 13932dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 13942dfb1e43SQu Wenruo u64 objectid, bool check_ref) 13952dfb1e43SQu Wenruo { 13962dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 13972dfb1e43SQu Wenruo } 13982dfb1e43SQu Wenruo 13992dfb1e43SQu Wenruo /* 14002dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 14012dfb1e43SQu Wenruo * the anonymous block device id 14022dfb1e43SQu Wenruo * 14032dfb1e43SQu Wenruo * @objectid: tree objectid 14042dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 14052dfb1e43SQu Wenruo * parameter value 14062dfb1e43SQu Wenruo */ 14072dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 14082dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 14092dfb1e43SQu Wenruo { 14102dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 14112dfb1e43SQu Wenruo } 14122dfb1e43SQu Wenruo 14138b712842SChris Mason /* 141449d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 141549d11beaSJosef Bacik * @fs_info: the fs_info 141649d11beaSJosef Bacik * @objectid: the objectid we need to lookup 141749d11beaSJosef Bacik * 141849d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 141949d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 142049d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 142149d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 142249d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 142349d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 142449d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 142549d11beaSJosef Bacik * backref code is finished using the root. 142649d11beaSJosef Bacik */ 142749d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 142849d11beaSJosef Bacik struct btrfs_path *path, 142949d11beaSJosef Bacik u64 objectid) 143049d11beaSJosef Bacik { 143149d11beaSJosef Bacik struct btrfs_root *root; 143249d11beaSJosef Bacik struct btrfs_key key; 143349d11beaSJosef Bacik 143449d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 143549d11beaSJosef Bacik 143649d11beaSJosef Bacik /* 143749d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 143849d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 143949d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 144049d11beaSJosef Bacik * != NULL just return it. 144149d11beaSJosef Bacik */ 144249d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 144349d11beaSJosef Bacik if (root) 144449d11beaSJosef Bacik return root; 144549d11beaSJosef Bacik 144649d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 144749d11beaSJosef Bacik if (root) 144849d11beaSJosef Bacik return root; 144949d11beaSJosef Bacik 145049d11beaSJosef Bacik key.objectid = objectid; 145149d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 145249d11beaSJosef Bacik key.offset = (u64)-1; 145349d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 145449d11beaSJosef Bacik btrfs_release_path(path); 145549d11beaSJosef Bacik 145649d11beaSJosef Bacik return root; 145749d11beaSJosef Bacik } 145849d11beaSJosef Bacik 1459a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1460a74a4b97SChris Mason { 14610d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1462d0278245SMiao Xie int again; 1463a74a4b97SChris Mason 1464d6fd0ae2SOmar Sandoval while (1) { 1465d0278245SMiao Xie again = 0; 14669d1a2a3aSDavid Sterba 1467fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1468fd340d0fSJosef Bacik 1469d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 14702ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1471d0278245SMiao Xie goto sleep; 1472d0278245SMiao Xie 147390c711abSZygo Blaxell /* 147490c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 147590c711abSZygo Blaxell * before we have finished mounting the filesystem. 147690c711abSZygo Blaxell */ 14770b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 147890c711abSZygo Blaxell goto sleep; 147990c711abSZygo Blaxell 14800b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1481d0278245SMiao Xie goto sleep; 1482d0278245SMiao Xie 1483dc7f370cSMiao Xie /* 1484dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1485dc7f370cSMiao Xie * during the above check and trylock. 1486dc7f370cSMiao Xie */ 14872ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 14880b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1489dc7f370cSMiao Xie goto sleep; 1490dc7f370cSMiao Xie } 1491dc7f370cSMiao Xie 1492b7625f46SQu Wenruo if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) 1493b7625f46SQu Wenruo btrfs_sysfs_feature_update(fs_info); 1494b7625f46SQu Wenruo 14952ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1496c2d6cb16SFilipe Manana 149733c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 14980b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1499a74a4b97SChris Mason 1500d0278245SMiao Xie /* 150105323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 150205323cd1SMiao Xie * needn't do anything special here. 1503d0278245SMiao Xie */ 15040b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 150567c5e7d4SFilipe Manana 150667c5e7d4SFilipe Manana /* 1507f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 150867c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 150967c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1510f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 151167c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 151267c5e7d4SFilipe Manana * unused block groups. 151367c5e7d4SFilipe Manana */ 15140b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 151518bb8bbfSJohannes Thumshirn 151618bb8bbfSJohannes Thumshirn /* 151718bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 151818bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 151918bb8bbfSJohannes Thumshirn * space. 152018bb8bbfSJohannes Thumshirn */ 152118bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1522d0278245SMiao Xie sleep: 1523a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1524d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1525d6fd0ae2SOmar Sandoval kthread_parkme(); 1526d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1527d6fd0ae2SOmar Sandoval return 0; 1528838fe188SJiri Kosina if (!again) { 1529a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1530a74a4b97SChris Mason schedule(); 1531a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1532a74a4b97SChris Mason } 1533da288d28SFilipe Manana } 1534a74a4b97SChris Mason } 1535a74a4b97SChris Mason 1536a74a4b97SChris Mason static int transaction_kthread(void *arg) 1537a74a4b97SChris Mason { 1538a74a4b97SChris Mason struct btrfs_root *root = arg; 15390b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1540a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1541a74a4b97SChris Mason struct btrfs_transaction *cur; 15428929ecfaSYan, Zheng u64 transid; 1543643900beSNikolay Borisov time64_t delta; 1544a74a4b97SChris Mason unsigned long delay; 1545914b2007SJan Kara bool cannot_commit; 1546a74a4b97SChris Mason 1547a74a4b97SChris Mason do { 1548914b2007SJan Kara cannot_commit = false; 1549ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 15500b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1551a74a4b97SChris Mason 15520b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 15530b246afaSJeff Mahoney cur = fs_info->running_transaction; 1554a74a4b97SChris Mason if (!cur) { 15550b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1556a74a4b97SChris Mason goto sleep; 1557a74a4b97SChris Mason } 155831153d81SYan Zheng 1559643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1560fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 156177d20c68SJosef Bacik cur->state < TRANS_STATE_COMMIT_PREP && 1562643900beSNikolay Borisov delta < fs_info->commit_interval) { 15630b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1564fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1565fb8a7e94SNikolay Borisov delay = min(delay, 1566fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1567a74a4b97SChris Mason goto sleep; 1568a74a4b97SChris Mason } 15698929ecfaSYan, Zheng transid = cur->transid; 15700b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 157156bec294SChris Mason 157279787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1573354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1574914b2007SJan Kara if (IS_ERR(trans)) { 1575354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1576914b2007SJan Kara cannot_commit = true; 157779787eaaSJeff Mahoney goto sleep; 1578914b2007SJan Kara } 15798929ecfaSYan, Zheng if (transid == trans->transid) { 15803a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 15818929ecfaSYan, Zheng } else { 15823a45bb20SJeff Mahoney btrfs_end_transaction(trans); 15838929ecfaSYan, Zheng } 1584a74a4b97SChris Mason sleep: 15850b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 15860b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1587a74a4b97SChris Mason 158884961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 15892ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 15908929ecfaSYan, Zheng if (!kthread_should_stop() && 15910b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1592914b2007SJan Kara cannot_commit)) 1593bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1594a74a4b97SChris Mason } while (!kthread_should_stop()); 1595a74a4b97SChris Mason return 0; 1596a74a4b97SChris Mason } 1597a74a4b97SChris Mason 1598af31f5e5SChris Mason /* 159901f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 160001f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 160101f0f9daSNikolay Borisov * anything. 1602af31f5e5SChris Mason * 1603af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1604af31f5e5SChris Mason * generation of the latest root in the array with the generation 1605af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1606af31f5e5SChris Mason */ 160701f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1608af31f5e5SChris Mason { 160901f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1610af31f5e5SChris Mason u64 cur; 1611af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1612af31f5e5SChris Mason int i; 1613af31f5e5SChris Mason 1614af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1615af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1616af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1617af31f5e5SChris Mason if (cur == newest_gen) 161801f0f9daSNikolay Borisov return i; 1619af31f5e5SChris Mason } 1620af31f5e5SChris Mason 162101f0f9daSNikolay Borisov return -EINVAL; 1622af31f5e5SChris Mason } 1623af31f5e5SChris Mason 1624af31f5e5SChris Mason /* 1625af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1626af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1627af31f5e5SChris Mason * done 1628af31f5e5SChris Mason */ 1629af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1630af31f5e5SChris Mason { 16316ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1632af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1633af31f5e5SChris Mason 1634af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1635af31f5e5SChris Mason 1636af31f5e5SChris Mason /* 1637af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1638af31f5e5SChris Mason * regardless of which ones we use today 1639af31f5e5SChris Mason */ 1640af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1641af31f5e5SChris Mason 1642af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1643af31f5e5SChris Mason 1644af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1645af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1646af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1647af31f5e5SChris Mason 1648af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1649af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1650af31f5e5SChris Mason 1651af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1652af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1653af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1654af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1655af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1656af31f5e5SChris Mason 16571c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 16589c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1659f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 16609c54e80dSJosef Bacik 16619c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 16629c54e80dSJosef Bacik extent_root->node->start); 1663af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 166429cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 1665af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 166629cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 1667af31f5e5SChris Mason 1668f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 1669f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 1670f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 1671f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 1672f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 16739c54e80dSJosef Bacik } 1674af31f5e5SChris Mason 16757c7e82a7SChris Mason /* 16767c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 16777c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 16787c7e82a7SChris Mason */ 16797c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 16807c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 16817c7e82a7SChris Mason info->fs_root->node->start); 1682af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1683af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1684af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1685af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 16867c7e82a7SChris Mason } 1687af31f5e5SChris Mason 1688af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1689af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1690af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1691af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1692af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1693af31f5e5SChris Mason 1694af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1695af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1696af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1697af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1698af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1699af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1700af31f5e5SChris Mason 1701af31f5e5SChris Mason /* 1702af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1703af31f5e5SChris Mason * for the next commit 1704af31f5e5SChris Mason */ 1705af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1706af31f5e5SChris Mason &info->super_for_commit->super_roots, 1707af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1708af31f5e5SChris Mason } 1709af31f5e5SChris Mason 1710af31f5e5SChris Mason /* 1711bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1712bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1713bd2336b2SNikolay Borisov * 1714bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1715bd2336b2SNikolay Borisov * priority - priority of backup root required 1716bd2336b2SNikolay Borisov * 1717bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1718bd2336b2SNikolay Borisov */ 1719bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1720bd2336b2SNikolay Borisov { 1721bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1722bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1723bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1724bd2336b2SNikolay Borisov 1725bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1726bd2336b2SNikolay Borisov if (priority == 0) 1727bd2336b2SNikolay Borisov return backup_index; 1728bd2336b2SNikolay Borisov 1729bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1730bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1731bd2336b2SNikolay Borisov } else { 1732bd2336b2SNikolay Borisov return -EINVAL; 1733bd2336b2SNikolay Borisov } 1734bd2336b2SNikolay Borisov 1735bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1736bd2336b2SNikolay Borisov 1737bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1738bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1739bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1740bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1741bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1742bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1743bd2336b2SNikolay Borisov 1744bd2336b2SNikolay Borisov /* 1745bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1746bd2336b2SNikolay Borisov * need a fsck 1747bd2336b2SNikolay Borisov */ 1748bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1749bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1750bd2336b2SNikolay Borisov 1751bd2336b2SNikolay Borisov return backup_index; 1752bd2336b2SNikolay Borisov } 1753bd2336b2SNikolay Borisov 17547abadb64SLiu Bo /* helper to cleanup workers */ 17557abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 17567abadb64SLiu Bo { 1757dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1758afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 17595cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1760d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 1761d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 1762385de0efSChristoph Hellwig if (fs_info->rmw_workers) 1763385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 1764fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 1765fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 1766fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1767fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 17685b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1769e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1770a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1771fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1772b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1773b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1774a9b9477dSFilipe Manana /* 1775a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 1776a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 1777a9b9477dSFilipe Manana * queues can do metadata I/O operations. 1778a9b9477dSFilipe Manana */ 1779d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 1780d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 17817abadb64SLiu Bo } 17827abadb64SLiu Bo 17832e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 17842e9f5954SRashika { 17852e9f5954SRashika if (root) { 17862e9f5954SRashika free_extent_buffer(root->node); 17872e9f5954SRashika free_extent_buffer(root->commit_root); 17882e9f5954SRashika root->node = NULL; 17892e9f5954SRashika root->commit_root = NULL; 17902e9f5954SRashika } 17912e9f5954SRashika } 17922e9f5954SRashika 1793abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 1794abed4aaaSJosef Bacik { 1795abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 1796abed4aaaSJosef Bacik 1797abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 1798abed4aaaSJosef Bacik &fs_info->global_root_tree, 1799abed4aaaSJosef Bacik rb_node) 1800abed4aaaSJosef Bacik free_root_extent_buffers(root); 1801abed4aaaSJosef Bacik } 1802abed4aaaSJosef Bacik 1803af31f5e5SChris Mason /* helper to cleanup tree roots */ 18044273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 1805af31f5e5SChris Mason { 18062e9f5954SRashika free_root_extent_buffers(info->tree_root); 1807655b09feSJosef Bacik 1808abed4aaaSJosef Bacik free_global_root_pointers(info); 18092e9f5954SRashika free_root_extent_buffers(info->dev_root); 18102e9f5954SRashika free_root_extent_buffers(info->quota_root); 18112e9f5954SRashika free_root_extent_buffers(info->uuid_root); 18128c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 1813aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 18149c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 18154273eaffSAnand Jain if (free_chunk_root) 18162e9f5954SRashika free_root_extent_buffers(info->chunk_root); 1817af31f5e5SChris Mason } 1818af31f5e5SChris Mason 18198c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 18208c38938cSJosef Bacik { 18218c38938cSJosef Bacik if (!root) 18228c38938cSJosef Bacik return; 18238c38938cSJosef Bacik 18248c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 18258c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 18261dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 18278c38938cSJosef Bacik if (root->anon_dev) 18288c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 1829923eb523SJohannes Thumshirn free_root_extent_buffers(root); 18308c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1831fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 18328c38938cSJosef Bacik list_del_init(&root->leak_list); 1833fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 18348c38938cSJosef Bacik #endif 18358c38938cSJosef Bacik kfree(root); 18368c38938cSJosef Bacik } 18378c38938cSJosef Bacik } 18388c38938cSJosef Bacik 1839faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 1840171f6537SJosef Bacik { 1841fc7cbcd4SDavid Sterba int ret; 1842fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 1843fc7cbcd4SDavid Sterba int i; 1844171f6537SJosef Bacik 1845171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 1846fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 1847171f6537SJosef Bacik struct btrfs_root, root_list); 1848fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 1849171f6537SJosef Bacik 1850fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 1851fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 1852fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 1853171f6537SJosef Bacik } 1854171f6537SJosef Bacik 1855fc7cbcd4SDavid Sterba while (1) { 1856fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 1857fc7cbcd4SDavid Sterba (void **)gang, 0, 1858fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 1859fc7cbcd4SDavid Sterba if (!ret) 1860fc7cbcd4SDavid Sterba break; 1861fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 1862fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 1863171f6537SJosef Bacik } 1864171f6537SJosef Bacik } 1865af31f5e5SChris Mason 1866638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 1867638aa7edSEric Sandeen { 1868638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 1869638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 1870638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 1871638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 1872638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 1873638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 1874ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 1875638aa7edSEric Sandeen } 1876638aa7edSEric Sandeen 1877779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 1878779a65a4SEric Sandeen { 1879779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 1880779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 1881779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 1882779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 1883779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 1884779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 1885907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 1886779a65a4SEric Sandeen } 1887779a65a4SEric Sandeen 1888dcb2137cSChristoph Hellwig static int btrfs_init_btree_inode(struct super_block *sb) 1889f37938e0SEric Sandeen { 1890dcb2137cSChristoph Hellwig struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1891e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 1892e256927bSJosef Bacik fs_info->tree_root); 1893dcb2137cSChristoph Hellwig struct inode *inode; 1894dcb2137cSChristoph Hellwig 1895dcb2137cSChristoph Hellwig inode = new_inode(sb); 1896dcb2137cSChristoph Hellwig if (!inode) 1897dcb2137cSChristoph Hellwig return -ENOMEM; 18982ff7e61eSJeff Mahoney 18992ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 19002ff7e61eSJeff Mahoney set_nlink(inode, 1); 1901f37938e0SEric Sandeen /* 1902f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 1903f37938e0SEric Sandeen * the real end of the address space is determined by all of 1904f37938e0SEric Sandeen * the devices in the system 1905f37938e0SEric Sandeen */ 19062ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 19072ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 1908dcb2137cSChristoph Hellwig mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); 1909f37938e0SEric Sandeen 19102ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 191143eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 191235da5a7eSDavid Sterba IO_TREE_BTREE_INODE_IO); 19132ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 1914f37938e0SEric Sandeen 19155c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 1916adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 1917adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 1918adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 19192ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 1920e256927bSJosef Bacik __insert_inode_hash(inode, hash); 1921dcb2137cSChristoph Hellwig fs_info->btree_inode = inode; 1922dcb2137cSChristoph Hellwig 1923dcb2137cSChristoph Hellwig return 0; 1924f37938e0SEric Sandeen } 1925f37938e0SEric Sandeen 1926ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 1927ad618368SEric Sandeen { 1928ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 1929129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 19307f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 1931ad618368SEric Sandeen } 1932ad618368SEric Sandeen 1933f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 1934f9e92e40SEric Sandeen { 1935f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 1936f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 1937f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 1938f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 1939f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 1940f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 1941d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 1942011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 1943f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 1944f9e92e40SEric Sandeen } 1945f9e92e40SEric Sandeen 1946d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 19472a458198SEric Sandeen { 1948f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 19496f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 195058e814fcSTejun Heo unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; 19512a458198SEric Sandeen 19522a458198SEric Sandeen fs_info->workers = 1953a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 19542a458198SEric Sandeen 19552a458198SEric Sandeen fs_info->delalloc_workers = 1956cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 1957cb001095SJeff Mahoney flags, max_active, 2); 19582a458198SEric Sandeen 19592a458198SEric Sandeen fs_info->flush_workers = 1960cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 1961cb001095SJeff Mahoney flags, max_active, 0); 19622a458198SEric Sandeen 19632a458198SEric Sandeen fs_info->caching_workers = 1964cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 19652a458198SEric Sandeen 19662a458198SEric Sandeen fs_info->fixup_workers = 196758e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); 19682a458198SEric Sandeen 19692a458198SEric Sandeen fs_info->endio_workers = 1970d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 19712a458198SEric Sandeen fs_info->endio_meta_workers = 1972d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 1973385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 19742a458198SEric Sandeen fs_info->endio_write_workers = 1975cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 1976cb001095SJeff Mahoney max_active, 2); 1977fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 1978fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 19792a458198SEric Sandeen fs_info->endio_freespace_worker = 1980cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 1981cb001095SJeff Mahoney max_active, 0); 19822a458198SEric Sandeen fs_info->delayed_workers = 1983cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 1984cb001095SJeff Mahoney max_active, 0); 19852a458198SEric Sandeen fs_info->qgroup_rescan_workers = 198658e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", 198758e814fcSTejun Heo ordered_flags); 1988b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 198958e814fcSTejun Heo alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); 19902a458198SEric Sandeen 19918bfec2e4SChristoph Hellwig if (!(fs_info->workers && 1992a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 19932a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 1994fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 19951a1a2851SQu Wenruo fs_info->endio_write_workers && 19962a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 1997f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 1998f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 1999b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 20002a458198SEric Sandeen return -ENOMEM; 20012a458198SEric Sandeen } 20022a458198SEric Sandeen 20032a458198SEric Sandeen return 0; 20042a458198SEric Sandeen } 20052a458198SEric Sandeen 20066d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 20076d97c6e3SJohannes Thumshirn { 20086d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2009b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 20106d97c6e3SJohannes Thumshirn 2011b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 20126d97c6e3SJohannes Thumshirn 20136d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 20146d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2015b4e967beSDavid Sterba csum_driver); 20166d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 20176d97c6e3SJohannes Thumshirn } 20186d97c6e3SJohannes Thumshirn 20196d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 20206d97c6e3SJohannes Thumshirn 202168d99ab0SChristoph Hellwig /* 202268d99ab0SChristoph Hellwig * Check if the checksum implementation is a fast accelerated one. 202368d99ab0SChristoph Hellwig * As-is this is a bit of a hack and should be replaced once the csum 202468d99ab0SChristoph Hellwig * implementations provide that information themselves. 202568d99ab0SChristoph Hellwig */ 202668d99ab0SChristoph Hellwig switch (csum_type) { 202768d99ab0SChristoph Hellwig case BTRFS_CSUM_TYPE_CRC32: 202868d99ab0SChristoph Hellwig if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) 202968d99ab0SChristoph Hellwig set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 203068d99ab0SChristoph Hellwig break; 2031efcfcbc6SDavid Sterba case BTRFS_CSUM_TYPE_XXHASH: 2032efcfcbc6SDavid Sterba set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 2033efcfcbc6SDavid Sterba break; 203468d99ab0SChristoph Hellwig default: 203568d99ab0SChristoph Hellwig break; 203668d99ab0SChristoph Hellwig } 203768d99ab0SChristoph Hellwig 2038c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2039c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2040c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 20416d97c6e3SJohannes Thumshirn return 0; 20426d97c6e3SJohannes Thumshirn } 20436d97c6e3SJohannes Thumshirn 204463443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 204563443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 204663443bf5SEric Sandeen { 204763443bf5SEric Sandeen int ret; 2048789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 204963443bf5SEric Sandeen struct btrfs_root *log_tree_root; 205063443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 205163443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2052581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 205363443bf5SEric Sandeen 205463443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2055f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 205663443bf5SEric Sandeen return -EIO; 205763443bf5SEric Sandeen } 205863443bf5SEric Sandeen 205996dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 206096dfcb46SJosef Bacik GFP_KERNEL); 206163443bf5SEric Sandeen if (!log_tree_root) 206263443bf5SEric Sandeen return -ENOMEM; 206363443bf5SEric Sandeen 2064789d6a3aSQu Wenruo check.level = level; 2065789d6a3aSQu Wenruo check.transid = fs_info->generation + 1; 2066789d6a3aSQu Wenruo check.owner_root = BTRFS_TREE_LOG_OBJECTID; 2067789d6a3aSQu Wenruo log_tree_root->node = read_tree_block(fs_info, bytenr, &check); 206864c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2069f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 20700eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 20718c38938cSJosef Bacik log_tree_root->node = NULL; 207200246528SJosef Bacik btrfs_put_root(log_tree_root); 20730eeff236SLiu Bo return ret; 20744eb150d6SQu Wenruo } 20754eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2076f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 207700246528SJosef Bacik btrfs_put_root(log_tree_root); 207863443bf5SEric Sandeen return -EIO; 207963443bf5SEric Sandeen } 20804eb150d6SQu Wenruo 208163443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 208263443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 208363443bf5SEric Sandeen if (ret) { 20840b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 208563443bf5SEric Sandeen "Failed to recover log tree"); 208600246528SJosef Bacik btrfs_put_root(log_tree_root); 208763443bf5SEric Sandeen return ret; 208863443bf5SEric Sandeen } 208963443bf5SEric Sandeen 2090bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 20916bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 209263443bf5SEric Sandeen if (ret) 209363443bf5SEric Sandeen return ret; 209463443bf5SEric Sandeen } 209563443bf5SEric Sandeen 209663443bf5SEric Sandeen return 0; 209763443bf5SEric Sandeen } 209863443bf5SEric Sandeen 2099abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2100abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2101abed4aaaSJosef Bacik const char *name) 2102abed4aaaSJosef Bacik { 2103abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2104abed4aaaSJosef Bacik struct btrfs_root *root; 2105f7238e50SJosef Bacik u64 max_global_id = 0; 2106abed4aaaSJosef Bacik int ret; 2107abed4aaaSJosef Bacik struct btrfs_key key = { 2108abed4aaaSJosef Bacik .objectid = objectid, 2109abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2110abed4aaaSJosef Bacik .offset = 0, 2111abed4aaaSJosef Bacik }; 2112abed4aaaSJosef Bacik bool found = false; 2113abed4aaaSJosef Bacik 2114abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2115abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2116abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2117abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2118abed4aaaSJosef Bacik return 0; 2119abed4aaaSJosef Bacik } 2120abed4aaaSJosef Bacik 2121abed4aaaSJosef Bacik while (1) { 2122abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2123abed4aaaSJosef Bacik if (ret < 0) 2124abed4aaaSJosef Bacik break; 2125abed4aaaSJosef Bacik 2126abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2127abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2128abed4aaaSJosef Bacik if (ret) { 2129abed4aaaSJosef Bacik if (ret > 0) 2130abed4aaaSJosef Bacik ret = 0; 2131abed4aaaSJosef Bacik break; 2132abed4aaaSJosef Bacik } 2133abed4aaaSJosef Bacik } 2134abed4aaaSJosef Bacik ret = 0; 2135abed4aaaSJosef Bacik 2136abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2137abed4aaaSJosef Bacik if (key.objectid != objectid) 2138abed4aaaSJosef Bacik break; 2139abed4aaaSJosef Bacik btrfs_release_path(path); 2140abed4aaaSJosef Bacik 2141f7238e50SJosef Bacik /* 2142f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2143f7238e50SJosef Bacik * everybody. 2144f7238e50SJosef Bacik */ 2145f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2146f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2147f7238e50SJosef Bacik 2148abed4aaaSJosef Bacik found = true; 2149abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2150abed4aaaSJosef Bacik if (IS_ERR(root)) { 2151abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2152abed4aaaSJosef Bacik ret = PTR_ERR(root); 2153abed4aaaSJosef Bacik break; 2154abed4aaaSJosef Bacik } 2155abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2156abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2157abed4aaaSJosef Bacik if (ret) { 2158abed4aaaSJosef Bacik btrfs_put_root(root); 2159abed4aaaSJosef Bacik break; 2160abed4aaaSJosef Bacik } 2161abed4aaaSJosef Bacik key.offset++; 2162abed4aaaSJosef Bacik } 2163abed4aaaSJosef Bacik btrfs_release_path(path); 2164abed4aaaSJosef Bacik 2165f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2166f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2167f7238e50SJosef Bacik 2168abed4aaaSJosef Bacik if (!found || ret) { 2169abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2170abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2171abed4aaaSJosef Bacik 2172abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2173abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2174abed4aaaSJosef Bacik else 2175abed4aaaSJosef Bacik ret = 0; 2176abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2177abed4aaaSJosef Bacik } 2178abed4aaaSJosef Bacik return ret; 2179abed4aaaSJosef Bacik } 2180abed4aaaSJosef Bacik 2181abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2182abed4aaaSJosef Bacik { 2183abed4aaaSJosef Bacik struct btrfs_path *path; 2184abed4aaaSJosef Bacik int ret = 0; 2185abed4aaaSJosef Bacik 2186abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2187abed4aaaSJosef Bacik if (!path) 2188abed4aaaSJosef Bacik return -ENOMEM; 2189abed4aaaSJosef Bacik 2190abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2191abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2192abed4aaaSJosef Bacik if (ret) 2193abed4aaaSJosef Bacik goto out; 2194abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2195abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2196abed4aaaSJosef Bacik if (ret) 2197abed4aaaSJosef Bacik goto out; 2198abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2199abed4aaaSJosef Bacik goto out; 2200abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2201abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2202abed4aaaSJosef Bacik "free space"); 2203abed4aaaSJosef Bacik out: 2204abed4aaaSJosef Bacik btrfs_free_path(path); 2205abed4aaaSJosef Bacik return ret; 2206abed4aaaSJosef Bacik } 2207abed4aaaSJosef Bacik 22086bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 22094bbcaa64SEric Sandeen { 22106bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2211a4f3d2c4SDavid Sterba struct btrfs_root *root; 22124bbcaa64SEric Sandeen struct btrfs_key location; 22134bbcaa64SEric Sandeen int ret; 22144bbcaa64SEric Sandeen 22156bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 22166bccf3abSJeff Mahoney 2217abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2218abed4aaaSJosef Bacik if (ret) 2219abed4aaaSJosef Bacik return ret; 2220abed4aaaSJosef Bacik 22214bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 22224bbcaa64SEric Sandeen location.offset = 0; 22234bbcaa64SEric Sandeen 22241c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 222514033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 222614033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 222714033b08SQu Wenruo if (IS_ERR(root)) { 222814033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 222914033b08SQu Wenruo ret = PTR_ERR(root); 223014033b08SQu Wenruo goto out; 223114033b08SQu Wenruo } 223214033b08SQu Wenruo } else { 223314033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 223414033b08SQu Wenruo fs_info->block_group_root = root; 223514033b08SQu Wenruo } 223614033b08SQu Wenruo } 223714033b08SQu Wenruo 223814033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2239a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2240f50f4353SLiu Bo if (IS_ERR(root)) { 224142437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2242f50f4353SLiu Bo ret = PTR_ERR(root); 2243f50f4353SLiu Bo goto out; 2244f50f4353SLiu Bo } 224542437a63SJosef Bacik } else { 2246a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2247a4f3d2c4SDavid Sterba fs_info->dev_root = root; 224842437a63SJosef Bacik } 2249820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2250a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2251a8d1b164SJohannes Thumshirn if (ret) 2252a8d1b164SJohannes Thumshirn goto out; 22534bbcaa64SEric Sandeen 2254aeb935a4SQu Wenruo /* 2255aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2256aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2257aeb935a4SQu Wenruo */ 225856e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 225956e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2260aeb935a4SQu Wenruo if (IS_ERR(root)) { 226142437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2262aeb935a4SQu Wenruo ret = PTR_ERR(root); 2263aeb935a4SQu Wenruo goto out; 2264aeb935a4SQu Wenruo } 226542437a63SJosef Bacik } else { 2266aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2267aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 226842437a63SJosef Bacik } 2269aeb935a4SQu Wenruo 22704bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2271a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2272a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2273a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2274afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2275a4f3d2c4SDavid Sterba fs_info->quota_root = root; 22764bbcaa64SEric Sandeen } 22774bbcaa64SEric Sandeen 22784bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2279a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2280a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 228142437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2282a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 22834bbcaa64SEric Sandeen if (ret != -ENOENT) 2284f50f4353SLiu Bo goto out; 228542437a63SJosef Bacik } 22864bbcaa64SEric Sandeen } else { 2287a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2288a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 22894bbcaa64SEric Sandeen } 22904bbcaa64SEric Sandeen 22914bbcaa64SEric Sandeen return 0; 2292f50f4353SLiu Bo out: 2293f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2294f50f4353SLiu Bo location.objectid, ret); 2295f50f4353SLiu Bo return ret; 22964bbcaa64SEric Sandeen } 22974bbcaa64SEric Sandeen 2298069ec957SQu Wenruo /* 2299069ec957SQu Wenruo * Real super block validation 2300069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2301069ec957SQu Wenruo * 2302069ec957SQu Wenruo * @sb: super block to check 2303069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2304069ec957SQu Wenruo * 0 the primary (1st) sb 2305069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2306069ec957SQu Wenruo * -1 skip bytenr check 2307069ec957SQu Wenruo */ 2308a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2309069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 231021a852b0SQu Wenruo { 231121a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 231221a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 231321a852b0SQu Wenruo int ret = 0; 231421a852b0SQu Wenruo 231521a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 231621a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 231721a852b0SQu Wenruo ret = -EINVAL; 231821a852b0SQu Wenruo } 231921a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 232021a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 232121a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 232221a852b0SQu Wenruo ret = -EINVAL; 232321a852b0SQu Wenruo } 232421a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 232521a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 232621a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 232721a852b0SQu Wenruo ret = -EINVAL; 232821a852b0SQu Wenruo } 232921a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 233021a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 233121a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 233221a852b0SQu Wenruo ret = -EINVAL; 233321a852b0SQu Wenruo } 233421a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 233521a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 233621a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 233721a852b0SQu Wenruo ret = -EINVAL; 233821a852b0SQu Wenruo } 233921a852b0SQu Wenruo 234021a852b0SQu Wenruo /* 234121a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 234221a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 234321a852b0SQu Wenruo */ 234421a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 234521a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 234621a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 234721a852b0SQu Wenruo ret = -EINVAL; 234821a852b0SQu Wenruo } 23490bb3eb3eSQu Wenruo 23500bb3eb3eSQu Wenruo /* 23511a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 23521a42daabSQu Wenruo * 23531a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 23541a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 23551a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 23561a42daabSQu Wenruo * beginning. 23570bb3eb3eSQu Wenruo */ 23581a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 235921a852b0SQu Wenruo btrfs_err(fs_info, 23600bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 236121a852b0SQu Wenruo sectorsize, PAGE_SIZE); 236221a852b0SQu Wenruo ret = -EINVAL; 236321a852b0SQu Wenruo } 23640bb3eb3eSQu Wenruo 236521a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 236621a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 236721a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 236821a852b0SQu Wenruo ret = -EINVAL; 236921a852b0SQu Wenruo } 237021a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 237121a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 237221a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 237321a852b0SQu Wenruo ret = -EINVAL; 237421a852b0SQu Wenruo } 237521a852b0SQu Wenruo 237621a852b0SQu Wenruo /* Root alignment check */ 237721a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 237821a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 237921a852b0SQu Wenruo btrfs_super_root(sb)); 238021a852b0SQu Wenruo ret = -EINVAL; 238121a852b0SQu Wenruo } 238221a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 238321a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 238421a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 238521a852b0SQu Wenruo ret = -EINVAL; 238621a852b0SQu Wenruo } 238721a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 238821a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 238921a852b0SQu Wenruo btrfs_super_log_root(sb)); 239021a852b0SQu Wenruo ret = -EINVAL; 239121a852b0SQu Wenruo } 239221a852b0SQu Wenruo 2393d167aa76SAnand Jain if (memcmp(fs_info->fs_devices->fsid, sb->fsid, BTRFS_FSID_SIZE) != 0) { 2394aefd7f70SNikolay Borisov btrfs_err(fs_info, 2395aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2396d167aa76SAnand Jain sb->fsid, fs_info->fs_devices->fsid); 2397aefd7f70SNikolay Borisov ret = -EINVAL; 2398aefd7f70SNikolay Borisov } 2399aefd7f70SNikolay Borisov 24006bfe3959SAnand Jain if (memcmp(fs_info->fs_devices->metadata_uuid, btrfs_sb_fsid_ptr(sb), 24016bfe3959SAnand Jain BTRFS_FSID_SIZE) != 0) { 2402aefd7f70SNikolay Borisov btrfs_err(fs_info, 2403aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 24046bfe3959SAnand Jain btrfs_sb_fsid_ptr(sb), fs_info->fs_devices->metadata_uuid); 2405aefd7f70SNikolay Borisov ret = -EINVAL; 2406aefd7f70SNikolay Borisov } 2407aefd7f70SNikolay Borisov 240825984a5aSAnand Jain if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 240925984a5aSAnand Jain BTRFS_FSID_SIZE) != 0) { 241025984a5aSAnand Jain btrfs_err(fs_info, 241125984a5aSAnand Jain "dev_item UUID does not match metadata fsid: %pU != %pU", 241225984a5aSAnand Jain fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 241325984a5aSAnand Jain ret = -EINVAL; 241425984a5aSAnand Jain } 241525984a5aSAnand Jain 24161c56ab99SQu Wenruo /* 24171c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 24181c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 24191c56ab99SQu Wenruo */ 24201c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 24211c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 24221c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 24231c56ab99SQu Wenruo btrfs_err(fs_info, 24241c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 24251c56ab99SQu Wenruo ret = -EINVAL; 24261c56ab99SQu Wenruo } 24271c56ab99SQu Wenruo 242821a852b0SQu Wenruo /* 242921a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 243021a852b0SQu Wenruo * done later 243121a852b0SQu Wenruo */ 243221a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 243321a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 243421a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 243521a852b0SQu Wenruo ret = -EINVAL; 243621a852b0SQu Wenruo } 243721a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 243821a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 243921a852b0SQu Wenruo btrfs_super_stripesize(sb)); 244021a852b0SQu Wenruo ret = -EINVAL; 244121a852b0SQu Wenruo } 244221a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 244321a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 244421a852b0SQu Wenruo btrfs_super_num_devices(sb)); 244521a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 244621a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 244721a852b0SQu Wenruo ret = -EINVAL; 244821a852b0SQu Wenruo } 244921a852b0SQu Wenruo 2450069ec957SQu Wenruo if (mirror_num >= 0 && 2451069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 245221a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 245321a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 245421a852b0SQu Wenruo ret = -EINVAL; 245521a852b0SQu Wenruo } 245621a852b0SQu Wenruo 245721a852b0SQu Wenruo /* 245821a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 245921a852b0SQu Wenruo * and one chunk 246021a852b0SQu Wenruo */ 246121a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 246221a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 246321a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 246421a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 246521a852b0SQu Wenruo ret = -EINVAL; 246621a852b0SQu Wenruo } 246721a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 246821a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 246921a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 247021a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 247121a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 247221a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 247321a852b0SQu Wenruo ret = -EINVAL; 247421a852b0SQu Wenruo } 247521a852b0SQu Wenruo 247621a852b0SQu Wenruo /* 247721a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 247821a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 247921a852b0SQu Wenruo */ 248021a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 248121a852b0SQu Wenruo btrfs_warn(fs_info, 248221a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 248321a852b0SQu Wenruo btrfs_super_generation(sb), 248421a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 248521a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 248621a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 248721a852b0SQu Wenruo btrfs_warn(fs_info, 248821a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 248921a852b0SQu Wenruo btrfs_super_generation(sb), 249021a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 249121a852b0SQu Wenruo 249221a852b0SQu Wenruo return ret; 249321a852b0SQu Wenruo } 249421a852b0SQu Wenruo 2495069ec957SQu Wenruo /* 2496069ec957SQu Wenruo * Validation of super block at mount time. 2497069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2498069ec957SQu Wenruo * flags will be skipped. 2499069ec957SQu Wenruo */ 2500069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2501069ec957SQu Wenruo { 2502a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2503069ec957SQu Wenruo } 2504069ec957SQu Wenruo 250575cb857dSQu Wenruo /* 250675cb857dSQu Wenruo * Validation of super block at write time. 250775cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 250875cb857dSQu Wenruo * overwritten soon. 250975cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 251075cb857dSQu Wenruo */ 251175cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 251275cb857dSQu Wenruo struct btrfs_super_block *sb) 251375cb857dSQu Wenruo { 251475cb857dSQu Wenruo int ret; 251575cb857dSQu Wenruo 2516a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 251775cb857dSQu Wenruo if (ret < 0) 251875cb857dSQu Wenruo goto out; 2519e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 252075cb857dSQu Wenruo ret = -EUCLEAN; 252175cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 252275cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 252375cb857dSQu Wenruo goto out; 252475cb857dSQu Wenruo } 252575cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 252675cb857dSQu Wenruo ret = -EUCLEAN; 252775cb857dSQu Wenruo btrfs_err(fs_info, 252875cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 252975cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 253075cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 253175cb857dSQu Wenruo goto out; 253275cb857dSQu Wenruo } 253375cb857dSQu Wenruo out: 253475cb857dSQu Wenruo if (ret < 0) 253575cb857dSQu Wenruo btrfs_err(fs_info, 253675cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 253775cb857dSQu Wenruo return ret; 253875cb857dSQu Wenruo } 253975cb857dSQu Wenruo 2540bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2541bd676446SJosef Bacik { 2542789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 2543789d6a3aSQu Wenruo .level = level, 2544789d6a3aSQu Wenruo .transid = gen, 2545789d6a3aSQu Wenruo .owner_root = root->root_key.objectid 2546789d6a3aSQu Wenruo }; 2547bd676446SJosef Bacik int ret = 0; 2548bd676446SJosef Bacik 2549789d6a3aSQu Wenruo root->node = read_tree_block(root->fs_info, bytenr, &check); 2550bd676446SJosef Bacik if (IS_ERR(root->node)) { 2551bd676446SJosef Bacik ret = PTR_ERR(root->node); 2552bd676446SJosef Bacik root->node = NULL; 25534eb150d6SQu Wenruo return ret; 25544eb150d6SQu Wenruo } 25554eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2556bd676446SJosef Bacik free_extent_buffer(root->node); 2557bd676446SJosef Bacik root->node = NULL; 25584eb150d6SQu Wenruo return -EIO; 2559bd676446SJosef Bacik } 2560bd676446SJosef Bacik 2561bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2562bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2563bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2564bd676446SJosef Bacik return ret; 2565bd676446SJosef Bacik } 2566bd676446SJosef Bacik 2567bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2568bd676446SJosef Bacik { 2569bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2570bd676446SJosef Bacik u64 gen, bytenr; 2571bd676446SJosef Bacik int level, ret; 2572bd676446SJosef Bacik 2573bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2574bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2575bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2576bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 25779c54e80dSJosef Bacik if (ret) { 2578bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2579bd676446SJosef Bacik return ret; 2580bd676446SJosef Bacik } 25819c54e80dSJosef Bacik return 0; 25829c54e80dSJosef Bacik } 25839c54e80dSJosef Bacik 25846ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2585b8522a1eSNikolay Borisov { 25866ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2587b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2588b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2589b8522a1eSNikolay Borisov bool handle_error = false; 2590b8522a1eSNikolay Borisov int ret = 0; 2591b8522a1eSNikolay Borisov int i; 2592b8522a1eSNikolay Borisov 2593b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2594b8522a1eSNikolay Borisov if (handle_error) { 2595b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2596b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2597b8522a1eSNikolay Borisov tree_root->node = NULL; 2598b8522a1eSNikolay Borisov 2599b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2600b8522a1eSNikolay Borisov break; 2601b8522a1eSNikolay Borisov 2602b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2603b8522a1eSNikolay Borisov 2604b8522a1eSNikolay Borisov /* 2605b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2606b8522a1eSNikolay Borisov * valid 2607b8522a1eSNikolay Borisov */ 2608b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2609b8522a1eSNikolay Borisov 2610b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2611b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2612b8522a1eSNikolay Borisov 2613745806fbSQu Wenruo btrfs_warn(fs_info, "try to load backup roots slot %d", i); 2614b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26156ef108ddSNikolay Borisov backup_index = ret; 2616b8522a1eSNikolay Borisov if (ret < 0) 2617b8522a1eSNikolay Borisov return ret; 2618b8522a1eSNikolay Borisov } 2619b8522a1eSNikolay Borisov 2620bd676446SJosef Bacik ret = load_important_roots(fs_info); 2621bd676446SJosef Bacik if (ret) { 2622217f5004SNikolay Borisov handle_error = true; 2623b8522a1eSNikolay Borisov continue; 2624b8522a1eSNikolay Borisov } 2625b8522a1eSNikolay Borisov 2626336a0d8dSNikolay Borisov /* 2627336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2628336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2629336a0d8dSNikolay Borisov */ 2630453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2631b8522a1eSNikolay Borisov if (ret < 0) { 2632b8522a1eSNikolay Borisov handle_error = true; 2633b8522a1eSNikolay Borisov continue; 2634b8522a1eSNikolay Borisov } 2635b8522a1eSNikolay Borisov 26366b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2637b8522a1eSNikolay Borisov 2638b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2639b8522a1eSNikolay Borisov if (ret < 0) { 2640b8522a1eSNikolay Borisov handle_error = true; 2641b8522a1eSNikolay Borisov continue; 2642b8522a1eSNikolay Borisov } 2643b8522a1eSNikolay Borisov 2644b8522a1eSNikolay Borisov /* All successful */ 2645bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2646bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2647d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 26486ef108ddSNikolay Borisov 26496ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26506ef108ddSNikolay Borisov if (backup_index < 0) { 26516ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26526ef108ddSNikolay Borisov } else { 26536ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26546ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26556ef108ddSNikolay Borisov } 2656b8522a1eSNikolay Borisov break; 2657b8522a1eSNikolay Borisov } 2658b8522a1eSNikolay Borisov 2659b8522a1eSNikolay Borisov return ret; 2660b8522a1eSNikolay Borisov } 2661b8522a1eSNikolay Borisov 26628260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2663eb60ceacSChris Mason { 2664fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 266501cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26668fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2667facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 266824bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2669eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 267011833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2671eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2672a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2673fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 267424bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26754cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2676ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2677f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 267847ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 267940ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2680afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2681c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2682f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2683abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2684d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2685f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 26867585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2687573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 26880bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 26895f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 2690de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 269119c00ddcSChris Mason 2692e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 26935a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 26948b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 26955f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 269677d20c68SJosef Bacik btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_prep, 269777d20c68SJosef Bacik BTRFS_LOCKDEP_TRANS_COMMIT_PREP); 26983e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 26993e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 27003e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 27013e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 27023e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 27033e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 2704e1489b4fSIoannis Angelakopoulos 27050b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 27066324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2707f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 270847ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 270918bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2710afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2711bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2712bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 27133fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27143fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2715bd647ce3SJosef Bacik #endif 2716c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 271766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 271866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 271966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 272066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 272166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 272266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 272366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2724ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2725ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2726ba2c4d4eSJosef Bacik 2727771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27284cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2729034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2730fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2731abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 273295ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27339ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27344cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2735a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2736f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27378b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2738fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2739c8b97818SChris Mason 2740b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2741b34b086cSChris Mason num_online_cpus() + 2, 8); 27420afbaf8cSChris Mason 2743199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2744199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 274569fe2d75SJosef Bacik 2746638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 274721adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 274821adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 274921adbd5cSStefan Behrens #endif 2750779a65a4SEric Sandeen btrfs_init_balance(fs_info); 275157056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2752a2de733cSArne Jansen 275316b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 275408dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 27550f9dd46cSJosef Bacik 2756fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 275735da5a7eSDavid Sterba IO_TREE_FS_EXCLUDED_EXTENTS); 275839279cc3SChris Mason 27595a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2760e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2761925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2762a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2763a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27641bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27659e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2766c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 276776dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2768803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2769fa9c0d79SChris Mason 2770ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2771f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2772b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2773416ac51dSArne Jansen 2774fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2775fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2776fa9c0d79SChris Mason 2777e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2778f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2779bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27804854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2781034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 27823768f368SChris Mason 2783da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2784da17066cSJeff Mahoney fs_info->nodesize = 4096; 2785da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2786ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2787da17066cSJeff Mahoney fs_info->stripesize = 4096; 2788da17066cSJeff Mahoney 2789f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 2790f7b12a62SNaohiro Aota 2791eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2792eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2793eede2bf3SOmar Sandoval 279418bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 279518bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 27968260edbaSJosef Bacik } 27978260edbaSJosef Bacik 27988260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 27998260edbaSJosef Bacik { 28008260edbaSJosef Bacik int ret; 28018260edbaSJosef Bacik 28028260edbaSJosef Bacik fs_info->sb = sb; 28038260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 28048260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 28059e967495SFilipe Manana 28065deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 2807ae18c37aSJosef Bacik if (ret) 2808ae18c37aSJosef Bacik return ret; 2809ae18c37aSJosef Bacik 2810ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2811ae18c37aSJosef Bacik if (ret) 2812c75e8394SJosef Bacik return ret; 2813ae18c37aSJosef Bacik 2814ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2815ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2816ae18c37aSJosef Bacik 2817ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2818ae18c37aSJosef Bacik if (ret) 2819c75e8394SJosef Bacik return ret; 2820ae18c37aSJosef Bacik 2821ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2822ae18c37aSJosef Bacik GFP_KERNEL); 2823ae18c37aSJosef Bacik if (ret) 2824c75e8394SJosef Bacik return ret; 2825ae18c37aSJosef Bacik 2826ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2827ae18c37aSJosef Bacik GFP_KERNEL); 2828c75e8394SJosef Bacik if (!fs_info->delayed_root) 2829c75e8394SJosef Bacik return -ENOMEM; 2830ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2831ae18c37aSJosef Bacik 2832a0a1db70SFilipe Manana if (sb_rdonly(sb)) 2833a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 2834a0a1db70SFilipe Manana 2835c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2836ae18c37aSJosef Bacik } 2837ae18c37aSJosef Bacik 283897f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 283997f4dd09SNikolay Borisov { 28400d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 284197f4dd09SNikolay Borisov int ret; 284297f4dd09SNikolay Borisov 284397f4dd09SNikolay Borisov /* 284497f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 284597f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 284697f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 284797f4dd09SNikolay Borisov */ 284897f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 284997f4dd09SNikolay Borisov if (ret < 0) { 2850c94bec2cSJosef Bacik if (ret != -EINTR) 2851c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2852c94bec2cSJosef Bacik ret); 285397f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 285497f4dd09SNikolay Borisov return ret; 285597f4dd09SNikolay Borisov } 285697f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 285797f4dd09SNikolay Borisov } 285897f4dd09SNikolay Borisov 285997f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 286097f4dd09SNikolay Borisov { 286197f4dd09SNikolay Borisov struct task_struct *task; 286297f4dd09SNikolay Borisov 286397f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 286497f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 286597f4dd09SNikolay Borisov if (IS_ERR(task)) { 286697f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 286797f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 286897f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 286997f4dd09SNikolay Borisov return PTR_ERR(task); 287097f4dd09SNikolay Borisov } 287197f4dd09SNikolay Borisov 287297f4dd09SNikolay Borisov return 0; 287397f4dd09SNikolay Borisov } 287497f4dd09SNikolay Borisov 2875504b1596SFilipe Manana static int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 2876504b1596SFilipe Manana { 2877504b1596SFilipe Manana u64 root_objectid = 0; 2878504b1596SFilipe Manana struct btrfs_root *gang[8]; 2879504b1596SFilipe Manana int i = 0; 2880504b1596SFilipe Manana int err = 0; 2881504b1596SFilipe Manana unsigned int ret = 0; 2882504b1596SFilipe Manana 2883504b1596SFilipe Manana while (1) { 2884504b1596SFilipe Manana spin_lock(&fs_info->fs_roots_radix_lock); 2885504b1596SFilipe Manana ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2886504b1596SFilipe Manana (void **)gang, root_objectid, 2887504b1596SFilipe Manana ARRAY_SIZE(gang)); 2888504b1596SFilipe Manana if (!ret) { 2889504b1596SFilipe Manana spin_unlock(&fs_info->fs_roots_radix_lock); 2890504b1596SFilipe Manana break; 2891504b1596SFilipe Manana } 2892504b1596SFilipe Manana root_objectid = gang[ret - 1]->root_key.objectid + 1; 2893504b1596SFilipe Manana 2894504b1596SFilipe Manana for (i = 0; i < ret; i++) { 2895504b1596SFilipe Manana /* Avoid to grab roots in dead_roots. */ 2896504b1596SFilipe Manana if (btrfs_root_refs(&gang[i]->root_item) == 0) { 2897504b1596SFilipe Manana gang[i] = NULL; 2898504b1596SFilipe Manana continue; 2899504b1596SFilipe Manana } 2900504b1596SFilipe Manana /* Grab all the search result for later use. */ 2901504b1596SFilipe Manana gang[i] = btrfs_grab_root(gang[i]); 2902504b1596SFilipe Manana } 2903504b1596SFilipe Manana spin_unlock(&fs_info->fs_roots_radix_lock); 2904504b1596SFilipe Manana 2905504b1596SFilipe Manana for (i = 0; i < ret; i++) { 2906504b1596SFilipe Manana if (!gang[i]) 2907504b1596SFilipe Manana continue; 2908504b1596SFilipe Manana root_objectid = gang[i]->root_key.objectid; 2909504b1596SFilipe Manana err = btrfs_orphan_cleanup(gang[i]); 2910504b1596SFilipe Manana if (err) 2911504b1596SFilipe Manana goto out; 2912504b1596SFilipe Manana btrfs_put_root(gang[i]); 2913504b1596SFilipe Manana } 2914504b1596SFilipe Manana root_objectid++; 2915504b1596SFilipe Manana } 2916504b1596SFilipe Manana out: 2917504b1596SFilipe Manana /* Release the uncleaned roots due to error. */ 2918504b1596SFilipe Manana for (; i < ret; i++) { 2919504b1596SFilipe Manana if (gang[i]) 2920504b1596SFilipe Manana btrfs_put_root(gang[i]); 2921504b1596SFilipe Manana } 2922504b1596SFilipe Manana return err; 2923504b1596SFilipe Manana } 2924504b1596SFilipe Manana 292544c0ca21SBoris Burkov /* 29268cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 29278cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 29288cd29088SBoris Burkov * code paths. 29298cd29088SBoris Burkov */ 29308cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 29318cd29088SBoris Burkov { 29328cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 29338b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 29348cd29088SBoris Burkov } 29358cd29088SBoris Burkov 29368cd29088SBoris Burkov /* 293744c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 293844c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 293944c0ca21SBoris Burkov */ 294044c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 294144c0ca21SBoris Burkov { 294244c0ca21SBoris Burkov int ret; 294394846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 29441d6a4fc8SQu Wenruo bool rebuild_free_space_tree = false; 29458b228324SBoris Burkov 29468b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 29478b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 29481d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 29498b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29508b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 29518b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 29521d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 29538b228324SBoris Burkov } 29548b228324SBoris Burkov 29551d6a4fc8SQu Wenruo if (rebuild_free_space_tree) { 29561d6a4fc8SQu Wenruo btrfs_info(fs_info, "rebuilding free space tree"); 29571d6a4fc8SQu Wenruo ret = btrfs_rebuild_free_space_tree(fs_info); 29588b228324SBoris Burkov if (ret) { 29598b228324SBoris Burkov btrfs_warn(fs_info, 29601d6a4fc8SQu Wenruo "failed to rebuild free space tree: %d", ret); 29611d6a4fc8SQu Wenruo goto out; 29621d6a4fc8SQu Wenruo } 29631d6a4fc8SQu Wenruo } 29641d6a4fc8SQu Wenruo 29651d6a4fc8SQu Wenruo if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29661d6a4fc8SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { 29671d6a4fc8SQu Wenruo btrfs_info(fs_info, "disabling free space tree"); 29681d6a4fc8SQu Wenruo ret = btrfs_delete_free_space_tree(fs_info); 29691d6a4fc8SQu Wenruo if (ret) { 29701d6a4fc8SQu Wenruo btrfs_warn(fs_info, 29711d6a4fc8SQu Wenruo "failed to disable free space tree: %d", ret); 29728b228324SBoris Burkov goto out; 29738b228324SBoris Burkov } 29748b228324SBoris Burkov } 297544c0ca21SBoris Burkov 29768d488a8cSFilipe Manana /* 29778d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 29788d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 2979fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 29808d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 29818d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 29828d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 29838d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 29848d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 29858d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 29868d488a8cSFilipe Manana */ 29878d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 29888d488a8cSFilipe Manana if (ret) 29898d488a8cSFilipe Manana goto out; 29908d488a8cSFilipe Manana 299144c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 299244c0ca21SBoris Burkov if (ret) 299344c0ca21SBoris Burkov goto out; 299444c0ca21SBoris Burkov 29958f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 29968f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29978f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29988f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29998f1c21d7SBoris Burkov goto out; 30008f1c21d7SBoris Burkov } 30018f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 30028f1c21d7SBoris Burkov 300344c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 30047eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 300544c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 300644c0ca21SBoris Burkov if (ret < 0) { 300744c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 300844c0ca21SBoris Burkov goto out; 300944c0ca21SBoris Burkov } 301044c0ca21SBoris Burkov 30115011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 30125011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 30135011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 30145011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 30155011139aSBoris Burkov if (ret) { 30165011139aSBoris Burkov btrfs_warn(fs_info, 30175011139aSBoris Burkov "failed to create free space tree: %d", ret); 30185011139aSBoris Burkov goto out; 30195011139aSBoris Burkov } 30205011139aSBoris Burkov } 30215011139aSBoris Burkov 302294846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 302394846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 302494846229SBoris Burkov if (ret) 302594846229SBoris Burkov goto out; 302694846229SBoris Burkov } 302794846229SBoris Burkov 302844c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 302944c0ca21SBoris Burkov if (ret) 303044c0ca21SBoris Burkov goto out; 303144c0ca21SBoris Burkov 303244c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 303344c0ca21SBoris Burkov if (ret) { 303444c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 303544c0ca21SBoris Burkov goto out; 303644c0ca21SBoris Burkov } 303744c0ca21SBoris Burkov 303844c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 303944c0ca21SBoris Burkov 304044c0ca21SBoris Burkov if (!fs_info->uuid_root) { 304144c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 304244c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 304344c0ca21SBoris Burkov if (ret) { 304444c0ca21SBoris Burkov btrfs_warn(fs_info, 304544c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 304644c0ca21SBoris Burkov goto out; 304744c0ca21SBoris Burkov } 304844c0ca21SBoris Burkov } 304944c0ca21SBoris Burkov 305044c0ca21SBoris Burkov out: 305144c0ca21SBoris Burkov return ret; 305244c0ca21SBoris Burkov } 305344c0ca21SBoris Burkov 3054d7f67ac9SQu Wenruo /* 3055d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3056d7f67ac9SQu Wenruo * 30572ba48b20SQu Wenruo * @is_rw_mount: If the mount is read-write. 30582ba48b20SQu Wenruo * 3059d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3060d7f67ac9SQu Wenruo * feature dependencies). 3061d7f67ac9SQu Wenruo * 3062d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3063d7f67ac9SQu Wenruo * btrfs_validate_super(). 3064d7f67ac9SQu Wenruo * 3065d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3066d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3067d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3068d7f67ac9SQu Wenruo */ 30692ba48b20SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) 3070d7f67ac9SQu Wenruo { 3071d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3072d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3073d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3074d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3075d7f67ac9SQu Wenruo 3076d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3077d7f67ac9SQu Wenruo btrfs_err(fs_info, 3078d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3079d7f67ac9SQu Wenruo incompat); 3080d7f67ac9SQu Wenruo return -EINVAL; 3081d7f67ac9SQu Wenruo } 3082d7f67ac9SQu Wenruo 3083d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3084d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3085d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3086d7f67ac9SQu Wenruo btrfs_err(fs_info, 3087d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3088d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3089d7f67ac9SQu Wenruo return -EINVAL; 3090d7f67ac9SQu Wenruo } 3091d7f67ac9SQu Wenruo 3092d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3093d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3094d7f67ac9SQu Wenruo 3095d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3096d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3097d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3098d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3099d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3100d7f67ac9SQu Wenruo 3101d7f67ac9SQu Wenruo /* 3102d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3103d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3104d7f67ac9SQu Wenruo */ 3105d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3106d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3107d7f67ac9SQu Wenruo 31082ba48b20SQu Wenruo if (compat_ro_unsupp && is_rw_mount) { 3109d7f67ac9SQu Wenruo btrfs_err(fs_info, 3110d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3111d7f67ac9SQu Wenruo compat_ro); 3112d7f67ac9SQu Wenruo return -EINVAL; 3113d7f67ac9SQu Wenruo } 3114d7f67ac9SQu Wenruo 3115d7f67ac9SQu Wenruo /* 3116d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3117d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3118d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3119d7f67ac9SQu Wenruo */ 3120d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3121d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3122d7f67ac9SQu Wenruo btrfs_err(fs_info, 3123d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3124d7f67ac9SQu Wenruo compat_ro); 3125d7f67ac9SQu Wenruo return -EINVAL; 3126d7f67ac9SQu Wenruo } 3127d7f67ac9SQu Wenruo 3128d7f67ac9SQu Wenruo /* 3129d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3130d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3131d7f67ac9SQu Wenruo */ 3132d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3133d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3134d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3135d7f67ac9SQu Wenruo btrfs_err(fs_info, 3136d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3137d7f67ac9SQu Wenruo return -EINVAL; 3138d7f67ac9SQu Wenruo } 3139d7f67ac9SQu Wenruo 3140d7f67ac9SQu Wenruo /* 3141d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3142d7f67ac9SQu Wenruo * 3143d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3144d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3145d7f67ac9SQu Wenruo * going to be deprecated anyway. 3146d7f67ac9SQu Wenruo */ 3147d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3148d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3149d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3150d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3151d7f67ac9SQu Wenruo return -EINVAL; 3152d7f67ac9SQu Wenruo } 3153d7f67ac9SQu Wenruo 3154d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3155d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3156d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3157d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3158d7f67ac9SQu Wenruo 3159d7f67ac9SQu Wenruo return 0; 3160d7f67ac9SQu Wenruo } 3161d7f67ac9SQu Wenruo 3162ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3163ae18c37aSJosef Bacik char *options) 3164ae18c37aSJosef Bacik { 3165ae18c37aSJosef Bacik u32 sectorsize; 3166ae18c37aSJosef Bacik u32 nodesize; 3167ae18c37aSJosef Bacik u32 stripesize; 3168ae18c37aSJosef Bacik u64 generation; 3169ae18c37aSJosef Bacik u64 features; 3170ae18c37aSJosef Bacik u16 csum_type; 3171ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3172ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3173ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3174ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3175ae18c37aSJosef Bacik int ret; 3176ae18c37aSJosef Bacik int level; 3177ae18c37aSJosef Bacik 31788260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 31794871c33bSQu Wenruo if (ret) 3180ae18c37aSJosef Bacik goto fail; 318153b381b3SDavid Woodhouse 3182ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3183ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3184ae18c37aSJosef Bacik GFP_KERNEL); 3185ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3186ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3187ae18c37aSJosef Bacik GFP_KERNEL); 3188ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3189ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 31904871c33bSQu Wenruo ret = -ENOMEM; 3191c75e8394SJosef Bacik goto fail; 3192ae18c37aSJosef Bacik } 3193ae18c37aSJosef Bacik 3194dcb2137cSChristoph Hellwig ret = btrfs_init_btree_inode(sb); 31954871c33bSQu Wenruo if (ret) 3196c75e8394SJosef Bacik goto fail; 3197ae18c37aSJosef Bacik 3198d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 31991104a885SDavid Sterba 32001104a885SDavid Sterba /* 32011104a885SDavid Sterba * Read super block and check the signature bytes only 32021104a885SDavid Sterba */ 3203d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 32048f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 32054871c33bSQu Wenruo ret = PTR_ERR(disk_super); 320616cdcec7SMiao Xie goto fail_alloc; 320720b45077SDave Young } 320839279cc3SChris Mason 32099f7894e2SQu Wenruo btrfs_info(fs_info, "first mount of filesystem %pU", disk_super->fsid); 32101104a885SDavid Sterba /* 3211260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 32128dc3f22cSJohannes Thumshirn * corrupted, we'll find out 32138dc3f22cSJohannes Thumshirn */ 32148f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 321551bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 32168dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 321751bce6c9SJohannes Thumshirn csum_type); 32184871c33bSQu Wenruo ret = -EINVAL; 32198f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32208dc3f22cSJohannes Thumshirn goto fail_alloc; 32218dc3f22cSJohannes Thumshirn } 32228dc3f22cSJohannes Thumshirn 322383c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 322483c68bbcSSu Yue 32256d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 32266d97c6e3SJohannes Thumshirn if (ret) { 32278f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32286d97c6e3SJohannes Thumshirn goto fail_alloc; 32296d97c6e3SJohannes Thumshirn } 32306d97c6e3SJohannes Thumshirn 32318dc3f22cSJohannes Thumshirn /* 32321104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 32331104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 32341104a885SDavid Sterba */ 32353d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 323605135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 32374871c33bSQu Wenruo ret = -EINVAL; 32388f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3239141386e1SJosef Bacik goto fail_alloc; 32401104a885SDavid Sterba } 32411104a885SDavid Sterba 32421104a885SDavid Sterba /* 32431104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 32441104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 32451104a885SDavid Sterba * the whole block of INFO_SIZE 32461104a885SDavid Sterba */ 32478f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 32488f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32495f39d397SChris Mason 3250fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3251fbc6feaeSNikolay Borisov 32520b86a832SChris Mason 3253fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3254fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3255fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3256fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3257fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3258fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3259fbc6feaeSNikolay Borisov } 3260fbc6feaeSNikolay Borisov 3261fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3262fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3263de37aa51SNikolay Borisov 3264069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 32651104a885SDavid Sterba if (ret) { 326605135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 32674871c33bSQu Wenruo ret = -EINVAL; 3268141386e1SJosef Bacik goto fail_alloc; 32691104a885SDavid Sterba } 32701104a885SDavid Sterba 32714871c33bSQu Wenruo if (!btrfs_super_root(disk_super)) { 32724871c33bSQu Wenruo btrfs_err(fs_info, "invalid superblock tree root bytenr"); 32734871c33bSQu Wenruo ret = -EINVAL; 3274141386e1SJosef Bacik goto fail_alloc; 32754871c33bSQu Wenruo } 32760f7d52f4SChris Mason 3277acce952bSliubo /* check FS state, whether FS is broken. */ 327887533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 3279ae3364e5SFilipe Manana WRITE_ONCE(fs_info->fs_error, -EUCLEAN); 3280acce952bSliubo 328175e7cb7fSLiu Bo /* 328275e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 328375e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 328475e7cb7fSLiu Bo */ 328575e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 328675e7cb7fSLiu Bo 32876f93e834SAnand Jain 32886f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 32896f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 32906f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 32916f93e834SAnand Jain stripesize = sectorsize; 32926f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 32936f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 32946f93e834SAnand Jain 32956f93e834SAnand Jain fs_info->nodesize = nodesize; 32966f93e834SAnand Jain fs_info->sectorsize = sectorsize; 32976f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 32986f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 32996f93e834SAnand Jain fs_info->stripesize = stripesize; 33006f93e834SAnand Jain 33012ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 33024871c33bSQu Wenruo if (ret) 3303141386e1SJosef Bacik goto fail_alloc; 3304dfe25020SChris Mason 33052ba48b20SQu Wenruo ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); 33064871c33bSQu Wenruo if (ret < 0) 3307141386e1SJosef Bacik goto fail_alloc; 3308f2b636e8SJosef Bacik 33098481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 33108481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 33118481dd80SQu Wenruo 33129f73f1aeSQu Wenruo /* 33139f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 33149f73f1aeSQu Wenruo * going to be deprecated. 33159f73f1aeSQu Wenruo * 33169f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 33179f73f1aeSQu Wenruo */ 33189f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 33199f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 33209f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 33219f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 33229f73f1aeSQu Wenruo 332395ea0486SQu Wenruo btrfs_warn(fs_info, 332495ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 33250bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 33268481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 33274871c33bSQu Wenruo if (!subpage_info) { 33284871c33bSQu Wenruo ret = -ENOMEM; 33298481dd80SQu Wenruo goto fail_alloc; 33304871c33bSQu Wenruo } 33318481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 33328481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3333c8050b3bSQu Wenruo } 33340bb3eb3eSQu Wenruo 3335d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 33364871c33bSQu Wenruo if (ret) 33370dc3b84aSJosef Bacik goto fail_sb_buffer; 33384543df7eSChris Mason 33399e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 33409e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 33414575c9ccSChris Mason 3342a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3343a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3344de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3345db94535dSChris Mason 3346925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 33476bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3348925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 334984eed90fSChris Mason if (ret) { 335005135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 33515d4f98a2SYan Zheng goto fail_sb_buffer; 335284eed90fSChris Mason } 33530b86a832SChris Mason 335484234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3355581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3356bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3357bd676446SJosef Bacik generation, level); 3358bd676446SJosef Bacik if (ret) { 335905135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3360af31f5e5SChris Mason goto fail_tree_roots; 336183121942SDavid Woodhouse } 33620b86a832SChris Mason 3363e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3364c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3365c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3366e17cade2SChris Mason 33675b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 33682b82032cSYan Zheng if (ret) { 336905135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3370af31f5e5SChris Mason goto fail_tree_roots; 33712b82032cSYan Zheng } 33720b86a832SChris Mason 33738dabb742SStefan Behrens /* 3374bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3375bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3376bacce86aSAnand Jain * target is required. So free devices that are not part of this 33771a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3378bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 33798dabb742SStefan Behrens */ 3380bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3381d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 338205135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 33834871c33bSQu Wenruo ret = -EIO; 3384a6b0d5c8SChris Mason goto fail_tree_roots; 3385a6b0d5c8SChris Mason } 3386a6b0d5c8SChris Mason 3387b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 33884bbcaa64SEric Sandeen if (ret) 3389b8522a1eSNikolay Borisov goto fail_tree_roots; 33908929ecfaSYan, Zheng 339175ec1db8SJosef Bacik /* 339273651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 339373651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 339473651042SNaohiro Aota * zoned incompat feature flag set. 339573651042SNaohiro Aota */ 339673651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 339773651042SNaohiro Aota if (ret) { 339873651042SNaohiro Aota btrfs_err(fs_info, 33994871c33bSQu Wenruo "zoned: failed to read device zone info: %d", ret); 340073651042SNaohiro Aota goto fail_block_groups; 340173651042SNaohiro Aota } 340273651042SNaohiro Aota 340373651042SNaohiro Aota /* 340475ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 340575ec1db8SJosef Bacik * check the generation here so we can set the 340675ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 340775ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 340875ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 340975ec1db8SJosef Bacik * even though it was perfectly fine. 341075ec1db8SJosef Bacik */ 341175ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 341275ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 341375ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 341475ec1db8SJosef Bacik 3415cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3416cf90d884SQu Wenruo if (ret) { 3417cf90d884SQu Wenruo btrfs_err(fs_info, 3418cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3419cf90d884SQu Wenruo ret); 3420cf90d884SQu Wenruo goto fail_block_groups; 3421cf90d884SQu Wenruo } 342268310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 342368310a5eSIlya Dryomov if (ret) { 342405135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 342568310a5eSIlya Dryomov goto fail_block_groups; 342668310a5eSIlya Dryomov } 342768310a5eSIlya Dryomov 3428733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3429733f4fbbSStefan Behrens if (ret) { 343005135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3431733f4fbbSStefan Behrens goto fail_block_groups; 3432733f4fbbSStefan Behrens } 3433733f4fbbSStefan Behrens 34348dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 34358dabb742SStefan Behrens if (ret) { 343605135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 34378dabb742SStefan Behrens goto fail_block_groups; 34388dabb742SStefan Behrens } 34398dabb742SStefan Behrens 3440b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3441b70f5097SNaohiro Aota if (ret) { 3442b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3443b70f5097SNaohiro Aota ret); 3444b70f5097SNaohiro Aota goto fail_block_groups; 3445b70f5097SNaohiro Aota } 3446b70f5097SNaohiro Aota 3447c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3448b7c35e81SAnand Jain if (ret) { 344905135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 345005135f59SDavid Sterba ret); 3451b7c35e81SAnand Jain goto fail_block_groups; 3452b7c35e81SAnand Jain } 3453b7c35e81SAnand Jain 345496f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 34555ac1d209SJeff Mahoney if (ret) { 345605135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3457b7c35e81SAnand Jain goto fail_fsdev_sysfs; 34585ac1d209SJeff Mahoney } 34595ac1d209SJeff Mahoney 3460c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3461c59021f8Sliubo if (ret) { 346205135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 34632365dd3cSAnand Jain goto fail_sysfs; 3464c59021f8Sliubo } 3465c59021f8Sliubo 34665b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 34671b1d1f66SJosef Bacik if (ret) { 346805135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 34692365dd3cSAnand Jain goto fail_sysfs; 34701b1d1f66SJosef Bacik } 34714330e183SQu Wenruo 347216beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 347316beac87SNaohiro Aota 3474a7e1ac7bSNaohiro Aota btrfs_check_active_zone_reservation(fs_info); 3475a7e1ac7bSNaohiro Aota 34765c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 34775c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 347805135f59SDavid Sterba btrfs_warn(fs_info, 347952042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 34804871c33bSQu Wenruo ret = -EINVAL; 34812365dd3cSAnand Jain goto fail_sysfs; 3482292fd7fcSStefan Behrens } 34839078a3e1SChris Mason 348433c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3485a74a4b97SChris Mason "btrfs-cleaner"); 34864871c33bSQu Wenruo if (IS_ERR(fs_info->cleaner_kthread)) { 34874871c33bSQu Wenruo ret = PTR_ERR(fs_info->cleaner_kthread); 34882365dd3cSAnand Jain goto fail_sysfs; 34894871c33bSQu Wenruo } 3490a74a4b97SChris Mason 3491a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3492a74a4b97SChris Mason tree_root, 3493a74a4b97SChris Mason "btrfs-transaction"); 34944871c33bSQu Wenruo if (IS_ERR(fs_info->transaction_kthread)) { 34954871c33bSQu Wenruo ret = PTR_ERR(fs_info->transaction_kthread); 34963f157a2fSChris Mason goto fail_cleaner; 34974871c33bSQu Wenruo } 3498a74a4b97SChris Mason 3499583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3500c289811cSChris Mason !fs_info->fs_devices->rotating) { 3501583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3502c289811cSChris Mason } 3503c289811cSChris Mason 3504572d9ab7SDavid Sterba /* 350563a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 350663a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 350763a7cb13SDavid Sterba * nodiscard for the same mount. 350895ca6599SNaohiro Aota * 350995ca6599SNaohiro Aota * The zoned mode piggy backs on the discard functionality for 351095ca6599SNaohiro Aota * resetting a zone. There is no reason to delay the zone reset as it is 351195ca6599SNaohiro Aota * fast enough. So, do not enable async discard for zoned mode. 351263a7cb13SDavid Sterba */ 351363a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 351463a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 351563a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 351695ca6599SNaohiro Aota fs_info->fs_devices->discardable && 351795ca6599SNaohiro Aota !btrfs_is_zoned(fs_info)) { 351863a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 351963a7cb13SDavid Sterba "auto enabling async discard"); 352063a7cb13SDavid Sterba } 352163a7cb13SDavid Sterba 352221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 35230b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 35242ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 35250b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3526cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 352721adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 352821adbd5cSStefan Behrens if (ret) 352905135f59SDavid Sterba btrfs_warn(fs_info, 353005135f59SDavid Sterba "failed to initialize integrity check module: %d", 353105135f59SDavid Sterba ret); 353221adbd5cSStefan Behrens } 353321adbd5cSStefan Behrens #endif 3534bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3535bcef60f2SArne Jansen if (ret) 3536bcef60f2SArne Jansen goto fail_trans_kthread; 353721adbd5cSStefan Behrens 3538fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3539fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3540fd708b81SJosef Bacik 354196da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 354296da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 35430b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3544e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 354563443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 35464871c33bSQu Wenruo if (ret) 354728c16cbbSWang Shilong goto fail_qgroup; 3548e556ce2cSYan Zheng } 35491a40e23bSZheng Yan 355056e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 35513140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 35524871c33bSQu Wenruo ret = PTR_ERR(fs_info->fs_root); 35534871c33bSQu Wenruo btrfs_warn(fs_info, "failed to read fs tree: %d", ret); 3554315bf8efSJosef Bacik fs_info->fs_root = NULL; 3555bcef60f2SArne Jansen goto fail_qgroup; 35563140c9a3SDan Carpenter } 3557c289811cSChris Mason 3558bc98a42cSDavid Howells if (sb_rdonly(sb)) 35598cd29088SBoris Burkov goto clear_oneshot; 35602b6ba629SIlya Dryomov 356144c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 35622b6ba629SIlya Dryomov if (ret) { 35636bccf3abSJeff Mahoney close_ctree(fs_info); 35642b6ba629SIlya Dryomov return ret; 3565e3acc2a6SJosef Bacik } 3566b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3567b382a324SJan Schmidt 356844c0ca21SBoris Burkov if (fs_info->uuid_root && 356944c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 357044c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 357105135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 357270f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 357370f80175SStefan Behrens if (ret) { 357405135f59SDavid Sterba btrfs_warn(fs_info, 357505135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 35766bccf3abSJeff Mahoney close_ctree(fs_info); 357770f80175SStefan Behrens return ret; 357870f80175SStefan Behrens } 3579f7a81ea4SStefan Behrens } 358094846229SBoris Burkov 3581afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 358247ab2a6cSJosef Bacik 3583b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3584b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3585b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3586b4be6aefSJosef Bacik 35878cd29088SBoris Burkov clear_oneshot: 35888cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3589ad2b2c80SAl Viro return 0; 359039279cc3SChris Mason 3591bcef60f2SArne Jansen fail_qgroup: 3592bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 35937c2ca468SChris Mason fail_trans_kthread: 35947c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 35952ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3596faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 35973f157a2fSChris Mason fail_cleaner: 3598a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 35997c2ca468SChris Mason 36007c2ca468SChris Mason /* 36017c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 36027c2ca468SChris Mason * kthreads 36037c2ca468SChris Mason */ 36047c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 36057c2ca468SChris Mason 36062365dd3cSAnand Jain fail_sysfs: 36076618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 36082365dd3cSAnand Jain 3609b7c35e81SAnand Jain fail_fsdev_sysfs: 3610b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3611b7c35e81SAnand Jain 36121b1d1f66SJosef Bacik fail_block_groups: 361354067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3614af31f5e5SChris Mason 3615af31f5e5SChris Mason fail_tree_roots: 36169e3aa805SJosef Bacik if (fs_info->data_reloc_root) 36179e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 36184273eaffSAnand Jain free_root_pointers(fs_info, true); 36192b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3620af31f5e5SChris Mason 362139279cc3SChris Mason fail_sb_buffer: 36227abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 36235cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 362416cdcec7SMiao Xie fail_alloc: 3625586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3626586e46e2SIlya Dryomov 36274543df7eSChris Mason iput(fs_info->btree_inode); 36287e662854SQinghuang Feng fail: 3629586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 36304871c33bSQu Wenruo ASSERT(ret < 0); 36314871c33bSQu Wenruo return ret; 3632eb60ceacSChris Mason } 3633663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3634eb60ceacSChris Mason 3635314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3636f2984462SChris Mason { 3637314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3638314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3639314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3640314b6dd0SJohannes Thumshirn struct page *page; 3641442a4f63SStefan Behrens 3642314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3643314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3644314b6dd0SJohannes Thumshirn 3645314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3646fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3647314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3648cb3e217bSQu Wenruo btrfs_dev_name(device), 3649314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3650314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3651314b6dd0SJohannes Thumshirn SetPageError(page); 3652314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3653314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3654314b6dd0SJohannes Thumshirn } else { 3655314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3656f2984462SChris Mason } 3657314b6dd0SJohannes Thumshirn 3658314b6dd0SJohannes Thumshirn put_page(page); 3659314b6dd0SJohannes Thumshirn unlock_page(page); 3660314b6dd0SJohannes Thumshirn } 3661314b6dd0SJohannes Thumshirn 3662314b6dd0SJohannes Thumshirn bio_put(bio); 3663f2984462SChris Mason } 3664f2984462SChris Mason 36658f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3666a05d3c91SQu Wenruo int copy_num, bool drop_cache) 366729c36d72SAnand Jain { 366829c36d72SAnand Jain struct btrfs_super_block *super; 36698f32380dSJohannes Thumshirn struct page *page; 367012659251SNaohiro Aota u64 bytenr, bytenr_orig; 36718f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 367212659251SNaohiro Aota int ret; 367329c36d72SAnand Jain 367412659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 367512659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 367612659251SNaohiro Aota if (ret == -ENOENT) 367712659251SNaohiro Aota return ERR_PTR(-EINVAL); 367812659251SNaohiro Aota else if (ret) 367912659251SNaohiro Aota return ERR_PTR(ret); 368012659251SNaohiro Aota 3681cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 36828f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 368329c36d72SAnand Jain 3684a05d3c91SQu Wenruo if (drop_cache) { 3685a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3686a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3687a05d3c91SQu Wenruo 3688a05d3c91SQu Wenruo /* 3689a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3690a05d3c91SQu Wenruo * always read from the device. 3691a05d3c91SQu Wenruo */ 3692a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3693a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3694a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3695a05d3c91SQu Wenruo } 3696a05d3c91SQu Wenruo 36978f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 36988f32380dSJohannes Thumshirn if (IS_ERR(page)) 36998f32380dSJohannes Thumshirn return ERR_CAST(page); 370029c36d72SAnand Jain 37018f32380dSJohannes Thumshirn super = page_address(page); 370296c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 370396c2e067SAnand Jain btrfs_release_disk_super(super); 370496c2e067SAnand Jain return ERR_PTR(-ENODATA); 370596c2e067SAnand Jain } 370696c2e067SAnand Jain 370712659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 37088f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 37098f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 371029c36d72SAnand Jain } 371129c36d72SAnand Jain 37128f32380dSJohannes Thumshirn return super; 371329c36d72SAnand Jain } 371429c36d72SAnand Jain 371529c36d72SAnand Jain 37168f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3717a512bbf8SYan Zheng { 37188f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3719a512bbf8SYan Zheng int i; 3720a512bbf8SYan Zheng u64 transid = 0; 3721a512bbf8SYan Zheng 3722a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3723a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3724a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3725a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3726a512bbf8SYan Zheng */ 3727a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3728a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 37298f32380dSJohannes Thumshirn if (IS_ERR(super)) 3730a512bbf8SYan Zheng continue; 3731a512bbf8SYan Zheng 3732a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 37338f32380dSJohannes Thumshirn if (latest) 37348f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 37358f32380dSJohannes Thumshirn 37368f32380dSJohannes Thumshirn latest = super; 3737a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3738a512bbf8SYan Zheng } 3739a512bbf8SYan Zheng } 374092fc03fbSAnand Jain 37418f32380dSJohannes Thumshirn return super; 3742a512bbf8SYan Zheng } 3743a512bbf8SYan Zheng 37444eedeb75SHisashi Hifumi /* 3745abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3746314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 37474eedeb75SHisashi Hifumi * 3748abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3749abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3750abbb3b8eSDavid Sterba * same for write and wait phases. 37514eedeb75SHisashi Hifumi * 3752314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 37534eedeb75SHisashi Hifumi */ 3754a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3755abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3756a512bbf8SYan Zheng { 3757d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3758314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3759d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3760a512bbf8SYan Zheng int i; 3761a512bbf8SYan Zheng int errors = 0; 376212659251SNaohiro Aota int ret; 376312659251SNaohiro Aota u64 bytenr, bytenr_orig; 3764a512bbf8SYan Zheng 3765a512bbf8SYan Zheng if (max_mirrors == 0) 3766a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3767a512bbf8SYan Zheng 3768d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3769d5178578SJohannes Thumshirn 3770a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3771314b6dd0SJohannes Thumshirn struct page *page; 3772314b6dd0SJohannes Thumshirn struct bio *bio; 3773314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3774314b6dd0SJohannes Thumshirn 377512659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 377612659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 377712659251SNaohiro Aota if (ret == -ENOENT) { 377812659251SNaohiro Aota continue; 377912659251SNaohiro Aota } else if (ret < 0) { 378012659251SNaohiro Aota btrfs_err(device->fs_info, 378112659251SNaohiro Aota "couldn't get super block location for mirror %d", 378212659251SNaohiro Aota i); 378312659251SNaohiro Aota errors++; 378412659251SNaohiro Aota continue; 378512659251SNaohiro Aota } 3786935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3787935e5cc9SMiao Xie device->commit_total_bytes) 3788a512bbf8SYan Zheng break; 3789a512bbf8SYan Zheng 379012659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3791a512bbf8SYan Zheng 3792fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3793fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3794fd08001fSEric Biggers sb->csum); 3795a512bbf8SYan Zheng 3796314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3797314b6dd0SJohannes Thumshirn GFP_NOFS); 3798314b6dd0SJohannes Thumshirn if (!page) { 3799fb456252SJeff Mahoney btrfs_err(device->fs_info, 3800314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3801f14d104dSDavid Sterba bytenr); 3802634554dcSJosef Bacik errors++; 3803634554dcSJosef Bacik continue; 3804634554dcSJosef Bacik } 3805634554dcSJosef Bacik 3806314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3807314b6dd0SJohannes Thumshirn get_page(page); 3808a512bbf8SYan Zheng 3809314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3810314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3811a512bbf8SYan Zheng 3812387125fcSChris Mason /* 3813314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3814314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3815314b6dd0SJohannes Thumshirn * checking. 3816387125fcSChris Mason */ 381707888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 381807888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 381907888c66SChristoph Hellwig GFP_NOFS); 3820314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3821314b6dd0SJohannes Thumshirn bio->bi_private = device; 3822314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3823314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3824314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3825314b6dd0SJohannes Thumshirn 3826314b6dd0SJohannes Thumshirn /* 3827314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3828314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3829314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3830314b6dd0SJohannes Thumshirn */ 38311b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3832314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3833314b6dd0SJohannes Thumshirn 383458ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 383558ff51f1SChristoph Hellwig submit_bio(bio); 38368376d9e1SNaohiro Aota 38378376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 38388376d9e1SNaohiro Aota errors++; 3839a512bbf8SYan Zheng } 3840a512bbf8SYan Zheng return errors < i ? 0 : -1; 3841a512bbf8SYan Zheng } 3842a512bbf8SYan Zheng 3843387125fcSChris Mason /* 3844abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3845abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3846abbb3b8eSDavid Sterba * 3847314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3848abbb3b8eSDavid Sterba * date. 3849abbb3b8eSDavid Sterba */ 3850abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3851abbb3b8eSDavid Sterba { 3852abbb3b8eSDavid Sterba int i; 3853abbb3b8eSDavid Sterba int errors = 0; 3854b6a535faSHoward McLauchlan bool primary_failed = false; 385512659251SNaohiro Aota int ret; 3856abbb3b8eSDavid Sterba u64 bytenr; 3857abbb3b8eSDavid Sterba 3858abbb3b8eSDavid Sterba if (max_mirrors == 0) 3859abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3860abbb3b8eSDavid Sterba 3861abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3862314b6dd0SJohannes Thumshirn struct page *page; 3863314b6dd0SJohannes Thumshirn 386412659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 386512659251SNaohiro Aota if (ret == -ENOENT) { 386612659251SNaohiro Aota break; 386712659251SNaohiro Aota } else if (ret < 0) { 386812659251SNaohiro Aota errors++; 386912659251SNaohiro Aota if (i == 0) 387012659251SNaohiro Aota primary_failed = true; 387112659251SNaohiro Aota continue; 387212659251SNaohiro Aota } 3873abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3874abbb3b8eSDavid Sterba device->commit_total_bytes) 3875abbb3b8eSDavid Sterba break; 3876abbb3b8eSDavid Sterba 3877314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3878314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3879314b6dd0SJohannes Thumshirn if (!page) { 3880abbb3b8eSDavid Sterba errors++; 3881b6a535faSHoward McLauchlan if (i == 0) 3882b6a535faSHoward McLauchlan primary_failed = true; 3883abbb3b8eSDavid Sterba continue; 3884abbb3b8eSDavid Sterba } 3885314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3886314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3887314b6dd0SJohannes Thumshirn if (PageError(page)) { 3888abbb3b8eSDavid Sterba errors++; 3889b6a535faSHoward McLauchlan if (i == 0) 3890b6a535faSHoward McLauchlan primary_failed = true; 3891b6a535faSHoward McLauchlan } 3892abbb3b8eSDavid Sterba 3893314b6dd0SJohannes Thumshirn /* Drop our reference */ 3894314b6dd0SJohannes Thumshirn put_page(page); 3895abbb3b8eSDavid Sterba 3896314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3897314b6dd0SJohannes Thumshirn put_page(page); 3898abbb3b8eSDavid Sterba } 3899abbb3b8eSDavid Sterba 3900b6a535faSHoward McLauchlan /* log error, force error return */ 3901b6a535faSHoward McLauchlan if (primary_failed) { 3902b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3903b6a535faSHoward McLauchlan device->devid); 3904b6a535faSHoward McLauchlan return -1; 3905b6a535faSHoward McLauchlan } 3906b6a535faSHoward McLauchlan 3907abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3908abbb3b8eSDavid Sterba } 3909abbb3b8eSDavid Sterba 3910abbb3b8eSDavid Sterba /* 3911387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3912387125fcSChris Mason * for the barrier when it is done 3913387125fcSChris Mason */ 39144246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3915387125fcSChris Mason { 3916f9e69aa9SChristoph Hellwig bio_uninit(bio); 3917387125fcSChris Mason complete(bio->bi_private); 3918387125fcSChris Mason } 3919387125fcSChris Mason 3920387125fcSChris Mason /* 39214fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 39224fc6441aSAnand Jain * done in the waiting counterpart. 3923387125fcSChris Mason */ 39244fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3925387125fcSChris Mason { 3926f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 3927387125fcSChris Mason 3928bfd3ea94SAnand Jain device->last_flush_error = BLK_STS_OK; 3929bfd3ea94SAnand Jain 3930a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3931a91cf0ffSWang Yugui /* 3932a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 3933a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 3934a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 3935a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 3936a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 3937a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 3938a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 3939a91cf0ffSWang Yugui * non-debug builds. 3940a91cf0ffSWang Yugui */ 394108e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 39424fc6441aSAnand Jain return; 3943a91cf0ffSWang Yugui #endif 3944387125fcSChris Mason 3945f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 3946f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 3947387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3948387125fcSChris Mason init_completion(&device->flush_wait); 3949387125fcSChris Mason bio->bi_private = &device->flush_wait; 3950387125fcSChris Mason 395158ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 395258ff51f1SChristoph Hellwig submit_bio(bio); 39531c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 39544fc6441aSAnand Jain } 3955387125fcSChris Mason 39564fc6441aSAnand Jain /* 39574fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 39581b465784SAnand Jain * Return true for any error, and false otherwise. 39594fc6441aSAnand Jain */ 39601b465784SAnand Jain static bool wait_dev_flush(struct btrfs_device *device) 39614fc6441aSAnand Jain { 3962f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 39634fc6441aSAnand Jain 39647e812f20SAnand Jain if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 39651b465784SAnand Jain return false; 39664fc6441aSAnand Jain 39672980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 39684fc6441aSAnand Jain 3969bfd3ea94SAnand Jain if (bio->bi_status) { 3970bfd3ea94SAnand Jain device->last_flush_error = bio->bi_status; 3971bfd3ea94SAnand Jain btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); 39721b465784SAnand Jain return true; 3973bfd3ea94SAnand Jain } 3974bfd3ea94SAnand Jain 39751b465784SAnand Jain return false; 3976387125fcSChris Mason } 3977387125fcSChris Mason 3978387125fcSChris Mason /* 3979387125fcSChris Mason * send an empty flush down to each device in parallel, 3980387125fcSChris Mason * then wait for them 3981387125fcSChris Mason */ 3982387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3983387125fcSChris Mason { 3984387125fcSChris Mason struct list_head *head; 3985387125fcSChris Mason struct btrfs_device *dev; 39865af3e8ccSStefan Behrens int errors_wait = 0; 3987387125fcSChris Mason 39881538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3989387125fcSChris Mason /* send down all the barriers */ 3990387125fcSChris Mason head = &info->fs_devices->devices; 39911538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3992e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3993f88ba6a2SHidetoshi Seto continue; 3994cea7c8bfSAnand Jain if (!dev->bdev) 3995387125fcSChris Mason continue; 3996e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3997ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3998387125fcSChris Mason continue; 3999387125fcSChris Mason 40004fc6441aSAnand Jain write_dev_flush(dev); 4001387125fcSChris Mason } 4002387125fcSChris Mason 4003387125fcSChris Mason /* wait for all the barriers */ 40041538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4005e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4006f88ba6a2SHidetoshi Seto continue; 4007387125fcSChris Mason if (!dev->bdev) { 40085af3e8ccSStefan Behrens errors_wait++; 4009387125fcSChris Mason continue; 4010387125fcSChris Mason } 4011e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4012ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4013387125fcSChris Mason continue; 4014387125fcSChris Mason 40151b465784SAnand Jain if (wait_dev_flush(dev)) 40165af3e8ccSStefan Behrens errors_wait++; 4017387125fcSChris Mason } 4018401b41e5SAnand Jain 4019401b41e5SAnand Jain /* 4020de38a206SAnand Jain * Checks last_flush_error of disks in order to determine the device 4021de38a206SAnand Jain * state. 4022401b41e5SAnand Jain */ 4023de38a206SAnand Jain if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) 4024de38a206SAnand Jain return -EIO; 4025de38a206SAnand Jain 4026387125fcSChris Mason return 0; 4027387125fcSChris Mason } 4028387125fcSChris Mason 4029943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4030943c6e99SZhao Lei { 40318789f4feSZhao Lei int raid_type; 40328789f4feSZhao Lei int min_tolerated = INT_MAX; 4033943c6e99SZhao Lei 40348789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 40358789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 40368c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 40378789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 40388789f4feSZhao Lei tolerated_failures); 4039943c6e99SZhao Lei 40408789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 40418789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 40428789f4feSZhao Lei continue; 404341a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 40448789f4feSZhao Lei continue; 40458c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 40468789f4feSZhao Lei btrfs_raid_array[raid_type]. 40478789f4feSZhao Lei tolerated_failures); 40488789f4feSZhao Lei } 4049943c6e99SZhao Lei 40508789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4051ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 40528789f4feSZhao Lei min_tolerated = 0; 40538789f4feSZhao Lei } 40548789f4feSZhao Lei 40558789f4feSZhao Lei return min_tolerated; 4056943c6e99SZhao Lei } 4057943c6e99SZhao Lei 4058eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4059f2984462SChris Mason { 4060e5e9a520SChris Mason struct list_head *head; 4061f2984462SChris Mason struct btrfs_device *dev; 4062a061fc8dSChris Mason struct btrfs_super_block *sb; 4063f2984462SChris Mason struct btrfs_dev_item *dev_item; 4064f2984462SChris Mason int ret; 4065f2984462SChris Mason int do_barriers; 4066a236aed1SChris Mason int max_errors; 4067a236aed1SChris Mason int total_errors = 0; 4068a061fc8dSChris Mason u64 flags; 4069f2984462SChris Mason 40700b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4071fed3b381SLiu Bo 4072fed3b381SLiu Bo /* 4073fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4074fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4075fed3b381SLiu Bo * been consistent on disk. 4076fed3b381SLiu Bo */ 4077fed3b381SLiu Bo if (max_mirrors == 0) 40780b246afaSJeff Mahoney backup_super_roots(fs_info); 4079f2984462SChris Mason 40800b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4081a061fc8dSChris Mason dev_item = &sb->dev_item; 4082e5e9a520SChris Mason 40830b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 40840b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 40850b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4086387125fcSChris Mason 40875af3e8ccSStefan Behrens if (do_barriers) { 40880b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 40895af3e8ccSStefan Behrens if (ret) { 40905af3e8ccSStefan Behrens mutex_unlock( 40910b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 40920b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 40935af3e8ccSStefan Behrens "errors while submitting device barriers."); 40945af3e8ccSStefan Behrens return ret; 40955af3e8ccSStefan Behrens } 40965af3e8ccSStefan Behrens } 4097387125fcSChris Mason 40981538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4099dfe25020SChris Mason if (!dev->bdev) { 4100dfe25020SChris Mason total_errors++; 4101dfe25020SChris Mason continue; 4102dfe25020SChris Mason } 4103e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4104ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4105dfe25020SChris Mason continue; 4106dfe25020SChris Mason 41072b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4108a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4109a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 41107df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4111935e5cc9SMiao Xie dev->commit_total_bytes); 4112ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4113ce7213c7SMiao Xie dev->commit_bytes_used); 4114a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4115a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4116a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4117a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 41187239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 41197239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4120a512bbf8SYan Zheng 4121a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4122a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4123f2984462SChris Mason 412475cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 412575cb857dSQu Wenruo if (ret < 0) { 412675cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 412775cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 412875cb857dSQu Wenruo "unexpected superblock corruption detected"); 412975cb857dSQu Wenruo return -EUCLEAN; 413075cb857dSQu Wenruo } 413175cb857dSQu Wenruo 4132abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4133a236aed1SChris Mason if (ret) 4134a236aed1SChris Mason total_errors++; 4135f2984462SChris Mason } 4136a236aed1SChris Mason if (total_errors > max_errors) { 41370b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4138d397712bSChris Mason total_errors); 41390b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 414079787eaaSJeff Mahoney 41419d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 41420b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 41430b246afaSJeff Mahoney "%d errors while writing supers", 41440b246afaSJeff Mahoney total_errors); 41459d565ba4SStefan Behrens return -EIO; 4146a236aed1SChris Mason } 4147f2984462SChris Mason 4148a512bbf8SYan Zheng total_errors = 0; 41491538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4150dfe25020SChris Mason if (!dev->bdev) 4151dfe25020SChris Mason continue; 4152e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4153ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4154dfe25020SChris Mason continue; 4155dfe25020SChris Mason 4156abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4157a512bbf8SYan Zheng if (ret) 41581259ab75SChris Mason total_errors++; 4159f2984462SChris Mason } 41600b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4161a236aed1SChris Mason if (total_errors > max_errors) { 41620b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 41630b246afaSJeff Mahoney "%d errors while writing supers", 41640b246afaSJeff Mahoney total_errors); 416579787eaaSJeff Mahoney return -EIO; 4166a236aed1SChris Mason } 4167f2984462SChris Mason return 0; 4168f2984462SChris Mason } 4169f2984462SChris Mason 4170cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4171cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4172cb517eabSMiao Xie struct btrfs_root *root) 41732619ba1fSChris Mason { 41744785e24fSJosef Bacik bool drop_ref = false; 41754785e24fSJosef Bacik 4176fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4177fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4178fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4179fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 41804785e24fSJosef Bacik drop_ref = true; 4181fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 418276dda93cSYan, Zheng 418384961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4184ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 41851c1ea4f7SLiu Bo if (root->reloc_root) { 418600246528SJosef Bacik btrfs_put_root(root->reloc_root); 41871c1ea4f7SLiu Bo root->reloc_root = NULL; 41881c1ea4f7SLiu Bo } 41891c1ea4f7SLiu Bo } 41903321719eSLiu Bo 41914785e24fSJosef Bacik if (drop_ref) 419200246528SJosef Bacik btrfs_put_root(root); 41932619ba1fSChris Mason } 41942619ba1fSChris Mason 41956bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4196c146afadSYan Zheng { 41976bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4198c146afadSYan Zheng struct btrfs_trans_handle *trans; 4199c146afadSYan Zheng 42000b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 42012ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 42020b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 42030b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4204c71bf099SYan, Zheng 4205c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 42060b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 42070b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4208c71bf099SYan, Zheng 42097a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 42103612b495STsutomu Itoh if (IS_ERR(trans)) 42113612b495STsutomu Itoh return PTR_ERR(trans); 42123a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4213c146afadSYan Zheng } 4214c146afadSYan Zheng 421536c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 421636c86a9eSQu Wenruo { 421736c86a9eSQu Wenruo struct btrfs_transaction *trans; 421836c86a9eSQu Wenruo struct btrfs_transaction *tmp; 421936c86a9eSQu Wenruo bool found = false; 422036c86a9eSQu Wenruo 422136c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 422236c86a9eSQu Wenruo return; 422336c86a9eSQu Wenruo 422436c86a9eSQu Wenruo /* 422536c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 422636c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 422736c86a9eSQu Wenruo */ 422836c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 422936c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 423036c86a9eSQu Wenruo struct extent_state *cached = NULL; 423136c86a9eSQu Wenruo u64 dirty_bytes = 0; 423236c86a9eSQu Wenruo u64 cur = 0; 423336c86a9eSQu Wenruo u64 found_start; 423436c86a9eSQu Wenruo u64 found_end; 423536c86a9eSQu Wenruo 423636c86a9eSQu Wenruo found = true; 4237e5860f82SFilipe Manana while (find_first_extent_bit(&trans->dirty_pages, cur, 423836c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 423936c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 424036c86a9eSQu Wenruo cur = found_end + 1; 424136c86a9eSQu Wenruo } 424236c86a9eSQu Wenruo btrfs_warn(fs_info, 424336c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 424436c86a9eSQu Wenruo trans->transid, dirty_bytes); 424536c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 424636c86a9eSQu Wenruo 424736c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 424836c86a9eSQu Wenruo fs_info->running_transaction = NULL; 424936c86a9eSQu Wenruo list_del_init(&trans->list); 425036c86a9eSQu Wenruo 425136c86a9eSQu Wenruo btrfs_put_transaction(trans); 425236c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 425336c86a9eSQu Wenruo } 425436c86a9eSQu Wenruo ASSERT(!found); 425536c86a9eSQu Wenruo } 425636c86a9eSQu Wenruo 4257b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4258eb60ceacSChris Mason { 4259c146afadSYan Zheng int ret; 4260e089f05cSChris Mason 4261afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 426231e70e52SFilipe Manana 426331e70e52SFilipe Manana /* 42648a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 42658a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 42668a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 42678a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 42688a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 42698a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 42708a1f1e3dSFilipe Manana * return 1. 42718a1f1e3dSFilipe Manana */ 42728a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 42738a1f1e3dSFilipe Manana 42748a1f1e3dSFilipe Manana /* 427531e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 427631e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 427731e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 427831e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 427931e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 428031e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 428131e70e52SFilipe Manana * trying to stop the async reclaim task. 428231e70e52SFilipe Manana */ 428331e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4284d6fd0ae2SOmar Sandoval /* 4285d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4286d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4287d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4288d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4289d6fd0ae2SOmar Sandoval */ 4290d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4291a2135011SChris Mason 42927343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4293d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 42947343dd61SJustin Maggard 4295803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4296803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4297803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4298803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4299803b2f54SStefan Behrens 4300837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4301aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4302837d5b6eSIlya Dryomov 43038dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 43048dabb742SStefan Behrens 4305aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 43064cb5300bSChris Mason 43074cb5300bSChris Mason /* wait for any defraggers to finish */ 43084cb5300bSChris Mason wait_event(fs_info->transaction_wait, 43094cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 43104cb5300bSChris Mason 43114cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 431226176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 43134cb5300bSChris Mason 4314a362bb86SFilipe Manana /* 4315a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4316a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4317a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4318a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4319a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4320a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4321a362bb86SFilipe Manana * no one else to run iputs. 4322a362bb86SFilipe Manana * 4323a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4324a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4325a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4326a362bb86SFilipe Manana * through some other means. 4327a362bb86SFilipe Manana * 4328a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4329a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4330a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4331a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4332a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4333a362bb86SFilipe Manana */ 4334a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4335a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4336a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4337a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4338a362bb86SFilipe Manana 433921c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 434057056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4341576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 434221c7e756SMiao Xie 4343b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4344b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4345b0643e59SDennis Zhou 4346bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4347e44163e1SJeff Mahoney /* 4348d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4349d6fd0ae2SOmar Sandoval * unused block groups. 4350e44163e1SJeff Mahoney */ 43510b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4352e44163e1SJeff Mahoney 4353f0cc2cd7SFilipe Manana /* 4354f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4355f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4356f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4357f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4358f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4359f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4360f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4361f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4362f0cc2cd7SFilipe Manana * This is a very rare case. 4363f0cc2cd7SFilipe Manana */ 4364f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4365f0cc2cd7SFilipe Manana 43666bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4367d397712bSChris Mason if (ret) 436804892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4369c146afadSYan Zheng } 4370ed2ff2cbSChris Mason 437184961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 43722ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4373acce952bSliubo 4374e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4375e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 43768929ecfaSYan, Zheng 4377e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4378afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4379f25784b3SYan Zheng 43805958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 43815958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 43825958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 43835958253cSQu Wenruo } 43845958253cSQu Wenruo 438504892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4386fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4387bcef60f2SArne Jansen 4388963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 438904892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4390963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4391b0c68f8bSChris Mason } 439231153d81SYan Zheng 43935deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 43944297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 43955deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 43964297ff84SJosef Bacik 43976618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4398b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 43995ac1d209SJeff Mahoney 44001a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 44011a4319ccSLiu Bo 4402de348ee0SWang Shilong /* 4403de348ee0SWang Shilong * we must make sure there is not any read request to 4404de348ee0SWang Shilong * submit after we stopping all workers. 4405de348ee0SWang Shilong */ 4406de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 440796192499SJosef Bacik btrfs_stop_all_workers(fs_info); 440896192499SJosef Bacik 44090a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 441036c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 44110a31daa4SFilipe Manana 4412afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 44134273eaffSAnand Jain free_root_pointers(fs_info, true); 44148c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 44159ad6b7bcSChris Mason 44164e19443dSJosef Bacik /* 44174e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 44184e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 44194e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 44204e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 44214e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 44224e19443dSJosef Bacik */ 44234e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 44244e19443dSJosef Bacik 442513e6c37bSJosef Bacik iput(fs_info->btree_inode); 4426d6bfde87SChris Mason 442721adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 44280b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 44292ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 443021adbd5cSStefan Behrens #endif 443121adbd5cSStefan Behrens 44320b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 443368c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4434eb60ceacSChris Mason } 4435eb60ceacSChris Mason 4436d5e09e38SFilipe Manana void btrfs_mark_buffer_dirty(struct btrfs_trans_handle *trans, 4437d5e09e38SFilipe Manana struct extent_buffer *buf) 44385f39d397SChris Mason { 44392f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 44405f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4441b4ce94deSChris Mason 444206ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 444306ea65a3SJosef Bacik /* 444406ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 444552042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 444606ea65a3SJosef Bacik * outside of the sanity tests. 444706ea65a3SJosef Bacik */ 4448b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 444906ea65a3SJosef Bacik return; 445006ea65a3SJosef Bacik #endif 4451d5e09e38SFilipe Manana /* This is an active transaction (its state < TRANS_STATE_UNBLOCKED). */ 4452d5e09e38SFilipe Manana ASSERT(trans->transid == fs_info->generation); 445349d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 4454d5e09e38SFilipe Manana if (transid != fs_info->generation) { 44555d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 44560b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 4457d5e09e38SFilipe Manana btrfs_abort_transaction(trans, -EUCLEAN); 4458d5e09e38SFilipe Manana } 4459f18cc978SChristoph Hellwig set_extent_buffer_dirty(buf); 44601f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 446169fc6cbbSQu Wenruo /* 446285d8a826SJosef Bacik * btrfs_check_leaf() won't check item data if we don't have WRITTEN 446385d8a826SJosef Bacik * set, so this will only validate the basic structure of the items. 446469fc6cbbSQu Wenruo */ 446585d8a826SJosef Bacik if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(buf)) { 4466a4f78750SDavid Sterba btrfs_print_leaf(buf); 44671f21ef0aSFilipe Manana ASSERT(0); 44681f21ef0aSFilipe Manana } 44691f21ef0aSFilipe Manana #endif 4470eb60ceacSChris Mason } 4471eb60ceacSChris Mason 44722ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4473b53d3f5dSLiu Bo int flush_delayed) 447435b7e476SChris Mason { 4475188de649SChris Mason /* 4476188de649SChris Mason * looks as though older kernels can get into trouble with 4477188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4478188de649SChris Mason */ 4479e2d84521SMiao Xie int ret; 4480d6bfde87SChris Mason 44816933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4482d6bfde87SChris Mason return; 4483d6bfde87SChris Mason 4484b53d3f5dSLiu Bo if (flush_delayed) 44852ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 448616cdcec7SMiao Xie 4487d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4488d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4489d814a491SEthan Lien fs_info->dirty_metadata_batch); 4490e2d84521SMiao Xie if (ret > 0) { 44910b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 449216cdcec7SMiao Xie } 449316cdcec7SMiao Xie } 449416cdcec7SMiao Xie 44952ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 449616cdcec7SMiao Xie { 44972ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 449835b7e476SChris Mason } 4499b53d3f5dSLiu Bo 45002ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4501b53d3f5dSLiu Bo { 45022ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4503d6bfde87SChris Mason } 45046b80053dSChris Mason 45052ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4506acce952bSliubo { 4507fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4508fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4509fe816d0fSNikolay Borisov 45100b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45112ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45120b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4513acce952bSliubo 45140b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45150b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4516acce952bSliubo } 4517acce952bSliubo 4518ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4519ef67963dSJosef Bacik { 4520fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4521fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4522fc7cbcd4SDavid Sterba int ret; 4523ef67963dSJosef Bacik 4524fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4525fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4526fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4527fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4528fc7cbcd4SDavid Sterba int i; 4529ef67963dSJosef Bacik 4530fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4531fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4532fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4533fc7cbcd4SDavid Sterba 4534fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4535fc7cbcd4SDavid Sterba if (!gang[i]) 4536ef67963dSJosef Bacik continue; 4537fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4538fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4539fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4540ef67963dSJosef Bacik } 4541fc7cbcd4SDavid Sterba root_objectid++; 4542fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4543ef67963dSJosef Bacik } 4544fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4545ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4546ef67963dSJosef Bacik } 4547ef67963dSJosef Bacik 4548143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4549acce952bSliubo { 4550acce952bSliubo struct btrfs_ordered_extent *ordered; 4551acce952bSliubo 4552199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4553779880efSJosef Bacik /* 4554779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4555779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4556779880efSJosef Bacik */ 4557199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4558779880efSJosef Bacik root_extent_list) 4559779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4560199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4561199c2a9cSMiao Xie } 4562199c2a9cSMiao Xie 4563199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4564199c2a9cSMiao Xie { 4565199c2a9cSMiao Xie struct btrfs_root *root; 456684af994bSRuan Jinjie LIST_HEAD(splice); 4567199c2a9cSMiao Xie 4568199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4569199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4570199c2a9cSMiao Xie while (!list_empty(&splice)) { 4571199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4572199c2a9cSMiao Xie ordered_root); 45731de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 45741de2cfdeSJosef Bacik &fs_info->ordered_roots); 4575199c2a9cSMiao Xie 45762a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4577199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4578199c2a9cSMiao Xie 45792a85d9caSLiu Bo cond_resched(); 45802a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4581199c2a9cSMiao Xie } 4582199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 458374d5d229SJosef Bacik 458474d5d229SJosef Bacik /* 458574d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 458674d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 458774d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 458874d5d229SJosef Bacik * actually get run and error out properly. 458974d5d229SJosef Bacik */ 459074d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4591acce952bSliubo } 4592acce952bSliubo 459399f09ce3SFilipe Manana static void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 45942ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4595acce952bSliubo { 4596acce952bSliubo struct rb_node *node; 4597acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4598acce952bSliubo struct btrfs_delayed_ref_node *ref; 4599acce952bSliubo 4600acce952bSliubo delayed_refs = &trans->delayed_refs; 4601acce952bSliubo 4602acce952bSliubo spin_lock(&delayed_refs->lock); 4603d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4604cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4605b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 460699f09ce3SFilipe Manana return; 4607acce952bSliubo } 4608acce952bSliubo 46095c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4610d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 46110e0adbcfSJosef Bacik struct rb_node *n; 4612e78417d1SJosef Bacik bool pin_bytes = false; 4613acce952bSliubo 4614d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4615d7df2c79SJosef Bacik href_node); 46163069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4617b939d1abSJosef Bacik continue; 46183069bd26SJosef Bacik 4619d7df2c79SJosef Bacik spin_lock(&head->lock); 4620e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 46210e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 46220e0adbcfSJosef Bacik ref_node); 4623e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 46240e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 46251d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 46261d57ee94SWang Xiaoguang list_del(&ref->add_list); 4627d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4628d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4629d7df2c79SJosef Bacik } 463054067ae9SJosef Bacik if (head->must_insert_reserved) 4631e78417d1SJosef Bacik pin_bytes = true; 463278a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4633fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4634d7df2c79SJosef Bacik spin_unlock(&head->lock); 4635acce952bSliubo spin_unlock(&delayed_refs->lock); 4636e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4637acce952bSliubo 4638f603bb94SNikolay Borisov if (pin_bytes) { 4639f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4640f603bb94SNikolay Borisov 4641f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4642f603bb94SNikolay Borisov BUG_ON(!cache); 4643f603bb94SNikolay Borisov 4644f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4645f603bb94SNikolay Borisov spin_lock(&cache->lock); 4646f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4647f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4648f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4649f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4650f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4651f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4652f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4653f603bb94SNikolay Borisov 4654f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4655f603bb94SNikolay Borisov 4656f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4657f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4658f603bb94SNikolay Borisov } 465931890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4660d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4661acce952bSliubo cond_resched(); 4662acce952bSliubo spin_lock(&delayed_refs->lock); 4663acce952bSliubo } 466481f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4665acce952bSliubo 4666acce952bSliubo spin_unlock(&delayed_refs->lock); 4667acce952bSliubo } 4668acce952bSliubo 4669143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4670acce952bSliubo { 4671acce952bSliubo struct btrfs_inode *btrfs_inode; 467284af994bSRuan Jinjie LIST_HEAD(splice); 4673acce952bSliubo 4674eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4675eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4676acce952bSliubo 4677acce952bSliubo while (!list_empty(&splice)) { 4678fe816d0fSNikolay Borisov struct inode *inode = NULL; 4679eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4680acce952bSliubo delalloc_inodes); 4681fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4682eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4683acce952bSliubo 4684fe816d0fSNikolay Borisov /* 4685fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4686fe816d0fSNikolay Borisov * meanwhile. 4687fe816d0fSNikolay Borisov */ 4688fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4689fe816d0fSNikolay Borisov if (inode) { 4690597441b3SJosef Bacik unsigned int nofs_flag; 4691597441b3SJosef Bacik 4692597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4693fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4694597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4695fe816d0fSNikolay Borisov iput(inode); 4696fe816d0fSNikolay Borisov } 4697eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4698acce952bSliubo } 4699eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4700eb73c1b7SMiao Xie } 4701eb73c1b7SMiao Xie 4702eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4703eb73c1b7SMiao Xie { 4704eb73c1b7SMiao Xie struct btrfs_root *root; 470584af994bSRuan Jinjie LIST_HEAD(splice); 4706eb73c1b7SMiao Xie 4707eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4708eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4709eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4710eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4711eb73c1b7SMiao Xie delalloc_root); 471200246528SJosef Bacik root = btrfs_grab_root(root); 4713eb73c1b7SMiao Xie BUG_ON(!root); 4714eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4715eb73c1b7SMiao Xie 4716eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 471700246528SJosef Bacik btrfs_put_root(root); 4718eb73c1b7SMiao Xie 4719eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4720eb73c1b7SMiao Xie } 4721eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4722acce952bSliubo } 4723acce952bSliubo 4724aec5716cSFilipe Manana static void btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4725acce952bSliubo struct extent_io_tree *dirty_pages, 4726acce952bSliubo int mark) 4727acce952bSliubo { 4728acce952bSliubo struct extent_buffer *eb; 4729acce952bSliubo u64 start = 0; 4730acce952bSliubo u64 end; 4731acce952bSliubo 4732e5860f82SFilipe Manana while (find_first_extent_bit(dirty_pages, start, &start, &end, 4733aec5716cSFilipe Manana mark, NULL)) { 473491166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4735acce952bSliubo while (start <= end) { 47360b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 47370b246afaSJeff Mahoney start += fs_info->nodesize; 4738fd8b2b61SJosef Bacik if (!eb) 4739acce952bSliubo continue; 4740acce952bSliubo 4741c4e54a65SJosef Bacik btrfs_tree_lock(eb); 4742c4e54a65SJosef Bacik wait_on_extent_buffer_writeback(eb); 4743190a8339SJosef Bacik btrfs_clear_buffer_dirty(NULL, eb); 4744c4e54a65SJosef Bacik btrfs_tree_unlock(eb); 4745c4e54a65SJosef Bacik 4746fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4747acce952bSliubo } 4748acce952bSliubo } 4749acce952bSliubo } 4750acce952bSliubo 475146d81ebdSFilipe Manana static void btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4752fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4753acce952bSliubo { 4754acce952bSliubo u64 start; 4755acce952bSliubo u64 end; 4756acce952bSliubo 4757acce952bSliubo while (1) { 47580e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 47590e6ec385SFilipe Manana 4760fcd5e742SLu Fengqi /* 4761fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4762fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4763fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4764fcd5e742SLu Fengqi * the same extent range. 4765fcd5e742SLu Fengqi */ 4766fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4767e5860f82SFilipe Manana if (!find_first_extent_bit(unpin, 0, &start, &end, 476846d81ebdSFilipe Manana EXTENT_DIRTY, &cached_state)) { 4769fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4770acce952bSliubo break; 4771fcd5e742SLu Fengqi } 4772acce952bSliubo 47730e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 47740e6ec385SFilipe Manana free_extent_state(cached_state); 47752ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4776fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4777acce952bSliubo cond_resched(); 4778acce952bSliubo } 4779acce952bSliubo } 4780acce952bSliubo 478132da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4782c79a1751SLiu Bo { 4783c79a1751SLiu Bo struct inode *inode; 4784c79a1751SLiu Bo 4785c79a1751SLiu Bo inode = cache->io_ctl.inode; 4786c79a1751SLiu Bo if (inode) { 4787597441b3SJosef Bacik unsigned int nofs_flag; 4788597441b3SJosef Bacik 4789597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4790c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4791597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4792597441b3SJosef Bacik 4793c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4794c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4795c79a1751SLiu Bo iput(inode); 4796c79a1751SLiu Bo } 4797bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4798c79a1751SLiu Bo btrfs_put_block_group(cache); 4799c79a1751SLiu Bo } 4800c79a1751SLiu Bo 4801c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 48022ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4803c79a1751SLiu Bo { 480432da5386SDavid Sterba struct btrfs_block_group *cache; 4805c79a1751SLiu Bo 4806c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4807c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4808c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 480932da5386SDavid Sterba struct btrfs_block_group, 4810c79a1751SLiu Bo dirty_list); 4811c79a1751SLiu Bo 4812c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4813c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4814c79a1751SLiu Bo list_del_init(&cache->io_list); 4815c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4816c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4817c79a1751SLiu Bo } 4818c79a1751SLiu Bo 4819c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4820c79a1751SLiu Bo spin_lock(&cache->lock); 4821c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4822c79a1751SLiu Bo spin_unlock(&cache->lock); 4823c79a1751SLiu Bo 4824c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4825c79a1751SLiu Bo btrfs_put_block_group(cache); 4826ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4827c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4828c79a1751SLiu Bo } 4829c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4830c79a1751SLiu Bo 483145ae2c18SNikolay Borisov /* 483245ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 483345ae2c18SNikolay Borisov * to use it without any locking 483445ae2c18SNikolay Borisov */ 4835c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4836c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 483732da5386SDavid Sterba struct btrfs_block_group, 4838c79a1751SLiu Bo io_list); 4839c79a1751SLiu Bo 4840c79a1751SLiu Bo list_del_init(&cache->io_list); 4841c79a1751SLiu Bo spin_lock(&cache->lock); 4842c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4843c79a1751SLiu Bo spin_unlock(&cache->lock); 4844c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4845c79a1751SLiu Bo } 4846c79a1751SLiu Bo } 4847c79a1751SLiu Bo 4848624bc6f6SBoris Burkov static void btrfs_free_all_qgroup_pertrans(struct btrfs_fs_info *fs_info) 4849624bc6f6SBoris Burkov { 4850624bc6f6SBoris Burkov struct btrfs_root *gang[8]; 4851624bc6f6SBoris Burkov int i; 4852624bc6f6SBoris Burkov int ret; 4853624bc6f6SBoris Burkov 4854624bc6f6SBoris Burkov spin_lock(&fs_info->fs_roots_radix_lock); 4855624bc6f6SBoris Burkov while (1) { 4856624bc6f6SBoris Burkov ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix, 4857624bc6f6SBoris Burkov (void **)gang, 0, 4858624bc6f6SBoris Burkov ARRAY_SIZE(gang), 4859624bc6f6SBoris Burkov BTRFS_ROOT_TRANS_TAG); 4860624bc6f6SBoris Burkov if (ret == 0) 4861624bc6f6SBoris Burkov break; 4862624bc6f6SBoris Burkov for (i = 0; i < ret; i++) { 4863624bc6f6SBoris Burkov struct btrfs_root *root = gang[i]; 4864624bc6f6SBoris Burkov 4865624bc6f6SBoris Burkov btrfs_qgroup_free_meta_all_pertrans(root); 4866624bc6f6SBoris Burkov radix_tree_tag_clear(&fs_info->fs_roots_radix, 4867624bc6f6SBoris Burkov (unsigned long)root->root_key.objectid, 4868624bc6f6SBoris Burkov BTRFS_ROOT_TRANS_TAG); 4869624bc6f6SBoris Burkov } 4870624bc6f6SBoris Burkov } 4871624bc6f6SBoris Burkov spin_unlock(&fs_info->fs_roots_radix_lock); 4872624bc6f6SBoris Burkov } 4873624bc6f6SBoris Burkov 487449b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 48752ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 487649b25e05SJeff Mahoney { 4877bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4878bbbf7243SNikolay Borisov 48792ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4880c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4881c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4882c79a1751SLiu Bo 4883bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4884bbbf7243SNikolay Borisov post_commit_list) { 4885bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4886bbbf7243SNikolay Borisov } 4887bbbf7243SNikolay Borisov 48882ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 488949b25e05SJeff Mahoney 48904a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 48910b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 489249b25e05SJeff Mahoney 48934a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 48940b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 489549b25e05SJeff Mahoney 4896ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 489767cde344SMiao Xie 48982ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 489949b25e05SJeff Mahoney EXTENT_DIRTY); 4900fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 490149b25e05SJeff Mahoney 4902624bc6f6SBoris Burkov btrfs_free_all_qgroup_pertrans(fs_info); 4903624bc6f6SBoris Burkov 49044a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 49054a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 490649b25e05SJeff Mahoney } 490749b25e05SJeff Mahoney 49082ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4909acce952bSliubo { 4910acce952bSliubo struct btrfs_transaction *t; 4911acce952bSliubo 49120b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4913acce952bSliubo 49140b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49150b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 49160b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4917724e2315SJosef Bacik struct btrfs_transaction, list); 491877d20c68SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_PREP) { 49199b64f57dSElena Reshetova refcount_inc(&t->use_count); 49200b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49212ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4922724e2315SJosef Bacik btrfs_put_transaction(t); 49230b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4924724e2315SJosef Bacik continue; 4925724e2315SJosef Bacik } 49260b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4927724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 49280b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4929724e2315SJosef Bacik /* 4930724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4931724e2315SJosef Bacik * handle open currently for this transaction. 4932724e2315SJosef Bacik */ 4933724e2315SJosef Bacik wait_event(t->writer_wait, 4934724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4935724e2315SJosef Bacik } else { 49360b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4937724e2315SJosef Bacik } 49382ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4939724e2315SJosef Bacik 49400b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49410b246afaSJeff Mahoney if (t == fs_info->running_transaction) 49420b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4943724e2315SJosef Bacik list_del_init(&t->list); 49440b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4945a4abeea4SJosef Bacik 4946724e2315SJosef Bacik btrfs_put_transaction(t); 49472e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 49480b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4949724e2315SJosef Bacik } 49500b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49510b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4952ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4953ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 49540b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4955ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 49560b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4957acce952bSliubo 4958acce952bSliubo return 0; 4959acce952bSliubo } 4960ec7d6dfdSNikolay Borisov 4961453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 4962ec7d6dfdSNikolay Borisov { 4963ec7d6dfdSNikolay Borisov struct btrfs_path *path; 4964ec7d6dfdSNikolay Borisov int ret; 4965ec7d6dfdSNikolay Borisov struct extent_buffer *l; 4966ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 4967ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 4968ec7d6dfdSNikolay Borisov int slot; 4969ec7d6dfdSNikolay Borisov 4970ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 4971ec7d6dfdSNikolay Borisov if (!path) 4972ec7d6dfdSNikolay Borisov return -ENOMEM; 4973ec7d6dfdSNikolay Borisov 4974ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 4975ec7d6dfdSNikolay Borisov search_key.type = -1; 4976ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 4977ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 4978ec7d6dfdSNikolay Borisov if (ret < 0) 4979ec7d6dfdSNikolay Borisov goto error; 4980ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 4981ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 4982ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 4983ec7d6dfdSNikolay Borisov l = path->nodes[0]; 4984ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 498523125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 498623125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 4987ec7d6dfdSNikolay Borisov } else { 498823125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 4989ec7d6dfdSNikolay Borisov } 4990ec7d6dfdSNikolay Borisov ret = 0; 4991ec7d6dfdSNikolay Borisov error: 4992ec7d6dfdSNikolay Borisov btrfs_free_path(path); 4993ec7d6dfdSNikolay Borisov return ret; 4994ec7d6dfdSNikolay Borisov } 4995ec7d6dfdSNikolay Borisov 4996543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 4997ec7d6dfdSNikolay Borisov { 4998ec7d6dfdSNikolay Borisov int ret; 4999ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5000ec7d6dfdSNikolay Borisov 50016b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5002ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5003ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5004ec7d6dfdSNikolay Borisov root->root_key.objectid); 5005ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5006ec7d6dfdSNikolay Borisov goto out; 5007ec7d6dfdSNikolay Borisov } 5008ec7d6dfdSNikolay Borisov 500923125104SNikolay Borisov *objectid = root->free_objectid++; 5010ec7d6dfdSNikolay Borisov ret = 0; 5011ec7d6dfdSNikolay Borisov out: 5012ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5013ec7d6dfdSNikolay Borisov return ret; 5014ec7d6dfdSNikolay Borisov } 5015