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 320*67bc5ad0SAnand Jain /* 321*67bc5ad0SAnand Jain * alloc_fs_devices() copies the fsid into metadata_uuid if the 322*67bc5ad0SAnand Jain * metadata_uuid is unset in the superblock, including for a seed device. 323*67bc5ad0SAnand Jain * So, we can use fs_devices->metadata_uuid. 324*67bc5ad0SAnand Jain */ 325*67bc5ad0SAnand 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); 523139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 5240b32f4bbSJosef Bacik struct extent_buffer *eb; 525139e8cd3SQu Wenruo int cur_bit = 0; 5260079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 5270b32f4bbSJosef Bacik 528139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 5290079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 5300b32f4bbSJosef Bacik BUG_ON(!eb); 5310b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 5320b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 53349d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 5340079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 535139e8cd3SQu Wenruo } 5360079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 537139e8cd3SQu Wenruo 538139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 539139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 540139e8cd3SQu Wenruo unsigned long flags; 541139e8cd3SQu Wenruo u64 cur; 542139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 543139e8cd3SQu Wenruo 544139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 545139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 546139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 547139e8cd3SQu Wenruo cur_bit++; 548139e8cd3SQu Wenruo continue; 549139e8cd3SQu Wenruo } 550139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 551139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 552139e8cd3SQu Wenruo 553139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 554139e8cd3SQu Wenruo ASSERT(eb); 555139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 556139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 55749d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 558139e8cd3SQu Wenruo free_extent_buffer(eb); 559139e8cd3SQu Wenruo 560139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 561139e8cd3SQu Wenruo } 5620079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 5630b32f4bbSJosef Bacik } 5640079c3b1SMatthew Wilcox (Oracle) #else 5650079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 5660079c3b1SMatthew Wilcox (Oracle) #endif 5670b32f4bbSJosef Bacik 5687f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 5690da5468fSChris Mason .writepages = btree_writepages, 570f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 571895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 5728958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 5730079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 574d98237b3SChris Mason }; 575123abc88SChris Mason 5762ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 5772ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 5783fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 5793fbaf258SJosef Bacik int level) 5800999df54SChris Mason { 5810b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 5820b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 5833fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 5840999df54SChris Mason } 5850999df54SChris Mason 586581c1760SQu Wenruo /* 587581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 588581c1760SQu Wenruo * verification. 589581c1760SQu Wenruo * 590789d6a3aSQu Wenruo * @check: expected tree parentness check, see comments of the 591789d6a3aSQu Wenruo * structure for details. 592581c1760SQu Wenruo */ 5932ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 594789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 595e20d96d6SChris Mason { 5965f39d397SChris Mason struct extent_buffer *buf = NULL; 59719c00ddcSChris Mason int ret; 59819c00ddcSChris Mason 599789d6a3aSQu Wenruo ASSERT(check); 600789d6a3aSQu Wenruo 601789d6a3aSQu Wenruo buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, 602789d6a3aSQu Wenruo check->level); 603c871b0f2SLiu Bo if (IS_ERR(buf)) 604c871b0f2SLiu Bo return buf; 605e4204dedSChris Mason 606789d6a3aSQu Wenruo ret = btrfs_read_extent_buffer(buf, check); 6070f0fe8f7SFilipe David Borba Manana if (ret) { 608537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 60964c043deSLiu Bo return ERR_PTR(ret); 6100f0fe8f7SFilipe David Borba Manana } 611789d6a3aSQu Wenruo if (btrfs_check_eb_owner(buf, check->owner_root)) { 61288c602abSQu Wenruo free_extent_buffer_stale(buf); 61388c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 61488c602abSQu Wenruo } 6155f39d397SChris Mason return buf; 616ce9adaa5SChris Mason 617eb60ceacSChris Mason } 618eb60ceacSChris Mason 619da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 620e20d96d6SChris Mason u64 objectid) 621d97e63b6SChris Mason { 6227c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 6232e608bd1SJosef Bacik 6242e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 6252e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 6262e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 62796dfcb46SJosef Bacik root->fs_info = fs_info; 6282e608bd1SJosef Bacik root->root_key.objectid = objectid; 629cfaa7295SChris Mason root->node = NULL; 630a28ec197SChris Mason root->commit_root = NULL; 63127cdeb70SMiao Xie root->state = 0; 632abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 6330b86a832SChris Mason 6340f7d52f4SChris Mason root->last_trans = 0; 6356b8fad57SNikolay Borisov root->free_objectid = 0; 636eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 637199c2a9cSMiao Xie root->nr_ordered_extents = 0; 6386bef4d31SEric Paris root->inode_tree = RB_ROOT; 639088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 6402e608bd1SJosef Bacik 6412e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 6420b86a832SChris Mason 6430b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 6445d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 645eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 646eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 647199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 648199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 649d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 6502ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 6512ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 6525d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 653eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 654199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 655f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 6562ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 6572ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 6588287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 659a2135011SChris Mason mutex_init(&root->objectid_mutex); 660e02119d5SChris Mason mutex_init(&root->log_mutex); 66131f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 662573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 663c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 6647237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 6657237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 6667237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 6678b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 6688b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 6697237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 6707237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 6717237f183SYan Zheng atomic_set(&root->log_writers, 0); 6722ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 6730700cea7SElena Reshetova refcount_set(&root->refs, 1); 6748ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 675eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 6767237f183SYan Zheng root->log_transid = 0; 677d1433debSMiao Xie root->log_transid_committed = -1; 678257c62e1SChris Mason root->last_log_commit = 0; 6792e608bd1SJosef Bacik root->anon_dev = 0; 680e289f03eSFilipe Manana if (!dummy) { 68143eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 68235da5a7eSDavid Sterba IO_TREE_ROOT_DIRTY_LOG_PAGES); 683e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 68435da5a7eSDavid Sterba IO_TREE_LOG_CSUM_RANGE); 685e289f03eSFilipe Manana } 686017e5369SChris Mason 6875f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 688370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 689bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 690bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 691fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 692bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 693fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 694bd647ce3SJosef Bacik #endif 6953768f368SChris Mason } 6963768f368SChris Mason 69774e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 69896dfcb46SJosef Bacik u64 objectid, gfp_t flags) 6996f07e42eSAl Viro { 70074e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 7016f07e42eSAl Viro if (root) 70296dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 7036f07e42eSAl Viro return root; 7046f07e42eSAl Viro } 7056f07e42eSAl Viro 70606ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 70706ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 708da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 70906ea65a3SJosef Bacik { 71006ea65a3SJosef Bacik struct btrfs_root *root; 71106ea65a3SJosef Bacik 7127c0260eeSJeff Mahoney if (!fs_info) 7137c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 7147c0260eeSJeff Mahoney 71596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 71606ea65a3SJosef Bacik if (!root) 71706ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 718da17066cSJeff Mahoney 719b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 720faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 72106ea65a3SJosef Bacik 72206ea65a3SJosef Bacik return root; 72306ea65a3SJosef Bacik } 72406ea65a3SJosef Bacik #endif 72506ea65a3SJosef Bacik 726abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 727abed4aaaSJosef Bacik { 728abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 729abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 730abed4aaaSJosef Bacik 731abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 732abed4aaaSJosef Bacik } 733abed4aaaSJosef Bacik 734abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 735abed4aaaSJosef Bacik { 736abed4aaaSJosef Bacik const struct btrfs_key *key = k; 737abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 738abed4aaaSJosef Bacik 739abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 740abed4aaaSJosef Bacik } 741abed4aaaSJosef Bacik 742abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 743abed4aaaSJosef Bacik { 744abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 745abed4aaaSJosef Bacik struct rb_node *tmp; 746745806fbSQu Wenruo int ret = 0; 747abed4aaaSJosef Bacik 748abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 749abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 750abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 751abed4aaaSJosef Bacik 752745806fbSQu Wenruo if (tmp) { 753745806fbSQu Wenruo ret = -EEXIST; 754745806fbSQu Wenruo btrfs_warn(fs_info, "global root %llu %llu already exists", 755745806fbSQu Wenruo root->root_key.objectid, root->root_key.offset); 756745806fbSQu Wenruo } 757745806fbSQu Wenruo return ret; 758abed4aaaSJosef Bacik } 759abed4aaaSJosef Bacik 760abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 761abed4aaaSJosef Bacik { 762abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 763abed4aaaSJosef Bacik 764abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 765abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 766abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 767abed4aaaSJosef Bacik } 768abed4aaaSJosef Bacik 769abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 770abed4aaaSJosef Bacik struct btrfs_key *key) 771abed4aaaSJosef Bacik { 772abed4aaaSJosef Bacik struct rb_node *node; 773abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 774abed4aaaSJosef Bacik 775abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 776abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 777abed4aaaSJosef Bacik if (node) 778abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 779abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 780abed4aaaSJosef Bacik 781abed4aaaSJosef Bacik return root; 782abed4aaaSJosef Bacik } 783abed4aaaSJosef Bacik 784f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 785f7238e50SJosef Bacik { 786f7238e50SJosef Bacik struct btrfs_block_group *block_group; 787f7238e50SJosef Bacik u64 ret; 788f7238e50SJosef Bacik 789f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 790f7238e50SJosef Bacik return 0; 791f7238e50SJosef Bacik 792f7238e50SJosef Bacik if (bytenr) 793f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 794f7238e50SJosef Bacik else 795f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 796f7238e50SJosef Bacik ASSERT(block_group); 797f7238e50SJosef Bacik if (!block_group) 798f7238e50SJosef Bacik return 0; 799f7238e50SJosef Bacik ret = block_group->global_root_id; 800f7238e50SJosef Bacik btrfs_put_block_group(block_group); 801f7238e50SJosef Bacik 802f7238e50SJosef Bacik return ret; 803f7238e50SJosef Bacik } 804f7238e50SJosef Bacik 805abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 806abed4aaaSJosef Bacik { 807abed4aaaSJosef Bacik struct btrfs_key key = { 808abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 809abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 810f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 811abed4aaaSJosef Bacik }; 812abed4aaaSJosef Bacik 813abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 814abed4aaaSJosef Bacik } 815abed4aaaSJosef Bacik 816abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 817abed4aaaSJosef Bacik { 818abed4aaaSJosef Bacik struct btrfs_key key = { 819abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 820abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 821f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 822abed4aaaSJosef Bacik }; 823abed4aaaSJosef Bacik 824abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 825abed4aaaSJosef Bacik } 826abed4aaaSJosef Bacik 82751129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 82851129b33SJosef Bacik { 82951129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 83051129b33SJosef Bacik return fs_info->block_group_root; 83151129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 83251129b33SJosef Bacik } 83351129b33SJosef Bacik 83420897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 83520897f5cSArne Jansen u64 objectid) 83620897f5cSArne Jansen { 8379b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 83820897f5cSArne Jansen struct extent_buffer *leaf; 83920897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 84020897f5cSArne Jansen struct btrfs_root *root; 84120897f5cSArne Jansen struct btrfs_key key; 842b89f6d1fSFilipe Manana unsigned int nofs_flag; 84320897f5cSArne Jansen int ret = 0; 84420897f5cSArne Jansen 845b89f6d1fSFilipe Manana /* 846b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 847b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 848b89f6d1fSFilipe Manana */ 849b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 85096dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 851b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 85220897f5cSArne Jansen if (!root) 85320897f5cSArne Jansen return ERR_PTR(-ENOMEM); 85420897f5cSArne Jansen 85520897f5cSArne Jansen root->root_key.objectid = objectid; 85620897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 85720897f5cSArne Jansen root->root_key.offset = 0; 85820897f5cSArne Jansen 8599631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 8609631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 86120897f5cSArne Jansen if (IS_ERR(leaf)) { 86220897f5cSArne Jansen ret = PTR_ERR(leaf); 8631dd05682STsutomu Itoh leaf = NULL; 864c1b07854SPeng Hao goto fail; 86520897f5cSArne Jansen } 86620897f5cSArne Jansen 86720897f5cSArne Jansen root->node = leaf; 86820897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 86920897f5cSArne Jansen 87020897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 87127cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 87220897f5cSArne Jansen 873f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 874f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 87520897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 87620897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 87720897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 87820897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 87920897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 88020897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 88120897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 88233d85fdaSQu Wenruo if (is_fstree(objectid)) 883807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 884807fc790SAndy Shevchenko else 885807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 886c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 88720897f5cSArne Jansen 8888a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 8898a6a87cdSBoris Burkov 89020897f5cSArne Jansen key.objectid = objectid; 89120897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 89220897f5cSArne Jansen key.offset = 0; 89320897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 89420897f5cSArne Jansen if (ret) 89520897f5cSArne Jansen goto fail; 89620897f5cSArne Jansen 89720897f5cSArne Jansen return root; 8981dd05682STsutomu Itoh 8998a6a87cdSBoris Burkov fail: 90000246528SJosef Bacik btrfs_put_root(root); 9011dd05682STsutomu Itoh 9021dd05682STsutomu Itoh return ERR_PTR(ret); 90320897f5cSArne Jansen } 90420897f5cSArne Jansen 9057237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 906e02119d5SChris Mason struct btrfs_fs_info *fs_info) 9070f7d52f4SChris Mason { 9080f7d52f4SChris Mason struct btrfs_root *root; 909e02119d5SChris Mason 91096dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 911e02119d5SChris Mason if (!root) 9127237f183SYan Zheng return ERR_PTR(-ENOMEM); 913e02119d5SChris Mason 914e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 915e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 916e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 91727cdeb70SMiao Xie 9186ab6ebb7SNaohiro Aota return root; 9196ab6ebb7SNaohiro Aota } 9206ab6ebb7SNaohiro Aota 9216ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 9226ab6ebb7SNaohiro Aota struct btrfs_root *root) 9236ab6ebb7SNaohiro Aota { 9246ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 9256ab6ebb7SNaohiro Aota 9267237f183SYan Zheng /* 92792a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 92827cdeb70SMiao Xie * 92992a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 93092a7cc42SQu Wenruo * and they go away before a real commit is actually done. 93192a7cc42SQu Wenruo * 93292a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 93392a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 9347237f183SYan Zheng */ 935e02119d5SChris Mason 9364d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 9379631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 9386ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 9396ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 940e02119d5SChris Mason 9417237f183SYan Zheng root->node = leaf; 942e02119d5SChris Mason 943e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 944e02119d5SChris Mason btrfs_tree_unlock(root->node); 9456ab6ebb7SNaohiro Aota 9466ab6ebb7SNaohiro Aota return 0; 9477237f183SYan Zheng } 9487237f183SYan Zheng 9497237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 9507237f183SYan Zheng struct btrfs_fs_info *fs_info) 9517237f183SYan Zheng { 9527237f183SYan Zheng struct btrfs_root *log_root; 9537237f183SYan Zheng 9547237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 9557237f183SYan Zheng if (IS_ERR(log_root)) 9567237f183SYan Zheng return PTR_ERR(log_root); 9576ab6ebb7SNaohiro Aota 9583ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 9593ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 9603ddebf27SNaohiro Aota 9616ab6ebb7SNaohiro Aota if (ret) { 9626ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 9636ab6ebb7SNaohiro Aota return ret; 9646ab6ebb7SNaohiro Aota } 9653ddebf27SNaohiro Aota } 9666ab6ebb7SNaohiro Aota 9677237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 9687237f183SYan Zheng fs_info->log_root_tree = log_root; 9697237f183SYan Zheng return 0; 9707237f183SYan Zheng } 9717237f183SYan Zheng 9727237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 9737237f183SYan Zheng struct btrfs_root *root) 9747237f183SYan Zheng { 9750b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 9767237f183SYan Zheng struct btrfs_root *log_root; 9777237f183SYan Zheng struct btrfs_inode_item *inode_item; 9786ab6ebb7SNaohiro Aota int ret; 9797237f183SYan Zheng 9800b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 9817237f183SYan Zheng if (IS_ERR(log_root)) 9827237f183SYan Zheng return PTR_ERR(log_root); 9837237f183SYan Zheng 9846ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 9856ab6ebb7SNaohiro Aota if (ret) { 9866ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 9876ab6ebb7SNaohiro Aota return ret; 9886ab6ebb7SNaohiro Aota } 9896ab6ebb7SNaohiro Aota 9907237f183SYan Zheng log_root->last_trans = trans->transid; 9917237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 9927237f183SYan Zheng 9937237f183SYan Zheng inode_item = &log_root->root_item.inode; 9943cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 9953cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 9963cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 997da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 9980b246afaSJeff Mahoney fs_info->nodesize); 9993cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 10007237f183SYan Zheng 10015d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 10027237f183SYan Zheng 10037237f183SYan Zheng WARN_ON(root->log_root); 10047237f183SYan Zheng root->log_root = log_root; 10057237f183SYan Zheng root->log_transid = 0; 1006d1433debSMiao Xie root->log_transid_committed = -1; 1007257c62e1SChris Mason root->last_log_commit = 0; 1008e02119d5SChris Mason return 0; 1009e02119d5SChris Mason } 1010e02119d5SChris Mason 101149d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 101249d11beaSJosef Bacik struct btrfs_path *path, 1013cb517eabSMiao Xie struct btrfs_key *key) 1014e02119d5SChris Mason { 1015e02119d5SChris Mason struct btrfs_root *root; 1016789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 1017e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 101884234f3aSYan Zheng u64 generation; 1019cb517eabSMiao Xie int ret; 1020581c1760SQu Wenruo int level; 1021cb517eabSMiao Xie 102296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 102349d11beaSJosef Bacik if (!root) 102449d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 10250f7d52f4SChris Mason 1026cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1027cb517eabSMiao Xie &root->root_item, &root->root_key); 10280f7d52f4SChris Mason if (ret) { 102913a8a7c8SYan, Zheng if (ret > 0) 103013a8a7c8SYan, Zheng ret = -ENOENT; 103149d11beaSJosef Bacik goto fail; 10320f7d52f4SChris Mason } 103313a8a7c8SYan, Zheng 103484234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1035581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 1036789d6a3aSQu Wenruo check.level = level; 1037789d6a3aSQu Wenruo check.transid = generation; 1038789d6a3aSQu Wenruo check.owner_root = key->objectid; 1039789d6a3aSQu Wenruo root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), 1040789d6a3aSQu Wenruo &check); 104164c043deSLiu Bo if (IS_ERR(root->node)) { 104264c043deSLiu Bo ret = PTR_ERR(root->node); 10438c38938cSJosef Bacik root->node = NULL; 104449d11beaSJosef Bacik goto fail; 10454eb150d6SQu Wenruo } 10464eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1047cb517eabSMiao Xie ret = -EIO; 104849d11beaSJosef Bacik goto fail; 1049416bc658SJosef Bacik } 105088c602abSQu Wenruo 105188c602abSQu Wenruo /* 105288c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 105388c602abSQu Wenruo * match its root node owner 105488c602abSQu Wenruo */ 105588c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 105688c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 105788c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 105888c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 105988c602abSQu Wenruo btrfs_crit(fs_info, 106088c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 106188c602abSQu Wenruo root->root_key.objectid, root->node->start, 106288c602abSQu Wenruo btrfs_header_owner(root->node), 106388c602abSQu Wenruo root->root_key.objectid); 106488c602abSQu Wenruo ret = -EUCLEAN; 106588c602abSQu Wenruo goto fail; 106688c602abSQu Wenruo } 10675d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1068cb517eabSMiao Xie return root; 106949d11beaSJosef Bacik fail: 107000246528SJosef Bacik btrfs_put_root(root); 107149d11beaSJosef Bacik return ERR_PTR(ret); 107249d11beaSJosef Bacik } 107349d11beaSJosef Bacik 107449d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 107549d11beaSJosef Bacik struct btrfs_key *key) 107649d11beaSJosef Bacik { 107749d11beaSJosef Bacik struct btrfs_root *root; 107849d11beaSJosef Bacik struct btrfs_path *path; 107949d11beaSJosef Bacik 108049d11beaSJosef Bacik path = btrfs_alloc_path(); 108149d11beaSJosef Bacik if (!path) 108249d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 108349d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 108449d11beaSJosef Bacik btrfs_free_path(path); 108549d11beaSJosef Bacik 108649d11beaSJosef Bacik return root; 1087cb517eabSMiao Xie } 1088cb517eabSMiao Xie 10892dfb1e43SQu Wenruo /* 10902dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 10912dfb1e43SQu Wenruo * 10922dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 10932dfb1e43SQu Wenruo */ 10942dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1095cb517eabSMiao Xie { 1096cb517eabSMiao Xie int ret; 1097cb517eabSMiao Xie 10980b548539SDavid Sterba btrfs_drew_lock_init(&root->snapshot_lock); 10998257b2dcSMiao Xie 1100aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 11016ebcd021SQu Wenruo !btrfs_is_data_reloc_root(root) && 11026ebcd021SQu Wenruo is_fstree(root->root_key.objectid)) { 110392a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1104f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1105f39e4571SJosef Bacik } 1106f39e4571SJosef Bacik 1107851fd730SQu Wenruo /* 1108851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1109851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1110851fd730SQu Wenruo */ 1111851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1112851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 11132dfb1e43SQu Wenruo if (!anon_dev) { 1114cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1115cb517eabSMiao Xie if (ret) 1116876d2cf1SLiu Bo goto fail; 11172dfb1e43SQu Wenruo } else { 11182dfb1e43SQu Wenruo root->anon_dev = anon_dev; 11192dfb1e43SQu Wenruo } 1120851fd730SQu Wenruo } 1121f32e48e9SChandan Rajendra 1122f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1123453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1124f32e48e9SChandan Rajendra if (ret) { 1125f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1126876d2cf1SLiu Bo goto fail; 1127f32e48e9SChandan Rajendra } 1128f32e48e9SChandan Rajendra 11296b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1130f32e48e9SChandan Rajendra 1131f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1132f32e48e9SChandan Rajendra 1133cb517eabSMiao Xie return 0; 1134cb517eabSMiao Xie fail: 113584db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1136cb517eabSMiao Xie return ret; 1137cb517eabSMiao Xie } 1138cb517eabSMiao Xie 1139a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1140cb517eabSMiao Xie u64 root_id) 1141cb517eabSMiao Xie { 1142cb517eabSMiao Xie struct btrfs_root *root; 1143cb517eabSMiao Xie 1144fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1145fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1146fc7cbcd4SDavid Sterba (unsigned long)root_id); 114700246528SJosef Bacik root = btrfs_grab_root(root); 1148fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1149cb517eabSMiao Xie return root; 1150cb517eabSMiao Xie } 1151cb517eabSMiao Xie 115249d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 115349d11beaSJosef Bacik u64 objectid) 115449d11beaSJosef Bacik { 1155abed4aaaSJosef Bacik struct btrfs_key key = { 1156abed4aaaSJosef Bacik .objectid = objectid, 1157abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1158abed4aaaSJosef Bacik .offset = 0, 1159abed4aaaSJosef Bacik }; 1160abed4aaaSJosef Bacik 1161e91909aaSChristoph Hellwig switch (objectid) { 1162e91909aaSChristoph Hellwig case BTRFS_ROOT_TREE_OBJECTID: 116349d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 1164e91909aaSChristoph Hellwig case BTRFS_EXTENT_TREE_OBJECTID: 1165abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1166e91909aaSChristoph Hellwig case BTRFS_CHUNK_TREE_OBJECTID: 116749d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 1168e91909aaSChristoph Hellwig case BTRFS_DEV_TREE_OBJECTID: 116949d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 1170e91909aaSChristoph Hellwig case BTRFS_CSUM_TREE_OBJECTID: 1171abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1172e91909aaSChristoph Hellwig case BTRFS_QUOTA_TREE_OBJECTID: 117385724171SChristoph Hellwig return btrfs_grab_root(fs_info->quota_root); 1174e91909aaSChristoph Hellwig case BTRFS_UUID_TREE_OBJECTID: 117585724171SChristoph Hellwig return btrfs_grab_root(fs_info->uuid_root); 1176e91909aaSChristoph Hellwig case BTRFS_BLOCK_GROUP_TREE_OBJECTID: 117785724171SChristoph Hellwig return btrfs_grab_root(fs_info->block_group_root); 1178e91909aaSChristoph Hellwig case BTRFS_FREE_SPACE_TREE_OBJECTID: 117985724171SChristoph Hellwig return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1180e91909aaSChristoph Hellwig default: 118149d11beaSJosef Bacik return NULL; 118249d11beaSJosef Bacik } 1183e91909aaSChristoph Hellwig } 118449d11beaSJosef Bacik 1185cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1186cb517eabSMiao Xie struct btrfs_root *root) 1187cb517eabSMiao Xie { 1188cb517eabSMiao Xie int ret; 1189cb517eabSMiao Xie 1190fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1191fc7cbcd4SDavid Sterba if (ret) 1192fc7cbcd4SDavid Sterba return ret; 1193fc7cbcd4SDavid Sterba 1194fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1195fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1196fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1197fc7cbcd4SDavid Sterba root); 1198af01d2e5SJosef Bacik if (ret == 0) { 119900246528SJosef Bacik btrfs_grab_root(root); 1200fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1201af01d2e5SJosef Bacik } 1202fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1203fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1204cb517eabSMiao Xie 1205cb517eabSMiao Xie return ret; 1206cb517eabSMiao Xie } 1207cb517eabSMiao Xie 1208bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1209bd647ce3SJosef Bacik { 1210bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1211bd647ce3SJosef Bacik struct btrfs_root *root; 1212bd647ce3SJosef Bacik 1213bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1214457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1215457f1864SJosef Bacik 1216bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1217bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1218457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 121971008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1220bd647ce3SJosef Bacik refcount_read(&root->refs)); 1221bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 122200246528SJosef Bacik btrfs_put_root(root); 122300246528SJosef Bacik btrfs_put_root(root); 1224bd647ce3SJosef Bacik } 1225bd647ce3SJosef Bacik #endif 1226bd647ce3SJosef Bacik } 1227bd647ce3SJosef Bacik 1228abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1229abed4aaaSJosef Bacik { 1230abed4aaaSJosef Bacik struct btrfs_root *root; 1231abed4aaaSJosef Bacik struct rb_node *node; 1232abed4aaaSJosef Bacik 1233abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1234abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1235abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1236abed4aaaSJosef Bacik btrfs_put_root(root); 1237abed4aaaSJosef Bacik } 1238abed4aaaSJosef Bacik } 1239abed4aaaSJosef Bacik 12400d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 12410d4b0463SJosef Bacik { 1242141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1243141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 12445deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1245141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1246141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1247141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1248141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 12490d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 12500d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1251abed4aaaSJosef Bacik free_global_roots(fs_info); 125200246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 125300246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 125400246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 125500246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 125600246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 125700246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1258aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 12599c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1260bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 12613fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 12620d4b0463SJosef Bacik kfree(fs_info->super_copy); 12630d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 12648481dd80SQu Wenruo kfree(fs_info->subpage_info); 12650d4b0463SJosef Bacik kvfree(fs_info); 12660d4b0463SJosef Bacik } 12670d4b0463SJosef Bacik 12680d4b0463SJosef Bacik 12692dfb1e43SQu Wenruo /* 12702dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 12712dfb1e43SQu Wenruo * 12722dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 12732dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 12742dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 12752dfb1e43SQu Wenruo * 12762dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 12772dfb1e43SQu Wenruo * 12782dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 12792dfb1e43SQu Wenruo * same root objectid. 12802dfb1e43SQu Wenruo * 12812dfb1e43SQu Wenruo * @objectid: root id 12822dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 12832dfb1e43SQu Wenruo * pass 0 for new allocation. 12842dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 12852dfb1e43SQu Wenruo * for orphan roots 12862dfb1e43SQu Wenruo */ 12872dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 12882dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 12892dfb1e43SQu Wenruo bool check_ref) 12905eda7b5eSChris Mason { 12915eda7b5eSChris Mason struct btrfs_root *root; 1292381cf658SDavid Sterba struct btrfs_path *path; 12931d4c08e0SDavid Sterba struct btrfs_key key; 12945eda7b5eSChris Mason int ret; 12955eda7b5eSChris Mason 129649d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 129749d11beaSJosef Bacik if (root) 129849d11beaSJosef Bacik return root; 1299773e722aSQu Wenruo 1300773e722aSQu Wenruo /* 1301773e722aSQu Wenruo * If we're called for non-subvolume trees, and above function didn't 1302773e722aSQu Wenruo * find one, do not try to read it from disk. 1303773e722aSQu Wenruo * 1304773e722aSQu Wenruo * This is namely for free-space-tree and quota tree, which can change 1305773e722aSQu Wenruo * at runtime and should only be grabbed from fs_info. 1306773e722aSQu Wenruo */ 1307773e722aSQu Wenruo if (!is_fstree(objectid) && objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) 1308773e722aSQu Wenruo return ERR_PTR(-ENOENT); 13094df27c4dSYan, Zheng again: 131056e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 131148475471SStefan Behrens if (root) { 13122dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 13132dfb1e43SQu Wenruo ASSERT(!anon_dev); 1314bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 131500246528SJosef Bacik btrfs_put_root(root); 131648475471SStefan Behrens return ERR_PTR(-ENOENT); 1317bc44d7c4SJosef Bacik } 13185eda7b5eSChris Mason return root; 131948475471SStefan Behrens } 13205eda7b5eSChris Mason 132156e9357aSDavid Sterba key.objectid = objectid; 132256e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 132356e9357aSDavid Sterba key.offset = (u64)-1; 132456e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 13255eda7b5eSChris Mason if (IS_ERR(root)) 13265eda7b5eSChris Mason return root; 13273394e160SChris Mason 1328c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1329d68fc57bSYan, Zheng ret = -ENOENT; 1330d68fc57bSYan, Zheng goto fail; 1331d68fc57bSYan, Zheng } 1332d68fc57bSYan, Zheng 13332dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1334cb517eabSMiao Xie if (ret) 1335cb517eabSMiao Xie goto fail; 1336cb517eabSMiao Xie 1337381cf658SDavid Sterba path = btrfs_alloc_path(); 1338381cf658SDavid Sterba if (!path) { 1339381cf658SDavid Sterba ret = -ENOMEM; 1340381cf658SDavid Sterba goto fail; 1341381cf658SDavid Sterba } 13421d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 13431d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 134456e9357aSDavid Sterba key.offset = objectid; 13451d4c08e0SDavid Sterba 13461d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1347381cf658SDavid Sterba btrfs_free_path(path); 1348d68fc57bSYan, Zheng if (ret < 0) 1349d68fc57bSYan, Zheng goto fail; 1350d68fc57bSYan, Zheng if (ret == 0) 135127cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1352d68fc57bSYan, Zheng 1353cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 13540f7d52f4SChris Mason if (ret) { 1355168a2f77SJia-Ju Bai if (ret == -EEXIST) { 135600246528SJosef Bacik btrfs_put_root(root); 13574df27c4dSYan, Zheng goto again; 1358168a2f77SJia-Ju Bai } 13594df27c4dSYan, Zheng goto fail; 13604df27c4dSYan, Zheng } 1361edbd8d4eSChris Mason return root; 13624df27c4dSYan, Zheng fail: 136333fab972SFilipe Manana /* 136433fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1365143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 136633fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 136733fab972SFilipe Manana * and once again by our caller. 136833fab972SFilipe Manana */ 136933fab972SFilipe Manana if (anon_dev) 137033fab972SFilipe Manana root->anon_dev = 0; 13718c38938cSJosef Bacik btrfs_put_root(root); 13724df27c4dSYan, Zheng return ERR_PTR(ret); 1373edbd8d4eSChris Mason } 1374edbd8d4eSChris Mason 13752dfb1e43SQu Wenruo /* 13762dfb1e43SQu Wenruo * Get in-memory reference of a root structure 13772dfb1e43SQu Wenruo * 13782dfb1e43SQu Wenruo * @objectid: tree objectid 13792dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 13802dfb1e43SQu Wenruo * one reference 13812dfb1e43SQu Wenruo */ 13822dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 13832dfb1e43SQu Wenruo u64 objectid, bool check_ref) 13842dfb1e43SQu Wenruo { 13852dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 13862dfb1e43SQu Wenruo } 13872dfb1e43SQu Wenruo 13882dfb1e43SQu Wenruo /* 13892dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 13902dfb1e43SQu Wenruo * the anonymous block device id 13912dfb1e43SQu Wenruo * 13922dfb1e43SQu Wenruo * @objectid: tree objectid 13932dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 13942dfb1e43SQu Wenruo * parameter value 13952dfb1e43SQu Wenruo */ 13962dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 13972dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 13982dfb1e43SQu Wenruo { 13992dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 14002dfb1e43SQu Wenruo } 14012dfb1e43SQu Wenruo 14028b712842SChris Mason /* 140349d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 140449d11beaSJosef Bacik * @fs_info: the fs_info 140549d11beaSJosef Bacik * @objectid: the objectid we need to lookup 140649d11beaSJosef Bacik * 140749d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 140849d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 140949d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 141049d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 141149d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 141249d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 141349d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 141449d11beaSJosef Bacik * backref code is finished using the root. 141549d11beaSJosef Bacik */ 141649d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 141749d11beaSJosef Bacik struct btrfs_path *path, 141849d11beaSJosef Bacik u64 objectid) 141949d11beaSJosef Bacik { 142049d11beaSJosef Bacik struct btrfs_root *root; 142149d11beaSJosef Bacik struct btrfs_key key; 142249d11beaSJosef Bacik 142349d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 142449d11beaSJosef Bacik 142549d11beaSJosef Bacik /* 142649d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 142749d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 142849d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 142949d11beaSJosef Bacik * != NULL just return it. 143049d11beaSJosef Bacik */ 143149d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 143249d11beaSJosef Bacik if (root) 143349d11beaSJosef Bacik return root; 143449d11beaSJosef Bacik 143549d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 143649d11beaSJosef Bacik if (root) 143749d11beaSJosef Bacik return root; 143849d11beaSJosef Bacik 143949d11beaSJosef Bacik key.objectid = objectid; 144049d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 144149d11beaSJosef Bacik key.offset = (u64)-1; 144249d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 144349d11beaSJosef Bacik btrfs_release_path(path); 144449d11beaSJosef Bacik 144549d11beaSJosef Bacik return root; 144649d11beaSJosef Bacik } 144749d11beaSJosef Bacik 1448a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1449a74a4b97SChris Mason { 14500d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1451d0278245SMiao Xie int again; 1452a74a4b97SChris Mason 1453d6fd0ae2SOmar Sandoval while (1) { 1454d0278245SMiao Xie again = 0; 14559d1a2a3aSDavid Sterba 1456fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1457fd340d0fSJosef Bacik 1458d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 14592ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1460d0278245SMiao Xie goto sleep; 1461d0278245SMiao Xie 146290c711abSZygo Blaxell /* 146390c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 146490c711abSZygo Blaxell * before we have finished mounting the filesystem. 146590c711abSZygo Blaxell */ 14660b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 146790c711abSZygo Blaxell goto sleep; 146890c711abSZygo Blaxell 14690b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1470d0278245SMiao Xie goto sleep; 1471d0278245SMiao Xie 1472dc7f370cSMiao Xie /* 1473dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1474dc7f370cSMiao Xie * during the above check and trylock. 1475dc7f370cSMiao Xie */ 14762ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 14770b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1478dc7f370cSMiao Xie goto sleep; 1479dc7f370cSMiao Xie } 1480dc7f370cSMiao Xie 1481b7625f46SQu Wenruo if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) 1482b7625f46SQu Wenruo btrfs_sysfs_feature_update(fs_info); 1483b7625f46SQu Wenruo 14842ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1485c2d6cb16SFilipe Manana 148633c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 14870b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1488a74a4b97SChris Mason 1489d0278245SMiao Xie /* 149005323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 149105323cd1SMiao Xie * needn't do anything special here. 1492d0278245SMiao Xie */ 14930b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 149467c5e7d4SFilipe Manana 149567c5e7d4SFilipe Manana /* 1496f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 149767c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 149867c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1499f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 150067c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 150167c5e7d4SFilipe Manana * unused block groups. 150267c5e7d4SFilipe Manana */ 15030b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 150418bb8bbfSJohannes Thumshirn 150518bb8bbfSJohannes Thumshirn /* 150618bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 150718bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 150818bb8bbfSJohannes Thumshirn * space. 150918bb8bbfSJohannes Thumshirn */ 151018bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1511d0278245SMiao Xie sleep: 1512a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1513d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1514d6fd0ae2SOmar Sandoval kthread_parkme(); 1515d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1516d6fd0ae2SOmar Sandoval return 0; 1517838fe188SJiri Kosina if (!again) { 1518a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1519a74a4b97SChris Mason schedule(); 1520a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1521a74a4b97SChris Mason } 1522da288d28SFilipe Manana } 1523a74a4b97SChris Mason } 1524a74a4b97SChris Mason 1525a74a4b97SChris Mason static int transaction_kthread(void *arg) 1526a74a4b97SChris Mason { 1527a74a4b97SChris Mason struct btrfs_root *root = arg; 15280b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1529a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1530a74a4b97SChris Mason struct btrfs_transaction *cur; 15318929ecfaSYan, Zheng u64 transid; 1532643900beSNikolay Borisov time64_t delta; 1533a74a4b97SChris Mason unsigned long delay; 1534914b2007SJan Kara bool cannot_commit; 1535a74a4b97SChris Mason 1536a74a4b97SChris Mason do { 1537914b2007SJan Kara cannot_commit = false; 1538ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 15390b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1540a74a4b97SChris Mason 15410b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 15420b246afaSJeff Mahoney cur = fs_info->running_transaction; 1543a74a4b97SChris Mason if (!cur) { 15440b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1545a74a4b97SChris Mason goto sleep; 1546a74a4b97SChris Mason } 154731153d81SYan Zheng 1548643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1549fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1550fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1551643900beSNikolay Borisov delta < fs_info->commit_interval) { 15520b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1553fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1554fb8a7e94SNikolay Borisov delay = min(delay, 1555fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1556a74a4b97SChris Mason goto sleep; 1557a74a4b97SChris Mason } 15588929ecfaSYan, Zheng transid = cur->transid; 15590b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 156056bec294SChris Mason 156179787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1562354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1563914b2007SJan Kara if (IS_ERR(trans)) { 1564354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1565914b2007SJan Kara cannot_commit = true; 156679787eaaSJeff Mahoney goto sleep; 1567914b2007SJan Kara } 15688929ecfaSYan, Zheng if (transid == trans->transid) { 15693a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 15708929ecfaSYan, Zheng } else { 15713a45bb20SJeff Mahoney btrfs_end_transaction(trans); 15728929ecfaSYan, Zheng } 1573a74a4b97SChris Mason sleep: 15740b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 15750b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1576a74a4b97SChris Mason 157784961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 15782ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 15798929ecfaSYan, Zheng if (!kthread_should_stop() && 15800b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1581914b2007SJan Kara cannot_commit)) 1582bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1583a74a4b97SChris Mason } while (!kthread_should_stop()); 1584a74a4b97SChris Mason return 0; 1585a74a4b97SChris Mason } 1586a74a4b97SChris Mason 1587af31f5e5SChris Mason /* 158801f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 158901f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 159001f0f9daSNikolay Borisov * anything. 1591af31f5e5SChris Mason * 1592af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1593af31f5e5SChris Mason * generation of the latest root in the array with the generation 1594af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1595af31f5e5SChris Mason */ 159601f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1597af31f5e5SChris Mason { 159801f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1599af31f5e5SChris Mason u64 cur; 1600af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1601af31f5e5SChris Mason int i; 1602af31f5e5SChris Mason 1603af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1604af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1605af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1606af31f5e5SChris Mason if (cur == newest_gen) 160701f0f9daSNikolay Borisov return i; 1608af31f5e5SChris Mason } 1609af31f5e5SChris Mason 161001f0f9daSNikolay Borisov return -EINVAL; 1611af31f5e5SChris Mason } 1612af31f5e5SChris Mason 1613af31f5e5SChris Mason /* 1614af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1615af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1616af31f5e5SChris Mason * done 1617af31f5e5SChris Mason */ 1618af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1619af31f5e5SChris Mason { 16206ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1621af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1622af31f5e5SChris Mason 1623af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1624af31f5e5SChris Mason 1625af31f5e5SChris Mason /* 1626af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1627af31f5e5SChris Mason * regardless of which ones we use today 1628af31f5e5SChris Mason */ 1629af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1630af31f5e5SChris Mason 1631af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1632af31f5e5SChris Mason 1633af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1634af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1635af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1636af31f5e5SChris Mason 1637af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1638af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1639af31f5e5SChris Mason 1640af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1641af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1642af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1643af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1644af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1645af31f5e5SChris Mason 16461c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 16479c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1648f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 16499c54e80dSJosef Bacik 16509c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 16519c54e80dSJosef Bacik extent_root->node->start); 1652af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 165329cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 1654af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 165529cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 1656af31f5e5SChris Mason 1657f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 1658f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 1659f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 1660f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 1661f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 16629c54e80dSJosef Bacik } 1663af31f5e5SChris Mason 16647c7e82a7SChris Mason /* 16657c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 16667c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 16677c7e82a7SChris Mason */ 16687c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 16697c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 16707c7e82a7SChris Mason info->fs_root->node->start); 1671af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1672af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1673af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1674af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 16757c7e82a7SChris Mason } 1676af31f5e5SChris Mason 1677af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1678af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1679af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1680af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1681af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1682af31f5e5SChris Mason 1683af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1684af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1685af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1686af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1687af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1688af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1689af31f5e5SChris Mason 1690af31f5e5SChris Mason /* 1691af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1692af31f5e5SChris Mason * for the next commit 1693af31f5e5SChris Mason */ 1694af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1695af31f5e5SChris Mason &info->super_for_commit->super_roots, 1696af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1697af31f5e5SChris Mason } 1698af31f5e5SChris Mason 1699af31f5e5SChris Mason /* 1700bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1701bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1702bd2336b2SNikolay Borisov * 1703bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1704bd2336b2SNikolay Borisov * priority - priority of backup root required 1705bd2336b2SNikolay Borisov * 1706bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1707bd2336b2SNikolay Borisov */ 1708bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1709bd2336b2SNikolay Borisov { 1710bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1711bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1712bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1713bd2336b2SNikolay Borisov 1714bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1715bd2336b2SNikolay Borisov if (priority == 0) 1716bd2336b2SNikolay Borisov return backup_index; 1717bd2336b2SNikolay Borisov 1718bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1719bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1720bd2336b2SNikolay Borisov } else { 1721bd2336b2SNikolay Borisov return -EINVAL; 1722bd2336b2SNikolay Borisov } 1723bd2336b2SNikolay Borisov 1724bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1725bd2336b2SNikolay Borisov 1726bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1727bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1728bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1729bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1730bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1731bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1732bd2336b2SNikolay Borisov 1733bd2336b2SNikolay Borisov /* 1734bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1735bd2336b2SNikolay Borisov * need a fsck 1736bd2336b2SNikolay Borisov */ 1737bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1738bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1739bd2336b2SNikolay Borisov 1740bd2336b2SNikolay Borisov return backup_index; 1741bd2336b2SNikolay Borisov } 1742bd2336b2SNikolay Borisov 17437abadb64SLiu Bo /* helper to cleanup workers */ 17447abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 17457abadb64SLiu Bo { 1746dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1747afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 17485cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1749d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 1750d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 1751385de0efSChristoph Hellwig if (fs_info->rmw_workers) 1752385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 1753fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 1754fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 1755fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1756fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 17575b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1758e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1759a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1760fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1761b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1762b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1763a9b9477dSFilipe Manana /* 1764a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 1765a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 1766a9b9477dSFilipe Manana * queues can do metadata I/O operations. 1767a9b9477dSFilipe Manana */ 1768d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 1769d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 17707abadb64SLiu Bo } 17717abadb64SLiu Bo 17722e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 17732e9f5954SRashika { 17742e9f5954SRashika if (root) { 17752e9f5954SRashika free_extent_buffer(root->node); 17762e9f5954SRashika free_extent_buffer(root->commit_root); 17772e9f5954SRashika root->node = NULL; 17782e9f5954SRashika root->commit_root = NULL; 17792e9f5954SRashika } 17802e9f5954SRashika } 17812e9f5954SRashika 1782abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 1783abed4aaaSJosef Bacik { 1784abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 1785abed4aaaSJosef Bacik 1786abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 1787abed4aaaSJosef Bacik &fs_info->global_root_tree, 1788abed4aaaSJosef Bacik rb_node) 1789abed4aaaSJosef Bacik free_root_extent_buffers(root); 1790abed4aaaSJosef Bacik } 1791abed4aaaSJosef Bacik 1792af31f5e5SChris Mason /* helper to cleanup tree roots */ 17934273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 1794af31f5e5SChris Mason { 17952e9f5954SRashika free_root_extent_buffers(info->tree_root); 1796655b09feSJosef Bacik 1797abed4aaaSJosef Bacik free_global_root_pointers(info); 17982e9f5954SRashika free_root_extent_buffers(info->dev_root); 17992e9f5954SRashika free_root_extent_buffers(info->quota_root); 18002e9f5954SRashika free_root_extent_buffers(info->uuid_root); 18018c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 1802aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 18039c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 18044273eaffSAnand Jain if (free_chunk_root) 18052e9f5954SRashika free_root_extent_buffers(info->chunk_root); 1806af31f5e5SChris Mason } 1807af31f5e5SChris Mason 18088c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 18098c38938cSJosef Bacik { 18108c38938cSJosef Bacik if (!root) 18118c38938cSJosef Bacik return; 18128c38938cSJosef Bacik 18138c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 18148c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 18151dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 18168c38938cSJosef Bacik if (root->anon_dev) 18178c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 1818923eb523SJohannes Thumshirn free_root_extent_buffers(root); 18198c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1820fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 18218c38938cSJosef Bacik list_del_init(&root->leak_list); 1822fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 18238c38938cSJosef Bacik #endif 18248c38938cSJosef Bacik kfree(root); 18258c38938cSJosef Bacik } 18268c38938cSJosef Bacik } 18278c38938cSJosef Bacik 1828faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 1829171f6537SJosef Bacik { 1830fc7cbcd4SDavid Sterba int ret; 1831fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 1832fc7cbcd4SDavid Sterba int i; 1833171f6537SJosef Bacik 1834171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 1835fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 1836171f6537SJosef Bacik struct btrfs_root, root_list); 1837fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 1838171f6537SJosef Bacik 1839fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 1840fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 1841fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 1842171f6537SJosef Bacik } 1843171f6537SJosef Bacik 1844fc7cbcd4SDavid Sterba while (1) { 1845fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 1846fc7cbcd4SDavid Sterba (void **)gang, 0, 1847fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 1848fc7cbcd4SDavid Sterba if (!ret) 1849fc7cbcd4SDavid Sterba break; 1850fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 1851fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 1852171f6537SJosef Bacik } 1853171f6537SJosef Bacik } 1854af31f5e5SChris Mason 1855638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 1856638aa7edSEric Sandeen { 1857638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 1858638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 1859638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 1860638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 1861638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 1862638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 1863ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 1864638aa7edSEric Sandeen } 1865638aa7edSEric Sandeen 1866779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 1867779a65a4SEric Sandeen { 1868779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 1869779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 1870779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 1871779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 1872779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 1873779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 1874907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 1875779a65a4SEric Sandeen } 1876779a65a4SEric Sandeen 1877dcb2137cSChristoph Hellwig static int btrfs_init_btree_inode(struct super_block *sb) 1878f37938e0SEric Sandeen { 1879dcb2137cSChristoph Hellwig struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1880e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 1881e256927bSJosef Bacik fs_info->tree_root); 1882dcb2137cSChristoph Hellwig struct inode *inode; 1883dcb2137cSChristoph Hellwig 1884dcb2137cSChristoph Hellwig inode = new_inode(sb); 1885dcb2137cSChristoph Hellwig if (!inode) 1886dcb2137cSChristoph Hellwig return -ENOMEM; 18872ff7e61eSJeff Mahoney 18882ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 18892ff7e61eSJeff Mahoney set_nlink(inode, 1); 1890f37938e0SEric Sandeen /* 1891f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 1892f37938e0SEric Sandeen * the real end of the address space is determined by all of 1893f37938e0SEric Sandeen * the devices in the system 1894f37938e0SEric Sandeen */ 18952ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 18962ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 1897dcb2137cSChristoph Hellwig mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); 1898f37938e0SEric Sandeen 18992ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 190043eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 190135da5a7eSDavid Sterba IO_TREE_BTREE_INODE_IO); 19022ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 1903f37938e0SEric Sandeen 19045c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 1905adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 1906adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 1907adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 19082ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 1909e256927bSJosef Bacik __insert_inode_hash(inode, hash); 1910dcb2137cSChristoph Hellwig fs_info->btree_inode = inode; 1911dcb2137cSChristoph Hellwig 1912dcb2137cSChristoph Hellwig return 0; 1913f37938e0SEric Sandeen } 1914f37938e0SEric Sandeen 1915ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 1916ad618368SEric Sandeen { 1917ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 1918129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 19197f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 1920ad618368SEric Sandeen } 1921ad618368SEric Sandeen 1922f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 1923f9e92e40SEric Sandeen { 1924f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 1925f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 1926f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 1927f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 1928f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 1929f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 1930d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 1931011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 1932f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 1933f9e92e40SEric Sandeen } 1934f9e92e40SEric Sandeen 1935d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 19362a458198SEric Sandeen { 1937f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 19386f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 193958e814fcSTejun Heo unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; 19402a458198SEric Sandeen 19412a458198SEric Sandeen fs_info->workers = 1942a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 19432a458198SEric Sandeen 19442a458198SEric Sandeen fs_info->delalloc_workers = 1945cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 1946cb001095SJeff Mahoney flags, max_active, 2); 19472a458198SEric Sandeen 19482a458198SEric Sandeen fs_info->flush_workers = 1949cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 1950cb001095SJeff Mahoney flags, max_active, 0); 19512a458198SEric Sandeen 19522a458198SEric Sandeen fs_info->caching_workers = 1953cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 19542a458198SEric Sandeen 19552a458198SEric Sandeen fs_info->fixup_workers = 195658e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); 19572a458198SEric Sandeen 19582a458198SEric Sandeen fs_info->endio_workers = 1959d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 19602a458198SEric Sandeen fs_info->endio_meta_workers = 1961d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 1962385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 19632a458198SEric Sandeen fs_info->endio_write_workers = 1964cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 1965cb001095SJeff Mahoney max_active, 2); 1966fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 1967fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 19682a458198SEric Sandeen fs_info->endio_freespace_worker = 1969cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 1970cb001095SJeff Mahoney max_active, 0); 19712a458198SEric Sandeen fs_info->delayed_workers = 1972cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 1973cb001095SJeff Mahoney max_active, 0); 19742a458198SEric Sandeen fs_info->qgroup_rescan_workers = 197558e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", 197658e814fcSTejun Heo ordered_flags); 1977b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 197858e814fcSTejun Heo alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); 19792a458198SEric Sandeen 19808bfec2e4SChristoph Hellwig if (!(fs_info->workers && 1981a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 19822a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 1983fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 19841a1a2851SQu Wenruo fs_info->endio_write_workers && 19852a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 1986f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 1987f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 1988b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 19892a458198SEric Sandeen return -ENOMEM; 19902a458198SEric Sandeen } 19912a458198SEric Sandeen 19922a458198SEric Sandeen return 0; 19932a458198SEric Sandeen } 19942a458198SEric Sandeen 19956d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 19966d97c6e3SJohannes Thumshirn { 19976d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 1998b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 19996d97c6e3SJohannes Thumshirn 2000b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 20016d97c6e3SJohannes Thumshirn 20026d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 20036d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2004b4e967beSDavid Sterba csum_driver); 20056d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 20066d97c6e3SJohannes Thumshirn } 20076d97c6e3SJohannes Thumshirn 20086d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 20096d97c6e3SJohannes Thumshirn 201068d99ab0SChristoph Hellwig /* 201168d99ab0SChristoph Hellwig * Check if the checksum implementation is a fast accelerated one. 201268d99ab0SChristoph Hellwig * As-is this is a bit of a hack and should be replaced once the csum 201368d99ab0SChristoph Hellwig * implementations provide that information themselves. 201468d99ab0SChristoph Hellwig */ 201568d99ab0SChristoph Hellwig switch (csum_type) { 201668d99ab0SChristoph Hellwig case BTRFS_CSUM_TYPE_CRC32: 201768d99ab0SChristoph Hellwig if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) 201868d99ab0SChristoph Hellwig set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 201968d99ab0SChristoph Hellwig break; 2020efcfcbc6SDavid Sterba case BTRFS_CSUM_TYPE_XXHASH: 2021efcfcbc6SDavid Sterba set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 2022efcfcbc6SDavid Sterba break; 202368d99ab0SChristoph Hellwig default: 202468d99ab0SChristoph Hellwig break; 202568d99ab0SChristoph Hellwig } 202668d99ab0SChristoph Hellwig 2027c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2028c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2029c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 20306d97c6e3SJohannes Thumshirn return 0; 20316d97c6e3SJohannes Thumshirn } 20326d97c6e3SJohannes Thumshirn 203363443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 203463443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 203563443bf5SEric Sandeen { 203663443bf5SEric Sandeen int ret; 2037789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 203863443bf5SEric Sandeen struct btrfs_root *log_tree_root; 203963443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 204063443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2041581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 204263443bf5SEric Sandeen 204363443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2044f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 204563443bf5SEric Sandeen return -EIO; 204663443bf5SEric Sandeen } 204763443bf5SEric Sandeen 204896dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 204996dfcb46SJosef Bacik GFP_KERNEL); 205063443bf5SEric Sandeen if (!log_tree_root) 205163443bf5SEric Sandeen return -ENOMEM; 205263443bf5SEric Sandeen 2053789d6a3aSQu Wenruo check.level = level; 2054789d6a3aSQu Wenruo check.transid = fs_info->generation + 1; 2055789d6a3aSQu Wenruo check.owner_root = BTRFS_TREE_LOG_OBJECTID; 2056789d6a3aSQu Wenruo log_tree_root->node = read_tree_block(fs_info, bytenr, &check); 205764c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2058f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 20590eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 20608c38938cSJosef Bacik log_tree_root->node = NULL; 206100246528SJosef Bacik btrfs_put_root(log_tree_root); 20620eeff236SLiu Bo return ret; 20634eb150d6SQu Wenruo } 20644eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2065f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 206600246528SJosef Bacik btrfs_put_root(log_tree_root); 206763443bf5SEric Sandeen return -EIO; 206863443bf5SEric Sandeen } 20694eb150d6SQu Wenruo 207063443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 207163443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 207263443bf5SEric Sandeen if (ret) { 20730b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 207463443bf5SEric Sandeen "Failed to recover log tree"); 207500246528SJosef Bacik btrfs_put_root(log_tree_root); 207663443bf5SEric Sandeen return ret; 207763443bf5SEric Sandeen } 207863443bf5SEric Sandeen 2079bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 20806bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 208163443bf5SEric Sandeen if (ret) 208263443bf5SEric Sandeen return ret; 208363443bf5SEric Sandeen } 208463443bf5SEric Sandeen 208563443bf5SEric Sandeen return 0; 208663443bf5SEric Sandeen } 208763443bf5SEric Sandeen 2088abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2089abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2090abed4aaaSJosef Bacik const char *name) 2091abed4aaaSJosef Bacik { 2092abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2093abed4aaaSJosef Bacik struct btrfs_root *root; 2094f7238e50SJosef Bacik u64 max_global_id = 0; 2095abed4aaaSJosef Bacik int ret; 2096abed4aaaSJosef Bacik struct btrfs_key key = { 2097abed4aaaSJosef Bacik .objectid = objectid, 2098abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2099abed4aaaSJosef Bacik .offset = 0, 2100abed4aaaSJosef Bacik }; 2101abed4aaaSJosef Bacik bool found = false; 2102abed4aaaSJosef Bacik 2103abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2104abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2105abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2106abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2107abed4aaaSJosef Bacik return 0; 2108abed4aaaSJosef Bacik } 2109abed4aaaSJosef Bacik 2110abed4aaaSJosef Bacik while (1) { 2111abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2112abed4aaaSJosef Bacik if (ret < 0) 2113abed4aaaSJosef Bacik break; 2114abed4aaaSJosef Bacik 2115abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2116abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2117abed4aaaSJosef Bacik if (ret) { 2118abed4aaaSJosef Bacik if (ret > 0) 2119abed4aaaSJosef Bacik ret = 0; 2120abed4aaaSJosef Bacik break; 2121abed4aaaSJosef Bacik } 2122abed4aaaSJosef Bacik } 2123abed4aaaSJosef Bacik ret = 0; 2124abed4aaaSJosef Bacik 2125abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2126abed4aaaSJosef Bacik if (key.objectid != objectid) 2127abed4aaaSJosef Bacik break; 2128abed4aaaSJosef Bacik btrfs_release_path(path); 2129abed4aaaSJosef Bacik 2130f7238e50SJosef Bacik /* 2131f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2132f7238e50SJosef Bacik * everybody. 2133f7238e50SJosef Bacik */ 2134f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2135f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2136f7238e50SJosef Bacik 2137abed4aaaSJosef Bacik found = true; 2138abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2139abed4aaaSJosef Bacik if (IS_ERR(root)) { 2140abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2141abed4aaaSJosef Bacik ret = PTR_ERR(root); 2142abed4aaaSJosef Bacik break; 2143abed4aaaSJosef Bacik } 2144abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2145abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2146abed4aaaSJosef Bacik if (ret) { 2147abed4aaaSJosef Bacik btrfs_put_root(root); 2148abed4aaaSJosef Bacik break; 2149abed4aaaSJosef Bacik } 2150abed4aaaSJosef Bacik key.offset++; 2151abed4aaaSJosef Bacik } 2152abed4aaaSJosef Bacik btrfs_release_path(path); 2153abed4aaaSJosef Bacik 2154f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2155f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2156f7238e50SJosef Bacik 2157abed4aaaSJosef Bacik if (!found || ret) { 2158abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2159abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2160abed4aaaSJosef Bacik 2161abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2162abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2163abed4aaaSJosef Bacik else 2164abed4aaaSJosef Bacik ret = 0; 2165abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2166abed4aaaSJosef Bacik } 2167abed4aaaSJosef Bacik return ret; 2168abed4aaaSJosef Bacik } 2169abed4aaaSJosef Bacik 2170abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2171abed4aaaSJosef Bacik { 2172abed4aaaSJosef Bacik struct btrfs_path *path; 2173abed4aaaSJosef Bacik int ret = 0; 2174abed4aaaSJosef Bacik 2175abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2176abed4aaaSJosef Bacik if (!path) 2177abed4aaaSJosef Bacik return -ENOMEM; 2178abed4aaaSJosef Bacik 2179abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2180abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2181abed4aaaSJosef Bacik if (ret) 2182abed4aaaSJosef Bacik goto out; 2183abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2184abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2185abed4aaaSJosef Bacik if (ret) 2186abed4aaaSJosef Bacik goto out; 2187abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2188abed4aaaSJosef Bacik goto out; 2189abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2190abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2191abed4aaaSJosef Bacik "free space"); 2192abed4aaaSJosef Bacik out: 2193abed4aaaSJosef Bacik btrfs_free_path(path); 2194abed4aaaSJosef Bacik return ret; 2195abed4aaaSJosef Bacik } 2196abed4aaaSJosef Bacik 21976bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 21984bbcaa64SEric Sandeen { 21996bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2200a4f3d2c4SDavid Sterba struct btrfs_root *root; 22014bbcaa64SEric Sandeen struct btrfs_key location; 22024bbcaa64SEric Sandeen int ret; 22034bbcaa64SEric Sandeen 22046bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 22056bccf3abSJeff Mahoney 2206abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2207abed4aaaSJosef Bacik if (ret) 2208abed4aaaSJosef Bacik return ret; 2209abed4aaaSJosef Bacik 22104bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 22114bbcaa64SEric Sandeen location.offset = 0; 22124bbcaa64SEric Sandeen 22131c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 221414033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 221514033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 221614033b08SQu Wenruo if (IS_ERR(root)) { 221714033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 221814033b08SQu Wenruo ret = PTR_ERR(root); 221914033b08SQu Wenruo goto out; 222014033b08SQu Wenruo } 222114033b08SQu Wenruo } else { 222214033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 222314033b08SQu Wenruo fs_info->block_group_root = root; 222414033b08SQu Wenruo } 222514033b08SQu Wenruo } 222614033b08SQu Wenruo 222714033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2228a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2229f50f4353SLiu Bo if (IS_ERR(root)) { 223042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2231f50f4353SLiu Bo ret = PTR_ERR(root); 2232f50f4353SLiu Bo goto out; 2233f50f4353SLiu Bo } 223442437a63SJosef Bacik } else { 2235a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2236a4f3d2c4SDavid Sterba fs_info->dev_root = root; 223742437a63SJosef Bacik } 2238820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2239a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2240a8d1b164SJohannes Thumshirn if (ret) 2241a8d1b164SJohannes Thumshirn goto out; 22424bbcaa64SEric Sandeen 2243aeb935a4SQu Wenruo /* 2244aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2245aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2246aeb935a4SQu Wenruo */ 224756e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 224856e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2249aeb935a4SQu Wenruo if (IS_ERR(root)) { 225042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2251aeb935a4SQu Wenruo ret = PTR_ERR(root); 2252aeb935a4SQu Wenruo goto out; 2253aeb935a4SQu Wenruo } 225442437a63SJosef Bacik } else { 2255aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2256aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 225742437a63SJosef Bacik } 2258aeb935a4SQu Wenruo 22594bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2260a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2261a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2262a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2263afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2264a4f3d2c4SDavid Sterba fs_info->quota_root = root; 22654bbcaa64SEric Sandeen } 22664bbcaa64SEric Sandeen 22674bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2268a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2269a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 227042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2271a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 22724bbcaa64SEric Sandeen if (ret != -ENOENT) 2273f50f4353SLiu Bo goto out; 227442437a63SJosef Bacik } 22754bbcaa64SEric Sandeen } else { 2276a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2277a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 22784bbcaa64SEric Sandeen } 22794bbcaa64SEric Sandeen 22804bbcaa64SEric Sandeen return 0; 2281f50f4353SLiu Bo out: 2282f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2283f50f4353SLiu Bo location.objectid, ret); 2284f50f4353SLiu Bo return ret; 22854bbcaa64SEric Sandeen } 22864bbcaa64SEric Sandeen 2287069ec957SQu Wenruo /* 2288069ec957SQu Wenruo * Real super block validation 2289069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2290069ec957SQu Wenruo * 2291069ec957SQu Wenruo * @sb: super block to check 2292069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2293069ec957SQu Wenruo * 0 the primary (1st) sb 2294069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2295069ec957SQu Wenruo * -1 skip bytenr check 2296069ec957SQu Wenruo */ 2297a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2298069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 229921a852b0SQu Wenruo { 230021a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 230121a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 230221a852b0SQu Wenruo int ret = 0; 230321a852b0SQu Wenruo 230421a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 230521a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 230621a852b0SQu Wenruo ret = -EINVAL; 230721a852b0SQu Wenruo } 230821a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 230921a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 231021a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 231121a852b0SQu Wenruo ret = -EINVAL; 231221a852b0SQu Wenruo } 231321a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 231421a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 231521a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 231621a852b0SQu Wenruo ret = -EINVAL; 231721a852b0SQu Wenruo } 231821a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 231921a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 232021a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 232121a852b0SQu Wenruo ret = -EINVAL; 232221a852b0SQu Wenruo } 232321a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 232421a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 232521a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 232621a852b0SQu Wenruo ret = -EINVAL; 232721a852b0SQu Wenruo } 232821a852b0SQu Wenruo 232921a852b0SQu Wenruo /* 233021a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 233121a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 233221a852b0SQu Wenruo */ 233321a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 233421a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 233521a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 233621a852b0SQu Wenruo ret = -EINVAL; 233721a852b0SQu Wenruo } 23380bb3eb3eSQu Wenruo 23390bb3eb3eSQu Wenruo /* 23401a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 23411a42daabSQu Wenruo * 23421a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 23431a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 23441a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 23451a42daabSQu Wenruo * beginning. 23460bb3eb3eSQu Wenruo */ 23471a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 234821a852b0SQu Wenruo btrfs_err(fs_info, 23490bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 235021a852b0SQu Wenruo sectorsize, PAGE_SIZE); 235121a852b0SQu Wenruo ret = -EINVAL; 235221a852b0SQu Wenruo } 23530bb3eb3eSQu Wenruo 235421a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 235521a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 235621a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 235721a852b0SQu Wenruo ret = -EINVAL; 235821a852b0SQu Wenruo } 235921a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 236021a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 236121a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 236221a852b0SQu Wenruo ret = -EINVAL; 236321a852b0SQu Wenruo } 236421a852b0SQu Wenruo 236521a852b0SQu Wenruo /* Root alignment check */ 236621a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 236721a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 236821a852b0SQu Wenruo btrfs_super_root(sb)); 236921a852b0SQu Wenruo ret = -EINVAL; 237021a852b0SQu Wenruo } 237121a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 237221a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 237321a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 237421a852b0SQu Wenruo ret = -EINVAL; 237521a852b0SQu Wenruo } 237621a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 237721a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 237821a852b0SQu Wenruo btrfs_super_log_root(sb)); 237921a852b0SQu Wenruo ret = -EINVAL; 238021a852b0SQu Wenruo } 238121a852b0SQu Wenruo 2382d167aa76SAnand Jain if (memcmp(fs_info->fs_devices->fsid, sb->fsid, BTRFS_FSID_SIZE) != 0) { 2383aefd7f70SNikolay Borisov btrfs_err(fs_info, 2384aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2385d167aa76SAnand Jain sb->fsid, fs_info->fs_devices->fsid); 2386aefd7f70SNikolay Borisov ret = -EINVAL; 2387aefd7f70SNikolay Borisov } 2388aefd7f70SNikolay Borisov 23896bfe3959SAnand Jain if (memcmp(fs_info->fs_devices->metadata_uuid, btrfs_sb_fsid_ptr(sb), 23906bfe3959SAnand Jain BTRFS_FSID_SIZE) != 0) { 2391aefd7f70SNikolay Borisov btrfs_err(fs_info, 2392aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 23936bfe3959SAnand Jain btrfs_sb_fsid_ptr(sb), fs_info->fs_devices->metadata_uuid); 2394aefd7f70SNikolay Borisov ret = -EINVAL; 2395aefd7f70SNikolay Borisov } 2396aefd7f70SNikolay Borisov 239725984a5aSAnand Jain if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 239825984a5aSAnand Jain BTRFS_FSID_SIZE) != 0) { 239925984a5aSAnand Jain btrfs_err(fs_info, 240025984a5aSAnand Jain "dev_item UUID does not match metadata fsid: %pU != %pU", 240125984a5aSAnand Jain fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 240225984a5aSAnand Jain ret = -EINVAL; 240325984a5aSAnand Jain } 240425984a5aSAnand Jain 24051c56ab99SQu Wenruo /* 24061c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 24071c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 24081c56ab99SQu Wenruo */ 24091c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 24101c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 24111c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 24121c56ab99SQu Wenruo btrfs_err(fs_info, 24131c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 24141c56ab99SQu Wenruo ret = -EINVAL; 24151c56ab99SQu Wenruo } 24161c56ab99SQu Wenruo 241721a852b0SQu Wenruo /* 241821a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 241921a852b0SQu Wenruo * done later 242021a852b0SQu Wenruo */ 242121a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 242221a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 242321a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 242421a852b0SQu Wenruo ret = -EINVAL; 242521a852b0SQu Wenruo } 242621a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 242721a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 242821a852b0SQu Wenruo btrfs_super_stripesize(sb)); 242921a852b0SQu Wenruo ret = -EINVAL; 243021a852b0SQu Wenruo } 243121a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 243221a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 243321a852b0SQu Wenruo btrfs_super_num_devices(sb)); 243421a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 243521a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 243621a852b0SQu Wenruo ret = -EINVAL; 243721a852b0SQu Wenruo } 243821a852b0SQu Wenruo 2439069ec957SQu Wenruo if (mirror_num >= 0 && 2440069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 244121a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 244221a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 244321a852b0SQu Wenruo ret = -EINVAL; 244421a852b0SQu Wenruo } 244521a852b0SQu Wenruo 244621a852b0SQu Wenruo /* 244721a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 244821a852b0SQu Wenruo * and one chunk 244921a852b0SQu Wenruo */ 245021a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 245121a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 245221a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 245321a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 245421a852b0SQu Wenruo ret = -EINVAL; 245521a852b0SQu Wenruo } 245621a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 245721a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 245821a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 245921a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 246021a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 246121a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 246221a852b0SQu Wenruo ret = -EINVAL; 246321a852b0SQu Wenruo } 246421a852b0SQu Wenruo 246521a852b0SQu Wenruo /* 246621a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 246721a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 246821a852b0SQu Wenruo */ 246921a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 247021a852b0SQu Wenruo btrfs_warn(fs_info, 247121a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 247221a852b0SQu Wenruo btrfs_super_generation(sb), 247321a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 247421a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 247521a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 247621a852b0SQu Wenruo btrfs_warn(fs_info, 247721a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 247821a852b0SQu Wenruo btrfs_super_generation(sb), 247921a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 248021a852b0SQu Wenruo 248121a852b0SQu Wenruo return ret; 248221a852b0SQu Wenruo } 248321a852b0SQu Wenruo 2484069ec957SQu Wenruo /* 2485069ec957SQu Wenruo * Validation of super block at mount time. 2486069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2487069ec957SQu Wenruo * flags will be skipped. 2488069ec957SQu Wenruo */ 2489069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2490069ec957SQu Wenruo { 2491a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2492069ec957SQu Wenruo } 2493069ec957SQu Wenruo 249475cb857dSQu Wenruo /* 249575cb857dSQu Wenruo * Validation of super block at write time. 249675cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 249775cb857dSQu Wenruo * overwritten soon. 249875cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 249975cb857dSQu Wenruo */ 250075cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 250175cb857dSQu Wenruo struct btrfs_super_block *sb) 250275cb857dSQu Wenruo { 250375cb857dSQu Wenruo int ret; 250475cb857dSQu Wenruo 2505a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 250675cb857dSQu Wenruo if (ret < 0) 250775cb857dSQu Wenruo goto out; 2508e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 250975cb857dSQu Wenruo ret = -EUCLEAN; 251075cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 251175cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 251275cb857dSQu Wenruo goto out; 251375cb857dSQu Wenruo } 251475cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 251575cb857dSQu Wenruo ret = -EUCLEAN; 251675cb857dSQu Wenruo btrfs_err(fs_info, 251775cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 251875cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 251975cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 252075cb857dSQu Wenruo goto out; 252175cb857dSQu Wenruo } 252275cb857dSQu Wenruo out: 252375cb857dSQu Wenruo if (ret < 0) 252475cb857dSQu Wenruo btrfs_err(fs_info, 252575cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 252675cb857dSQu Wenruo return ret; 252775cb857dSQu Wenruo } 252875cb857dSQu Wenruo 2529bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2530bd676446SJosef Bacik { 2531789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 2532789d6a3aSQu Wenruo .level = level, 2533789d6a3aSQu Wenruo .transid = gen, 2534789d6a3aSQu Wenruo .owner_root = root->root_key.objectid 2535789d6a3aSQu Wenruo }; 2536bd676446SJosef Bacik int ret = 0; 2537bd676446SJosef Bacik 2538789d6a3aSQu Wenruo root->node = read_tree_block(root->fs_info, bytenr, &check); 2539bd676446SJosef Bacik if (IS_ERR(root->node)) { 2540bd676446SJosef Bacik ret = PTR_ERR(root->node); 2541bd676446SJosef Bacik root->node = NULL; 25424eb150d6SQu Wenruo return ret; 25434eb150d6SQu Wenruo } 25444eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2545bd676446SJosef Bacik free_extent_buffer(root->node); 2546bd676446SJosef Bacik root->node = NULL; 25474eb150d6SQu Wenruo return -EIO; 2548bd676446SJosef Bacik } 2549bd676446SJosef Bacik 2550bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2551bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2552bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2553bd676446SJosef Bacik return ret; 2554bd676446SJosef Bacik } 2555bd676446SJosef Bacik 2556bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2557bd676446SJosef Bacik { 2558bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2559bd676446SJosef Bacik u64 gen, bytenr; 2560bd676446SJosef Bacik int level, ret; 2561bd676446SJosef Bacik 2562bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2563bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2564bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2565bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 25669c54e80dSJosef Bacik if (ret) { 2567bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2568bd676446SJosef Bacik return ret; 2569bd676446SJosef Bacik } 25709c54e80dSJosef Bacik return 0; 25719c54e80dSJosef Bacik } 25729c54e80dSJosef Bacik 25736ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2574b8522a1eSNikolay Borisov { 25756ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2576b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2577b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2578b8522a1eSNikolay Borisov bool handle_error = false; 2579b8522a1eSNikolay Borisov int ret = 0; 2580b8522a1eSNikolay Borisov int i; 2581b8522a1eSNikolay Borisov 2582b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2583b8522a1eSNikolay Borisov if (handle_error) { 2584b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2585b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2586b8522a1eSNikolay Borisov tree_root->node = NULL; 2587b8522a1eSNikolay Borisov 2588b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2589b8522a1eSNikolay Borisov break; 2590b8522a1eSNikolay Borisov 2591b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2592b8522a1eSNikolay Borisov 2593b8522a1eSNikolay Borisov /* 2594b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2595b8522a1eSNikolay Borisov * valid 2596b8522a1eSNikolay Borisov */ 2597b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2598b8522a1eSNikolay Borisov 2599b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2600b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2601b8522a1eSNikolay Borisov 2602745806fbSQu Wenruo btrfs_warn(fs_info, "try to load backup roots slot %d", i); 2603b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26046ef108ddSNikolay Borisov backup_index = ret; 2605b8522a1eSNikolay Borisov if (ret < 0) 2606b8522a1eSNikolay Borisov return ret; 2607b8522a1eSNikolay Borisov } 2608b8522a1eSNikolay Borisov 2609bd676446SJosef Bacik ret = load_important_roots(fs_info); 2610bd676446SJosef Bacik if (ret) { 2611217f5004SNikolay Borisov handle_error = true; 2612b8522a1eSNikolay Borisov continue; 2613b8522a1eSNikolay Borisov } 2614b8522a1eSNikolay Borisov 2615336a0d8dSNikolay Borisov /* 2616336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2617336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2618336a0d8dSNikolay Borisov */ 2619453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2620b8522a1eSNikolay Borisov if (ret < 0) { 2621b8522a1eSNikolay Borisov handle_error = true; 2622b8522a1eSNikolay Borisov continue; 2623b8522a1eSNikolay Borisov } 2624b8522a1eSNikolay Borisov 26256b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2626b8522a1eSNikolay Borisov 2627b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2628b8522a1eSNikolay Borisov if (ret < 0) { 2629b8522a1eSNikolay Borisov handle_error = true; 2630b8522a1eSNikolay Borisov continue; 2631b8522a1eSNikolay Borisov } 2632b8522a1eSNikolay Borisov 2633b8522a1eSNikolay Borisov /* All successful */ 2634bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2635bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2636d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 26376ef108ddSNikolay Borisov 26386ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26396ef108ddSNikolay Borisov if (backup_index < 0) { 26406ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26416ef108ddSNikolay Borisov } else { 26426ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26436ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26446ef108ddSNikolay Borisov } 2645b8522a1eSNikolay Borisov break; 2646b8522a1eSNikolay Borisov } 2647b8522a1eSNikolay Borisov 2648b8522a1eSNikolay Borisov return ret; 2649b8522a1eSNikolay Borisov } 2650b8522a1eSNikolay Borisov 26518260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2652eb60ceacSChris Mason { 2653fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 265401cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26558fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2656facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 265724bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2658eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 265911833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2660eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2661a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2662fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 266324bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26644cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2665ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2666f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 266747ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 266840ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2669afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2670c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2671f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2672abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2673d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2674f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 26757585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2676573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 26770bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 26785f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 2679de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 268019c00ddcSChris Mason 2681e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 26825a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 26838b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 26845f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 26853e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 26863e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 26873e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 26883e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 26893e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 26903e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 26913e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 26923e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 2693e1489b4fSIoannis Angelakopoulos 26940b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 26956324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2696f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 269747ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 269818bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2699afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2700bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2701bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 27023fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27033fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2704bd647ce3SJosef Bacik #endif 2705c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 270666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 270766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 270866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 270966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 271066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 271166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 271266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2713ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2714ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2715ba2c4d4eSJosef Bacik 2716771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27174cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2718034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2719fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2720abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 272195ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27229ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27234cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2724a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2725f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27268b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2727fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2728c8b97818SChris Mason 2729b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2730b34b086cSChris Mason num_online_cpus() + 2, 8); 27310afbaf8cSChris Mason 2732199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2733199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 273469fe2d75SJosef Bacik 2735638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 273621adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 273721adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 273821adbd5cSStefan Behrens #endif 2739779a65a4SEric Sandeen btrfs_init_balance(fs_info); 274057056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2741a2de733cSArne Jansen 274216b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 274308dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 27440f9dd46cSJosef Bacik 2745fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 274635da5a7eSDavid Sterba IO_TREE_FS_EXCLUDED_EXTENTS); 274739279cc3SChris Mason 27485a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2749e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2750925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2751a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2752a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27531bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27549e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2755c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 275676dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2757803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2758fa9c0d79SChris Mason 2759ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2760f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2761b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2762416ac51dSArne Jansen 2763fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2764fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2765fa9c0d79SChris Mason 2766e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2767f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2768bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27694854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2770034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 27713768f368SChris Mason 2772da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2773da17066cSJeff Mahoney fs_info->nodesize = 4096; 2774da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2775ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2776da17066cSJeff Mahoney fs_info->stripesize = 4096; 2777da17066cSJeff Mahoney 2778f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 2779f7b12a62SNaohiro Aota 2780eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2781eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2782eede2bf3SOmar Sandoval 278318bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 278418bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 27858260edbaSJosef Bacik } 27868260edbaSJosef Bacik 27878260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 27888260edbaSJosef Bacik { 27898260edbaSJosef Bacik int ret; 27908260edbaSJosef Bacik 27918260edbaSJosef Bacik fs_info->sb = sb; 27928260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 27938260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 27949e967495SFilipe Manana 27955deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 2796ae18c37aSJosef Bacik if (ret) 2797ae18c37aSJosef Bacik return ret; 2798ae18c37aSJosef Bacik 2799ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2800ae18c37aSJosef Bacik if (ret) 2801c75e8394SJosef Bacik return ret; 2802ae18c37aSJosef Bacik 2803ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2804ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2805ae18c37aSJosef Bacik 2806ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2807ae18c37aSJosef Bacik if (ret) 2808c75e8394SJosef Bacik return ret; 2809ae18c37aSJosef Bacik 2810ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2811ae18c37aSJosef Bacik GFP_KERNEL); 2812ae18c37aSJosef Bacik if (ret) 2813c75e8394SJosef Bacik return ret; 2814ae18c37aSJosef Bacik 2815ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2816ae18c37aSJosef Bacik GFP_KERNEL); 2817c75e8394SJosef Bacik if (!fs_info->delayed_root) 2818c75e8394SJosef Bacik return -ENOMEM; 2819ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2820ae18c37aSJosef Bacik 2821a0a1db70SFilipe Manana if (sb_rdonly(sb)) 2822a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 2823a0a1db70SFilipe Manana 2824c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2825ae18c37aSJosef Bacik } 2826ae18c37aSJosef Bacik 282797f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 282897f4dd09SNikolay Borisov { 28290d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 283097f4dd09SNikolay Borisov int ret; 283197f4dd09SNikolay Borisov 283297f4dd09SNikolay Borisov /* 283397f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 283497f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 283597f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 283697f4dd09SNikolay Borisov */ 283797f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 283897f4dd09SNikolay Borisov if (ret < 0) { 2839c94bec2cSJosef Bacik if (ret != -EINTR) 2840c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2841c94bec2cSJosef Bacik ret); 284297f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 284397f4dd09SNikolay Borisov return ret; 284497f4dd09SNikolay Borisov } 284597f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 284697f4dd09SNikolay Borisov } 284797f4dd09SNikolay Borisov 284897f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 284997f4dd09SNikolay Borisov { 285097f4dd09SNikolay Borisov struct task_struct *task; 285197f4dd09SNikolay Borisov 285297f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 285397f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 285497f4dd09SNikolay Borisov if (IS_ERR(task)) { 285597f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 285697f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 285797f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 285897f4dd09SNikolay Borisov return PTR_ERR(task); 285997f4dd09SNikolay Borisov } 286097f4dd09SNikolay Borisov 286197f4dd09SNikolay Borisov return 0; 286297f4dd09SNikolay Borisov } 286397f4dd09SNikolay Borisov 2864504b1596SFilipe Manana static int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 2865504b1596SFilipe Manana { 2866504b1596SFilipe Manana u64 root_objectid = 0; 2867504b1596SFilipe Manana struct btrfs_root *gang[8]; 2868504b1596SFilipe Manana int i = 0; 2869504b1596SFilipe Manana int err = 0; 2870504b1596SFilipe Manana unsigned int ret = 0; 2871504b1596SFilipe Manana 2872504b1596SFilipe Manana while (1) { 2873504b1596SFilipe Manana spin_lock(&fs_info->fs_roots_radix_lock); 2874504b1596SFilipe Manana ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2875504b1596SFilipe Manana (void **)gang, root_objectid, 2876504b1596SFilipe Manana ARRAY_SIZE(gang)); 2877504b1596SFilipe Manana if (!ret) { 2878504b1596SFilipe Manana spin_unlock(&fs_info->fs_roots_radix_lock); 2879504b1596SFilipe Manana break; 2880504b1596SFilipe Manana } 2881504b1596SFilipe Manana root_objectid = gang[ret - 1]->root_key.objectid + 1; 2882504b1596SFilipe Manana 2883504b1596SFilipe Manana for (i = 0; i < ret; i++) { 2884504b1596SFilipe Manana /* Avoid to grab roots in dead_roots. */ 2885504b1596SFilipe Manana if (btrfs_root_refs(&gang[i]->root_item) == 0) { 2886504b1596SFilipe Manana gang[i] = NULL; 2887504b1596SFilipe Manana continue; 2888504b1596SFilipe Manana } 2889504b1596SFilipe Manana /* Grab all the search result for later use. */ 2890504b1596SFilipe Manana gang[i] = btrfs_grab_root(gang[i]); 2891504b1596SFilipe Manana } 2892504b1596SFilipe Manana spin_unlock(&fs_info->fs_roots_radix_lock); 2893504b1596SFilipe Manana 2894504b1596SFilipe Manana for (i = 0; i < ret; i++) { 2895504b1596SFilipe Manana if (!gang[i]) 2896504b1596SFilipe Manana continue; 2897504b1596SFilipe Manana root_objectid = gang[i]->root_key.objectid; 2898504b1596SFilipe Manana err = btrfs_orphan_cleanup(gang[i]); 2899504b1596SFilipe Manana if (err) 2900504b1596SFilipe Manana goto out; 2901504b1596SFilipe Manana btrfs_put_root(gang[i]); 2902504b1596SFilipe Manana } 2903504b1596SFilipe Manana root_objectid++; 2904504b1596SFilipe Manana } 2905504b1596SFilipe Manana out: 2906504b1596SFilipe Manana /* Release the uncleaned roots due to error. */ 2907504b1596SFilipe Manana for (; i < ret; i++) { 2908504b1596SFilipe Manana if (gang[i]) 2909504b1596SFilipe Manana btrfs_put_root(gang[i]); 2910504b1596SFilipe Manana } 2911504b1596SFilipe Manana return err; 2912504b1596SFilipe Manana } 2913504b1596SFilipe Manana 291444c0ca21SBoris Burkov /* 29158cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 29168cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 29178cd29088SBoris Burkov * code paths. 29188cd29088SBoris Burkov */ 29198cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 29208cd29088SBoris Burkov { 29218cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 29228b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 29238cd29088SBoris Burkov } 29248cd29088SBoris Burkov 29258cd29088SBoris Burkov /* 292644c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 292744c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 292844c0ca21SBoris Burkov */ 292944c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 293044c0ca21SBoris Burkov { 293144c0ca21SBoris Burkov int ret; 293294846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 29331d6a4fc8SQu Wenruo bool rebuild_free_space_tree = false; 29348b228324SBoris Burkov 29358b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 29368b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 29371d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 29388b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29398b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 29408b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 29411d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 29428b228324SBoris Burkov } 29438b228324SBoris Burkov 29441d6a4fc8SQu Wenruo if (rebuild_free_space_tree) { 29451d6a4fc8SQu Wenruo btrfs_info(fs_info, "rebuilding free space tree"); 29461d6a4fc8SQu Wenruo ret = btrfs_rebuild_free_space_tree(fs_info); 29478b228324SBoris Burkov if (ret) { 29488b228324SBoris Burkov btrfs_warn(fs_info, 29491d6a4fc8SQu Wenruo "failed to rebuild free space tree: %d", ret); 29501d6a4fc8SQu Wenruo goto out; 29511d6a4fc8SQu Wenruo } 29521d6a4fc8SQu Wenruo } 29531d6a4fc8SQu Wenruo 29541d6a4fc8SQu Wenruo if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29551d6a4fc8SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { 29561d6a4fc8SQu Wenruo btrfs_info(fs_info, "disabling free space tree"); 29571d6a4fc8SQu Wenruo ret = btrfs_delete_free_space_tree(fs_info); 29581d6a4fc8SQu Wenruo if (ret) { 29591d6a4fc8SQu Wenruo btrfs_warn(fs_info, 29601d6a4fc8SQu Wenruo "failed to disable free space tree: %d", ret); 29618b228324SBoris Burkov goto out; 29628b228324SBoris Burkov } 29638b228324SBoris Burkov } 296444c0ca21SBoris Burkov 29658d488a8cSFilipe Manana /* 29668d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 29678d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 2968fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 29698d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 29708d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 29718d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 29728d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 29738d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 29748d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 29758d488a8cSFilipe Manana */ 29768d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 29778d488a8cSFilipe Manana if (ret) 29788d488a8cSFilipe Manana goto out; 29798d488a8cSFilipe Manana 298044c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 298144c0ca21SBoris Burkov if (ret) 298244c0ca21SBoris Burkov goto out; 298344c0ca21SBoris Burkov 29848f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 29858f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29868f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29878f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29888f1c21d7SBoris Burkov goto out; 29898f1c21d7SBoris Burkov } 29908f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29918f1c21d7SBoris Burkov 299244c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 29937eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 299444c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 299544c0ca21SBoris Burkov if (ret < 0) { 299644c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 299744c0ca21SBoris Burkov goto out; 299844c0ca21SBoris Burkov } 299944c0ca21SBoris Burkov 30005011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 30015011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 30025011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 30035011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 30045011139aSBoris Burkov if (ret) { 30055011139aSBoris Burkov btrfs_warn(fs_info, 30065011139aSBoris Burkov "failed to create free space tree: %d", ret); 30075011139aSBoris Burkov goto out; 30085011139aSBoris Burkov } 30095011139aSBoris Burkov } 30105011139aSBoris Burkov 301194846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 301294846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 301394846229SBoris Burkov if (ret) 301494846229SBoris Burkov goto out; 301594846229SBoris Burkov } 301694846229SBoris Burkov 301744c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 301844c0ca21SBoris Burkov if (ret) 301944c0ca21SBoris Burkov goto out; 302044c0ca21SBoris Burkov 302144c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 302244c0ca21SBoris Burkov if (ret) { 302344c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 302444c0ca21SBoris Burkov goto out; 302544c0ca21SBoris Burkov } 302644c0ca21SBoris Burkov 302744c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 302844c0ca21SBoris Burkov 302944c0ca21SBoris Burkov if (!fs_info->uuid_root) { 303044c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 303144c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 303244c0ca21SBoris Burkov if (ret) { 303344c0ca21SBoris Burkov btrfs_warn(fs_info, 303444c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 303544c0ca21SBoris Burkov goto out; 303644c0ca21SBoris Burkov } 303744c0ca21SBoris Burkov } 303844c0ca21SBoris Burkov 303944c0ca21SBoris Burkov out: 304044c0ca21SBoris Burkov return ret; 304144c0ca21SBoris Burkov } 304244c0ca21SBoris Burkov 3043d7f67ac9SQu Wenruo /* 3044d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3045d7f67ac9SQu Wenruo * 30462ba48b20SQu Wenruo * @is_rw_mount: If the mount is read-write. 30472ba48b20SQu Wenruo * 3048d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3049d7f67ac9SQu Wenruo * feature dependencies). 3050d7f67ac9SQu Wenruo * 3051d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3052d7f67ac9SQu Wenruo * btrfs_validate_super(). 3053d7f67ac9SQu Wenruo * 3054d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3055d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3056d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3057d7f67ac9SQu Wenruo */ 30582ba48b20SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) 3059d7f67ac9SQu Wenruo { 3060d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3061d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3062d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3063d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3064d7f67ac9SQu Wenruo 3065d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3066d7f67ac9SQu Wenruo btrfs_err(fs_info, 3067d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3068d7f67ac9SQu Wenruo incompat); 3069d7f67ac9SQu Wenruo return -EINVAL; 3070d7f67ac9SQu Wenruo } 3071d7f67ac9SQu Wenruo 3072d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3073d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3074d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3075d7f67ac9SQu Wenruo btrfs_err(fs_info, 3076d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3077d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3078d7f67ac9SQu Wenruo return -EINVAL; 3079d7f67ac9SQu Wenruo } 3080d7f67ac9SQu Wenruo 3081d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3082d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3083d7f67ac9SQu Wenruo 3084d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3085d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3086d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3087d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3088d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3089d7f67ac9SQu Wenruo 3090d7f67ac9SQu Wenruo /* 3091d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3092d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3093d7f67ac9SQu Wenruo */ 3094d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3095d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3096d7f67ac9SQu Wenruo 30972ba48b20SQu Wenruo if (compat_ro_unsupp && is_rw_mount) { 3098d7f67ac9SQu Wenruo btrfs_err(fs_info, 3099d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3100d7f67ac9SQu Wenruo compat_ro); 3101d7f67ac9SQu Wenruo return -EINVAL; 3102d7f67ac9SQu Wenruo } 3103d7f67ac9SQu Wenruo 3104d7f67ac9SQu Wenruo /* 3105d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3106d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3107d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3108d7f67ac9SQu Wenruo */ 3109d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3110d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3111d7f67ac9SQu Wenruo btrfs_err(fs_info, 3112d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3113d7f67ac9SQu Wenruo compat_ro); 3114d7f67ac9SQu Wenruo return -EINVAL; 3115d7f67ac9SQu Wenruo } 3116d7f67ac9SQu Wenruo 3117d7f67ac9SQu Wenruo /* 3118d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3119d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3120d7f67ac9SQu Wenruo */ 3121d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3122d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3123d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3124d7f67ac9SQu Wenruo btrfs_err(fs_info, 3125d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3126d7f67ac9SQu Wenruo return -EINVAL; 3127d7f67ac9SQu Wenruo } 3128d7f67ac9SQu Wenruo 3129d7f67ac9SQu Wenruo /* 3130d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3131d7f67ac9SQu Wenruo * 3132d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3133d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3134d7f67ac9SQu Wenruo * going to be deprecated anyway. 3135d7f67ac9SQu Wenruo */ 3136d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3137d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3138d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3139d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3140d7f67ac9SQu Wenruo return -EINVAL; 3141d7f67ac9SQu Wenruo } 3142d7f67ac9SQu Wenruo 3143d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3144d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3145d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3146d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3147d7f67ac9SQu Wenruo 3148d7f67ac9SQu Wenruo return 0; 3149d7f67ac9SQu Wenruo } 3150d7f67ac9SQu Wenruo 3151ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3152ae18c37aSJosef Bacik char *options) 3153ae18c37aSJosef Bacik { 3154ae18c37aSJosef Bacik u32 sectorsize; 3155ae18c37aSJosef Bacik u32 nodesize; 3156ae18c37aSJosef Bacik u32 stripesize; 3157ae18c37aSJosef Bacik u64 generation; 3158ae18c37aSJosef Bacik u64 features; 3159ae18c37aSJosef Bacik u16 csum_type; 3160ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3161ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3162ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3163ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3164ae18c37aSJosef Bacik int ret; 3165ae18c37aSJosef Bacik int level; 3166ae18c37aSJosef Bacik 31678260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 31684871c33bSQu Wenruo if (ret) 3169ae18c37aSJosef Bacik goto fail; 317053b381b3SDavid Woodhouse 3171ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3172ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3173ae18c37aSJosef Bacik GFP_KERNEL); 3174ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3175ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3176ae18c37aSJosef Bacik GFP_KERNEL); 3177ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3178ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 31794871c33bSQu Wenruo ret = -ENOMEM; 3180c75e8394SJosef Bacik goto fail; 3181ae18c37aSJosef Bacik } 3182ae18c37aSJosef Bacik 3183dcb2137cSChristoph Hellwig ret = btrfs_init_btree_inode(sb); 31844871c33bSQu Wenruo if (ret) 3185c75e8394SJosef Bacik goto fail; 3186ae18c37aSJosef Bacik 3187d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 31881104a885SDavid Sterba 31891104a885SDavid Sterba /* 31901104a885SDavid Sterba * Read super block and check the signature bytes only 31911104a885SDavid Sterba */ 3192d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 31938f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 31944871c33bSQu Wenruo ret = PTR_ERR(disk_super); 319516cdcec7SMiao Xie goto fail_alloc; 319620b45077SDave Young } 319739279cc3SChris Mason 31981104a885SDavid Sterba /* 3199260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 32008dc3f22cSJohannes Thumshirn * corrupted, we'll find out 32018dc3f22cSJohannes Thumshirn */ 32028f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 320351bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 32048dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 320551bce6c9SJohannes Thumshirn csum_type); 32064871c33bSQu Wenruo ret = -EINVAL; 32078f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32088dc3f22cSJohannes Thumshirn goto fail_alloc; 32098dc3f22cSJohannes Thumshirn } 32108dc3f22cSJohannes Thumshirn 321183c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 321283c68bbcSSu Yue 32136d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 32146d97c6e3SJohannes Thumshirn if (ret) { 32158f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32166d97c6e3SJohannes Thumshirn goto fail_alloc; 32176d97c6e3SJohannes Thumshirn } 32186d97c6e3SJohannes Thumshirn 32198dc3f22cSJohannes Thumshirn /* 32201104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 32211104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 32221104a885SDavid Sterba */ 32233d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 322405135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 32254871c33bSQu Wenruo ret = -EINVAL; 32268f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3227141386e1SJosef Bacik goto fail_alloc; 32281104a885SDavid Sterba } 32291104a885SDavid Sterba 32301104a885SDavid Sterba /* 32311104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 32321104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 32331104a885SDavid Sterba * the whole block of INFO_SIZE 32341104a885SDavid Sterba */ 32358f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 32368f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 32375f39d397SChris Mason 3238fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3239fbc6feaeSNikolay Borisov 32400b86a832SChris Mason 3241fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3242fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3243fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3244fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3245fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3246fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3247fbc6feaeSNikolay Borisov } 3248fbc6feaeSNikolay Borisov 3249fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3250fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3251de37aa51SNikolay Borisov 3252069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 32531104a885SDavid Sterba if (ret) { 325405135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 32554871c33bSQu Wenruo ret = -EINVAL; 3256141386e1SJosef Bacik goto fail_alloc; 32571104a885SDavid Sterba } 32581104a885SDavid Sterba 32594871c33bSQu Wenruo if (!btrfs_super_root(disk_super)) { 32604871c33bSQu Wenruo btrfs_err(fs_info, "invalid superblock tree root bytenr"); 32614871c33bSQu Wenruo ret = -EINVAL; 3262141386e1SJosef Bacik goto fail_alloc; 32634871c33bSQu Wenruo } 32640f7d52f4SChris Mason 3265acce952bSliubo /* check FS state, whether FS is broken. */ 326687533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 3267ae3364e5SFilipe Manana WRITE_ONCE(fs_info->fs_error, -EUCLEAN); 3268acce952bSliubo 326975e7cb7fSLiu Bo /* 327075e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 327175e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 327275e7cb7fSLiu Bo */ 327375e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 327475e7cb7fSLiu Bo 32756f93e834SAnand Jain 32766f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 32776f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 32786f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 32796f93e834SAnand Jain stripesize = sectorsize; 32806f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 32816f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 32826f93e834SAnand Jain 32836f93e834SAnand Jain fs_info->nodesize = nodesize; 32846f93e834SAnand Jain fs_info->sectorsize = sectorsize; 32856f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 32866f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 32876f93e834SAnand Jain fs_info->stripesize = stripesize; 32886f93e834SAnand Jain 32892ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 32904871c33bSQu Wenruo if (ret) 3291141386e1SJosef Bacik goto fail_alloc; 3292dfe25020SChris Mason 32932ba48b20SQu Wenruo ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); 32944871c33bSQu Wenruo if (ret < 0) 3295141386e1SJosef Bacik goto fail_alloc; 3296f2b636e8SJosef Bacik 32978481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 32988481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 32998481dd80SQu Wenruo 33009f73f1aeSQu Wenruo /* 33019f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 33029f73f1aeSQu Wenruo * going to be deprecated. 33039f73f1aeSQu Wenruo * 33049f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 33059f73f1aeSQu Wenruo */ 33069f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 33079f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 33089f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 33099f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 33109f73f1aeSQu Wenruo 331195ea0486SQu Wenruo btrfs_warn(fs_info, 331295ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 33130bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 33148481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 33154871c33bSQu Wenruo if (!subpage_info) { 33164871c33bSQu Wenruo ret = -ENOMEM; 33178481dd80SQu Wenruo goto fail_alloc; 33184871c33bSQu Wenruo } 33198481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 33208481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3321c8050b3bSQu Wenruo } 33220bb3eb3eSQu Wenruo 3323d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 33244871c33bSQu Wenruo if (ret) 33250dc3b84aSJosef Bacik goto fail_sb_buffer; 33264543df7eSChris Mason 33279e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 33289e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 33294575c9ccSChris Mason 3330a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3331a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3332de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3333db94535dSChris Mason 3334925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 33356bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3336925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 333784eed90fSChris Mason if (ret) { 333805135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 33395d4f98a2SYan Zheng goto fail_sb_buffer; 334084eed90fSChris Mason } 33410b86a832SChris Mason 334284234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3343581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3344bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3345bd676446SJosef Bacik generation, level); 3346bd676446SJosef Bacik if (ret) { 334705135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3348af31f5e5SChris Mason goto fail_tree_roots; 334983121942SDavid Woodhouse } 33500b86a832SChris Mason 3351e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3352c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3353c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3354e17cade2SChris Mason 33555b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 33562b82032cSYan Zheng if (ret) { 335705135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3358af31f5e5SChris Mason goto fail_tree_roots; 33592b82032cSYan Zheng } 33600b86a832SChris Mason 33618dabb742SStefan Behrens /* 3362bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3363bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3364bacce86aSAnand Jain * target is required. So free devices that are not part of this 33651a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3366bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 33678dabb742SStefan Behrens */ 3368bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3369d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 337005135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 33714871c33bSQu Wenruo ret = -EIO; 3372a6b0d5c8SChris Mason goto fail_tree_roots; 3373a6b0d5c8SChris Mason } 3374a6b0d5c8SChris Mason 3375b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 33764bbcaa64SEric Sandeen if (ret) 3377b8522a1eSNikolay Borisov goto fail_tree_roots; 33788929ecfaSYan, Zheng 337975ec1db8SJosef Bacik /* 338073651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 338173651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 338273651042SNaohiro Aota * zoned incompat feature flag set. 338373651042SNaohiro Aota */ 338473651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 338573651042SNaohiro Aota if (ret) { 338673651042SNaohiro Aota btrfs_err(fs_info, 33874871c33bSQu Wenruo "zoned: failed to read device zone info: %d", ret); 338873651042SNaohiro Aota goto fail_block_groups; 338973651042SNaohiro Aota } 339073651042SNaohiro Aota 339173651042SNaohiro Aota /* 339275ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 339375ec1db8SJosef Bacik * check the generation here so we can set the 339475ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 339575ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 339675ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 339775ec1db8SJosef Bacik * even though it was perfectly fine. 339875ec1db8SJosef Bacik */ 339975ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 340075ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 340175ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 340275ec1db8SJosef Bacik 3403cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3404cf90d884SQu Wenruo if (ret) { 3405cf90d884SQu Wenruo btrfs_err(fs_info, 3406cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3407cf90d884SQu Wenruo ret); 3408cf90d884SQu Wenruo goto fail_block_groups; 3409cf90d884SQu Wenruo } 341068310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 341168310a5eSIlya Dryomov if (ret) { 341205135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 341368310a5eSIlya Dryomov goto fail_block_groups; 341468310a5eSIlya Dryomov } 341568310a5eSIlya Dryomov 3416733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3417733f4fbbSStefan Behrens if (ret) { 341805135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3419733f4fbbSStefan Behrens goto fail_block_groups; 3420733f4fbbSStefan Behrens } 3421733f4fbbSStefan Behrens 34228dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 34238dabb742SStefan Behrens if (ret) { 342405135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 34258dabb742SStefan Behrens goto fail_block_groups; 34268dabb742SStefan Behrens } 34278dabb742SStefan Behrens 3428b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3429b70f5097SNaohiro Aota if (ret) { 3430b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3431b70f5097SNaohiro Aota ret); 3432b70f5097SNaohiro Aota goto fail_block_groups; 3433b70f5097SNaohiro Aota } 3434b70f5097SNaohiro Aota 3435c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3436b7c35e81SAnand Jain if (ret) { 343705135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 343805135f59SDavid Sterba ret); 3439b7c35e81SAnand Jain goto fail_block_groups; 3440b7c35e81SAnand Jain } 3441b7c35e81SAnand Jain 344296f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 34435ac1d209SJeff Mahoney if (ret) { 344405135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3445b7c35e81SAnand Jain goto fail_fsdev_sysfs; 34465ac1d209SJeff Mahoney } 34475ac1d209SJeff Mahoney 3448c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3449c59021f8Sliubo if (ret) { 345005135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 34512365dd3cSAnand Jain goto fail_sysfs; 3452c59021f8Sliubo } 3453c59021f8Sliubo 34545b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 34551b1d1f66SJosef Bacik if (ret) { 345605135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 34572365dd3cSAnand Jain goto fail_sysfs; 34581b1d1f66SJosef Bacik } 34594330e183SQu Wenruo 346016beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 346116beac87SNaohiro Aota 3462a7e1ac7bSNaohiro Aota btrfs_check_active_zone_reservation(fs_info); 3463a7e1ac7bSNaohiro Aota 34645c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 34655c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 346605135f59SDavid Sterba btrfs_warn(fs_info, 346752042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 34684871c33bSQu Wenruo ret = -EINVAL; 34692365dd3cSAnand Jain goto fail_sysfs; 3470292fd7fcSStefan Behrens } 34719078a3e1SChris Mason 347233c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3473a74a4b97SChris Mason "btrfs-cleaner"); 34744871c33bSQu Wenruo if (IS_ERR(fs_info->cleaner_kthread)) { 34754871c33bSQu Wenruo ret = PTR_ERR(fs_info->cleaner_kthread); 34762365dd3cSAnand Jain goto fail_sysfs; 34774871c33bSQu Wenruo } 3478a74a4b97SChris Mason 3479a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3480a74a4b97SChris Mason tree_root, 3481a74a4b97SChris Mason "btrfs-transaction"); 34824871c33bSQu Wenruo if (IS_ERR(fs_info->transaction_kthread)) { 34834871c33bSQu Wenruo ret = PTR_ERR(fs_info->transaction_kthread); 34843f157a2fSChris Mason goto fail_cleaner; 34854871c33bSQu Wenruo } 3486a74a4b97SChris Mason 3487583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3488c289811cSChris Mason !fs_info->fs_devices->rotating) { 3489583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3490c289811cSChris Mason } 3491c289811cSChris Mason 3492572d9ab7SDavid Sterba /* 349363a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 349463a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 349563a7cb13SDavid Sterba * nodiscard for the same mount. 349695ca6599SNaohiro Aota * 349795ca6599SNaohiro Aota * The zoned mode piggy backs on the discard functionality for 349895ca6599SNaohiro Aota * resetting a zone. There is no reason to delay the zone reset as it is 349995ca6599SNaohiro Aota * fast enough. So, do not enable async discard for zoned mode. 350063a7cb13SDavid Sterba */ 350163a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 350263a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 350363a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 350495ca6599SNaohiro Aota fs_info->fs_devices->discardable && 350595ca6599SNaohiro Aota !btrfs_is_zoned(fs_info)) { 350663a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 350763a7cb13SDavid Sterba "auto enabling async discard"); 350863a7cb13SDavid Sterba } 350963a7cb13SDavid Sterba 351021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 35110b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 35122ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 35130b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3514cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 351521adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 351621adbd5cSStefan Behrens if (ret) 351705135f59SDavid Sterba btrfs_warn(fs_info, 351805135f59SDavid Sterba "failed to initialize integrity check module: %d", 351905135f59SDavid Sterba ret); 352021adbd5cSStefan Behrens } 352121adbd5cSStefan Behrens #endif 3522bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3523bcef60f2SArne Jansen if (ret) 3524bcef60f2SArne Jansen goto fail_trans_kthread; 352521adbd5cSStefan Behrens 3526fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3527fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3528fd708b81SJosef Bacik 352996da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 353096da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 35310b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3532e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 353363443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 35344871c33bSQu Wenruo if (ret) 353528c16cbbSWang Shilong goto fail_qgroup; 3536e556ce2cSYan Zheng } 35371a40e23bSZheng Yan 353856e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 35393140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 35404871c33bSQu Wenruo ret = PTR_ERR(fs_info->fs_root); 35414871c33bSQu Wenruo btrfs_warn(fs_info, "failed to read fs tree: %d", ret); 3542315bf8efSJosef Bacik fs_info->fs_root = NULL; 3543bcef60f2SArne Jansen goto fail_qgroup; 35443140c9a3SDan Carpenter } 3545c289811cSChris Mason 3546bc98a42cSDavid Howells if (sb_rdonly(sb)) 35478cd29088SBoris Burkov goto clear_oneshot; 35482b6ba629SIlya Dryomov 354944c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 35502b6ba629SIlya Dryomov if (ret) { 35516bccf3abSJeff Mahoney close_ctree(fs_info); 35522b6ba629SIlya Dryomov return ret; 3553e3acc2a6SJosef Bacik } 3554b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3555b382a324SJan Schmidt 355644c0ca21SBoris Burkov if (fs_info->uuid_root && 355744c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 355844c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 355905135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 356070f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 356170f80175SStefan Behrens if (ret) { 356205135f59SDavid Sterba btrfs_warn(fs_info, 356305135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 35646bccf3abSJeff Mahoney close_ctree(fs_info); 356570f80175SStefan Behrens return ret; 356670f80175SStefan Behrens } 3567f7a81ea4SStefan Behrens } 356894846229SBoris Burkov 3569afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 357047ab2a6cSJosef Bacik 3571b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3572b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3573b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3574b4be6aefSJosef Bacik 35758cd29088SBoris Burkov clear_oneshot: 35768cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3577ad2b2c80SAl Viro return 0; 357839279cc3SChris Mason 3579bcef60f2SArne Jansen fail_qgroup: 3580bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 35817c2ca468SChris Mason fail_trans_kthread: 35827c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 35832ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3584faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 35853f157a2fSChris Mason fail_cleaner: 3586a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 35877c2ca468SChris Mason 35887c2ca468SChris Mason /* 35897c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 35907c2ca468SChris Mason * kthreads 35917c2ca468SChris Mason */ 35927c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 35937c2ca468SChris Mason 35942365dd3cSAnand Jain fail_sysfs: 35956618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 35962365dd3cSAnand Jain 3597b7c35e81SAnand Jain fail_fsdev_sysfs: 3598b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3599b7c35e81SAnand Jain 36001b1d1f66SJosef Bacik fail_block_groups: 360154067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3602af31f5e5SChris Mason 3603af31f5e5SChris Mason fail_tree_roots: 36049e3aa805SJosef Bacik if (fs_info->data_reloc_root) 36059e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 36064273eaffSAnand Jain free_root_pointers(fs_info, true); 36072b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3608af31f5e5SChris Mason 360939279cc3SChris Mason fail_sb_buffer: 36107abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 36115cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 361216cdcec7SMiao Xie fail_alloc: 3613586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3614586e46e2SIlya Dryomov 36154543df7eSChris Mason iput(fs_info->btree_inode); 36167e662854SQinghuang Feng fail: 3617586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 36184871c33bSQu Wenruo ASSERT(ret < 0); 36194871c33bSQu Wenruo return ret; 3620eb60ceacSChris Mason } 3621663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3622eb60ceacSChris Mason 3623314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3624f2984462SChris Mason { 3625314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3626314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3627314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3628314b6dd0SJohannes Thumshirn struct page *page; 3629442a4f63SStefan Behrens 3630314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3631314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3632314b6dd0SJohannes Thumshirn 3633314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3634fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3635314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3636cb3e217bSQu Wenruo btrfs_dev_name(device), 3637314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3638314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3639314b6dd0SJohannes Thumshirn SetPageError(page); 3640314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3641314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3642314b6dd0SJohannes Thumshirn } else { 3643314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3644f2984462SChris Mason } 3645314b6dd0SJohannes Thumshirn 3646314b6dd0SJohannes Thumshirn put_page(page); 3647314b6dd0SJohannes Thumshirn unlock_page(page); 3648314b6dd0SJohannes Thumshirn } 3649314b6dd0SJohannes Thumshirn 3650314b6dd0SJohannes Thumshirn bio_put(bio); 3651f2984462SChris Mason } 3652f2984462SChris Mason 36538f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3654a05d3c91SQu Wenruo int copy_num, bool drop_cache) 365529c36d72SAnand Jain { 365629c36d72SAnand Jain struct btrfs_super_block *super; 36578f32380dSJohannes Thumshirn struct page *page; 365812659251SNaohiro Aota u64 bytenr, bytenr_orig; 36598f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 366012659251SNaohiro Aota int ret; 366129c36d72SAnand Jain 366212659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 366312659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 366412659251SNaohiro Aota if (ret == -ENOENT) 366512659251SNaohiro Aota return ERR_PTR(-EINVAL); 366612659251SNaohiro Aota else if (ret) 366712659251SNaohiro Aota return ERR_PTR(ret); 366812659251SNaohiro Aota 3669cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 36708f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 367129c36d72SAnand Jain 3672a05d3c91SQu Wenruo if (drop_cache) { 3673a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3674a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3675a05d3c91SQu Wenruo 3676a05d3c91SQu Wenruo /* 3677a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3678a05d3c91SQu Wenruo * always read from the device. 3679a05d3c91SQu Wenruo */ 3680a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3681a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3682a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3683a05d3c91SQu Wenruo } 3684a05d3c91SQu Wenruo 36858f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 36868f32380dSJohannes Thumshirn if (IS_ERR(page)) 36878f32380dSJohannes Thumshirn return ERR_CAST(page); 368829c36d72SAnand Jain 36898f32380dSJohannes Thumshirn super = page_address(page); 369096c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 369196c2e067SAnand Jain btrfs_release_disk_super(super); 369296c2e067SAnand Jain return ERR_PTR(-ENODATA); 369396c2e067SAnand Jain } 369496c2e067SAnand Jain 369512659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 36968f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36978f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 369829c36d72SAnand Jain } 369929c36d72SAnand Jain 37008f32380dSJohannes Thumshirn return super; 370129c36d72SAnand Jain } 370229c36d72SAnand Jain 370329c36d72SAnand Jain 37048f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3705a512bbf8SYan Zheng { 37068f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3707a512bbf8SYan Zheng int i; 3708a512bbf8SYan Zheng u64 transid = 0; 3709a512bbf8SYan Zheng 3710a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3711a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3712a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3713a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3714a512bbf8SYan Zheng */ 3715a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3716a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 37178f32380dSJohannes Thumshirn if (IS_ERR(super)) 3718a512bbf8SYan Zheng continue; 3719a512bbf8SYan Zheng 3720a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 37218f32380dSJohannes Thumshirn if (latest) 37228f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 37238f32380dSJohannes Thumshirn 37248f32380dSJohannes Thumshirn latest = super; 3725a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3726a512bbf8SYan Zheng } 3727a512bbf8SYan Zheng } 372892fc03fbSAnand Jain 37298f32380dSJohannes Thumshirn return super; 3730a512bbf8SYan Zheng } 3731a512bbf8SYan Zheng 37324eedeb75SHisashi Hifumi /* 3733abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3734314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 37354eedeb75SHisashi Hifumi * 3736abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3737abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3738abbb3b8eSDavid Sterba * same for write and wait phases. 37394eedeb75SHisashi Hifumi * 3740314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 37414eedeb75SHisashi Hifumi */ 3742a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3743abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3744a512bbf8SYan Zheng { 3745d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3746314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3747d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3748a512bbf8SYan Zheng int i; 3749a512bbf8SYan Zheng int errors = 0; 375012659251SNaohiro Aota int ret; 375112659251SNaohiro Aota u64 bytenr, bytenr_orig; 3752a512bbf8SYan Zheng 3753a512bbf8SYan Zheng if (max_mirrors == 0) 3754a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3755a512bbf8SYan Zheng 3756d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3757d5178578SJohannes Thumshirn 3758a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3759314b6dd0SJohannes Thumshirn struct page *page; 3760314b6dd0SJohannes Thumshirn struct bio *bio; 3761314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3762314b6dd0SJohannes Thumshirn 376312659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 376412659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 376512659251SNaohiro Aota if (ret == -ENOENT) { 376612659251SNaohiro Aota continue; 376712659251SNaohiro Aota } else if (ret < 0) { 376812659251SNaohiro Aota btrfs_err(device->fs_info, 376912659251SNaohiro Aota "couldn't get super block location for mirror %d", 377012659251SNaohiro Aota i); 377112659251SNaohiro Aota errors++; 377212659251SNaohiro Aota continue; 377312659251SNaohiro Aota } 3774935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3775935e5cc9SMiao Xie device->commit_total_bytes) 3776a512bbf8SYan Zheng break; 3777a512bbf8SYan Zheng 377812659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3779a512bbf8SYan Zheng 3780fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3781fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3782fd08001fSEric Biggers sb->csum); 3783a512bbf8SYan Zheng 3784314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3785314b6dd0SJohannes Thumshirn GFP_NOFS); 3786314b6dd0SJohannes Thumshirn if (!page) { 3787fb456252SJeff Mahoney btrfs_err(device->fs_info, 3788314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3789f14d104dSDavid Sterba bytenr); 3790634554dcSJosef Bacik errors++; 3791634554dcSJosef Bacik continue; 3792634554dcSJosef Bacik } 3793634554dcSJosef Bacik 3794314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3795314b6dd0SJohannes Thumshirn get_page(page); 3796a512bbf8SYan Zheng 3797314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3798314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3799a512bbf8SYan Zheng 3800387125fcSChris Mason /* 3801314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3802314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3803314b6dd0SJohannes Thumshirn * checking. 3804387125fcSChris Mason */ 380507888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 380607888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 380707888c66SChristoph Hellwig GFP_NOFS); 3808314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3809314b6dd0SJohannes Thumshirn bio->bi_private = device; 3810314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3811314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3812314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3813314b6dd0SJohannes Thumshirn 3814314b6dd0SJohannes Thumshirn /* 3815314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3816314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3817314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3818314b6dd0SJohannes Thumshirn */ 38191b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3820314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3821314b6dd0SJohannes Thumshirn 382258ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 382358ff51f1SChristoph Hellwig submit_bio(bio); 38248376d9e1SNaohiro Aota 38258376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 38268376d9e1SNaohiro Aota errors++; 3827a512bbf8SYan Zheng } 3828a512bbf8SYan Zheng return errors < i ? 0 : -1; 3829a512bbf8SYan Zheng } 3830a512bbf8SYan Zheng 3831387125fcSChris Mason /* 3832abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3833abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3834abbb3b8eSDavid Sterba * 3835314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3836abbb3b8eSDavid Sterba * date. 3837abbb3b8eSDavid Sterba */ 3838abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3839abbb3b8eSDavid Sterba { 3840abbb3b8eSDavid Sterba int i; 3841abbb3b8eSDavid Sterba int errors = 0; 3842b6a535faSHoward McLauchlan bool primary_failed = false; 384312659251SNaohiro Aota int ret; 3844abbb3b8eSDavid Sterba u64 bytenr; 3845abbb3b8eSDavid Sterba 3846abbb3b8eSDavid Sterba if (max_mirrors == 0) 3847abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3848abbb3b8eSDavid Sterba 3849abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3850314b6dd0SJohannes Thumshirn struct page *page; 3851314b6dd0SJohannes Thumshirn 385212659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 385312659251SNaohiro Aota if (ret == -ENOENT) { 385412659251SNaohiro Aota break; 385512659251SNaohiro Aota } else if (ret < 0) { 385612659251SNaohiro Aota errors++; 385712659251SNaohiro Aota if (i == 0) 385812659251SNaohiro Aota primary_failed = true; 385912659251SNaohiro Aota continue; 386012659251SNaohiro Aota } 3861abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3862abbb3b8eSDavid Sterba device->commit_total_bytes) 3863abbb3b8eSDavid Sterba break; 3864abbb3b8eSDavid Sterba 3865314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3866314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3867314b6dd0SJohannes Thumshirn if (!page) { 3868abbb3b8eSDavid Sterba errors++; 3869b6a535faSHoward McLauchlan if (i == 0) 3870b6a535faSHoward McLauchlan primary_failed = true; 3871abbb3b8eSDavid Sterba continue; 3872abbb3b8eSDavid Sterba } 3873314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3874314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3875314b6dd0SJohannes Thumshirn if (PageError(page)) { 3876abbb3b8eSDavid Sterba errors++; 3877b6a535faSHoward McLauchlan if (i == 0) 3878b6a535faSHoward McLauchlan primary_failed = true; 3879b6a535faSHoward McLauchlan } 3880abbb3b8eSDavid Sterba 3881314b6dd0SJohannes Thumshirn /* Drop our reference */ 3882314b6dd0SJohannes Thumshirn put_page(page); 3883abbb3b8eSDavid Sterba 3884314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3885314b6dd0SJohannes Thumshirn put_page(page); 3886abbb3b8eSDavid Sterba } 3887abbb3b8eSDavid Sterba 3888b6a535faSHoward McLauchlan /* log error, force error return */ 3889b6a535faSHoward McLauchlan if (primary_failed) { 3890b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3891b6a535faSHoward McLauchlan device->devid); 3892b6a535faSHoward McLauchlan return -1; 3893b6a535faSHoward McLauchlan } 3894b6a535faSHoward McLauchlan 3895abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3896abbb3b8eSDavid Sterba } 3897abbb3b8eSDavid Sterba 3898abbb3b8eSDavid Sterba /* 3899387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3900387125fcSChris Mason * for the barrier when it is done 3901387125fcSChris Mason */ 39024246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3903387125fcSChris Mason { 3904f9e69aa9SChristoph Hellwig bio_uninit(bio); 3905387125fcSChris Mason complete(bio->bi_private); 3906387125fcSChris Mason } 3907387125fcSChris Mason 3908387125fcSChris Mason /* 39094fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 39104fc6441aSAnand Jain * done in the waiting counterpart. 3911387125fcSChris Mason */ 39124fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3913387125fcSChris Mason { 3914f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 3915387125fcSChris Mason 3916bfd3ea94SAnand Jain device->last_flush_error = BLK_STS_OK; 3917bfd3ea94SAnand Jain 3918a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3919a91cf0ffSWang Yugui /* 3920a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 3921a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 3922a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 3923a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 3924a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 3925a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 3926a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 3927a91cf0ffSWang Yugui * non-debug builds. 3928a91cf0ffSWang Yugui */ 392908e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 39304fc6441aSAnand Jain return; 3931a91cf0ffSWang Yugui #endif 3932387125fcSChris Mason 3933f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 3934f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 3935387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3936387125fcSChris Mason init_completion(&device->flush_wait); 3937387125fcSChris Mason bio->bi_private = &device->flush_wait; 3938387125fcSChris Mason 393958ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 394058ff51f1SChristoph Hellwig submit_bio(bio); 39411c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 39424fc6441aSAnand Jain } 3943387125fcSChris Mason 39444fc6441aSAnand Jain /* 39454fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 39461b465784SAnand Jain * Return true for any error, and false otherwise. 39474fc6441aSAnand Jain */ 39481b465784SAnand Jain static bool wait_dev_flush(struct btrfs_device *device) 39494fc6441aSAnand Jain { 3950f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 39514fc6441aSAnand Jain 39527e812f20SAnand Jain if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 39531b465784SAnand Jain return false; 39544fc6441aSAnand Jain 39552980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 39564fc6441aSAnand Jain 3957bfd3ea94SAnand Jain if (bio->bi_status) { 3958bfd3ea94SAnand Jain device->last_flush_error = bio->bi_status; 3959bfd3ea94SAnand Jain btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); 39601b465784SAnand Jain return true; 3961bfd3ea94SAnand Jain } 3962bfd3ea94SAnand Jain 39631b465784SAnand Jain return false; 3964387125fcSChris Mason } 3965387125fcSChris Mason 3966387125fcSChris Mason /* 3967387125fcSChris Mason * send an empty flush down to each device in parallel, 3968387125fcSChris Mason * then wait for them 3969387125fcSChris Mason */ 3970387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3971387125fcSChris Mason { 3972387125fcSChris Mason struct list_head *head; 3973387125fcSChris Mason struct btrfs_device *dev; 39745af3e8ccSStefan Behrens int errors_wait = 0; 3975387125fcSChris Mason 39761538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3977387125fcSChris Mason /* send down all the barriers */ 3978387125fcSChris Mason head = &info->fs_devices->devices; 39791538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3980e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3981f88ba6a2SHidetoshi Seto continue; 3982cea7c8bfSAnand Jain if (!dev->bdev) 3983387125fcSChris Mason continue; 3984e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3985ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3986387125fcSChris Mason continue; 3987387125fcSChris Mason 39884fc6441aSAnand Jain write_dev_flush(dev); 3989387125fcSChris Mason } 3990387125fcSChris Mason 3991387125fcSChris Mason /* wait for all the barriers */ 39921538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3993e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3994f88ba6a2SHidetoshi Seto continue; 3995387125fcSChris Mason if (!dev->bdev) { 39965af3e8ccSStefan Behrens errors_wait++; 3997387125fcSChris Mason continue; 3998387125fcSChris Mason } 3999e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4000ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4001387125fcSChris Mason continue; 4002387125fcSChris Mason 40031b465784SAnand Jain if (wait_dev_flush(dev)) 40045af3e8ccSStefan Behrens errors_wait++; 4005387125fcSChris Mason } 4006401b41e5SAnand Jain 4007401b41e5SAnand Jain /* 4008de38a206SAnand Jain * Checks last_flush_error of disks in order to determine the device 4009de38a206SAnand Jain * state. 4010401b41e5SAnand Jain */ 4011de38a206SAnand Jain if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) 4012de38a206SAnand Jain return -EIO; 4013de38a206SAnand Jain 4014387125fcSChris Mason return 0; 4015387125fcSChris Mason } 4016387125fcSChris Mason 4017943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4018943c6e99SZhao Lei { 40198789f4feSZhao Lei int raid_type; 40208789f4feSZhao Lei int min_tolerated = INT_MAX; 4021943c6e99SZhao Lei 40228789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 40238789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 40248c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 40258789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 40268789f4feSZhao Lei tolerated_failures); 4027943c6e99SZhao Lei 40288789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 40298789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 40308789f4feSZhao Lei continue; 403141a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 40328789f4feSZhao Lei continue; 40338c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 40348789f4feSZhao Lei btrfs_raid_array[raid_type]. 40358789f4feSZhao Lei tolerated_failures); 40368789f4feSZhao Lei } 4037943c6e99SZhao Lei 40388789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4039ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 40408789f4feSZhao Lei min_tolerated = 0; 40418789f4feSZhao Lei } 40428789f4feSZhao Lei 40438789f4feSZhao Lei return min_tolerated; 4044943c6e99SZhao Lei } 4045943c6e99SZhao Lei 4046eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4047f2984462SChris Mason { 4048e5e9a520SChris Mason struct list_head *head; 4049f2984462SChris Mason struct btrfs_device *dev; 4050a061fc8dSChris Mason struct btrfs_super_block *sb; 4051f2984462SChris Mason struct btrfs_dev_item *dev_item; 4052f2984462SChris Mason int ret; 4053f2984462SChris Mason int do_barriers; 4054a236aed1SChris Mason int max_errors; 4055a236aed1SChris Mason int total_errors = 0; 4056a061fc8dSChris Mason u64 flags; 4057f2984462SChris Mason 40580b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4059fed3b381SLiu Bo 4060fed3b381SLiu Bo /* 4061fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4062fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4063fed3b381SLiu Bo * been consistent on disk. 4064fed3b381SLiu Bo */ 4065fed3b381SLiu Bo if (max_mirrors == 0) 40660b246afaSJeff Mahoney backup_super_roots(fs_info); 4067f2984462SChris Mason 40680b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4069a061fc8dSChris Mason dev_item = &sb->dev_item; 4070e5e9a520SChris Mason 40710b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 40720b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 40730b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4074387125fcSChris Mason 40755af3e8ccSStefan Behrens if (do_barriers) { 40760b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 40775af3e8ccSStefan Behrens if (ret) { 40785af3e8ccSStefan Behrens mutex_unlock( 40790b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 40800b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 40815af3e8ccSStefan Behrens "errors while submitting device barriers."); 40825af3e8ccSStefan Behrens return ret; 40835af3e8ccSStefan Behrens } 40845af3e8ccSStefan Behrens } 4085387125fcSChris Mason 40861538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4087dfe25020SChris Mason if (!dev->bdev) { 4088dfe25020SChris Mason total_errors++; 4089dfe25020SChris Mason continue; 4090dfe25020SChris Mason } 4091e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4092ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4093dfe25020SChris Mason continue; 4094dfe25020SChris Mason 40952b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4096a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4097a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 40987df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4099935e5cc9SMiao Xie dev->commit_total_bytes); 4100ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4101ce7213c7SMiao Xie dev->commit_bytes_used); 4102a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4103a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4104a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4105a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 41067239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 41077239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4108a512bbf8SYan Zheng 4109a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4110a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4111f2984462SChris Mason 411275cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 411375cb857dSQu Wenruo if (ret < 0) { 411475cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 411575cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 411675cb857dSQu Wenruo "unexpected superblock corruption detected"); 411775cb857dSQu Wenruo return -EUCLEAN; 411875cb857dSQu Wenruo } 411975cb857dSQu Wenruo 4120abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4121a236aed1SChris Mason if (ret) 4122a236aed1SChris Mason total_errors++; 4123f2984462SChris Mason } 4124a236aed1SChris Mason if (total_errors > max_errors) { 41250b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4126d397712bSChris Mason total_errors); 41270b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 412879787eaaSJeff Mahoney 41299d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 41300b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 41310b246afaSJeff Mahoney "%d errors while writing supers", 41320b246afaSJeff Mahoney total_errors); 41339d565ba4SStefan Behrens return -EIO; 4134a236aed1SChris Mason } 4135f2984462SChris Mason 4136a512bbf8SYan Zheng total_errors = 0; 41371538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4138dfe25020SChris Mason if (!dev->bdev) 4139dfe25020SChris Mason continue; 4140e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4141ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4142dfe25020SChris Mason continue; 4143dfe25020SChris Mason 4144abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4145a512bbf8SYan Zheng if (ret) 41461259ab75SChris Mason total_errors++; 4147f2984462SChris Mason } 41480b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4149a236aed1SChris Mason if (total_errors > max_errors) { 41500b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 41510b246afaSJeff Mahoney "%d errors while writing supers", 41520b246afaSJeff Mahoney total_errors); 415379787eaaSJeff Mahoney return -EIO; 4154a236aed1SChris Mason } 4155f2984462SChris Mason return 0; 4156f2984462SChris Mason } 4157f2984462SChris Mason 4158cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4159cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4160cb517eabSMiao Xie struct btrfs_root *root) 41612619ba1fSChris Mason { 41624785e24fSJosef Bacik bool drop_ref = false; 41634785e24fSJosef Bacik 4164fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4165fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4166fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4167fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 41684785e24fSJosef Bacik drop_ref = true; 4169fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 417076dda93cSYan, Zheng 417184961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4172ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 41731c1ea4f7SLiu Bo if (root->reloc_root) { 417400246528SJosef Bacik btrfs_put_root(root->reloc_root); 41751c1ea4f7SLiu Bo root->reloc_root = NULL; 41761c1ea4f7SLiu Bo } 41771c1ea4f7SLiu Bo } 41783321719eSLiu Bo 41794785e24fSJosef Bacik if (drop_ref) 418000246528SJosef Bacik btrfs_put_root(root); 41812619ba1fSChris Mason } 41822619ba1fSChris Mason 41836bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4184c146afadSYan Zheng { 41856bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4186c146afadSYan Zheng struct btrfs_trans_handle *trans; 4187c146afadSYan Zheng 41880b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 41892ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 41900b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 41910b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4192c71bf099SYan, Zheng 4193c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 41940b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 41950b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4196c71bf099SYan, Zheng 41977a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 41983612b495STsutomu Itoh if (IS_ERR(trans)) 41993612b495STsutomu Itoh return PTR_ERR(trans); 42003a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4201c146afadSYan Zheng } 4202c146afadSYan Zheng 420336c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 420436c86a9eSQu Wenruo { 420536c86a9eSQu Wenruo struct btrfs_transaction *trans; 420636c86a9eSQu Wenruo struct btrfs_transaction *tmp; 420736c86a9eSQu Wenruo bool found = false; 420836c86a9eSQu Wenruo 420936c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 421036c86a9eSQu Wenruo return; 421136c86a9eSQu Wenruo 421236c86a9eSQu Wenruo /* 421336c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 421436c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 421536c86a9eSQu Wenruo */ 421636c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 421736c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 421836c86a9eSQu Wenruo struct extent_state *cached = NULL; 421936c86a9eSQu Wenruo u64 dirty_bytes = 0; 422036c86a9eSQu Wenruo u64 cur = 0; 422136c86a9eSQu Wenruo u64 found_start; 422236c86a9eSQu Wenruo u64 found_end; 422336c86a9eSQu Wenruo 422436c86a9eSQu Wenruo found = true; 4225e5860f82SFilipe Manana while (find_first_extent_bit(&trans->dirty_pages, cur, 422636c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 422736c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 422836c86a9eSQu Wenruo cur = found_end + 1; 422936c86a9eSQu Wenruo } 423036c86a9eSQu Wenruo btrfs_warn(fs_info, 423136c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 423236c86a9eSQu Wenruo trans->transid, dirty_bytes); 423336c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 423436c86a9eSQu Wenruo 423536c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 423636c86a9eSQu Wenruo fs_info->running_transaction = NULL; 423736c86a9eSQu Wenruo list_del_init(&trans->list); 423836c86a9eSQu Wenruo 423936c86a9eSQu Wenruo btrfs_put_transaction(trans); 424036c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 424136c86a9eSQu Wenruo } 424236c86a9eSQu Wenruo ASSERT(!found); 424336c86a9eSQu Wenruo } 424436c86a9eSQu Wenruo 4245b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4246eb60ceacSChris Mason { 4247c146afadSYan Zheng int ret; 4248e089f05cSChris Mason 4249afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 425031e70e52SFilipe Manana 425131e70e52SFilipe Manana /* 42528a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 42538a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 42548a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 42558a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 42568a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 42578a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 42588a1f1e3dSFilipe Manana * return 1. 42598a1f1e3dSFilipe Manana */ 42608a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 42618a1f1e3dSFilipe Manana 42628a1f1e3dSFilipe Manana /* 426331e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 426431e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 426531e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 426631e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 426731e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 426831e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 426931e70e52SFilipe Manana * trying to stop the async reclaim task. 427031e70e52SFilipe Manana */ 427131e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4272d6fd0ae2SOmar Sandoval /* 4273d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4274d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4275d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4276d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4277d6fd0ae2SOmar Sandoval */ 4278d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4279a2135011SChris Mason 42807343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4281d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 42827343dd61SJustin Maggard 4283803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4284803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4285803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4286803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4287803b2f54SStefan Behrens 4288837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4289aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4290837d5b6eSIlya Dryomov 42918dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 42928dabb742SStefan Behrens 4293aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 42944cb5300bSChris Mason 42954cb5300bSChris Mason /* wait for any defraggers to finish */ 42964cb5300bSChris Mason wait_event(fs_info->transaction_wait, 42974cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 42984cb5300bSChris Mason 42994cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 430026176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 43014cb5300bSChris Mason 4302a362bb86SFilipe Manana /* 4303a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4304a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4305a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4306a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4307a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4308a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4309a362bb86SFilipe Manana * no one else to run iputs. 4310a362bb86SFilipe Manana * 4311a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4312a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4313a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4314a362bb86SFilipe Manana * through some other means. 4315a362bb86SFilipe Manana * 4316a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4317a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4318a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4319a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4320a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4321a362bb86SFilipe Manana */ 4322a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4323a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4324a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4325a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4326a362bb86SFilipe Manana 432721c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 432857056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4329576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 433021c7e756SMiao Xie 4331b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4332b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4333b0643e59SDennis Zhou 4334bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4335e44163e1SJeff Mahoney /* 4336d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4337d6fd0ae2SOmar Sandoval * unused block groups. 4338e44163e1SJeff Mahoney */ 43390b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4340e44163e1SJeff Mahoney 4341f0cc2cd7SFilipe Manana /* 4342f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4343f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4344f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4345f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4346f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4347f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4348f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4349f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4350f0cc2cd7SFilipe Manana * This is a very rare case. 4351f0cc2cd7SFilipe Manana */ 4352f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4353f0cc2cd7SFilipe Manana 43546bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4355d397712bSChris Mason if (ret) 435604892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4357c146afadSYan Zheng } 4358ed2ff2cbSChris Mason 435984961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 43602ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4361acce952bSliubo 4362e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4363e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 43648929ecfaSYan, Zheng 4365e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4366afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4367f25784b3SYan Zheng 43685958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 43695958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 43705958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 43715958253cSQu Wenruo } 43725958253cSQu Wenruo 437304892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4374fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4375bcef60f2SArne Jansen 4376963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 437704892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4378963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4379b0c68f8bSChris Mason } 438031153d81SYan Zheng 43815deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 43824297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 43835deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 43844297ff84SJosef Bacik 43856618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4386b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 43875ac1d209SJeff Mahoney 43881a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 43891a4319ccSLiu Bo 4390de348ee0SWang Shilong /* 4391de348ee0SWang Shilong * we must make sure there is not any read request to 4392de348ee0SWang Shilong * submit after we stopping all workers. 4393de348ee0SWang Shilong */ 4394de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 439596192499SJosef Bacik btrfs_stop_all_workers(fs_info); 439696192499SJosef Bacik 43970a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 439836c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 43990a31daa4SFilipe Manana 4400afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 44014273eaffSAnand Jain free_root_pointers(fs_info, true); 44028c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 44039ad6b7bcSChris Mason 44044e19443dSJosef Bacik /* 44054e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 44064e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 44074e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 44084e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 44094e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 44104e19443dSJosef Bacik */ 44114e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 44124e19443dSJosef Bacik 441313e6c37bSJosef Bacik iput(fs_info->btree_inode); 4414d6bfde87SChris Mason 441521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 44160b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 44172ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 441821adbd5cSStefan Behrens #endif 441921adbd5cSStefan Behrens 44200b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 442168c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4422eb60ceacSChris Mason } 4423eb60ceacSChris Mason 44245f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 44255f39d397SChris Mason { 44262f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 44275f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4428b4ce94deSChris Mason 442906ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 443006ea65a3SJosef Bacik /* 443106ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 443252042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 443306ea65a3SJosef Bacik * outside of the sanity tests. 443406ea65a3SJosef Bacik */ 4435b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 443606ea65a3SJosef Bacik return; 443706ea65a3SJosef Bacik #endif 443849d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 44390b246afaSJeff Mahoney if (transid != fs_info->generation) 44405d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 44410b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 4442f18cc978SChristoph Hellwig set_extent_buffer_dirty(buf); 44431f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 444469fc6cbbSQu Wenruo /* 444585d8a826SJosef Bacik * btrfs_check_leaf() won't check item data if we don't have WRITTEN 444685d8a826SJosef Bacik * set, so this will only validate the basic structure of the items. 444769fc6cbbSQu Wenruo */ 444885d8a826SJosef Bacik if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(buf)) { 4449a4f78750SDavid Sterba btrfs_print_leaf(buf); 44501f21ef0aSFilipe Manana ASSERT(0); 44511f21ef0aSFilipe Manana } 44521f21ef0aSFilipe Manana #endif 4453eb60ceacSChris Mason } 4454eb60ceacSChris Mason 44552ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4456b53d3f5dSLiu Bo int flush_delayed) 445735b7e476SChris Mason { 4458188de649SChris Mason /* 4459188de649SChris Mason * looks as though older kernels can get into trouble with 4460188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4461188de649SChris Mason */ 4462e2d84521SMiao Xie int ret; 4463d6bfde87SChris Mason 44646933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4465d6bfde87SChris Mason return; 4466d6bfde87SChris Mason 4467b53d3f5dSLiu Bo if (flush_delayed) 44682ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 446916cdcec7SMiao Xie 4470d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4471d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4472d814a491SEthan Lien fs_info->dirty_metadata_batch); 4473e2d84521SMiao Xie if (ret > 0) { 44740b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 447516cdcec7SMiao Xie } 447616cdcec7SMiao Xie } 447716cdcec7SMiao Xie 44782ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 447916cdcec7SMiao Xie { 44802ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 448135b7e476SChris Mason } 4482b53d3f5dSLiu Bo 44832ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4484b53d3f5dSLiu Bo { 44852ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4486d6bfde87SChris Mason } 44876b80053dSChris Mason 44882ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4489acce952bSliubo { 4490fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4491fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4492fe816d0fSNikolay Borisov 44930b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 44942ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 44950b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4496acce952bSliubo 44970b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 44980b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4499acce952bSliubo } 4500acce952bSliubo 4501ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4502ef67963dSJosef Bacik { 4503fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4504fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4505fc7cbcd4SDavid Sterba int ret; 4506ef67963dSJosef Bacik 4507fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4508fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4509fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4510fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4511fc7cbcd4SDavid Sterba int i; 4512ef67963dSJosef Bacik 4513fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4514fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4515fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4516fc7cbcd4SDavid Sterba 4517fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4518fc7cbcd4SDavid Sterba if (!gang[i]) 4519ef67963dSJosef Bacik continue; 4520fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4521fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4522fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4523ef67963dSJosef Bacik } 4524fc7cbcd4SDavid Sterba root_objectid++; 4525fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4526ef67963dSJosef Bacik } 4527fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4528ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4529ef67963dSJosef Bacik } 4530ef67963dSJosef Bacik 4531143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4532acce952bSliubo { 4533acce952bSliubo struct btrfs_ordered_extent *ordered; 4534acce952bSliubo 4535199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4536779880efSJosef Bacik /* 4537779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4538779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4539779880efSJosef Bacik */ 4540199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4541779880efSJosef Bacik root_extent_list) 4542779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4543199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4544199c2a9cSMiao Xie } 4545199c2a9cSMiao Xie 4546199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4547199c2a9cSMiao Xie { 4548199c2a9cSMiao Xie struct btrfs_root *root; 454984af994bSRuan Jinjie LIST_HEAD(splice); 4550199c2a9cSMiao Xie 4551199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4552199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4553199c2a9cSMiao Xie while (!list_empty(&splice)) { 4554199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4555199c2a9cSMiao Xie ordered_root); 45561de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 45571de2cfdeSJosef Bacik &fs_info->ordered_roots); 4558199c2a9cSMiao Xie 45592a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4560199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4561199c2a9cSMiao Xie 45622a85d9caSLiu Bo cond_resched(); 45632a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4564199c2a9cSMiao Xie } 4565199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 456674d5d229SJosef Bacik 456774d5d229SJosef Bacik /* 456874d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 456974d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 457074d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 457174d5d229SJosef Bacik * actually get run and error out properly. 457274d5d229SJosef Bacik */ 457374d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4574acce952bSliubo } 4575acce952bSliubo 457699f09ce3SFilipe Manana static void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 45772ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4578acce952bSliubo { 4579acce952bSliubo struct rb_node *node; 4580acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4581acce952bSliubo struct btrfs_delayed_ref_node *ref; 4582acce952bSliubo 4583acce952bSliubo delayed_refs = &trans->delayed_refs; 4584acce952bSliubo 4585acce952bSliubo spin_lock(&delayed_refs->lock); 4586d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4587cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4588b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 458999f09ce3SFilipe Manana return; 4590acce952bSliubo } 4591acce952bSliubo 45925c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4593d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 45940e0adbcfSJosef Bacik struct rb_node *n; 4595e78417d1SJosef Bacik bool pin_bytes = false; 4596acce952bSliubo 4597d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4598d7df2c79SJosef Bacik href_node); 45993069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4600b939d1abSJosef Bacik continue; 46013069bd26SJosef Bacik 4602d7df2c79SJosef Bacik spin_lock(&head->lock); 4603e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 46040e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 46050e0adbcfSJosef Bacik ref_node); 4606e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 46070e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 46081d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 46091d57ee94SWang Xiaoguang list_del(&ref->add_list); 4610d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4611d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4612d7df2c79SJosef Bacik } 461354067ae9SJosef Bacik if (head->must_insert_reserved) 4614e78417d1SJosef Bacik pin_bytes = true; 461578a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4616fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4617d7df2c79SJosef Bacik spin_unlock(&head->lock); 4618acce952bSliubo spin_unlock(&delayed_refs->lock); 4619e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4620acce952bSliubo 4621f603bb94SNikolay Borisov if (pin_bytes) { 4622f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4623f603bb94SNikolay Borisov 4624f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4625f603bb94SNikolay Borisov BUG_ON(!cache); 4626f603bb94SNikolay Borisov 4627f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4628f603bb94SNikolay Borisov spin_lock(&cache->lock); 4629f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4630f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4631f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4632f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4633f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4634f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4635f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4636f603bb94SNikolay Borisov 4637f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4638f603bb94SNikolay Borisov 4639f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4640f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4641f603bb94SNikolay Borisov } 464231890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4643d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4644acce952bSliubo cond_resched(); 4645acce952bSliubo spin_lock(&delayed_refs->lock); 4646acce952bSliubo } 464781f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4648acce952bSliubo 4649acce952bSliubo spin_unlock(&delayed_refs->lock); 4650acce952bSliubo } 4651acce952bSliubo 4652143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4653acce952bSliubo { 4654acce952bSliubo struct btrfs_inode *btrfs_inode; 465584af994bSRuan Jinjie LIST_HEAD(splice); 4656acce952bSliubo 4657eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4658eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4659acce952bSliubo 4660acce952bSliubo while (!list_empty(&splice)) { 4661fe816d0fSNikolay Borisov struct inode *inode = NULL; 4662eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4663acce952bSliubo delalloc_inodes); 4664fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4665eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4666acce952bSliubo 4667fe816d0fSNikolay Borisov /* 4668fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4669fe816d0fSNikolay Borisov * meanwhile. 4670fe816d0fSNikolay Borisov */ 4671fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4672fe816d0fSNikolay Borisov if (inode) { 4673597441b3SJosef Bacik unsigned int nofs_flag; 4674597441b3SJosef Bacik 4675597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4676fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4677597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4678fe816d0fSNikolay Borisov iput(inode); 4679fe816d0fSNikolay Borisov } 4680eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4681acce952bSliubo } 4682eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4683eb73c1b7SMiao Xie } 4684eb73c1b7SMiao Xie 4685eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4686eb73c1b7SMiao Xie { 4687eb73c1b7SMiao Xie struct btrfs_root *root; 468884af994bSRuan Jinjie LIST_HEAD(splice); 4689eb73c1b7SMiao Xie 4690eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4691eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4692eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4693eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4694eb73c1b7SMiao Xie delalloc_root); 469500246528SJosef Bacik root = btrfs_grab_root(root); 4696eb73c1b7SMiao Xie BUG_ON(!root); 4697eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4698eb73c1b7SMiao Xie 4699eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 470000246528SJosef Bacik btrfs_put_root(root); 4701eb73c1b7SMiao Xie 4702eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4703eb73c1b7SMiao Xie } 4704eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4705acce952bSliubo } 4706acce952bSliubo 4707aec5716cSFilipe Manana static void btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4708acce952bSliubo struct extent_io_tree *dirty_pages, 4709acce952bSliubo int mark) 4710acce952bSliubo { 4711acce952bSliubo struct extent_buffer *eb; 4712acce952bSliubo u64 start = 0; 4713acce952bSliubo u64 end; 4714acce952bSliubo 4715e5860f82SFilipe Manana while (find_first_extent_bit(dirty_pages, start, &start, &end, 4716aec5716cSFilipe Manana mark, NULL)) { 471791166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4718acce952bSliubo while (start <= end) { 47190b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 47200b246afaSJeff Mahoney start += fs_info->nodesize; 4721fd8b2b61SJosef Bacik if (!eb) 4722acce952bSliubo continue; 4723acce952bSliubo 4724c4e54a65SJosef Bacik btrfs_tree_lock(eb); 4725c4e54a65SJosef Bacik wait_on_extent_buffer_writeback(eb); 4726190a8339SJosef Bacik btrfs_clear_buffer_dirty(NULL, eb); 4727c4e54a65SJosef Bacik btrfs_tree_unlock(eb); 4728c4e54a65SJosef Bacik 4729fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4730acce952bSliubo } 4731acce952bSliubo } 4732acce952bSliubo } 4733acce952bSliubo 473446d81ebdSFilipe Manana static void btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4735fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4736acce952bSliubo { 4737acce952bSliubo u64 start; 4738acce952bSliubo u64 end; 4739acce952bSliubo 4740acce952bSliubo while (1) { 47410e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 47420e6ec385SFilipe Manana 4743fcd5e742SLu Fengqi /* 4744fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4745fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4746fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4747fcd5e742SLu Fengqi * the same extent range. 4748fcd5e742SLu Fengqi */ 4749fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4750e5860f82SFilipe Manana if (!find_first_extent_bit(unpin, 0, &start, &end, 475146d81ebdSFilipe Manana EXTENT_DIRTY, &cached_state)) { 4752fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4753acce952bSliubo break; 4754fcd5e742SLu Fengqi } 4755acce952bSliubo 47560e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 47570e6ec385SFilipe Manana free_extent_state(cached_state); 47582ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4759fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4760acce952bSliubo cond_resched(); 4761acce952bSliubo } 4762acce952bSliubo } 4763acce952bSliubo 476432da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4765c79a1751SLiu Bo { 4766c79a1751SLiu Bo struct inode *inode; 4767c79a1751SLiu Bo 4768c79a1751SLiu Bo inode = cache->io_ctl.inode; 4769c79a1751SLiu Bo if (inode) { 4770597441b3SJosef Bacik unsigned int nofs_flag; 4771597441b3SJosef Bacik 4772597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4773c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4774597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4775597441b3SJosef Bacik 4776c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4777c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4778c79a1751SLiu Bo iput(inode); 4779c79a1751SLiu Bo } 4780bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4781c79a1751SLiu Bo btrfs_put_block_group(cache); 4782c79a1751SLiu Bo } 4783c79a1751SLiu Bo 4784c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 47852ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4786c79a1751SLiu Bo { 478732da5386SDavid Sterba struct btrfs_block_group *cache; 4788c79a1751SLiu Bo 4789c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4790c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4791c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 479232da5386SDavid Sterba struct btrfs_block_group, 4793c79a1751SLiu Bo dirty_list); 4794c79a1751SLiu Bo 4795c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4796c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4797c79a1751SLiu Bo list_del_init(&cache->io_list); 4798c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4799c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4800c79a1751SLiu Bo } 4801c79a1751SLiu Bo 4802c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4803c79a1751SLiu Bo spin_lock(&cache->lock); 4804c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4805c79a1751SLiu Bo spin_unlock(&cache->lock); 4806c79a1751SLiu Bo 4807c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4808c79a1751SLiu Bo btrfs_put_block_group(cache); 4809ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4810c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4811c79a1751SLiu Bo } 4812c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4813c79a1751SLiu Bo 481445ae2c18SNikolay Borisov /* 481545ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 481645ae2c18SNikolay Borisov * to use it without any locking 481745ae2c18SNikolay Borisov */ 4818c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4819c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 482032da5386SDavid Sterba struct btrfs_block_group, 4821c79a1751SLiu Bo io_list); 4822c79a1751SLiu Bo 4823c79a1751SLiu Bo list_del_init(&cache->io_list); 4824c79a1751SLiu Bo spin_lock(&cache->lock); 4825c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4826c79a1751SLiu Bo spin_unlock(&cache->lock); 4827c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4828c79a1751SLiu Bo } 4829c79a1751SLiu Bo } 4830c79a1751SLiu Bo 483149b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 48322ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 483349b25e05SJeff Mahoney { 4834bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4835bbbf7243SNikolay Borisov 48362ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4837c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4838c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4839c79a1751SLiu Bo 4840bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4841bbbf7243SNikolay Borisov post_commit_list) { 4842bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4843bbbf7243SNikolay Borisov } 4844bbbf7243SNikolay Borisov 48452ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 484649b25e05SJeff Mahoney 48474a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 48480b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 484949b25e05SJeff Mahoney 48504a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 48510b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 485249b25e05SJeff Mahoney 4853ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 485467cde344SMiao Xie 48552ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 485649b25e05SJeff Mahoney EXTENT_DIRTY); 4857fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 485849b25e05SJeff Mahoney 48594a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 48604a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 486149b25e05SJeff Mahoney } 486249b25e05SJeff Mahoney 48632ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4864acce952bSliubo { 4865acce952bSliubo struct btrfs_transaction *t; 4866acce952bSliubo 48670b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4868acce952bSliubo 48690b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 48700b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 48710b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4872724e2315SJosef Bacik struct btrfs_transaction, list); 4873724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 48749b64f57dSElena Reshetova refcount_inc(&t->use_count); 48750b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 48762ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4877724e2315SJosef Bacik btrfs_put_transaction(t); 48780b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4879724e2315SJosef Bacik continue; 4880724e2315SJosef Bacik } 48810b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4882724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 48830b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4884724e2315SJosef Bacik /* 4885724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4886724e2315SJosef Bacik * handle open currently for this transaction. 4887724e2315SJosef Bacik */ 4888724e2315SJosef Bacik wait_event(t->writer_wait, 4889724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4890724e2315SJosef Bacik } else { 48910b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4892724e2315SJosef Bacik } 48932ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4894724e2315SJosef Bacik 48950b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 48960b246afaSJeff Mahoney if (t == fs_info->running_transaction) 48970b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4898724e2315SJosef Bacik list_del_init(&t->list); 48990b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4900a4abeea4SJosef Bacik 4901724e2315SJosef Bacik btrfs_put_transaction(t); 49022e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 49030b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4904724e2315SJosef Bacik } 49050b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49060b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4907ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4908ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 49090b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4910ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 49110b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4912acce952bSliubo 4913acce952bSliubo return 0; 4914acce952bSliubo } 4915ec7d6dfdSNikolay Borisov 4916453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 4917ec7d6dfdSNikolay Borisov { 4918ec7d6dfdSNikolay Borisov struct btrfs_path *path; 4919ec7d6dfdSNikolay Borisov int ret; 4920ec7d6dfdSNikolay Borisov struct extent_buffer *l; 4921ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 4922ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 4923ec7d6dfdSNikolay Borisov int slot; 4924ec7d6dfdSNikolay Borisov 4925ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 4926ec7d6dfdSNikolay Borisov if (!path) 4927ec7d6dfdSNikolay Borisov return -ENOMEM; 4928ec7d6dfdSNikolay Borisov 4929ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 4930ec7d6dfdSNikolay Borisov search_key.type = -1; 4931ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 4932ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 4933ec7d6dfdSNikolay Borisov if (ret < 0) 4934ec7d6dfdSNikolay Borisov goto error; 4935ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 4936ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 4937ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 4938ec7d6dfdSNikolay Borisov l = path->nodes[0]; 4939ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 494023125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 494123125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 4942ec7d6dfdSNikolay Borisov } else { 494323125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 4944ec7d6dfdSNikolay Borisov } 4945ec7d6dfdSNikolay Borisov ret = 0; 4946ec7d6dfdSNikolay Borisov error: 4947ec7d6dfdSNikolay Borisov btrfs_free_path(path); 4948ec7d6dfdSNikolay Borisov return ret; 4949ec7d6dfdSNikolay Borisov } 4950ec7d6dfdSNikolay Borisov 4951543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 4952ec7d6dfdSNikolay Borisov { 4953ec7d6dfdSNikolay Borisov int ret; 4954ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 4955ec7d6dfdSNikolay Borisov 49566b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 4957ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 4958ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 4959ec7d6dfdSNikolay Borisov root->root_key.objectid); 4960ec7d6dfdSNikolay Borisov ret = -ENOSPC; 4961ec7d6dfdSNikolay Borisov goto out; 4962ec7d6dfdSNikolay Borisov } 4963ec7d6dfdSNikolay Borisov 496423125104SNikolay Borisov *objectid = root->free_objectid++; 4965ec7d6dfdSNikolay Borisov ret = 0; 4966ec7d6dfdSNikolay Borisov out: 4967ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 4968ec7d6dfdSNikolay Borisov return ret; 4969ec7d6dfdSNikolay Borisov } 4970