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]; 316944d3f9fSNikolay Borisov u8 *metadata_uuid; 3172b82032cSYan Zheng 3189a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 3199a8658e3SDavid Sterba BTRFS_FSID_SIZE); 3207239ff4bSNikolay Borisov /* 321944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 322944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 323944d3f9fSNikolay Borisov * ->fsid in this case is fine 3247239ff4bSNikolay Borisov */ 325944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 3267239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 3277239ff4bSNikolay Borisov else 3287239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 3297239ff4bSNikolay Borisov 330944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 331413fb1bcSAnand Jain return false; 332944d3f9fSNikolay Borisov 333944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 334944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 335413fb1bcSAnand Jain return false; 336944d3f9fSNikolay Borisov 337413fb1bcSAnand Jain return true; 3382b82032cSYan Zheng } 3392b82032cSYan Zheng 34077bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 341046b562bSChristoph Hellwig int btrfs_validate_extent_buffer(struct extent_buffer *eb, 342947a6299SQu Wenruo struct btrfs_tree_parent_check *check) 343ce9adaa5SChris Mason { 34477bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 345ce9adaa5SChris Mason u64 found_start; 34677bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 34777bf40a2SQu Wenruo u8 found_level; 3482996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 349dfd29eedSDavid Sterba const u8 *header_csum; 35077bf40a2SQu Wenruo int ret = 0; 351ea466794SJosef Bacik 352947a6299SQu Wenruo ASSERT(check); 353947a6299SQu Wenruo 354ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 355727011e0SChris Mason if (found_start != eb->start) { 3568f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, 3578f0ed7d4SQu Wenruo "bad tree block start, mirror %u want %llu have %llu", 3588f0ed7d4SQu Wenruo eb->read_mirror, eb->start, found_start); 359f188591eSChris Mason ret = -EIO; 36077bf40a2SQu Wenruo goto out; 361ce9adaa5SChris Mason } 362b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 3638f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", 3648f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 3651259ab75SChris Mason ret = -EIO; 36677bf40a2SQu Wenruo goto out; 3671259ab75SChris Mason } 368ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 3691c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 3708f0ed7d4SQu Wenruo btrfs_err(fs_info, 3718f0ed7d4SQu Wenruo "bad tree block level, mirror %u level %d on logical %llu", 3728f0ed7d4SQu Wenruo eb->read_mirror, btrfs_header_level(eb), eb->start); 3731c24c3ceSJosef Bacik ret = -EIO; 37477bf40a2SQu Wenruo goto out; 3751c24c3ceSJosef Bacik } 376ce9adaa5SChris Mason 377c67b3892SDavid Sterba csum_tree_block(eb, result); 378dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 379dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 380a826d6dcSJosef Bacik 381dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 3822996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 3838f0ed7d4SQu Wenruo "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", 3848f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 385dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 38635be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 38735be8851SJohannes Thumshirn btrfs_header_level(eb)); 3882996e1f8SJohannes Thumshirn ret = -EUCLEAN; 38977bf40a2SQu Wenruo goto out; 3902996e1f8SJohannes Thumshirn } 3912996e1f8SJohannes Thumshirn 392947a6299SQu Wenruo if (found_level != check->level) { 39377177ed1SQu Wenruo btrfs_err(fs_info, 39477177ed1SQu Wenruo "level verify failed on logical %llu mirror %u wanted %u found %u", 39577177ed1SQu Wenruo eb->start, eb->read_mirror, check->level, found_level); 396947a6299SQu Wenruo ret = -EIO; 397947a6299SQu Wenruo goto out; 398947a6299SQu Wenruo } 399947a6299SQu Wenruo if (unlikely(check->transid && 400947a6299SQu Wenruo btrfs_header_generation(eb) != check->transid)) { 401947a6299SQu Wenruo btrfs_err_rl(eb->fs_info, 402947a6299SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 403947a6299SQu Wenruo eb->start, eb->read_mirror, check->transid, 404947a6299SQu Wenruo btrfs_header_generation(eb)); 405947a6299SQu Wenruo ret = -EIO; 406947a6299SQu Wenruo goto out; 407947a6299SQu Wenruo } 408947a6299SQu Wenruo if (check->has_first_key) { 409947a6299SQu Wenruo struct btrfs_key *expect_key = &check->first_key; 410947a6299SQu Wenruo struct btrfs_key found_key; 411947a6299SQu Wenruo 412947a6299SQu Wenruo if (found_level) 413947a6299SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 414947a6299SQu Wenruo else 415947a6299SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 416947a6299SQu Wenruo if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { 417947a6299SQu Wenruo btrfs_err(fs_info, 418947a6299SQu Wenruo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 419947a6299SQu Wenruo eb->start, check->transid, 420947a6299SQu Wenruo expect_key->objectid, 421947a6299SQu Wenruo expect_key->type, expect_key->offset, 422947a6299SQu Wenruo found_key.objectid, found_key.type, 423947a6299SQu Wenruo found_key.offset); 424947a6299SQu Wenruo ret = -EUCLEAN; 425947a6299SQu Wenruo goto out; 426947a6299SQu Wenruo } 427947a6299SQu Wenruo } 428947a6299SQu Wenruo if (check->owner_root) { 429947a6299SQu Wenruo ret = btrfs_check_eb_owner(eb, check->owner_root); 430947a6299SQu Wenruo if (ret < 0) 431947a6299SQu Wenruo goto out; 432947a6299SQu Wenruo } 433947a6299SQu Wenruo 434a826d6dcSJosef Bacik /* 435a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 436a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 437a826d6dcSJosef Bacik * return -EIO. 438a826d6dcSJosef Bacik */ 43985d8a826SJosef Bacik if (found_level == 0 && btrfs_check_leaf(eb)) { 440a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 441a826d6dcSJosef Bacik ret = -EIO; 442a826d6dcSJosef Bacik } 443ce9adaa5SChris Mason 444813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 445053ab70fSLiu Bo ret = -EIO; 446053ab70fSLiu Bo 447aebcc159SChristoph Hellwig if (ret) 44875391f0dSQu Wenruo btrfs_err(fs_info, 4498f0ed7d4SQu Wenruo "read time tree block corruption detected on logical %llu mirror %u", 4508f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 45177bf40a2SQu Wenruo out: 45277bf40a2SQu Wenruo return ret; 45377bf40a2SQu Wenruo } 45477bf40a2SQu Wenruo 4553dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 4568958b551SMatthew Wilcox (Oracle) static int btree_migrate_folio(struct address_space *mapping, 4578958b551SMatthew Wilcox (Oracle) struct folio *dst, struct folio *src, enum migrate_mode mode) 458784b4e29SChris Mason { 459784b4e29SChris Mason /* 460784b4e29SChris Mason * we can't safely write a btree page from here, 461784b4e29SChris Mason * we haven't done the locking hook 462784b4e29SChris Mason */ 4638958b551SMatthew Wilcox (Oracle) if (folio_test_dirty(src)) 464784b4e29SChris Mason return -EAGAIN; 465784b4e29SChris Mason /* 466784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 467784b4e29SChris Mason * We must have no buffers or drop them. 468784b4e29SChris Mason */ 4698958b551SMatthew Wilcox (Oracle) if (folio_get_private(src) && 4708958b551SMatthew Wilcox (Oracle) !filemap_release_folio(src, GFP_KERNEL)) 471784b4e29SChris Mason return -EAGAIN; 47254184650SMatthew Wilcox (Oracle) return migrate_folio(mapping, dst, src, mode); 473784b4e29SChris Mason } 4748958b551SMatthew Wilcox (Oracle) #else 4758958b551SMatthew Wilcox (Oracle) #define btree_migrate_folio NULL 4763dd1462eSJan Beulich #endif 477784b4e29SChris Mason 4780da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 4790da5468fSChris Mason struct writeback_control *wbc) 4800da5468fSChris Mason { 481e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 482e2d84521SMiao Xie int ret; 483e2d84521SMiao Xie 484d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 485448d640bSChris Mason 486448d640bSChris Mason if (wbc->for_kupdate) 487448d640bSChris Mason return 0; 488448d640bSChris Mason 489e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 490b9473439SChris Mason /* this is a bit racy, but that's ok */ 491d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 492d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 493d814a491SEthan Lien fs_info->dirty_metadata_batch); 494e2d84521SMiao Xie if (ret < 0) 495793955bcSChris Mason return 0; 496793955bcSChris Mason } 4970b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 4980da5468fSChris Mason } 4990da5468fSChris Mason 500f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 5015f39d397SChris Mason { 502f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 503f913cff3SMatthew Wilcox (Oracle) return false; 5040c4e538bSDavid Sterba 505f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 506d98237b3SChris Mason } 507d98237b3SChris Mason 508895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 509895586ebSMatthew Wilcox (Oracle) size_t length) 510d98237b3SChris Mason { 511d1310b2eSChris Mason struct extent_io_tree *tree; 512895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 513895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 514f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 515895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 516895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 517895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 518895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 519895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 5209ad6b7bcSChris Mason } 521d98237b3SChris Mason } 522d98237b3SChris Mason 523bb146eb2SJosef Bacik #ifdef DEBUG 5240079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 5250079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 5260079c3b1SMatthew Wilcox (Oracle) { 5270079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 528139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 5290b32f4bbSJosef Bacik struct extent_buffer *eb; 530139e8cd3SQu Wenruo int cur_bit = 0; 5310079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 5320b32f4bbSJosef Bacik 533139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 5340079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 5350b32f4bbSJosef Bacik BUG_ON(!eb); 5360b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 5370b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 53849d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 5390079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 540139e8cd3SQu Wenruo } 5410079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 542139e8cd3SQu Wenruo 543139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 544139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 545139e8cd3SQu Wenruo unsigned long flags; 546139e8cd3SQu Wenruo u64 cur; 547139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 548139e8cd3SQu Wenruo 549139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 550139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 551139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 552139e8cd3SQu Wenruo cur_bit++; 553139e8cd3SQu Wenruo continue; 554139e8cd3SQu Wenruo } 555139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 556139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 557139e8cd3SQu Wenruo 558139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 559139e8cd3SQu Wenruo ASSERT(eb); 560139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 561139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 56249d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 563139e8cd3SQu Wenruo free_extent_buffer(eb); 564139e8cd3SQu Wenruo 565139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 566139e8cd3SQu Wenruo } 5670079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 5680b32f4bbSJosef Bacik } 5690079c3b1SMatthew Wilcox (Oracle) #else 5700079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 5710079c3b1SMatthew Wilcox (Oracle) #endif 5720b32f4bbSJosef Bacik 5737f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 5740da5468fSChris Mason .writepages = btree_writepages, 575f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 576895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 5778958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 5780079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 579d98237b3SChris Mason }; 580123abc88SChris Mason 5812ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 5822ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 5833fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 5843fbaf258SJosef Bacik int level) 5850999df54SChris Mason { 5860b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 5870b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 5883fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 5890999df54SChris Mason } 5900999df54SChris Mason 591581c1760SQu Wenruo /* 592581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 593581c1760SQu Wenruo * verification. 594581c1760SQu Wenruo * 595789d6a3aSQu Wenruo * @check: expected tree parentness check, see comments of the 596789d6a3aSQu Wenruo * structure for details. 597581c1760SQu Wenruo */ 5982ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 599789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 600e20d96d6SChris Mason { 6015f39d397SChris Mason struct extent_buffer *buf = NULL; 60219c00ddcSChris Mason int ret; 60319c00ddcSChris Mason 604789d6a3aSQu Wenruo ASSERT(check); 605789d6a3aSQu Wenruo 606789d6a3aSQu Wenruo buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, 607789d6a3aSQu Wenruo check->level); 608c871b0f2SLiu Bo if (IS_ERR(buf)) 609c871b0f2SLiu Bo return buf; 610e4204dedSChris Mason 611789d6a3aSQu Wenruo ret = btrfs_read_extent_buffer(buf, check); 6120f0fe8f7SFilipe David Borba Manana if (ret) { 613537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 61464c043deSLiu Bo return ERR_PTR(ret); 6150f0fe8f7SFilipe David Borba Manana } 616789d6a3aSQu Wenruo if (btrfs_check_eb_owner(buf, check->owner_root)) { 61788c602abSQu Wenruo free_extent_buffer_stale(buf); 61888c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 61988c602abSQu Wenruo } 6205f39d397SChris Mason return buf; 621ce9adaa5SChris Mason 622eb60ceacSChris Mason } 623eb60ceacSChris Mason 624da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 625e20d96d6SChris Mason u64 objectid) 626d97e63b6SChris Mason { 6277c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 6282e608bd1SJosef Bacik 6292e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 6302e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 6312e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 63296dfcb46SJosef Bacik root->fs_info = fs_info; 6332e608bd1SJosef Bacik root->root_key.objectid = objectid; 634cfaa7295SChris Mason root->node = NULL; 635a28ec197SChris Mason root->commit_root = NULL; 63627cdeb70SMiao Xie root->state = 0; 637abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 6380b86a832SChris Mason 6390f7d52f4SChris Mason root->last_trans = 0; 6406b8fad57SNikolay Borisov root->free_objectid = 0; 641eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 642199c2a9cSMiao Xie root->nr_ordered_extents = 0; 6436bef4d31SEric Paris root->inode_tree = RB_ROOT; 644088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 6452e608bd1SJosef Bacik 6462e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 6470b86a832SChris Mason 6480b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 6495d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 650eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 651eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 652199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 653199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 654d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 6552ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 6562ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 6575d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 658eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 659199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 660f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 6612ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 6622ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 6638287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 664a2135011SChris Mason mutex_init(&root->objectid_mutex); 665e02119d5SChris Mason mutex_init(&root->log_mutex); 66631f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 667573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 668c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 6697237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 6707237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 6717237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 6728b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 6738b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 6747237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 6757237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 6767237f183SYan Zheng atomic_set(&root->log_writers, 0); 6772ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 6780700cea7SElena Reshetova refcount_set(&root->refs, 1); 6798ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 680eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 6817237f183SYan Zheng root->log_transid = 0; 682d1433debSMiao Xie root->log_transid_committed = -1; 683257c62e1SChris Mason root->last_log_commit = 0; 6842e608bd1SJosef Bacik root->anon_dev = 0; 685e289f03eSFilipe Manana if (!dummy) { 68643eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 68735da5a7eSDavid Sterba IO_TREE_ROOT_DIRTY_LOG_PAGES); 688e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 68935da5a7eSDavid Sterba IO_TREE_LOG_CSUM_RANGE); 690e289f03eSFilipe Manana } 691017e5369SChris Mason 6925f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 693370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 694bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 695bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 696fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 697bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 698fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 699bd647ce3SJosef Bacik #endif 7003768f368SChris Mason } 7013768f368SChris Mason 70274e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 70396dfcb46SJosef Bacik u64 objectid, gfp_t flags) 7046f07e42eSAl Viro { 70574e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 7066f07e42eSAl Viro if (root) 70796dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 7086f07e42eSAl Viro return root; 7096f07e42eSAl Viro } 7106f07e42eSAl Viro 71106ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 71206ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 713da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 71406ea65a3SJosef Bacik { 71506ea65a3SJosef Bacik struct btrfs_root *root; 71606ea65a3SJosef Bacik 7177c0260eeSJeff Mahoney if (!fs_info) 7187c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 7197c0260eeSJeff Mahoney 72096dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 72106ea65a3SJosef Bacik if (!root) 72206ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 723da17066cSJeff Mahoney 724b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 725faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 72606ea65a3SJosef Bacik 72706ea65a3SJosef Bacik return root; 72806ea65a3SJosef Bacik } 72906ea65a3SJosef Bacik #endif 73006ea65a3SJosef Bacik 731abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 732abed4aaaSJosef Bacik { 733abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 734abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 735abed4aaaSJosef Bacik 736abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 737abed4aaaSJosef Bacik } 738abed4aaaSJosef Bacik 739abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 740abed4aaaSJosef Bacik { 741abed4aaaSJosef Bacik const struct btrfs_key *key = k; 742abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 743abed4aaaSJosef Bacik 744abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 745abed4aaaSJosef Bacik } 746abed4aaaSJosef Bacik 747abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 748abed4aaaSJosef Bacik { 749abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 750abed4aaaSJosef Bacik struct rb_node *tmp; 751745806fbSQu Wenruo int ret = 0; 752abed4aaaSJosef Bacik 753abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 754abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 755abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 756abed4aaaSJosef Bacik 757745806fbSQu Wenruo if (tmp) { 758745806fbSQu Wenruo ret = -EEXIST; 759745806fbSQu Wenruo btrfs_warn(fs_info, "global root %llu %llu already exists", 760745806fbSQu Wenruo root->root_key.objectid, root->root_key.offset); 761745806fbSQu Wenruo } 762745806fbSQu Wenruo return ret; 763abed4aaaSJosef Bacik } 764abed4aaaSJosef Bacik 765abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 766abed4aaaSJosef Bacik { 767abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 768abed4aaaSJosef Bacik 769abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 770abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 771abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 772abed4aaaSJosef Bacik } 773abed4aaaSJosef Bacik 774abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 775abed4aaaSJosef Bacik struct btrfs_key *key) 776abed4aaaSJosef Bacik { 777abed4aaaSJosef Bacik struct rb_node *node; 778abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 779abed4aaaSJosef Bacik 780abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 781abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 782abed4aaaSJosef Bacik if (node) 783abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 784abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 785abed4aaaSJosef Bacik 786abed4aaaSJosef Bacik return root; 787abed4aaaSJosef Bacik } 788abed4aaaSJosef Bacik 789f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 790f7238e50SJosef Bacik { 791f7238e50SJosef Bacik struct btrfs_block_group *block_group; 792f7238e50SJosef Bacik u64 ret; 793f7238e50SJosef Bacik 794f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 795f7238e50SJosef Bacik return 0; 796f7238e50SJosef Bacik 797f7238e50SJosef Bacik if (bytenr) 798f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 799f7238e50SJosef Bacik else 800f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 801f7238e50SJosef Bacik ASSERT(block_group); 802f7238e50SJosef Bacik if (!block_group) 803f7238e50SJosef Bacik return 0; 804f7238e50SJosef Bacik ret = block_group->global_root_id; 805f7238e50SJosef Bacik btrfs_put_block_group(block_group); 806f7238e50SJosef Bacik 807f7238e50SJosef Bacik return ret; 808f7238e50SJosef Bacik } 809f7238e50SJosef Bacik 810abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 811abed4aaaSJosef Bacik { 812abed4aaaSJosef Bacik struct btrfs_key key = { 813abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 814abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 815f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 816abed4aaaSJosef Bacik }; 817abed4aaaSJosef Bacik 818abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 819abed4aaaSJosef Bacik } 820abed4aaaSJosef Bacik 821abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 822abed4aaaSJosef Bacik { 823abed4aaaSJosef Bacik struct btrfs_key key = { 824abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 825abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 826f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 827abed4aaaSJosef Bacik }; 828abed4aaaSJosef Bacik 829abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 830abed4aaaSJosef Bacik } 831abed4aaaSJosef Bacik 83251129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 83351129b33SJosef Bacik { 83451129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 83551129b33SJosef Bacik return fs_info->block_group_root; 83651129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 83751129b33SJosef Bacik } 83851129b33SJosef Bacik 83920897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 84020897f5cSArne Jansen u64 objectid) 84120897f5cSArne Jansen { 8429b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 84320897f5cSArne Jansen struct extent_buffer *leaf; 84420897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 84520897f5cSArne Jansen struct btrfs_root *root; 84620897f5cSArne Jansen struct btrfs_key key; 847b89f6d1fSFilipe Manana unsigned int nofs_flag; 84820897f5cSArne Jansen int ret = 0; 84920897f5cSArne Jansen 850b89f6d1fSFilipe Manana /* 851b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 852b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 853b89f6d1fSFilipe Manana */ 854b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 85596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 856b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 85720897f5cSArne Jansen if (!root) 85820897f5cSArne Jansen return ERR_PTR(-ENOMEM); 85920897f5cSArne Jansen 86020897f5cSArne Jansen root->root_key.objectid = objectid; 86120897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 86220897f5cSArne Jansen root->root_key.offset = 0; 86320897f5cSArne Jansen 8649631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 8659631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 86620897f5cSArne Jansen if (IS_ERR(leaf)) { 86720897f5cSArne Jansen ret = PTR_ERR(leaf); 8681dd05682STsutomu Itoh leaf = NULL; 869c1b07854SPeng Hao goto fail; 87020897f5cSArne Jansen } 87120897f5cSArne Jansen 87220897f5cSArne Jansen root->node = leaf; 87320897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 87420897f5cSArne Jansen 87520897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 87627cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 87720897f5cSArne Jansen 878f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 879f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 88020897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 88120897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 88220897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 88320897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 88420897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 88520897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 88620897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 88733d85fdaSQu Wenruo if (is_fstree(objectid)) 888807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 889807fc790SAndy Shevchenko else 890807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 891c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 89220897f5cSArne Jansen 8938a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 8948a6a87cdSBoris Burkov 89520897f5cSArne Jansen key.objectid = objectid; 89620897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 89720897f5cSArne Jansen key.offset = 0; 89820897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 89920897f5cSArne Jansen if (ret) 90020897f5cSArne Jansen goto fail; 90120897f5cSArne Jansen 90220897f5cSArne Jansen return root; 9031dd05682STsutomu Itoh 9048a6a87cdSBoris Burkov fail: 90500246528SJosef Bacik btrfs_put_root(root); 9061dd05682STsutomu Itoh 9071dd05682STsutomu Itoh return ERR_PTR(ret); 90820897f5cSArne Jansen } 90920897f5cSArne Jansen 9107237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 911e02119d5SChris Mason struct btrfs_fs_info *fs_info) 9120f7d52f4SChris Mason { 9130f7d52f4SChris Mason struct btrfs_root *root; 914e02119d5SChris Mason 91596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 916e02119d5SChris Mason if (!root) 9177237f183SYan Zheng return ERR_PTR(-ENOMEM); 918e02119d5SChris Mason 919e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 920e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 921e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 92227cdeb70SMiao Xie 9236ab6ebb7SNaohiro Aota return root; 9246ab6ebb7SNaohiro Aota } 9256ab6ebb7SNaohiro Aota 9266ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 9276ab6ebb7SNaohiro Aota struct btrfs_root *root) 9286ab6ebb7SNaohiro Aota { 9296ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 9306ab6ebb7SNaohiro Aota 9317237f183SYan Zheng /* 93292a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 93327cdeb70SMiao Xie * 93492a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 93592a7cc42SQu Wenruo * and they go away before a real commit is actually done. 93692a7cc42SQu Wenruo * 93792a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 93892a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 9397237f183SYan Zheng */ 940e02119d5SChris Mason 9414d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 9429631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 9436ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 9446ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 945e02119d5SChris Mason 9467237f183SYan Zheng root->node = leaf; 947e02119d5SChris Mason 948e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 949e02119d5SChris Mason btrfs_tree_unlock(root->node); 9506ab6ebb7SNaohiro Aota 9516ab6ebb7SNaohiro Aota return 0; 9527237f183SYan Zheng } 9537237f183SYan Zheng 9547237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 9557237f183SYan Zheng struct btrfs_fs_info *fs_info) 9567237f183SYan Zheng { 9577237f183SYan Zheng struct btrfs_root *log_root; 9587237f183SYan Zheng 9597237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 9607237f183SYan Zheng if (IS_ERR(log_root)) 9617237f183SYan Zheng return PTR_ERR(log_root); 9626ab6ebb7SNaohiro Aota 9633ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 9643ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 9653ddebf27SNaohiro Aota 9666ab6ebb7SNaohiro Aota if (ret) { 9676ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 9686ab6ebb7SNaohiro Aota return ret; 9696ab6ebb7SNaohiro Aota } 9703ddebf27SNaohiro Aota } 9716ab6ebb7SNaohiro Aota 9727237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 9737237f183SYan Zheng fs_info->log_root_tree = log_root; 9747237f183SYan Zheng return 0; 9757237f183SYan Zheng } 9767237f183SYan Zheng 9777237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 9787237f183SYan Zheng struct btrfs_root *root) 9797237f183SYan Zheng { 9800b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 9817237f183SYan Zheng struct btrfs_root *log_root; 9827237f183SYan Zheng struct btrfs_inode_item *inode_item; 9836ab6ebb7SNaohiro Aota int ret; 9847237f183SYan Zheng 9850b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 9867237f183SYan Zheng if (IS_ERR(log_root)) 9877237f183SYan Zheng return PTR_ERR(log_root); 9887237f183SYan Zheng 9896ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 9906ab6ebb7SNaohiro Aota if (ret) { 9916ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 9926ab6ebb7SNaohiro Aota return ret; 9936ab6ebb7SNaohiro Aota } 9946ab6ebb7SNaohiro Aota 9957237f183SYan Zheng log_root->last_trans = trans->transid; 9967237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 9977237f183SYan Zheng 9987237f183SYan Zheng inode_item = &log_root->root_item.inode; 9993cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 10003cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 10013cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1002da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 10030b246afaSJeff Mahoney fs_info->nodesize); 10043cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 10057237f183SYan Zheng 10065d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 10077237f183SYan Zheng 10087237f183SYan Zheng WARN_ON(root->log_root); 10097237f183SYan Zheng root->log_root = log_root; 10107237f183SYan Zheng root->log_transid = 0; 1011d1433debSMiao Xie root->log_transid_committed = -1; 1012257c62e1SChris Mason root->last_log_commit = 0; 1013e02119d5SChris Mason return 0; 1014e02119d5SChris Mason } 1015e02119d5SChris Mason 101649d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 101749d11beaSJosef Bacik struct btrfs_path *path, 1018cb517eabSMiao Xie struct btrfs_key *key) 1019e02119d5SChris Mason { 1020e02119d5SChris Mason struct btrfs_root *root; 1021789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 1022e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 102384234f3aSYan Zheng u64 generation; 1024cb517eabSMiao Xie int ret; 1025581c1760SQu Wenruo int level; 1026cb517eabSMiao Xie 102796dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 102849d11beaSJosef Bacik if (!root) 102949d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 10300f7d52f4SChris Mason 1031cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1032cb517eabSMiao Xie &root->root_item, &root->root_key); 10330f7d52f4SChris Mason if (ret) { 103413a8a7c8SYan, Zheng if (ret > 0) 103513a8a7c8SYan, Zheng ret = -ENOENT; 103649d11beaSJosef Bacik goto fail; 10370f7d52f4SChris Mason } 103813a8a7c8SYan, Zheng 103984234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1040581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 1041789d6a3aSQu Wenruo check.level = level; 1042789d6a3aSQu Wenruo check.transid = generation; 1043789d6a3aSQu Wenruo check.owner_root = key->objectid; 1044789d6a3aSQu Wenruo root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), 1045789d6a3aSQu Wenruo &check); 104664c043deSLiu Bo if (IS_ERR(root->node)) { 104764c043deSLiu Bo ret = PTR_ERR(root->node); 10488c38938cSJosef Bacik root->node = NULL; 104949d11beaSJosef Bacik goto fail; 10504eb150d6SQu Wenruo } 10514eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1052cb517eabSMiao Xie ret = -EIO; 105349d11beaSJosef Bacik goto fail; 1054416bc658SJosef Bacik } 105588c602abSQu Wenruo 105688c602abSQu Wenruo /* 105788c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 105888c602abSQu Wenruo * match its root node owner 105988c602abSQu Wenruo */ 106088c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 106188c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 106288c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 106388c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 106488c602abSQu Wenruo btrfs_crit(fs_info, 106588c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 106688c602abSQu Wenruo root->root_key.objectid, root->node->start, 106788c602abSQu Wenruo btrfs_header_owner(root->node), 106888c602abSQu Wenruo root->root_key.objectid); 106988c602abSQu Wenruo ret = -EUCLEAN; 107088c602abSQu Wenruo goto fail; 107188c602abSQu Wenruo } 10725d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1073cb517eabSMiao Xie return root; 107449d11beaSJosef Bacik fail: 107500246528SJosef Bacik btrfs_put_root(root); 107649d11beaSJosef Bacik return ERR_PTR(ret); 107749d11beaSJosef Bacik } 107849d11beaSJosef Bacik 107949d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 108049d11beaSJosef Bacik struct btrfs_key *key) 108149d11beaSJosef Bacik { 108249d11beaSJosef Bacik struct btrfs_root *root; 108349d11beaSJosef Bacik struct btrfs_path *path; 108449d11beaSJosef Bacik 108549d11beaSJosef Bacik path = btrfs_alloc_path(); 108649d11beaSJosef Bacik if (!path) 108749d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 108849d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 108949d11beaSJosef Bacik btrfs_free_path(path); 109049d11beaSJosef Bacik 109149d11beaSJosef Bacik return root; 1092cb517eabSMiao Xie } 1093cb517eabSMiao Xie 10942dfb1e43SQu Wenruo /* 10952dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 10962dfb1e43SQu Wenruo * 10972dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 10982dfb1e43SQu Wenruo */ 10992dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1100cb517eabSMiao Xie { 1101cb517eabSMiao Xie int ret; 1102cb517eabSMiao Xie 11030b548539SDavid Sterba btrfs_drew_lock_init(&root->snapshot_lock); 11048257b2dcSMiao Xie 1105aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 11066ebcd021SQu Wenruo !btrfs_is_data_reloc_root(root) && 11076ebcd021SQu Wenruo is_fstree(root->root_key.objectid)) { 110892a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1109f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1110f39e4571SJosef Bacik } 1111f39e4571SJosef Bacik 1112851fd730SQu Wenruo /* 1113851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1114851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1115851fd730SQu Wenruo */ 1116851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1117851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 11182dfb1e43SQu Wenruo if (!anon_dev) { 1119cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1120cb517eabSMiao Xie if (ret) 1121876d2cf1SLiu Bo goto fail; 11222dfb1e43SQu Wenruo } else { 11232dfb1e43SQu Wenruo root->anon_dev = anon_dev; 11242dfb1e43SQu Wenruo } 1125851fd730SQu Wenruo } 1126f32e48e9SChandan Rajendra 1127f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1128453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1129f32e48e9SChandan Rajendra if (ret) { 1130f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1131876d2cf1SLiu Bo goto fail; 1132f32e48e9SChandan Rajendra } 1133f32e48e9SChandan Rajendra 11346b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1135f32e48e9SChandan Rajendra 1136f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1137f32e48e9SChandan Rajendra 1138cb517eabSMiao Xie return 0; 1139cb517eabSMiao Xie fail: 114084db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1141cb517eabSMiao Xie return ret; 1142cb517eabSMiao Xie } 1143cb517eabSMiao Xie 1144a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1145cb517eabSMiao Xie u64 root_id) 1146cb517eabSMiao Xie { 1147cb517eabSMiao Xie struct btrfs_root *root; 1148cb517eabSMiao Xie 1149fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1150fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1151fc7cbcd4SDavid Sterba (unsigned long)root_id); 115200246528SJosef Bacik root = btrfs_grab_root(root); 1153fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1154cb517eabSMiao Xie return root; 1155cb517eabSMiao Xie } 1156cb517eabSMiao Xie 115749d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 115849d11beaSJosef Bacik u64 objectid) 115949d11beaSJosef Bacik { 1160abed4aaaSJosef Bacik struct btrfs_key key = { 1161abed4aaaSJosef Bacik .objectid = objectid, 1162abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1163abed4aaaSJosef Bacik .offset = 0, 1164abed4aaaSJosef Bacik }; 1165abed4aaaSJosef Bacik 1166e91909aaSChristoph Hellwig switch (objectid) { 1167e91909aaSChristoph Hellwig case BTRFS_ROOT_TREE_OBJECTID: 116849d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 1169e91909aaSChristoph Hellwig case BTRFS_EXTENT_TREE_OBJECTID: 1170abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1171e91909aaSChristoph Hellwig case BTRFS_CHUNK_TREE_OBJECTID: 117249d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 1173e91909aaSChristoph Hellwig case BTRFS_DEV_TREE_OBJECTID: 117449d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 1175e91909aaSChristoph Hellwig case BTRFS_CSUM_TREE_OBJECTID: 1176abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1177e91909aaSChristoph Hellwig case BTRFS_QUOTA_TREE_OBJECTID: 117885724171SChristoph Hellwig return btrfs_grab_root(fs_info->quota_root); 1179e91909aaSChristoph Hellwig case BTRFS_UUID_TREE_OBJECTID: 118085724171SChristoph Hellwig return btrfs_grab_root(fs_info->uuid_root); 1181e91909aaSChristoph Hellwig case BTRFS_BLOCK_GROUP_TREE_OBJECTID: 118285724171SChristoph Hellwig return btrfs_grab_root(fs_info->block_group_root); 1183e91909aaSChristoph Hellwig case BTRFS_FREE_SPACE_TREE_OBJECTID: 118485724171SChristoph Hellwig return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1185e91909aaSChristoph Hellwig default: 118649d11beaSJosef Bacik return NULL; 118749d11beaSJosef Bacik } 1188e91909aaSChristoph Hellwig } 118949d11beaSJosef Bacik 1190cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1191cb517eabSMiao Xie struct btrfs_root *root) 1192cb517eabSMiao Xie { 1193cb517eabSMiao Xie int ret; 1194cb517eabSMiao Xie 1195fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1196fc7cbcd4SDavid Sterba if (ret) 1197fc7cbcd4SDavid Sterba return ret; 1198fc7cbcd4SDavid Sterba 1199fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1200fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1201fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1202fc7cbcd4SDavid Sterba root); 1203af01d2e5SJosef Bacik if (ret == 0) { 120400246528SJosef Bacik btrfs_grab_root(root); 1205fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1206af01d2e5SJosef Bacik } 1207fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1208fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1209cb517eabSMiao Xie 1210cb517eabSMiao Xie return ret; 1211cb517eabSMiao Xie } 1212cb517eabSMiao Xie 1213bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1214bd647ce3SJosef Bacik { 1215bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1216bd647ce3SJosef Bacik struct btrfs_root *root; 1217bd647ce3SJosef Bacik 1218bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1219457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1220457f1864SJosef Bacik 1221bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1222bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1223457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 122471008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1225bd647ce3SJosef Bacik refcount_read(&root->refs)); 1226bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 122700246528SJosef Bacik btrfs_put_root(root); 122800246528SJosef Bacik btrfs_put_root(root); 1229bd647ce3SJosef Bacik } 1230bd647ce3SJosef Bacik #endif 1231bd647ce3SJosef Bacik } 1232bd647ce3SJosef Bacik 1233abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1234abed4aaaSJosef Bacik { 1235abed4aaaSJosef Bacik struct btrfs_root *root; 1236abed4aaaSJosef Bacik struct rb_node *node; 1237abed4aaaSJosef Bacik 1238abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1239abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1240abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1241abed4aaaSJosef Bacik btrfs_put_root(root); 1242abed4aaaSJosef Bacik } 1243abed4aaaSJosef Bacik } 1244abed4aaaSJosef Bacik 12450d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 12460d4b0463SJosef Bacik { 1247141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1248141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 12495deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1250141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1251141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1252141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1253141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 12540d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 12550d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1256abed4aaaSJosef Bacik free_global_roots(fs_info); 125700246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 125800246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 125900246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 126000246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 126100246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 126200246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1263aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 12649c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1265bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 12663fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 12670d4b0463SJosef Bacik kfree(fs_info->super_copy); 12680d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 12698481dd80SQu Wenruo kfree(fs_info->subpage_info); 12700d4b0463SJosef Bacik kvfree(fs_info); 12710d4b0463SJosef Bacik } 12720d4b0463SJosef Bacik 12730d4b0463SJosef Bacik 12742dfb1e43SQu Wenruo /* 12752dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 12762dfb1e43SQu Wenruo * 12772dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 12782dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 12792dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 12802dfb1e43SQu Wenruo * 12812dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 12822dfb1e43SQu Wenruo * 12832dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 12842dfb1e43SQu Wenruo * same root objectid. 12852dfb1e43SQu Wenruo * 12862dfb1e43SQu Wenruo * @objectid: root id 12872dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 12882dfb1e43SQu Wenruo * pass 0 for new allocation. 12892dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 12902dfb1e43SQu Wenruo * for orphan roots 12912dfb1e43SQu Wenruo */ 12922dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 12932dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 12942dfb1e43SQu Wenruo bool check_ref) 12955eda7b5eSChris Mason { 12965eda7b5eSChris Mason struct btrfs_root *root; 1297381cf658SDavid Sterba struct btrfs_path *path; 12981d4c08e0SDavid Sterba struct btrfs_key key; 12995eda7b5eSChris Mason int ret; 13005eda7b5eSChris Mason 130149d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 130249d11beaSJosef Bacik if (root) 130349d11beaSJosef Bacik return root; 1304773e722aSQu Wenruo 1305773e722aSQu Wenruo /* 1306773e722aSQu Wenruo * If we're called for non-subvolume trees, and above function didn't 1307773e722aSQu Wenruo * find one, do not try to read it from disk. 1308773e722aSQu Wenruo * 1309773e722aSQu Wenruo * This is namely for free-space-tree and quota tree, which can change 1310773e722aSQu Wenruo * at runtime and should only be grabbed from fs_info. 1311773e722aSQu Wenruo */ 1312773e722aSQu Wenruo if (!is_fstree(objectid) && objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) 1313773e722aSQu Wenruo return ERR_PTR(-ENOENT); 13144df27c4dSYan, Zheng again: 131556e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 131648475471SStefan Behrens if (root) { 13172dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 13182dfb1e43SQu Wenruo ASSERT(!anon_dev); 1319bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 132000246528SJosef Bacik btrfs_put_root(root); 132148475471SStefan Behrens return ERR_PTR(-ENOENT); 1322bc44d7c4SJosef Bacik } 13235eda7b5eSChris Mason return root; 132448475471SStefan Behrens } 13255eda7b5eSChris Mason 132656e9357aSDavid Sterba key.objectid = objectid; 132756e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 132856e9357aSDavid Sterba key.offset = (u64)-1; 132956e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 13305eda7b5eSChris Mason if (IS_ERR(root)) 13315eda7b5eSChris Mason return root; 13323394e160SChris Mason 1333c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1334d68fc57bSYan, Zheng ret = -ENOENT; 1335d68fc57bSYan, Zheng goto fail; 1336d68fc57bSYan, Zheng } 1337d68fc57bSYan, Zheng 13382dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1339cb517eabSMiao Xie if (ret) 1340cb517eabSMiao Xie goto fail; 1341cb517eabSMiao Xie 1342381cf658SDavid Sterba path = btrfs_alloc_path(); 1343381cf658SDavid Sterba if (!path) { 1344381cf658SDavid Sterba ret = -ENOMEM; 1345381cf658SDavid Sterba goto fail; 1346381cf658SDavid Sterba } 13471d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 13481d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 134956e9357aSDavid Sterba key.offset = objectid; 13501d4c08e0SDavid Sterba 13511d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1352381cf658SDavid Sterba btrfs_free_path(path); 1353d68fc57bSYan, Zheng if (ret < 0) 1354d68fc57bSYan, Zheng goto fail; 1355d68fc57bSYan, Zheng if (ret == 0) 135627cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1357d68fc57bSYan, Zheng 1358cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 13590f7d52f4SChris Mason if (ret) { 1360168a2f77SJia-Ju Bai if (ret == -EEXIST) { 136100246528SJosef Bacik btrfs_put_root(root); 13624df27c4dSYan, Zheng goto again; 1363168a2f77SJia-Ju Bai } 13644df27c4dSYan, Zheng goto fail; 13654df27c4dSYan, Zheng } 1366edbd8d4eSChris Mason return root; 13674df27c4dSYan, Zheng fail: 136833fab972SFilipe Manana /* 136933fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1370143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 137133fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 137233fab972SFilipe Manana * and once again by our caller. 137333fab972SFilipe Manana */ 137433fab972SFilipe Manana if (anon_dev) 137533fab972SFilipe Manana root->anon_dev = 0; 13768c38938cSJosef Bacik btrfs_put_root(root); 13774df27c4dSYan, Zheng return ERR_PTR(ret); 1378edbd8d4eSChris Mason } 1379edbd8d4eSChris Mason 13802dfb1e43SQu Wenruo /* 13812dfb1e43SQu Wenruo * Get in-memory reference of a root structure 13822dfb1e43SQu Wenruo * 13832dfb1e43SQu Wenruo * @objectid: tree objectid 13842dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 13852dfb1e43SQu Wenruo * one reference 13862dfb1e43SQu Wenruo */ 13872dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 13882dfb1e43SQu Wenruo u64 objectid, bool check_ref) 13892dfb1e43SQu Wenruo { 13902dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 13912dfb1e43SQu Wenruo } 13922dfb1e43SQu Wenruo 13932dfb1e43SQu Wenruo /* 13942dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 13952dfb1e43SQu Wenruo * the anonymous block device id 13962dfb1e43SQu Wenruo * 13972dfb1e43SQu Wenruo * @objectid: tree objectid 13982dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 13992dfb1e43SQu Wenruo * parameter value 14002dfb1e43SQu Wenruo */ 14012dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 14022dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 14032dfb1e43SQu Wenruo { 14042dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 14052dfb1e43SQu Wenruo } 14062dfb1e43SQu Wenruo 14078b712842SChris Mason /* 140849d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 140949d11beaSJosef Bacik * @fs_info: the fs_info 141049d11beaSJosef Bacik * @objectid: the objectid we need to lookup 141149d11beaSJosef Bacik * 141249d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 141349d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 141449d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 141549d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 141649d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 141749d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 141849d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 141949d11beaSJosef Bacik * backref code is finished using the root. 142049d11beaSJosef Bacik */ 142149d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 142249d11beaSJosef Bacik struct btrfs_path *path, 142349d11beaSJosef Bacik u64 objectid) 142449d11beaSJosef Bacik { 142549d11beaSJosef Bacik struct btrfs_root *root; 142649d11beaSJosef Bacik struct btrfs_key key; 142749d11beaSJosef Bacik 142849d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 142949d11beaSJosef Bacik 143049d11beaSJosef Bacik /* 143149d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 143249d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 143349d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 143449d11beaSJosef Bacik * != NULL just return it. 143549d11beaSJosef Bacik */ 143649d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 143749d11beaSJosef Bacik if (root) 143849d11beaSJosef Bacik return root; 143949d11beaSJosef Bacik 144049d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 144149d11beaSJosef Bacik if (root) 144249d11beaSJosef Bacik return root; 144349d11beaSJosef Bacik 144449d11beaSJosef Bacik key.objectid = objectid; 144549d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 144649d11beaSJosef Bacik key.offset = (u64)-1; 144749d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 144849d11beaSJosef Bacik btrfs_release_path(path); 144949d11beaSJosef Bacik 145049d11beaSJosef Bacik return root; 145149d11beaSJosef Bacik } 145249d11beaSJosef Bacik 1453a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1454a74a4b97SChris Mason { 14550d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1456d0278245SMiao Xie int again; 1457a74a4b97SChris Mason 1458d6fd0ae2SOmar Sandoval while (1) { 1459d0278245SMiao Xie again = 0; 14609d1a2a3aSDavid Sterba 1461fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1462fd340d0fSJosef Bacik 1463d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 14642ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1465d0278245SMiao Xie goto sleep; 1466d0278245SMiao Xie 146790c711abSZygo Blaxell /* 146890c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 146990c711abSZygo Blaxell * before we have finished mounting the filesystem. 147090c711abSZygo Blaxell */ 14710b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 147290c711abSZygo Blaxell goto sleep; 147390c711abSZygo Blaxell 14740b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1475d0278245SMiao Xie goto sleep; 1476d0278245SMiao Xie 1477dc7f370cSMiao Xie /* 1478dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1479dc7f370cSMiao Xie * during the above check and trylock. 1480dc7f370cSMiao Xie */ 14812ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 14820b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1483dc7f370cSMiao Xie goto sleep; 1484dc7f370cSMiao Xie } 1485dc7f370cSMiao Xie 1486b7625f46SQu Wenruo if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) 1487b7625f46SQu Wenruo btrfs_sysfs_feature_update(fs_info); 1488b7625f46SQu Wenruo 14892ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1490c2d6cb16SFilipe Manana 149133c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 14920b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1493a74a4b97SChris Mason 1494d0278245SMiao Xie /* 149505323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 149605323cd1SMiao Xie * needn't do anything special here. 1497d0278245SMiao Xie */ 14980b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 149967c5e7d4SFilipe Manana 150067c5e7d4SFilipe Manana /* 1501f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 150267c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 150367c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1504f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 150567c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 150667c5e7d4SFilipe Manana * unused block groups. 150767c5e7d4SFilipe Manana */ 15080b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 150918bb8bbfSJohannes Thumshirn 151018bb8bbfSJohannes Thumshirn /* 151118bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 151218bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 151318bb8bbfSJohannes Thumshirn * space. 151418bb8bbfSJohannes Thumshirn */ 151518bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1516d0278245SMiao Xie sleep: 1517a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1518d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1519d6fd0ae2SOmar Sandoval kthread_parkme(); 1520d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1521d6fd0ae2SOmar Sandoval return 0; 1522838fe188SJiri Kosina if (!again) { 1523a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1524a74a4b97SChris Mason schedule(); 1525a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1526a74a4b97SChris Mason } 1527da288d28SFilipe Manana } 1528a74a4b97SChris Mason } 1529a74a4b97SChris Mason 1530a74a4b97SChris Mason static int transaction_kthread(void *arg) 1531a74a4b97SChris Mason { 1532a74a4b97SChris Mason struct btrfs_root *root = arg; 15330b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1534a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1535a74a4b97SChris Mason struct btrfs_transaction *cur; 15368929ecfaSYan, Zheng u64 transid; 1537643900beSNikolay Borisov time64_t delta; 1538a74a4b97SChris Mason unsigned long delay; 1539914b2007SJan Kara bool cannot_commit; 1540a74a4b97SChris Mason 1541a74a4b97SChris Mason do { 1542914b2007SJan Kara cannot_commit = false; 1543ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 15440b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1545a74a4b97SChris Mason 15460b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 15470b246afaSJeff Mahoney cur = fs_info->running_transaction; 1548a74a4b97SChris Mason if (!cur) { 15490b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1550a74a4b97SChris Mason goto sleep; 1551a74a4b97SChris Mason } 155231153d81SYan Zheng 1553643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1554fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1555fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1556643900beSNikolay Borisov delta < fs_info->commit_interval) { 15570b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1558fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1559fb8a7e94SNikolay Borisov delay = min(delay, 1560fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1561a74a4b97SChris Mason goto sleep; 1562a74a4b97SChris Mason } 15638929ecfaSYan, Zheng transid = cur->transid; 15640b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 156556bec294SChris Mason 156679787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1567354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1568914b2007SJan Kara if (IS_ERR(trans)) { 1569354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1570914b2007SJan Kara cannot_commit = true; 157179787eaaSJeff Mahoney goto sleep; 1572914b2007SJan Kara } 15738929ecfaSYan, Zheng if (transid == trans->transid) { 15743a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 15758929ecfaSYan, Zheng } else { 15763a45bb20SJeff Mahoney btrfs_end_transaction(trans); 15778929ecfaSYan, Zheng } 1578a74a4b97SChris Mason sleep: 15790b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 15800b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1581a74a4b97SChris Mason 158284961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 15832ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 15848929ecfaSYan, Zheng if (!kthread_should_stop() && 15850b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1586914b2007SJan Kara cannot_commit)) 1587bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1588a74a4b97SChris Mason } while (!kthread_should_stop()); 1589a74a4b97SChris Mason return 0; 1590a74a4b97SChris Mason } 1591a74a4b97SChris Mason 1592af31f5e5SChris Mason /* 159301f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 159401f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 159501f0f9daSNikolay Borisov * anything. 1596af31f5e5SChris Mason * 1597af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1598af31f5e5SChris Mason * generation of the latest root in the array with the generation 1599af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1600af31f5e5SChris Mason */ 160101f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1602af31f5e5SChris Mason { 160301f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1604af31f5e5SChris Mason u64 cur; 1605af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1606af31f5e5SChris Mason int i; 1607af31f5e5SChris Mason 1608af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1609af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1610af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1611af31f5e5SChris Mason if (cur == newest_gen) 161201f0f9daSNikolay Borisov return i; 1613af31f5e5SChris Mason } 1614af31f5e5SChris Mason 161501f0f9daSNikolay Borisov return -EINVAL; 1616af31f5e5SChris Mason } 1617af31f5e5SChris Mason 1618af31f5e5SChris Mason /* 1619af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1620af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1621af31f5e5SChris Mason * done 1622af31f5e5SChris Mason */ 1623af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1624af31f5e5SChris Mason { 16256ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1626af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1627af31f5e5SChris Mason 1628af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1629af31f5e5SChris Mason 1630af31f5e5SChris Mason /* 1631af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1632af31f5e5SChris Mason * regardless of which ones we use today 1633af31f5e5SChris Mason */ 1634af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1635af31f5e5SChris Mason 1636af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1637af31f5e5SChris Mason 1638af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1639af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1640af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1641af31f5e5SChris Mason 1642af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1643af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1644af31f5e5SChris Mason 1645af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1646af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1647af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1648af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1649af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1650af31f5e5SChris Mason 16511c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 16529c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1653f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 16549c54e80dSJosef Bacik 16559c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 16569c54e80dSJosef Bacik extent_root->node->start); 1657af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 165829cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 1659af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 166029cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 1661af31f5e5SChris Mason 1662f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 1663f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 1664f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 1665f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 1666f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 16679c54e80dSJosef Bacik } 1668af31f5e5SChris Mason 16697c7e82a7SChris Mason /* 16707c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 16717c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 16727c7e82a7SChris Mason */ 16737c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 16747c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 16757c7e82a7SChris Mason info->fs_root->node->start); 1676af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1677af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1678af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1679af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 16807c7e82a7SChris Mason } 1681af31f5e5SChris Mason 1682af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1683af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1684af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1685af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1686af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1687af31f5e5SChris Mason 1688af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1689af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1690af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1691af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1692af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1693af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1694af31f5e5SChris Mason 1695af31f5e5SChris Mason /* 1696af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1697af31f5e5SChris Mason * for the next commit 1698af31f5e5SChris Mason */ 1699af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1700af31f5e5SChris Mason &info->super_for_commit->super_roots, 1701af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1702af31f5e5SChris Mason } 1703af31f5e5SChris Mason 1704af31f5e5SChris Mason /* 1705bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1706bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1707bd2336b2SNikolay Borisov * 1708bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1709bd2336b2SNikolay Borisov * priority - priority of backup root required 1710bd2336b2SNikolay Borisov * 1711bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1712bd2336b2SNikolay Borisov */ 1713bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1714bd2336b2SNikolay Borisov { 1715bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1716bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1717bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1718bd2336b2SNikolay Borisov 1719bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1720bd2336b2SNikolay Borisov if (priority == 0) 1721bd2336b2SNikolay Borisov return backup_index; 1722bd2336b2SNikolay Borisov 1723bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1724bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1725bd2336b2SNikolay Borisov } else { 1726bd2336b2SNikolay Borisov return -EINVAL; 1727bd2336b2SNikolay Borisov } 1728bd2336b2SNikolay Borisov 1729bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1730bd2336b2SNikolay Borisov 1731bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1732bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1733bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1734bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1735bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1736bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1737bd2336b2SNikolay Borisov 1738bd2336b2SNikolay Borisov /* 1739bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1740bd2336b2SNikolay Borisov * need a fsck 1741bd2336b2SNikolay Borisov */ 1742bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1743bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1744bd2336b2SNikolay Borisov 1745bd2336b2SNikolay Borisov return backup_index; 1746bd2336b2SNikolay Borisov } 1747bd2336b2SNikolay Borisov 17487abadb64SLiu Bo /* helper to cleanup workers */ 17497abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 17507abadb64SLiu Bo { 1751dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1752afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 17535cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1754d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 1755d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 1756385de0efSChristoph Hellwig if (fs_info->rmw_workers) 1757385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 1758fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 1759fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 1760fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1761fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 17625b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1763e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1764a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1765fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1766b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1767b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1768a9b9477dSFilipe Manana /* 1769a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 1770a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 1771a9b9477dSFilipe Manana * queues can do metadata I/O operations. 1772a9b9477dSFilipe Manana */ 1773d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 1774d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 17757abadb64SLiu Bo } 17767abadb64SLiu Bo 17772e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 17782e9f5954SRashika { 17792e9f5954SRashika if (root) { 17802e9f5954SRashika free_extent_buffer(root->node); 17812e9f5954SRashika free_extent_buffer(root->commit_root); 17822e9f5954SRashika root->node = NULL; 17832e9f5954SRashika root->commit_root = NULL; 17842e9f5954SRashika } 17852e9f5954SRashika } 17862e9f5954SRashika 1787abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 1788abed4aaaSJosef Bacik { 1789abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 1790abed4aaaSJosef Bacik 1791abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 1792abed4aaaSJosef Bacik &fs_info->global_root_tree, 1793abed4aaaSJosef Bacik rb_node) 1794abed4aaaSJosef Bacik free_root_extent_buffers(root); 1795abed4aaaSJosef Bacik } 1796abed4aaaSJosef Bacik 1797af31f5e5SChris Mason /* helper to cleanup tree roots */ 17984273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 1799af31f5e5SChris Mason { 18002e9f5954SRashika free_root_extent_buffers(info->tree_root); 1801655b09feSJosef Bacik 1802abed4aaaSJosef Bacik free_global_root_pointers(info); 18032e9f5954SRashika free_root_extent_buffers(info->dev_root); 18042e9f5954SRashika free_root_extent_buffers(info->quota_root); 18052e9f5954SRashika free_root_extent_buffers(info->uuid_root); 18068c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 1807aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 18089c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 18094273eaffSAnand Jain if (free_chunk_root) 18102e9f5954SRashika free_root_extent_buffers(info->chunk_root); 1811af31f5e5SChris Mason } 1812af31f5e5SChris Mason 18138c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 18148c38938cSJosef Bacik { 18158c38938cSJosef Bacik if (!root) 18168c38938cSJosef Bacik return; 18178c38938cSJosef Bacik 18188c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 18198c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 18201dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 18218c38938cSJosef Bacik if (root->anon_dev) 18228c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 1823923eb523SJohannes Thumshirn free_root_extent_buffers(root); 18248c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1825fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 18268c38938cSJosef Bacik list_del_init(&root->leak_list); 1827fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 18288c38938cSJosef Bacik #endif 18298c38938cSJosef Bacik kfree(root); 18308c38938cSJosef Bacik } 18318c38938cSJosef Bacik } 18328c38938cSJosef Bacik 1833faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 1834171f6537SJosef Bacik { 1835fc7cbcd4SDavid Sterba int ret; 1836fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 1837fc7cbcd4SDavid Sterba int i; 1838171f6537SJosef Bacik 1839171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 1840fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 1841171f6537SJosef Bacik struct btrfs_root, root_list); 1842fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 1843171f6537SJosef Bacik 1844fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 1845fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 1846fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 1847171f6537SJosef Bacik } 1848171f6537SJosef Bacik 1849fc7cbcd4SDavid Sterba while (1) { 1850fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 1851fc7cbcd4SDavid Sterba (void **)gang, 0, 1852fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 1853fc7cbcd4SDavid Sterba if (!ret) 1854fc7cbcd4SDavid Sterba break; 1855fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 1856fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 1857171f6537SJosef Bacik } 1858171f6537SJosef Bacik } 1859af31f5e5SChris Mason 1860638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 1861638aa7edSEric Sandeen { 1862638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 1863638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 1864638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 1865638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 1866638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 1867638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 1868ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 1869638aa7edSEric Sandeen } 1870638aa7edSEric Sandeen 1871779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 1872779a65a4SEric Sandeen { 1873779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 1874779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 1875779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 1876779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 1877779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 1878779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 1879907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 1880779a65a4SEric Sandeen } 1881779a65a4SEric Sandeen 1882dcb2137cSChristoph Hellwig static int btrfs_init_btree_inode(struct super_block *sb) 1883f37938e0SEric Sandeen { 1884dcb2137cSChristoph Hellwig struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1885e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 1886e256927bSJosef Bacik fs_info->tree_root); 1887dcb2137cSChristoph Hellwig struct inode *inode; 1888dcb2137cSChristoph Hellwig 1889dcb2137cSChristoph Hellwig inode = new_inode(sb); 1890dcb2137cSChristoph Hellwig if (!inode) 1891dcb2137cSChristoph Hellwig return -ENOMEM; 18922ff7e61eSJeff Mahoney 18932ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 18942ff7e61eSJeff Mahoney set_nlink(inode, 1); 1895f37938e0SEric Sandeen /* 1896f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 1897f37938e0SEric Sandeen * the real end of the address space is determined by all of 1898f37938e0SEric Sandeen * the devices in the system 1899f37938e0SEric Sandeen */ 19002ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 19012ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 1902dcb2137cSChristoph Hellwig mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); 1903f37938e0SEric Sandeen 19042ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 190543eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 190635da5a7eSDavid Sterba IO_TREE_BTREE_INODE_IO); 19072ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 1908f37938e0SEric Sandeen 19095c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 1910adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 1911adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 1912adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 19132ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 1914e256927bSJosef Bacik __insert_inode_hash(inode, hash); 1915dcb2137cSChristoph Hellwig fs_info->btree_inode = inode; 1916dcb2137cSChristoph Hellwig 1917dcb2137cSChristoph Hellwig return 0; 1918f37938e0SEric Sandeen } 1919f37938e0SEric Sandeen 1920ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 1921ad618368SEric Sandeen { 1922ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 1923129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 19247f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 1925ad618368SEric Sandeen } 1926ad618368SEric Sandeen 1927f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 1928f9e92e40SEric Sandeen { 1929f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 1930f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 1931f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 1932f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 1933f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 1934f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 1935d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 1936011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 1937f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 1938f9e92e40SEric Sandeen } 1939f9e92e40SEric Sandeen 1940d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 19412a458198SEric Sandeen { 1942f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 19436f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 194458e814fcSTejun Heo unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; 19452a458198SEric Sandeen 19462a458198SEric Sandeen fs_info->workers = 1947a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 19482a458198SEric Sandeen 19492a458198SEric Sandeen fs_info->delalloc_workers = 1950cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 1951cb001095SJeff Mahoney flags, max_active, 2); 19522a458198SEric Sandeen 19532a458198SEric Sandeen fs_info->flush_workers = 1954cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 1955cb001095SJeff Mahoney flags, max_active, 0); 19562a458198SEric Sandeen 19572a458198SEric Sandeen fs_info->caching_workers = 1958cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 19592a458198SEric Sandeen 19602a458198SEric Sandeen fs_info->fixup_workers = 196158e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); 19622a458198SEric Sandeen 19632a458198SEric Sandeen fs_info->endio_workers = 1964d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 19652a458198SEric Sandeen fs_info->endio_meta_workers = 1966d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 1967385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 19682a458198SEric Sandeen fs_info->endio_write_workers = 1969cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 1970cb001095SJeff Mahoney max_active, 2); 1971fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 1972fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 19732a458198SEric Sandeen fs_info->endio_freespace_worker = 1974cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 1975cb001095SJeff Mahoney max_active, 0); 19762a458198SEric Sandeen fs_info->delayed_workers = 1977cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 1978cb001095SJeff Mahoney max_active, 0); 19792a458198SEric Sandeen fs_info->qgroup_rescan_workers = 198058e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", 198158e814fcSTejun Heo ordered_flags); 1982b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 198358e814fcSTejun Heo alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); 19842a458198SEric Sandeen 19858bfec2e4SChristoph Hellwig if (!(fs_info->workers && 1986a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 19872a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 1988fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 19891a1a2851SQu Wenruo fs_info->endio_write_workers && 19902a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 1991f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 1992f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 1993b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 19942a458198SEric Sandeen return -ENOMEM; 19952a458198SEric Sandeen } 19962a458198SEric Sandeen 19972a458198SEric Sandeen return 0; 19982a458198SEric Sandeen } 19992a458198SEric Sandeen 20006d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 20016d97c6e3SJohannes Thumshirn { 20026d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2003b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 20046d97c6e3SJohannes Thumshirn 2005b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 20066d97c6e3SJohannes Thumshirn 20076d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 20086d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2009b4e967beSDavid Sterba csum_driver); 20106d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 20116d97c6e3SJohannes Thumshirn } 20126d97c6e3SJohannes Thumshirn 20136d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 20146d97c6e3SJohannes Thumshirn 201568d99ab0SChristoph Hellwig /* 201668d99ab0SChristoph Hellwig * Check if the checksum implementation is a fast accelerated one. 201768d99ab0SChristoph Hellwig * As-is this is a bit of a hack and should be replaced once the csum 201868d99ab0SChristoph Hellwig * implementations provide that information themselves. 201968d99ab0SChristoph Hellwig */ 202068d99ab0SChristoph Hellwig switch (csum_type) { 202168d99ab0SChristoph Hellwig case BTRFS_CSUM_TYPE_CRC32: 202268d99ab0SChristoph Hellwig if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) 202368d99ab0SChristoph Hellwig set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 202468d99ab0SChristoph Hellwig break; 2025efcfcbc6SDavid Sterba case BTRFS_CSUM_TYPE_XXHASH: 2026efcfcbc6SDavid Sterba set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 2027efcfcbc6SDavid Sterba break; 202868d99ab0SChristoph Hellwig default: 202968d99ab0SChristoph Hellwig break; 203068d99ab0SChristoph Hellwig } 203168d99ab0SChristoph Hellwig 2032c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2033c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2034c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 20356d97c6e3SJohannes Thumshirn return 0; 20366d97c6e3SJohannes Thumshirn } 20376d97c6e3SJohannes Thumshirn 203863443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 203963443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 204063443bf5SEric Sandeen { 204163443bf5SEric Sandeen int ret; 2042789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 204363443bf5SEric Sandeen struct btrfs_root *log_tree_root; 204463443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 204563443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2046581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 204763443bf5SEric Sandeen 204863443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2049f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 205063443bf5SEric Sandeen return -EIO; 205163443bf5SEric Sandeen } 205263443bf5SEric Sandeen 205396dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 205496dfcb46SJosef Bacik GFP_KERNEL); 205563443bf5SEric Sandeen if (!log_tree_root) 205663443bf5SEric Sandeen return -ENOMEM; 205763443bf5SEric Sandeen 2058789d6a3aSQu Wenruo check.level = level; 2059789d6a3aSQu Wenruo check.transid = fs_info->generation + 1; 2060789d6a3aSQu Wenruo check.owner_root = BTRFS_TREE_LOG_OBJECTID; 2061789d6a3aSQu Wenruo log_tree_root->node = read_tree_block(fs_info, bytenr, &check); 206264c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2063f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 20640eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 20658c38938cSJosef Bacik log_tree_root->node = NULL; 206600246528SJosef Bacik btrfs_put_root(log_tree_root); 20670eeff236SLiu Bo return ret; 20684eb150d6SQu Wenruo } 20694eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2070f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 207100246528SJosef Bacik btrfs_put_root(log_tree_root); 207263443bf5SEric Sandeen return -EIO; 207363443bf5SEric Sandeen } 20744eb150d6SQu Wenruo 207563443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 207663443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 207763443bf5SEric Sandeen if (ret) { 20780b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 207963443bf5SEric Sandeen "Failed to recover log tree"); 208000246528SJosef Bacik btrfs_put_root(log_tree_root); 208163443bf5SEric Sandeen return ret; 208263443bf5SEric Sandeen } 208363443bf5SEric Sandeen 2084bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 20856bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 208663443bf5SEric Sandeen if (ret) 208763443bf5SEric Sandeen return ret; 208863443bf5SEric Sandeen } 208963443bf5SEric Sandeen 209063443bf5SEric Sandeen return 0; 209163443bf5SEric Sandeen } 209263443bf5SEric Sandeen 2093abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2094abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2095abed4aaaSJosef Bacik const char *name) 2096abed4aaaSJosef Bacik { 2097abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2098abed4aaaSJosef Bacik struct btrfs_root *root; 2099f7238e50SJosef Bacik u64 max_global_id = 0; 2100abed4aaaSJosef Bacik int ret; 2101abed4aaaSJosef Bacik struct btrfs_key key = { 2102abed4aaaSJosef Bacik .objectid = objectid, 2103abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2104abed4aaaSJosef Bacik .offset = 0, 2105abed4aaaSJosef Bacik }; 2106abed4aaaSJosef Bacik bool found = false; 2107abed4aaaSJosef Bacik 2108abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2109abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2110abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2111abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2112abed4aaaSJosef Bacik return 0; 2113abed4aaaSJosef Bacik } 2114abed4aaaSJosef Bacik 2115abed4aaaSJosef Bacik while (1) { 2116abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2117abed4aaaSJosef Bacik if (ret < 0) 2118abed4aaaSJosef Bacik break; 2119abed4aaaSJosef Bacik 2120abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2121abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2122abed4aaaSJosef Bacik if (ret) { 2123abed4aaaSJosef Bacik if (ret > 0) 2124abed4aaaSJosef Bacik ret = 0; 2125abed4aaaSJosef Bacik break; 2126abed4aaaSJosef Bacik } 2127abed4aaaSJosef Bacik } 2128abed4aaaSJosef Bacik ret = 0; 2129abed4aaaSJosef Bacik 2130abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2131abed4aaaSJosef Bacik if (key.objectid != objectid) 2132abed4aaaSJosef Bacik break; 2133abed4aaaSJosef Bacik btrfs_release_path(path); 2134abed4aaaSJosef Bacik 2135f7238e50SJosef Bacik /* 2136f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2137f7238e50SJosef Bacik * everybody. 2138f7238e50SJosef Bacik */ 2139f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2140f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2141f7238e50SJosef Bacik 2142abed4aaaSJosef Bacik found = true; 2143abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2144abed4aaaSJosef Bacik if (IS_ERR(root)) { 2145abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2146abed4aaaSJosef Bacik ret = PTR_ERR(root); 2147abed4aaaSJosef Bacik break; 2148abed4aaaSJosef Bacik } 2149abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2150abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2151abed4aaaSJosef Bacik if (ret) { 2152abed4aaaSJosef Bacik btrfs_put_root(root); 2153abed4aaaSJosef Bacik break; 2154abed4aaaSJosef Bacik } 2155abed4aaaSJosef Bacik key.offset++; 2156abed4aaaSJosef Bacik } 2157abed4aaaSJosef Bacik btrfs_release_path(path); 2158abed4aaaSJosef Bacik 2159f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2160f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2161f7238e50SJosef Bacik 2162abed4aaaSJosef Bacik if (!found || ret) { 2163abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2164abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2165abed4aaaSJosef Bacik 2166abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2167abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2168abed4aaaSJosef Bacik else 2169abed4aaaSJosef Bacik ret = 0; 2170abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2171abed4aaaSJosef Bacik } 2172abed4aaaSJosef Bacik return ret; 2173abed4aaaSJosef Bacik } 2174abed4aaaSJosef Bacik 2175abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2176abed4aaaSJosef Bacik { 2177abed4aaaSJosef Bacik struct btrfs_path *path; 2178abed4aaaSJosef Bacik int ret = 0; 2179abed4aaaSJosef Bacik 2180abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2181abed4aaaSJosef Bacik if (!path) 2182abed4aaaSJosef Bacik return -ENOMEM; 2183abed4aaaSJosef Bacik 2184abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2185abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2186abed4aaaSJosef Bacik if (ret) 2187abed4aaaSJosef Bacik goto out; 2188abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2189abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2190abed4aaaSJosef Bacik if (ret) 2191abed4aaaSJosef Bacik goto out; 2192abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2193abed4aaaSJosef Bacik goto out; 2194abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2195abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2196abed4aaaSJosef Bacik "free space"); 2197abed4aaaSJosef Bacik out: 2198abed4aaaSJosef Bacik btrfs_free_path(path); 2199abed4aaaSJosef Bacik return ret; 2200abed4aaaSJosef Bacik } 2201abed4aaaSJosef Bacik 22026bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 22034bbcaa64SEric Sandeen { 22046bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2205a4f3d2c4SDavid Sterba struct btrfs_root *root; 22064bbcaa64SEric Sandeen struct btrfs_key location; 22074bbcaa64SEric Sandeen int ret; 22084bbcaa64SEric Sandeen 22096bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 22106bccf3abSJeff Mahoney 2211abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2212abed4aaaSJosef Bacik if (ret) 2213abed4aaaSJosef Bacik return ret; 2214abed4aaaSJosef Bacik 22154bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 22164bbcaa64SEric Sandeen location.offset = 0; 22174bbcaa64SEric Sandeen 22181c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 221914033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 222014033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 222114033b08SQu Wenruo if (IS_ERR(root)) { 222214033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 222314033b08SQu Wenruo ret = PTR_ERR(root); 222414033b08SQu Wenruo goto out; 222514033b08SQu Wenruo } 222614033b08SQu Wenruo } else { 222714033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 222814033b08SQu Wenruo fs_info->block_group_root = root; 222914033b08SQu Wenruo } 223014033b08SQu Wenruo } 223114033b08SQu Wenruo 223214033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2233a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2234f50f4353SLiu Bo if (IS_ERR(root)) { 223542437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2236f50f4353SLiu Bo ret = PTR_ERR(root); 2237f50f4353SLiu Bo goto out; 2238f50f4353SLiu Bo } 223942437a63SJosef Bacik } else { 2240a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2241a4f3d2c4SDavid Sterba fs_info->dev_root = root; 224242437a63SJosef Bacik } 2243820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2244a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2245a8d1b164SJohannes Thumshirn if (ret) 2246a8d1b164SJohannes Thumshirn goto out; 22474bbcaa64SEric Sandeen 2248aeb935a4SQu Wenruo /* 2249aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2250aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2251aeb935a4SQu Wenruo */ 225256e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 225356e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2254aeb935a4SQu Wenruo if (IS_ERR(root)) { 225542437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2256aeb935a4SQu Wenruo ret = PTR_ERR(root); 2257aeb935a4SQu Wenruo goto out; 2258aeb935a4SQu Wenruo } 225942437a63SJosef Bacik } else { 2260aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2261aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 226242437a63SJosef Bacik } 2263aeb935a4SQu Wenruo 22644bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2265a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2266a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2267a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2268afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2269a4f3d2c4SDavid Sterba fs_info->quota_root = root; 22704bbcaa64SEric Sandeen } 22714bbcaa64SEric Sandeen 22724bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2273a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2274a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 227542437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2276a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 22774bbcaa64SEric Sandeen if (ret != -ENOENT) 2278f50f4353SLiu Bo goto out; 227942437a63SJosef Bacik } 22804bbcaa64SEric Sandeen } else { 2281a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2282a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 22834bbcaa64SEric Sandeen } 22844bbcaa64SEric Sandeen 22854bbcaa64SEric Sandeen return 0; 2286f50f4353SLiu Bo out: 2287f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2288f50f4353SLiu Bo location.objectid, ret); 2289f50f4353SLiu Bo return ret; 22904bbcaa64SEric Sandeen } 22914bbcaa64SEric Sandeen 2292069ec957SQu Wenruo /* 2293069ec957SQu Wenruo * Real super block validation 2294069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2295069ec957SQu Wenruo * 2296069ec957SQu Wenruo * @sb: super block to check 2297069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2298069ec957SQu Wenruo * 0 the primary (1st) sb 2299069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2300069ec957SQu Wenruo * -1 skip bytenr check 2301069ec957SQu Wenruo */ 2302a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2303069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 230421a852b0SQu Wenruo { 230521a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 230621a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 230721a852b0SQu Wenruo int ret = 0; 230821a852b0SQu Wenruo 230921a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 231021a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 231121a852b0SQu Wenruo ret = -EINVAL; 231221a852b0SQu Wenruo } 231321a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 231421a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 231521a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 231621a852b0SQu Wenruo ret = -EINVAL; 231721a852b0SQu Wenruo } 231821a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 231921a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 232021a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 232121a852b0SQu Wenruo ret = -EINVAL; 232221a852b0SQu Wenruo } 232321a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 232421a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 232521a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 232621a852b0SQu Wenruo ret = -EINVAL; 232721a852b0SQu Wenruo } 232821a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 232921a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 233021a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 233121a852b0SQu Wenruo ret = -EINVAL; 233221a852b0SQu Wenruo } 233321a852b0SQu Wenruo 233421a852b0SQu Wenruo /* 233521a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 233621a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 233721a852b0SQu Wenruo */ 233821a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 233921a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 234021a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 234121a852b0SQu Wenruo ret = -EINVAL; 234221a852b0SQu Wenruo } 23430bb3eb3eSQu Wenruo 23440bb3eb3eSQu Wenruo /* 23451a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 23461a42daabSQu Wenruo * 23471a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 23481a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 23491a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 23501a42daabSQu Wenruo * beginning. 23510bb3eb3eSQu Wenruo */ 23521a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 235321a852b0SQu Wenruo btrfs_err(fs_info, 23540bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 235521a852b0SQu Wenruo sectorsize, PAGE_SIZE); 235621a852b0SQu Wenruo ret = -EINVAL; 235721a852b0SQu Wenruo } 23580bb3eb3eSQu Wenruo 235921a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 236021a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 236121a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 236221a852b0SQu Wenruo ret = -EINVAL; 236321a852b0SQu Wenruo } 236421a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 236521a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 236621a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 236721a852b0SQu Wenruo ret = -EINVAL; 236821a852b0SQu Wenruo } 236921a852b0SQu Wenruo 237021a852b0SQu Wenruo /* Root alignment check */ 237121a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 237221a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 237321a852b0SQu Wenruo btrfs_super_root(sb)); 237421a852b0SQu Wenruo ret = -EINVAL; 237521a852b0SQu Wenruo } 237621a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 237721a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 237821a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 237921a852b0SQu Wenruo ret = -EINVAL; 238021a852b0SQu Wenruo } 238121a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 238221a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 238321a852b0SQu Wenruo btrfs_super_log_root(sb)); 238421a852b0SQu Wenruo ret = -EINVAL; 238521a852b0SQu Wenruo } 238621a852b0SQu Wenruo 2387aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2388aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2389aefd7f70SNikolay Borisov btrfs_err(fs_info, 2390aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2391aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2392aefd7f70SNikolay Borisov ret = -EINVAL; 2393aefd7f70SNikolay Borisov } 2394aefd7f70SNikolay Borisov 2395aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2396aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2397aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2398aefd7f70SNikolay Borisov btrfs_err(fs_info, 2399aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2400aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2401aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2402aefd7f70SNikolay Borisov ret = -EINVAL; 2403aefd7f70SNikolay Borisov } 2404aefd7f70SNikolay Borisov 240525984a5aSAnand Jain if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 240625984a5aSAnand Jain BTRFS_FSID_SIZE) != 0) { 240725984a5aSAnand Jain btrfs_err(fs_info, 240825984a5aSAnand Jain "dev_item UUID does not match metadata fsid: %pU != %pU", 240925984a5aSAnand Jain fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 241025984a5aSAnand Jain ret = -EINVAL; 241125984a5aSAnand Jain } 241225984a5aSAnand Jain 24131c56ab99SQu Wenruo /* 24141c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 24151c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 24161c56ab99SQu Wenruo */ 24171c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 24181c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 24191c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 24201c56ab99SQu Wenruo btrfs_err(fs_info, 24211c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 24221c56ab99SQu Wenruo ret = -EINVAL; 24231c56ab99SQu Wenruo } 24241c56ab99SQu Wenruo 242521a852b0SQu Wenruo /* 242621a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 242721a852b0SQu Wenruo * done later 242821a852b0SQu Wenruo */ 242921a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 243021a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 243121a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 243221a852b0SQu Wenruo ret = -EINVAL; 243321a852b0SQu Wenruo } 243421a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 243521a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 243621a852b0SQu Wenruo btrfs_super_stripesize(sb)); 243721a852b0SQu Wenruo ret = -EINVAL; 243821a852b0SQu Wenruo } 243921a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 244021a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 244121a852b0SQu Wenruo btrfs_super_num_devices(sb)); 244221a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 244321a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 244421a852b0SQu Wenruo ret = -EINVAL; 244521a852b0SQu Wenruo } 244621a852b0SQu Wenruo 2447069ec957SQu Wenruo if (mirror_num >= 0 && 2448069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 244921a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 245021a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 245121a852b0SQu Wenruo ret = -EINVAL; 245221a852b0SQu Wenruo } 245321a852b0SQu Wenruo 245421a852b0SQu Wenruo /* 245521a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 245621a852b0SQu Wenruo * and one chunk 245721a852b0SQu Wenruo */ 245821a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 245921a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 246021a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 246121a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 246221a852b0SQu Wenruo ret = -EINVAL; 246321a852b0SQu Wenruo } 246421a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 246521a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 246621a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 246721a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 246821a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 246921a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 247021a852b0SQu Wenruo ret = -EINVAL; 247121a852b0SQu Wenruo } 247221a852b0SQu Wenruo 247321a852b0SQu Wenruo /* 247421a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 247521a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 247621a852b0SQu Wenruo */ 247721a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 247821a852b0SQu Wenruo btrfs_warn(fs_info, 247921a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 248021a852b0SQu Wenruo btrfs_super_generation(sb), 248121a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 248221a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 248321a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 248421a852b0SQu Wenruo btrfs_warn(fs_info, 248521a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 248621a852b0SQu Wenruo btrfs_super_generation(sb), 248721a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 248821a852b0SQu Wenruo 248921a852b0SQu Wenruo return ret; 249021a852b0SQu Wenruo } 249121a852b0SQu Wenruo 2492069ec957SQu Wenruo /* 2493069ec957SQu Wenruo * Validation of super block at mount time. 2494069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2495069ec957SQu Wenruo * flags will be skipped. 2496069ec957SQu Wenruo */ 2497069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2498069ec957SQu Wenruo { 2499a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2500069ec957SQu Wenruo } 2501069ec957SQu Wenruo 250275cb857dSQu Wenruo /* 250375cb857dSQu Wenruo * Validation of super block at write time. 250475cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 250575cb857dSQu Wenruo * overwritten soon. 250675cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 250775cb857dSQu Wenruo */ 250875cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 250975cb857dSQu Wenruo struct btrfs_super_block *sb) 251075cb857dSQu Wenruo { 251175cb857dSQu Wenruo int ret; 251275cb857dSQu Wenruo 2513a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 251475cb857dSQu Wenruo if (ret < 0) 251575cb857dSQu Wenruo goto out; 2516e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 251775cb857dSQu Wenruo ret = -EUCLEAN; 251875cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 251975cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 252075cb857dSQu Wenruo goto out; 252175cb857dSQu Wenruo } 252275cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 252375cb857dSQu Wenruo ret = -EUCLEAN; 252475cb857dSQu Wenruo btrfs_err(fs_info, 252575cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 252675cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 252775cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 252875cb857dSQu Wenruo goto out; 252975cb857dSQu Wenruo } 253075cb857dSQu Wenruo out: 253175cb857dSQu Wenruo if (ret < 0) 253275cb857dSQu Wenruo btrfs_err(fs_info, 253375cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 253475cb857dSQu Wenruo return ret; 253575cb857dSQu Wenruo } 253675cb857dSQu Wenruo 2537bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2538bd676446SJosef Bacik { 2539789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 2540789d6a3aSQu Wenruo .level = level, 2541789d6a3aSQu Wenruo .transid = gen, 2542789d6a3aSQu Wenruo .owner_root = root->root_key.objectid 2543789d6a3aSQu Wenruo }; 2544bd676446SJosef Bacik int ret = 0; 2545bd676446SJosef Bacik 2546789d6a3aSQu Wenruo root->node = read_tree_block(root->fs_info, bytenr, &check); 2547bd676446SJosef Bacik if (IS_ERR(root->node)) { 2548bd676446SJosef Bacik ret = PTR_ERR(root->node); 2549bd676446SJosef Bacik root->node = NULL; 25504eb150d6SQu Wenruo return ret; 25514eb150d6SQu Wenruo } 25524eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2553bd676446SJosef Bacik free_extent_buffer(root->node); 2554bd676446SJosef Bacik root->node = NULL; 25554eb150d6SQu Wenruo return -EIO; 2556bd676446SJosef Bacik } 2557bd676446SJosef Bacik 2558bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2559bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2560bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2561bd676446SJosef Bacik return ret; 2562bd676446SJosef Bacik } 2563bd676446SJosef Bacik 2564bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2565bd676446SJosef Bacik { 2566bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2567bd676446SJosef Bacik u64 gen, bytenr; 2568bd676446SJosef Bacik int level, ret; 2569bd676446SJosef Bacik 2570bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2571bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2572bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2573bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 25749c54e80dSJosef Bacik if (ret) { 2575bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2576bd676446SJosef Bacik return ret; 2577bd676446SJosef Bacik } 25789c54e80dSJosef Bacik return 0; 25799c54e80dSJosef Bacik } 25809c54e80dSJosef Bacik 25816ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2582b8522a1eSNikolay Borisov { 25836ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2584b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2585b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2586b8522a1eSNikolay Borisov bool handle_error = false; 2587b8522a1eSNikolay Borisov int ret = 0; 2588b8522a1eSNikolay Borisov int i; 2589b8522a1eSNikolay Borisov 2590b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2591b8522a1eSNikolay Borisov if (handle_error) { 2592b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2593b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2594b8522a1eSNikolay Borisov tree_root->node = NULL; 2595b8522a1eSNikolay Borisov 2596b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2597b8522a1eSNikolay Borisov break; 2598b8522a1eSNikolay Borisov 2599b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2600b8522a1eSNikolay Borisov 2601b8522a1eSNikolay Borisov /* 2602b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2603b8522a1eSNikolay Borisov * valid 2604b8522a1eSNikolay Borisov */ 2605b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2606b8522a1eSNikolay Borisov 2607b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2608b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2609b8522a1eSNikolay Borisov 2610745806fbSQu Wenruo btrfs_warn(fs_info, "try to load backup roots slot %d", i); 2611b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26126ef108ddSNikolay Borisov backup_index = ret; 2613b8522a1eSNikolay Borisov if (ret < 0) 2614b8522a1eSNikolay Borisov return ret; 2615b8522a1eSNikolay Borisov } 2616b8522a1eSNikolay Borisov 2617bd676446SJosef Bacik ret = load_important_roots(fs_info); 2618bd676446SJosef Bacik if (ret) { 2619217f5004SNikolay Borisov handle_error = true; 2620b8522a1eSNikolay Borisov continue; 2621b8522a1eSNikolay Borisov } 2622b8522a1eSNikolay Borisov 2623336a0d8dSNikolay Borisov /* 2624336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2625336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2626336a0d8dSNikolay Borisov */ 2627453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2628b8522a1eSNikolay Borisov if (ret < 0) { 2629b8522a1eSNikolay Borisov handle_error = true; 2630b8522a1eSNikolay Borisov continue; 2631b8522a1eSNikolay Borisov } 2632b8522a1eSNikolay Borisov 26336b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2634b8522a1eSNikolay Borisov 2635b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2636b8522a1eSNikolay Borisov if (ret < 0) { 2637b8522a1eSNikolay Borisov handle_error = true; 2638b8522a1eSNikolay Borisov continue; 2639b8522a1eSNikolay Borisov } 2640b8522a1eSNikolay Borisov 2641b8522a1eSNikolay Borisov /* All successful */ 2642bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2643bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2644d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 26456ef108ddSNikolay Borisov 26466ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26476ef108ddSNikolay Borisov if (backup_index < 0) { 26486ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26496ef108ddSNikolay Borisov } else { 26506ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26516ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26526ef108ddSNikolay Borisov } 2653b8522a1eSNikolay Borisov break; 2654b8522a1eSNikolay Borisov } 2655b8522a1eSNikolay Borisov 2656b8522a1eSNikolay Borisov return ret; 2657b8522a1eSNikolay Borisov } 2658b8522a1eSNikolay Borisov 26598260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2660eb60ceacSChris Mason { 2661fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 266201cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26638fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2664facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 266524bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2666eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 266711833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2668eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2669a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2670fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 267124bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26724cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2673ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2674f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 267547ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 267640ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2677afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2678c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2679f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2680abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2681d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2682f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 26837585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2684573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 26850bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 26865f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 2687de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 268819c00ddcSChris Mason 2689e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 26905a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 26918b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 26925f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 26933e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 26943e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 26953e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 26963e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 26973e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 26983e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 26993e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 27003e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 2701e1489b4fSIoannis Angelakopoulos 27020b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 27036324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2704f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 270547ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 270618bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2707afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2708bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2709bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 27103fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27113fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2712bd647ce3SJosef Bacik #endif 2713c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 271466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 271566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 271666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 271766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 271866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 271966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 272066d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2721ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2722ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2723ba2c4d4eSJosef Bacik 2724771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27254cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2726034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2727fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2728abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 272995ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27309ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27314cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2732a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2733f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27348b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2735fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2736c8b97818SChris Mason 2737b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2738b34b086cSChris Mason num_online_cpus() + 2, 8); 27390afbaf8cSChris Mason 2740199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2741199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 274269fe2d75SJosef Bacik 2743638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 274421adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 274521adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 274621adbd5cSStefan Behrens #endif 2747779a65a4SEric Sandeen btrfs_init_balance(fs_info); 274857056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2749a2de733cSArne Jansen 275016b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 275108dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 27520f9dd46cSJosef Bacik 2753fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 275435da5a7eSDavid Sterba IO_TREE_FS_EXCLUDED_EXTENTS); 275539279cc3SChris Mason 27565a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2757e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2758925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2759a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2760a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27611bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27629e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2763c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 276476dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2765803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2766fa9c0d79SChris Mason 2767ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2768f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2769b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2770416ac51dSArne Jansen 2771fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2772fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2773fa9c0d79SChris Mason 2774e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2775f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2776bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27774854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2778034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 27793768f368SChris Mason 2780da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2781da17066cSJeff Mahoney fs_info->nodesize = 4096; 2782da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2783ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2784da17066cSJeff Mahoney fs_info->stripesize = 4096; 2785da17066cSJeff Mahoney 2786f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 2787f7b12a62SNaohiro Aota 2788eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2789eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2790eede2bf3SOmar Sandoval 279118bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 279218bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 27938260edbaSJosef Bacik } 27948260edbaSJosef Bacik 27958260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 27968260edbaSJosef Bacik { 27978260edbaSJosef Bacik int ret; 27988260edbaSJosef Bacik 27998260edbaSJosef Bacik fs_info->sb = sb; 28008260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 28018260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 28029e967495SFilipe Manana 28035deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 2804ae18c37aSJosef Bacik if (ret) 2805ae18c37aSJosef Bacik return ret; 2806ae18c37aSJosef Bacik 2807ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2808ae18c37aSJosef Bacik if (ret) 2809c75e8394SJosef Bacik return ret; 2810ae18c37aSJosef Bacik 2811ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2812ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2813ae18c37aSJosef Bacik 2814ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2815ae18c37aSJosef Bacik if (ret) 2816c75e8394SJosef Bacik return ret; 2817ae18c37aSJosef Bacik 2818ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2819ae18c37aSJosef Bacik GFP_KERNEL); 2820ae18c37aSJosef Bacik if (ret) 2821c75e8394SJosef Bacik return ret; 2822ae18c37aSJosef Bacik 2823ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2824ae18c37aSJosef Bacik GFP_KERNEL); 2825c75e8394SJosef Bacik if (!fs_info->delayed_root) 2826c75e8394SJosef Bacik return -ENOMEM; 2827ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2828ae18c37aSJosef Bacik 2829a0a1db70SFilipe Manana if (sb_rdonly(sb)) 2830a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 2831a0a1db70SFilipe Manana 2832c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2833ae18c37aSJosef Bacik } 2834ae18c37aSJosef Bacik 283597f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 283697f4dd09SNikolay Borisov { 28370d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 283897f4dd09SNikolay Borisov int ret; 283997f4dd09SNikolay Borisov 284097f4dd09SNikolay Borisov /* 284197f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 284297f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 284397f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 284497f4dd09SNikolay Borisov */ 284597f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 284697f4dd09SNikolay Borisov if (ret < 0) { 2847c94bec2cSJosef Bacik if (ret != -EINTR) 2848c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2849c94bec2cSJosef Bacik ret); 285097f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 285197f4dd09SNikolay Borisov return ret; 285297f4dd09SNikolay Borisov } 285397f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 285497f4dd09SNikolay Borisov } 285597f4dd09SNikolay Borisov 285697f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 285797f4dd09SNikolay Borisov { 285897f4dd09SNikolay Borisov struct task_struct *task; 285997f4dd09SNikolay Borisov 286097f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 286197f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 286297f4dd09SNikolay Borisov if (IS_ERR(task)) { 286397f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 286497f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 286597f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 286697f4dd09SNikolay Borisov return PTR_ERR(task); 286797f4dd09SNikolay Borisov } 286897f4dd09SNikolay Borisov 286997f4dd09SNikolay Borisov return 0; 287097f4dd09SNikolay Borisov } 287197f4dd09SNikolay Borisov 287244c0ca21SBoris Burkov /* 28738cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 28748cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 28758cd29088SBoris Burkov * code paths. 28768cd29088SBoris Burkov */ 28778cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 28788cd29088SBoris Burkov { 28798cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 28808b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 28818cd29088SBoris Burkov } 28828cd29088SBoris Burkov 28838cd29088SBoris Burkov /* 288444c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 288544c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 288644c0ca21SBoris Burkov */ 288744c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 288844c0ca21SBoris Burkov { 288944c0ca21SBoris Burkov int ret; 289094846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 28911d6a4fc8SQu Wenruo bool rebuild_free_space_tree = false; 28928b228324SBoris Burkov 28938b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 28948b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 28951d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 28968b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 28978b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 28988b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 28991d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 29008b228324SBoris Burkov } 29018b228324SBoris Burkov 29021d6a4fc8SQu Wenruo if (rebuild_free_space_tree) { 29031d6a4fc8SQu Wenruo btrfs_info(fs_info, "rebuilding free space tree"); 29041d6a4fc8SQu Wenruo ret = btrfs_rebuild_free_space_tree(fs_info); 29058b228324SBoris Burkov if (ret) { 29068b228324SBoris Burkov btrfs_warn(fs_info, 29071d6a4fc8SQu Wenruo "failed to rebuild free space tree: %d", ret); 29081d6a4fc8SQu Wenruo goto out; 29091d6a4fc8SQu Wenruo } 29101d6a4fc8SQu Wenruo } 29111d6a4fc8SQu Wenruo 29121d6a4fc8SQu Wenruo if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29131d6a4fc8SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { 29141d6a4fc8SQu Wenruo btrfs_info(fs_info, "disabling free space tree"); 29151d6a4fc8SQu Wenruo ret = btrfs_delete_free_space_tree(fs_info); 29161d6a4fc8SQu Wenruo if (ret) { 29171d6a4fc8SQu Wenruo btrfs_warn(fs_info, 29181d6a4fc8SQu Wenruo "failed to disable free space tree: %d", ret); 29198b228324SBoris Burkov goto out; 29208b228324SBoris Burkov } 29218b228324SBoris Burkov } 292244c0ca21SBoris Burkov 29238d488a8cSFilipe Manana /* 29248d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 29258d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 2926fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 29278d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 29288d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 29298d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 29308d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 29318d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 29328d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 29338d488a8cSFilipe Manana */ 29348d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 29358d488a8cSFilipe Manana if (ret) 29368d488a8cSFilipe Manana goto out; 29378d488a8cSFilipe Manana 293844c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 293944c0ca21SBoris Burkov if (ret) 294044c0ca21SBoris Burkov goto out; 294144c0ca21SBoris Burkov 29428f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 29438f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29448f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29458f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29468f1c21d7SBoris Burkov goto out; 29478f1c21d7SBoris Burkov } 29488f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29498f1c21d7SBoris Burkov 295044c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 29517eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 295244c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 295344c0ca21SBoris Burkov if (ret < 0) { 295444c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 295544c0ca21SBoris Burkov goto out; 295644c0ca21SBoris Burkov } 295744c0ca21SBoris Burkov 29585011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 29595011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 29605011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 29615011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 29625011139aSBoris Burkov if (ret) { 29635011139aSBoris Burkov btrfs_warn(fs_info, 29645011139aSBoris Burkov "failed to create free space tree: %d", ret); 29655011139aSBoris Burkov goto out; 29665011139aSBoris Burkov } 29675011139aSBoris Burkov } 29685011139aSBoris Burkov 296994846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 297094846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 297194846229SBoris Burkov if (ret) 297294846229SBoris Burkov goto out; 297394846229SBoris Burkov } 297494846229SBoris Burkov 297544c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 297644c0ca21SBoris Burkov if (ret) 297744c0ca21SBoris Burkov goto out; 297844c0ca21SBoris Burkov 297944c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 298044c0ca21SBoris Burkov if (ret) { 298144c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 298244c0ca21SBoris Burkov goto out; 298344c0ca21SBoris Burkov } 298444c0ca21SBoris Burkov 298544c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 298644c0ca21SBoris Burkov 298744c0ca21SBoris Burkov if (!fs_info->uuid_root) { 298844c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 298944c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 299044c0ca21SBoris Burkov if (ret) { 299144c0ca21SBoris Burkov btrfs_warn(fs_info, 299244c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 299344c0ca21SBoris Burkov goto out; 299444c0ca21SBoris Burkov } 299544c0ca21SBoris Burkov } 299644c0ca21SBoris Burkov 299744c0ca21SBoris Burkov out: 299844c0ca21SBoris Burkov return ret; 299944c0ca21SBoris Burkov } 300044c0ca21SBoris Burkov 3001d7f67ac9SQu Wenruo /* 3002d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3003d7f67ac9SQu Wenruo * 30042ba48b20SQu Wenruo * @is_rw_mount: If the mount is read-write. 30052ba48b20SQu Wenruo * 3006d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3007d7f67ac9SQu Wenruo * feature dependencies). 3008d7f67ac9SQu Wenruo * 3009d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3010d7f67ac9SQu Wenruo * btrfs_validate_super(). 3011d7f67ac9SQu Wenruo * 3012d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3013d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3014d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3015d7f67ac9SQu Wenruo */ 30162ba48b20SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) 3017d7f67ac9SQu Wenruo { 3018d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3019d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3020d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3021d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3022d7f67ac9SQu Wenruo 3023d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3024d7f67ac9SQu Wenruo btrfs_err(fs_info, 3025d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3026d7f67ac9SQu Wenruo incompat); 3027d7f67ac9SQu Wenruo return -EINVAL; 3028d7f67ac9SQu Wenruo } 3029d7f67ac9SQu Wenruo 3030d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3031d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3032d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3033d7f67ac9SQu Wenruo btrfs_err(fs_info, 3034d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3035d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3036d7f67ac9SQu Wenruo return -EINVAL; 3037d7f67ac9SQu Wenruo } 3038d7f67ac9SQu Wenruo 3039d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3040d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3041d7f67ac9SQu Wenruo 3042d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3043d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3044d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3045d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3046d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3047d7f67ac9SQu Wenruo 3048d7f67ac9SQu Wenruo /* 3049d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3050d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3051d7f67ac9SQu Wenruo */ 3052d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3053d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3054d7f67ac9SQu Wenruo 30552ba48b20SQu Wenruo if (compat_ro_unsupp && is_rw_mount) { 3056d7f67ac9SQu Wenruo btrfs_err(fs_info, 3057d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3058d7f67ac9SQu Wenruo compat_ro); 3059d7f67ac9SQu Wenruo return -EINVAL; 3060d7f67ac9SQu Wenruo } 3061d7f67ac9SQu Wenruo 3062d7f67ac9SQu Wenruo /* 3063d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3064d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3065d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3066d7f67ac9SQu Wenruo */ 3067d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3068d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3069d7f67ac9SQu Wenruo btrfs_err(fs_info, 3070d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3071d7f67ac9SQu Wenruo compat_ro); 3072d7f67ac9SQu Wenruo return -EINVAL; 3073d7f67ac9SQu Wenruo } 3074d7f67ac9SQu Wenruo 3075d7f67ac9SQu Wenruo /* 3076d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3077d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3078d7f67ac9SQu Wenruo */ 3079d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3080d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3081d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3082d7f67ac9SQu Wenruo btrfs_err(fs_info, 3083d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3084d7f67ac9SQu Wenruo return -EINVAL; 3085d7f67ac9SQu Wenruo } 3086d7f67ac9SQu Wenruo 3087d7f67ac9SQu Wenruo /* 3088d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3089d7f67ac9SQu Wenruo * 3090d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3091d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3092d7f67ac9SQu Wenruo * going to be deprecated anyway. 3093d7f67ac9SQu Wenruo */ 3094d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3095d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3096d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3097d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3098d7f67ac9SQu Wenruo return -EINVAL; 3099d7f67ac9SQu Wenruo } 3100d7f67ac9SQu Wenruo 3101d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3102d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3103d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3104d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3105d7f67ac9SQu Wenruo 3106d7f67ac9SQu Wenruo return 0; 3107d7f67ac9SQu Wenruo } 3108d7f67ac9SQu Wenruo 3109ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3110ae18c37aSJosef Bacik char *options) 3111ae18c37aSJosef Bacik { 3112ae18c37aSJosef Bacik u32 sectorsize; 3113ae18c37aSJosef Bacik u32 nodesize; 3114ae18c37aSJosef Bacik u32 stripesize; 3115ae18c37aSJosef Bacik u64 generation; 3116ae18c37aSJosef Bacik u64 features; 3117ae18c37aSJosef Bacik u16 csum_type; 3118ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3119ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3120ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3121ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3122ae18c37aSJosef Bacik int ret; 3123ae18c37aSJosef Bacik int level; 3124ae18c37aSJosef Bacik 31258260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 31264871c33bSQu Wenruo if (ret) 3127ae18c37aSJosef Bacik goto fail; 312853b381b3SDavid Woodhouse 3129ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3130ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3131ae18c37aSJosef Bacik GFP_KERNEL); 3132ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3133ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3134ae18c37aSJosef Bacik GFP_KERNEL); 3135ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3136ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 31374871c33bSQu Wenruo ret = -ENOMEM; 3138c75e8394SJosef Bacik goto fail; 3139ae18c37aSJosef Bacik } 3140ae18c37aSJosef Bacik 3141dcb2137cSChristoph Hellwig ret = btrfs_init_btree_inode(sb); 31424871c33bSQu Wenruo if (ret) 3143c75e8394SJosef Bacik goto fail; 3144ae18c37aSJosef Bacik 3145d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 31461104a885SDavid Sterba 31471104a885SDavid Sterba /* 31481104a885SDavid Sterba * Read super block and check the signature bytes only 31491104a885SDavid Sterba */ 3150d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 31518f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 31524871c33bSQu Wenruo ret = PTR_ERR(disk_super); 315316cdcec7SMiao Xie goto fail_alloc; 315420b45077SDave Young } 315539279cc3SChris Mason 31561104a885SDavid Sterba /* 3157260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 31588dc3f22cSJohannes Thumshirn * corrupted, we'll find out 31598dc3f22cSJohannes Thumshirn */ 31608f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 316151bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 31628dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 316351bce6c9SJohannes Thumshirn csum_type); 31644871c33bSQu Wenruo ret = -EINVAL; 31658f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31668dc3f22cSJohannes Thumshirn goto fail_alloc; 31678dc3f22cSJohannes Thumshirn } 31688dc3f22cSJohannes Thumshirn 316983c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 317083c68bbcSSu Yue 31716d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 31726d97c6e3SJohannes Thumshirn if (ret) { 31738f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31746d97c6e3SJohannes Thumshirn goto fail_alloc; 31756d97c6e3SJohannes Thumshirn } 31766d97c6e3SJohannes Thumshirn 31778dc3f22cSJohannes Thumshirn /* 31781104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 31791104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 31801104a885SDavid Sterba */ 31813d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 318205135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 31834871c33bSQu Wenruo ret = -EINVAL; 31848f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3185141386e1SJosef Bacik goto fail_alloc; 31861104a885SDavid Sterba } 31871104a885SDavid Sterba 31881104a885SDavid Sterba /* 31891104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 31901104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 31911104a885SDavid Sterba * the whole block of INFO_SIZE 31921104a885SDavid Sterba */ 31938f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 31948f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31955f39d397SChris Mason 3196fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3197fbc6feaeSNikolay Borisov 31980b86a832SChris Mason 3199fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3200fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3201fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3202fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3203fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3204fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3205fbc6feaeSNikolay Borisov } 3206fbc6feaeSNikolay Borisov 3207fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3208fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3209de37aa51SNikolay Borisov 3210069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 32111104a885SDavid Sterba if (ret) { 321205135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 32134871c33bSQu Wenruo ret = -EINVAL; 3214141386e1SJosef Bacik goto fail_alloc; 32151104a885SDavid Sterba } 32161104a885SDavid Sterba 32174871c33bSQu Wenruo if (!btrfs_super_root(disk_super)) { 32184871c33bSQu Wenruo btrfs_err(fs_info, "invalid superblock tree root bytenr"); 32194871c33bSQu Wenruo ret = -EINVAL; 3220141386e1SJosef Bacik goto fail_alloc; 32214871c33bSQu Wenruo } 32220f7d52f4SChris Mason 3223acce952bSliubo /* check FS state, whether FS is broken. */ 322487533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 3225*ae3364e5SFilipe Manana WRITE_ONCE(fs_info->fs_error, -EUCLEAN); 3226acce952bSliubo 322775e7cb7fSLiu Bo /* 322875e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 322975e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 323075e7cb7fSLiu Bo */ 323175e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 323275e7cb7fSLiu Bo 32336f93e834SAnand Jain 32346f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 32356f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 32366f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 32376f93e834SAnand Jain stripesize = sectorsize; 32386f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 32396f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 32406f93e834SAnand Jain 32416f93e834SAnand Jain fs_info->nodesize = nodesize; 32426f93e834SAnand Jain fs_info->sectorsize = sectorsize; 32436f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 32446f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 32456f93e834SAnand Jain fs_info->stripesize = stripesize; 32466f93e834SAnand Jain 32472ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 32484871c33bSQu Wenruo if (ret) 3249141386e1SJosef Bacik goto fail_alloc; 3250dfe25020SChris Mason 32512ba48b20SQu Wenruo ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); 32524871c33bSQu Wenruo if (ret < 0) 3253141386e1SJosef Bacik goto fail_alloc; 3254f2b636e8SJosef Bacik 32558481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 32568481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 32578481dd80SQu Wenruo 32589f73f1aeSQu Wenruo /* 32599f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 32609f73f1aeSQu Wenruo * going to be deprecated. 32619f73f1aeSQu Wenruo * 32629f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 32639f73f1aeSQu Wenruo */ 32649f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 32659f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 32669f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 32679f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 32689f73f1aeSQu Wenruo 326995ea0486SQu Wenruo btrfs_warn(fs_info, 327095ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 32710bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 32728481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 32734871c33bSQu Wenruo if (!subpage_info) { 32744871c33bSQu Wenruo ret = -ENOMEM; 32758481dd80SQu Wenruo goto fail_alloc; 32764871c33bSQu Wenruo } 32778481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 32788481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3279c8050b3bSQu Wenruo } 32800bb3eb3eSQu Wenruo 3281d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 32824871c33bSQu Wenruo if (ret) 32830dc3b84aSJosef Bacik goto fail_sb_buffer; 32844543df7eSChris Mason 32859e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 32869e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 32874575c9ccSChris Mason 3288a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3289a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3290de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3291db94535dSChris Mason 3292925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 32936bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3294925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 329584eed90fSChris Mason if (ret) { 329605135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 32975d4f98a2SYan Zheng goto fail_sb_buffer; 329884eed90fSChris Mason } 32990b86a832SChris Mason 330084234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3301581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3302bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3303bd676446SJosef Bacik generation, level); 3304bd676446SJosef Bacik if (ret) { 330505135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3306af31f5e5SChris Mason goto fail_tree_roots; 330783121942SDavid Woodhouse } 33080b86a832SChris Mason 3309e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3310c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3311c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3312e17cade2SChris Mason 33135b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 33142b82032cSYan Zheng if (ret) { 331505135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3316af31f5e5SChris Mason goto fail_tree_roots; 33172b82032cSYan Zheng } 33180b86a832SChris Mason 33198dabb742SStefan Behrens /* 3320bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3321bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3322bacce86aSAnand Jain * target is required. So free devices that are not part of this 33231a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3324bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 33258dabb742SStefan Behrens */ 3326bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3327d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 332805135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 33294871c33bSQu Wenruo ret = -EIO; 3330a6b0d5c8SChris Mason goto fail_tree_roots; 3331a6b0d5c8SChris Mason } 3332a6b0d5c8SChris Mason 3333b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 33344bbcaa64SEric Sandeen if (ret) 3335b8522a1eSNikolay Borisov goto fail_tree_roots; 33368929ecfaSYan, Zheng 333775ec1db8SJosef Bacik /* 333873651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 333973651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 334073651042SNaohiro Aota * zoned incompat feature flag set. 334173651042SNaohiro Aota */ 334273651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 334373651042SNaohiro Aota if (ret) { 334473651042SNaohiro Aota btrfs_err(fs_info, 33454871c33bSQu Wenruo "zoned: failed to read device zone info: %d", ret); 334673651042SNaohiro Aota goto fail_block_groups; 334773651042SNaohiro Aota } 334873651042SNaohiro Aota 334973651042SNaohiro Aota /* 335075ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 335175ec1db8SJosef Bacik * check the generation here so we can set the 335275ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 335375ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 335475ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 335575ec1db8SJosef Bacik * even though it was perfectly fine. 335675ec1db8SJosef Bacik */ 335775ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 335875ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 335975ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 336075ec1db8SJosef Bacik 3361cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3362cf90d884SQu Wenruo if (ret) { 3363cf90d884SQu Wenruo btrfs_err(fs_info, 3364cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3365cf90d884SQu Wenruo ret); 3366cf90d884SQu Wenruo goto fail_block_groups; 3367cf90d884SQu Wenruo } 336868310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 336968310a5eSIlya Dryomov if (ret) { 337005135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 337168310a5eSIlya Dryomov goto fail_block_groups; 337268310a5eSIlya Dryomov } 337368310a5eSIlya Dryomov 3374733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3375733f4fbbSStefan Behrens if (ret) { 337605135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3377733f4fbbSStefan Behrens goto fail_block_groups; 3378733f4fbbSStefan Behrens } 3379733f4fbbSStefan Behrens 33808dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 33818dabb742SStefan Behrens if (ret) { 338205135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 33838dabb742SStefan Behrens goto fail_block_groups; 33848dabb742SStefan Behrens } 33858dabb742SStefan Behrens 3386b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3387b70f5097SNaohiro Aota if (ret) { 3388b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3389b70f5097SNaohiro Aota ret); 3390b70f5097SNaohiro Aota goto fail_block_groups; 3391b70f5097SNaohiro Aota } 3392b70f5097SNaohiro Aota 3393c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3394b7c35e81SAnand Jain if (ret) { 339505135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 339605135f59SDavid Sterba ret); 3397b7c35e81SAnand Jain goto fail_block_groups; 3398b7c35e81SAnand Jain } 3399b7c35e81SAnand Jain 340096f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 34015ac1d209SJeff Mahoney if (ret) { 340205135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3403b7c35e81SAnand Jain goto fail_fsdev_sysfs; 34045ac1d209SJeff Mahoney } 34055ac1d209SJeff Mahoney 3406c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3407c59021f8Sliubo if (ret) { 340805135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 34092365dd3cSAnand Jain goto fail_sysfs; 3410c59021f8Sliubo } 3411c59021f8Sliubo 34125b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 34131b1d1f66SJosef Bacik if (ret) { 341405135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 34152365dd3cSAnand Jain goto fail_sysfs; 34161b1d1f66SJosef Bacik } 34174330e183SQu Wenruo 341816beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 341916beac87SNaohiro Aota 34205c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 34215c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 342205135f59SDavid Sterba btrfs_warn(fs_info, 342352042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 34244871c33bSQu Wenruo ret = -EINVAL; 34252365dd3cSAnand Jain goto fail_sysfs; 3426292fd7fcSStefan Behrens } 34279078a3e1SChris Mason 342833c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3429a74a4b97SChris Mason "btrfs-cleaner"); 34304871c33bSQu Wenruo if (IS_ERR(fs_info->cleaner_kthread)) { 34314871c33bSQu Wenruo ret = PTR_ERR(fs_info->cleaner_kthread); 34322365dd3cSAnand Jain goto fail_sysfs; 34334871c33bSQu Wenruo } 3434a74a4b97SChris Mason 3435a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3436a74a4b97SChris Mason tree_root, 3437a74a4b97SChris Mason "btrfs-transaction"); 34384871c33bSQu Wenruo if (IS_ERR(fs_info->transaction_kthread)) { 34394871c33bSQu Wenruo ret = PTR_ERR(fs_info->transaction_kthread); 34403f157a2fSChris Mason goto fail_cleaner; 34414871c33bSQu Wenruo } 3442a74a4b97SChris Mason 3443583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3444c289811cSChris Mason !fs_info->fs_devices->rotating) { 3445583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3446c289811cSChris Mason } 3447c289811cSChris Mason 3448572d9ab7SDavid Sterba /* 344963a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 345063a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 345163a7cb13SDavid Sterba * nodiscard for the same mount. 345295ca6599SNaohiro Aota * 345395ca6599SNaohiro Aota * The zoned mode piggy backs on the discard functionality for 345495ca6599SNaohiro Aota * resetting a zone. There is no reason to delay the zone reset as it is 345595ca6599SNaohiro Aota * fast enough. So, do not enable async discard for zoned mode. 345663a7cb13SDavid Sterba */ 345763a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 345863a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 345963a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 346095ca6599SNaohiro Aota fs_info->fs_devices->discardable && 346195ca6599SNaohiro Aota !btrfs_is_zoned(fs_info)) { 346263a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 346363a7cb13SDavid Sterba "auto enabling async discard"); 346463a7cb13SDavid Sterba } 346563a7cb13SDavid Sterba 346621adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 34670b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 34682ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 34690b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3470cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 347121adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 347221adbd5cSStefan Behrens if (ret) 347305135f59SDavid Sterba btrfs_warn(fs_info, 347405135f59SDavid Sterba "failed to initialize integrity check module: %d", 347505135f59SDavid Sterba ret); 347621adbd5cSStefan Behrens } 347721adbd5cSStefan Behrens #endif 3478bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3479bcef60f2SArne Jansen if (ret) 3480bcef60f2SArne Jansen goto fail_trans_kthread; 348121adbd5cSStefan Behrens 3482fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3483fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3484fd708b81SJosef Bacik 348596da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 348696da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 34870b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3488e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 348963443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 34904871c33bSQu Wenruo if (ret) 349128c16cbbSWang Shilong goto fail_qgroup; 3492e556ce2cSYan Zheng } 34931a40e23bSZheng Yan 349456e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 34953140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 34964871c33bSQu Wenruo ret = PTR_ERR(fs_info->fs_root); 34974871c33bSQu Wenruo btrfs_warn(fs_info, "failed to read fs tree: %d", ret); 3498315bf8efSJosef Bacik fs_info->fs_root = NULL; 3499bcef60f2SArne Jansen goto fail_qgroup; 35003140c9a3SDan Carpenter } 3501c289811cSChris Mason 3502bc98a42cSDavid Howells if (sb_rdonly(sb)) 35038cd29088SBoris Burkov goto clear_oneshot; 35042b6ba629SIlya Dryomov 350544c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 35062b6ba629SIlya Dryomov if (ret) { 35076bccf3abSJeff Mahoney close_ctree(fs_info); 35082b6ba629SIlya Dryomov return ret; 3509e3acc2a6SJosef Bacik } 3510b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3511b382a324SJan Schmidt 351244c0ca21SBoris Burkov if (fs_info->uuid_root && 351344c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 351444c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 351505135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 351670f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 351770f80175SStefan Behrens if (ret) { 351805135f59SDavid Sterba btrfs_warn(fs_info, 351905135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 35206bccf3abSJeff Mahoney close_ctree(fs_info); 352170f80175SStefan Behrens return ret; 352270f80175SStefan Behrens } 3523f7a81ea4SStefan Behrens } 352494846229SBoris Burkov 3525afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 352647ab2a6cSJosef Bacik 3527b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3528b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3529b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3530b4be6aefSJosef Bacik 35318cd29088SBoris Burkov clear_oneshot: 35328cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3533ad2b2c80SAl Viro return 0; 353439279cc3SChris Mason 3535bcef60f2SArne Jansen fail_qgroup: 3536bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 35377c2ca468SChris Mason fail_trans_kthread: 35387c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 35392ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3540faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 35413f157a2fSChris Mason fail_cleaner: 3542a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 35437c2ca468SChris Mason 35447c2ca468SChris Mason /* 35457c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 35467c2ca468SChris Mason * kthreads 35477c2ca468SChris Mason */ 35487c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 35497c2ca468SChris Mason 35502365dd3cSAnand Jain fail_sysfs: 35516618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 35522365dd3cSAnand Jain 3553b7c35e81SAnand Jain fail_fsdev_sysfs: 3554b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3555b7c35e81SAnand Jain 35561b1d1f66SJosef Bacik fail_block_groups: 355754067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3558af31f5e5SChris Mason 3559af31f5e5SChris Mason fail_tree_roots: 35609e3aa805SJosef Bacik if (fs_info->data_reloc_root) 35619e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 35624273eaffSAnand Jain free_root_pointers(fs_info, true); 35632b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3564af31f5e5SChris Mason 356539279cc3SChris Mason fail_sb_buffer: 35667abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 35675cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 356816cdcec7SMiao Xie fail_alloc: 3569586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3570586e46e2SIlya Dryomov 35714543df7eSChris Mason iput(fs_info->btree_inode); 35727e662854SQinghuang Feng fail: 3573586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 35744871c33bSQu Wenruo ASSERT(ret < 0); 35754871c33bSQu Wenruo return ret; 3576eb60ceacSChris Mason } 3577663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3578eb60ceacSChris Mason 3579314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3580f2984462SChris Mason { 3581314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3582314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3583314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3584314b6dd0SJohannes Thumshirn struct page *page; 3585442a4f63SStefan Behrens 3586314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3587314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3588314b6dd0SJohannes Thumshirn 3589314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3590fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3591314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3592cb3e217bSQu Wenruo btrfs_dev_name(device), 3593314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3594314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3595314b6dd0SJohannes Thumshirn SetPageError(page); 3596314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3597314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3598314b6dd0SJohannes Thumshirn } else { 3599314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3600f2984462SChris Mason } 3601314b6dd0SJohannes Thumshirn 3602314b6dd0SJohannes Thumshirn put_page(page); 3603314b6dd0SJohannes Thumshirn unlock_page(page); 3604314b6dd0SJohannes Thumshirn } 3605314b6dd0SJohannes Thumshirn 3606314b6dd0SJohannes Thumshirn bio_put(bio); 3607f2984462SChris Mason } 3608f2984462SChris Mason 36098f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3610a05d3c91SQu Wenruo int copy_num, bool drop_cache) 361129c36d72SAnand Jain { 361229c36d72SAnand Jain struct btrfs_super_block *super; 36138f32380dSJohannes Thumshirn struct page *page; 361412659251SNaohiro Aota u64 bytenr, bytenr_orig; 36158f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 361612659251SNaohiro Aota int ret; 361729c36d72SAnand Jain 361812659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 361912659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 362012659251SNaohiro Aota if (ret == -ENOENT) 362112659251SNaohiro Aota return ERR_PTR(-EINVAL); 362212659251SNaohiro Aota else if (ret) 362312659251SNaohiro Aota return ERR_PTR(ret); 362412659251SNaohiro Aota 3625cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 36268f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 362729c36d72SAnand Jain 3628a05d3c91SQu Wenruo if (drop_cache) { 3629a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3630a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3631a05d3c91SQu Wenruo 3632a05d3c91SQu Wenruo /* 3633a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3634a05d3c91SQu Wenruo * always read from the device. 3635a05d3c91SQu Wenruo */ 3636a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3637a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3638a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3639a05d3c91SQu Wenruo } 3640a05d3c91SQu Wenruo 36418f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 36428f32380dSJohannes Thumshirn if (IS_ERR(page)) 36438f32380dSJohannes Thumshirn return ERR_CAST(page); 364429c36d72SAnand Jain 36458f32380dSJohannes Thumshirn super = page_address(page); 364696c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 364796c2e067SAnand Jain btrfs_release_disk_super(super); 364896c2e067SAnand Jain return ERR_PTR(-ENODATA); 364996c2e067SAnand Jain } 365096c2e067SAnand Jain 365112659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 36528f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36538f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 365429c36d72SAnand Jain } 365529c36d72SAnand Jain 36568f32380dSJohannes Thumshirn return super; 365729c36d72SAnand Jain } 365829c36d72SAnand Jain 365929c36d72SAnand Jain 36608f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3661a512bbf8SYan Zheng { 36628f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3663a512bbf8SYan Zheng int i; 3664a512bbf8SYan Zheng u64 transid = 0; 3665a512bbf8SYan Zheng 3666a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3667a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3668a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3669a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3670a512bbf8SYan Zheng */ 3671a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3672a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 36738f32380dSJohannes Thumshirn if (IS_ERR(super)) 3674a512bbf8SYan Zheng continue; 3675a512bbf8SYan Zheng 3676a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 36778f32380dSJohannes Thumshirn if (latest) 36788f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36798f32380dSJohannes Thumshirn 36808f32380dSJohannes Thumshirn latest = super; 3681a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3682a512bbf8SYan Zheng } 3683a512bbf8SYan Zheng } 368492fc03fbSAnand Jain 36858f32380dSJohannes Thumshirn return super; 3686a512bbf8SYan Zheng } 3687a512bbf8SYan Zheng 36884eedeb75SHisashi Hifumi /* 3689abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3690314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 36914eedeb75SHisashi Hifumi * 3692abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3693abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3694abbb3b8eSDavid Sterba * same for write and wait phases. 36954eedeb75SHisashi Hifumi * 3696314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 36974eedeb75SHisashi Hifumi */ 3698a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3699abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3700a512bbf8SYan Zheng { 3701d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3702314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3703d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3704a512bbf8SYan Zheng int i; 3705a512bbf8SYan Zheng int errors = 0; 370612659251SNaohiro Aota int ret; 370712659251SNaohiro Aota u64 bytenr, bytenr_orig; 3708a512bbf8SYan Zheng 3709a512bbf8SYan Zheng if (max_mirrors == 0) 3710a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3711a512bbf8SYan Zheng 3712d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3713d5178578SJohannes Thumshirn 3714a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3715314b6dd0SJohannes Thumshirn struct page *page; 3716314b6dd0SJohannes Thumshirn struct bio *bio; 3717314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3718314b6dd0SJohannes Thumshirn 371912659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 372012659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 372112659251SNaohiro Aota if (ret == -ENOENT) { 372212659251SNaohiro Aota continue; 372312659251SNaohiro Aota } else if (ret < 0) { 372412659251SNaohiro Aota btrfs_err(device->fs_info, 372512659251SNaohiro Aota "couldn't get super block location for mirror %d", 372612659251SNaohiro Aota i); 372712659251SNaohiro Aota errors++; 372812659251SNaohiro Aota continue; 372912659251SNaohiro Aota } 3730935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3731935e5cc9SMiao Xie device->commit_total_bytes) 3732a512bbf8SYan Zheng break; 3733a512bbf8SYan Zheng 373412659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3735a512bbf8SYan Zheng 3736fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3737fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3738fd08001fSEric Biggers sb->csum); 3739a512bbf8SYan Zheng 3740314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3741314b6dd0SJohannes Thumshirn GFP_NOFS); 3742314b6dd0SJohannes Thumshirn if (!page) { 3743fb456252SJeff Mahoney btrfs_err(device->fs_info, 3744314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3745f14d104dSDavid Sterba bytenr); 3746634554dcSJosef Bacik errors++; 3747634554dcSJosef Bacik continue; 3748634554dcSJosef Bacik } 3749634554dcSJosef Bacik 3750314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3751314b6dd0SJohannes Thumshirn get_page(page); 3752a512bbf8SYan Zheng 3753314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3754314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3755a512bbf8SYan Zheng 3756387125fcSChris Mason /* 3757314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3758314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3759314b6dd0SJohannes Thumshirn * checking. 3760387125fcSChris Mason */ 376107888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 376207888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 376307888c66SChristoph Hellwig GFP_NOFS); 3764314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3765314b6dd0SJohannes Thumshirn bio->bi_private = device; 3766314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3767314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3768314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3769314b6dd0SJohannes Thumshirn 3770314b6dd0SJohannes Thumshirn /* 3771314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3772314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3773314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3774314b6dd0SJohannes Thumshirn */ 37751b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3776314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3777314b6dd0SJohannes Thumshirn 377858ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 377958ff51f1SChristoph Hellwig submit_bio(bio); 37808376d9e1SNaohiro Aota 37818376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 37828376d9e1SNaohiro Aota errors++; 3783a512bbf8SYan Zheng } 3784a512bbf8SYan Zheng return errors < i ? 0 : -1; 3785a512bbf8SYan Zheng } 3786a512bbf8SYan Zheng 3787387125fcSChris Mason /* 3788abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3789abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3790abbb3b8eSDavid Sterba * 3791314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3792abbb3b8eSDavid Sterba * date. 3793abbb3b8eSDavid Sterba */ 3794abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3795abbb3b8eSDavid Sterba { 3796abbb3b8eSDavid Sterba int i; 3797abbb3b8eSDavid Sterba int errors = 0; 3798b6a535faSHoward McLauchlan bool primary_failed = false; 379912659251SNaohiro Aota int ret; 3800abbb3b8eSDavid Sterba u64 bytenr; 3801abbb3b8eSDavid Sterba 3802abbb3b8eSDavid Sterba if (max_mirrors == 0) 3803abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3804abbb3b8eSDavid Sterba 3805abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3806314b6dd0SJohannes Thumshirn struct page *page; 3807314b6dd0SJohannes Thumshirn 380812659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 380912659251SNaohiro Aota if (ret == -ENOENT) { 381012659251SNaohiro Aota break; 381112659251SNaohiro Aota } else if (ret < 0) { 381212659251SNaohiro Aota errors++; 381312659251SNaohiro Aota if (i == 0) 381412659251SNaohiro Aota primary_failed = true; 381512659251SNaohiro Aota continue; 381612659251SNaohiro Aota } 3817abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3818abbb3b8eSDavid Sterba device->commit_total_bytes) 3819abbb3b8eSDavid Sterba break; 3820abbb3b8eSDavid Sterba 3821314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3822314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3823314b6dd0SJohannes Thumshirn if (!page) { 3824abbb3b8eSDavid Sterba errors++; 3825b6a535faSHoward McLauchlan if (i == 0) 3826b6a535faSHoward McLauchlan primary_failed = true; 3827abbb3b8eSDavid Sterba continue; 3828abbb3b8eSDavid Sterba } 3829314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3830314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3831314b6dd0SJohannes Thumshirn if (PageError(page)) { 3832abbb3b8eSDavid Sterba errors++; 3833b6a535faSHoward McLauchlan if (i == 0) 3834b6a535faSHoward McLauchlan primary_failed = true; 3835b6a535faSHoward McLauchlan } 3836abbb3b8eSDavid Sterba 3837314b6dd0SJohannes Thumshirn /* Drop our reference */ 3838314b6dd0SJohannes Thumshirn put_page(page); 3839abbb3b8eSDavid Sterba 3840314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3841314b6dd0SJohannes Thumshirn put_page(page); 3842abbb3b8eSDavid Sterba } 3843abbb3b8eSDavid Sterba 3844b6a535faSHoward McLauchlan /* log error, force error return */ 3845b6a535faSHoward McLauchlan if (primary_failed) { 3846b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3847b6a535faSHoward McLauchlan device->devid); 3848b6a535faSHoward McLauchlan return -1; 3849b6a535faSHoward McLauchlan } 3850b6a535faSHoward McLauchlan 3851abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3852abbb3b8eSDavid Sterba } 3853abbb3b8eSDavid Sterba 3854abbb3b8eSDavid Sterba /* 3855387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3856387125fcSChris Mason * for the barrier when it is done 3857387125fcSChris Mason */ 38584246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3859387125fcSChris Mason { 3860f9e69aa9SChristoph Hellwig bio_uninit(bio); 3861387125fcSChris Mason complete(bio->bi_private); 3862387125fcSChris Mason } 3863387125fcSChris Mason 3864387125fcSChris Mason /* 38654fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 38664fc6441aSAnand Jain * done in the waiting counterpart. 3867387125fcSChris Mason */ 38684fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3869387125fcSChris Mason { 3870f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 3871387125fcSChris Mason 3872bfd3ea94SAnand Jain device->last_flush_error = BLK_STS_OK; 3873bfd3ea94SAnand Jain 3874a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3875a91cf0ffSWang Yugui /* 3876a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 3877a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 3878a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 3879a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 3880a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 3881a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 3882a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 3883a91cf0ffSWang Yugui * non-debug builds. 3884a91cf0ffSWang Yugui */ 388508e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 38864fc6441aSAnand Jain return; 3887a91cf0ffSWang Yugui #endif 3888387125fcSChris Mason 3889f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 3890f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 3891387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3892387125fcSChris Mason init_completion(&device->flush_wait); 3893387125fcSChris Mason bio->bi_private = &device->flush_wait; 3894387125fcSChris Mason 389558ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 389658ff51f1SChristoph Hellwig submit_bio(bio); 38971c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 38984fc6441aSAnand Jain } 3899387125fcSChris Mason 39004fc6441aSAnand Jain /* 39014fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 39021b465784SAnand Jain * Return true for any error, and false otherwise. 39034fc6441aSAnand Jain */ 39041b465784SAnand Jain static bool wait_dev_flush(struct btrfs_device *device) 39054fc6441aSAnand Jain { 3906f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 39074fc6441aSAnand Jain 39087e812f20SAnand Jain if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 39091b465784SAnand Jain return false; 39104fc6441aSAnand Jain 39112980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 39124fc6441aSAnand Jain 3913bfd3ea94SAnand Jain if (bio->bi_status) { 3914bfd3ea94SAnand Jain device->last_flush_error = bio->bi_status; 3915bfd3ea94SAnand Jain btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); 39161b465784SAnand Jain return true; 3917bfd3ea94SAnand Jain } 3918bfd3ea94SAnand Jain 39191b465784SAnand Jain return false; 3920387125fcSChris Mason } 3921387125fcSChris Mason 3922387125fcSChris Mason /* 3923387125fcSChris Mason * send an empty flush down to each device in parallel, 3924387125fcSChris Mason * then wait for them 3925387125fcSChris Mason */ 3926387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3927387125fcSChris Mason { 3928387125fcSChris Mason struct list_head *head; 3929387125fcSChris Mason struct btrfs_device *dev; 39305af3e8ccSStefan Behrens int errors_wait = 0; 3931387125fcSChris Mason 39321538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3933387125fcSChris Mason /* send down all the barriers */ 3934387125fcSChris Mason head = &info->fs_devices->devices; 39351538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3936e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3937f88ba6a2SHidetoshi Seto continue; 3938cea7c8bfSAnand Jain if (!dev->bdev) 3939387125fcSChris Mason continue; 3940e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3941ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3942387125fcSChris Mason continue; 3943387125fcSChris Mason 39444fc6441aSAnand Jain write_dev_flush(dev); 3945387125fcSChris Mason } 3946387125fcSChris Mason 3947387125fcSChris Mason /* wait for all the barriers */ 39481538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3949e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3950f88ba6a2SHidetoshi Seto continue; 3951387125fcSChris Mason if (!dev->bdev) { 39525af3e8ccSStefan Behrens errors_wait++; 3953387125fcSChris Mason continue; 3954387125fcSChris Mason } 3955e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3956ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3957387125fcSChris Mason continue; 3958387125fcSChris Mason 39591b465784SAnand Jain if (wait_dev_flush(dev)) 39605af3e8ccSStefan Behrens errors_wait++; 3961387125fcSChris Mason } 3962401b41e5SAnand Jain 3963401b41e5SAnand Jain /* 3964de38a206SAnand Jain * Checks last_flush_error of disks in order to determine the device 3965de38a206SAnand Jain * state. 3966401b41e5SAnand Jain */ 3967de38a206SAnand Jain if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) 3968de38a206SAnand Jain return -EIO; 3969de38a206SAnand Jain 3970387125fcSChris Mason return 0; 3971387125fcSChris Mason } 3972387125fcSChris Mason 3973943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3974943c6e99SZhao Lei { 39758789f4feSZhao Lei int raid_type; 39768789f4feSZhao Lei int min_tolerated = INT_MAX; 3977943c6e99SZhao Lei 39788789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 39798789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 39808c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 39818789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 39828789f4feSZhao Lei tolerated_failures); 3983943c6e99SZhao Lei 39848789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 39858789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 39868789f4feSZhao Lei continue; 398741a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 39888789f4feSZhao Lei continue; 39898c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 39908789f4feSZhao Lei btrfs_raid_array[raid_type]. 39918789f4feSZhao Lei tolerated_failures); 39928789f4feSZhao Lei } 3993943c6e99SZhao Lei 39948789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3995ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 39968789f4feSZhao Lei min_tolerated = 0; 39978789f4feSZhao Lei } 39988789f4feSZhao Lei 39998789f4feSZhao Lei return min_tolerated; 4000943c6e99SZhao Lei } 4001943c6e99SZhao Lei 4002eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4003f2984462SChris Mason { 4004e5e9a520SChris Mason struct list_head *head; 4005f2984462SChris Mason struct btrfs_device *dev; 4006a061fc8dSChris Mason struct btrfs_super_block *sb; 4007f2984462SChris Mason struct btrfs_dev_item *dev_item; 4008f2984462SChris Mason int ret; 4009f2984462SChris Mason int do_barriers; 4010a236aed1SChris Mason int max_errors; 4011a236aed1SChris Mason int total_errors = 0; 4012a061fc8dSChris Mason u64 flags; 4013f2984462SChris Mason 40140b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4015fed3b381SLiu Bo 4016fed3b381SLiu Bo /* 4017fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4018fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4019fed3b381SLiu Bo * been consistent on disk. 4020fed3b381SLiu Bo */ 4021fed3b381SLiu Bo if (max_mirrors == 0) 40220b246afaSJeff Mahoney backup_super_roots(fs_info); 4023f2984462SChris Mason 40240b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4025a061fc8dSChris Mason dev_item = &sb->dev_item; 4026e5e9a520SChris Mason 40270b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 40280b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 40290b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4030387125fcSChris Mason 40315af3e8ccSStefan Behrens if (do_barriers) { 40320b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 40335af3e8ccSStefan Behrens if (ret) { 40345af3e8ccSStefan Behrens mutex_unlock( 40350b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 40360b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 40375af3e8ccSStefan Behrens "errors while submitting device barriers."); 40385af3e8ccSStefan Behrens return ret; 40395af3e8ccSStefan Behrens } 40405af3e8ccSStefan Behrens } 4041387125fcSChris Mason 40421538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4043dfe25020SChris Mason if (!dev->bdev) { 4044dfe25020SChris Mason total_errors++; 4045dfe25020SChris Mason continue; 4046dfe25020SChris Mason } 4047e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4048ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4049dfe25020SChris Mason continue; 4050dfe25020SChris Mason 40512b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4052a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4053a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 40547df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4055935e5cc9SMiao Xie dev->commit_total_bytes); 4056ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4057ce7213c7SMiao Xie dev->commit_bytes_used); 4058a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4059a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4060a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4061a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 40627239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 40637239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4064a512bbf8SYan Zheng 4065a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4066a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4067f2984462SChris Mason 406875cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 406975cb857dSQu Wenruo if (ret < 0) { 407075cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 407175cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 407275cb857dSQu Wenruo "unexpected superblock corruption detected"); 407375cb857dSQu Wenruo return -EUCLEAN; 407475cb857dSQu Wenruo } 407575cb857dSQu Wenruo 4076abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4077a236aed1SChris Mason if (ret) 4078a236aed1SChris Mason total_errors++; 4079f2984462SChris Mason } 4080a236aed1SChris Mason if (total_errors > max_errors) { 40810b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4082d397712bSChris Mason total_errors); 40830b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 408479787eaaSJeff Mahoney 40859d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 40860b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 40870b246afaSJeff Mahoney "%d errors while writing supers", 40880b246afaSJeff Mahoney total_errors); 40899d565ba4SStefan Behrens return -EIO; 4090a236aed1SChris Mason } 4091f2984462SChris Mason 4092a512bbf8SYan Zheng total_errors = 0; 40931538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4094dfe25020SChris Mason if (!dev->bdev) 4095dfe25020SChris Mason continue; 4096e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4097ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4098dfe25020SChris Mason continue; 4099dfe25020SChris Mason 4100abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4101a512bbf8SYan Zheng if (ret) 41021259ab75SChris Mason total_errors++; 4103f2984462SChris Mason } 41040b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4105a236aed1SChris Mason if (total_errors > max_errors) { 41060b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 41070b246afaSJeff Mahoney "%d errors while writing supers", 41080b246afaSJeff Mahoney total_errors); 410979787eaaSJeff Mahoney return -EIO; 4110a236aed1SChris Mason } 4111f2984462SChris Mason return 0; 4112f2984462SChris Mason } 4113f2984462SChris Mason 4114cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4115cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4116cb517eabSMiao Xie struct btrfs_root *root) 41172619ba1fSChris Mason { 41184785e24fSJosef Bacik bool drop_ref = false; 41194785e24fSJosef Bacik 4120fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4121fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4122fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4123fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 41244785e24fSJosef Bacik drop_ref = true; 4125fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 412676dda93cSYan, Zheng 412784961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4128ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 41291c1ea4f7SLiu Bo if (root->reloc_root) { 413000246528SJosef Bacik btrfs_put_root(root->reloc_root); 41311c1ea4f7SLiu Bo root->reloc_root = NULL; 41321c1ea4f7SLiu Bo } 41331c1ea4f7SLiu Bo } 41343321719eSLiu Bo 41354785e24fSJosef Bacik if (drop_ref) 413600246528SJosef Bacik btrfs_put_root(root); 41372619ba1fSChris Mason } 41382619ba1fSChris Mason 4139c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4140c146afadSYan Zheng { 4141fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4142fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4143fc7cbcd4SDavid Sterba int i = 0; 414465d33fd7SQu Wenruo int err = 0; 4145fc7cbcd4SDavid Sterba unsigned int ret = 0; 4146c146afadSYan Zheng 4147c146afadSYan Zheng while (1) { 4148fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4149fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4150fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4151fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4152fc7cbcd4SDavid Sterba if (!ret) { 4153fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4154c146afadSYan Zheng break; 415565d33fd7SQu Wenruo } 4156fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 415766b4ffd1SJosef Bacik 4158fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4159fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4160fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4161fc7cbcd4SDavid Sterba gang[i] = NULL; 416265d33fd7SQu Wenruo continue; 4163c146afadSYan Zheng } 4164fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4165fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4166fc7cbcd4SDavid Sterba } 4167fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4168fc7cbcd4SDavid Sterba 4169fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4170fc7cbcd4SDavid Sterba if (!gang[i]) 4171fc7cbcd4SDavid Sterba continue; 4172fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4173fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4174fc7cbcd4SDavid Sterba if (err) 41756989627dSJosef Bacik goto out; 4176fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4177fc7cbcd4SDavid Sterba } 4178fc7cbcd4SDavid Sterba root_objectid++; 4179c146afadSYan Zheng } 41806989627dSJosef Bacik out: 4181fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4182fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4183fc7cbcd4SDavid Sterba if (gang[i]) 4184fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 418565d33fd7SQu Wenruo } 418665d33fd7SQu Wenruo return err; 4187c146afadSYan Zheng } 4188c146afadSYan Zheng 41896bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4190c146afadSYan Zheng { 41916bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4192c146afadSYan Zheng struct btrfs_trans_handle *trans; 4193c146afadSYan Zheng 41940b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 41952ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 41960b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 41970b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4198c71bf099SYan, Zheng 4199c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 42000b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 42010b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4202c71bf099SYan, Zheng 42037a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 42043612b495STsutomu Itoh if (IS_ERR(trans)) 42053612b495STsutomu Itoh return PTR_ERR(trans); 42063a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4207c146afadSYan Zheng } 4208c146afadSYan Zheng 420936c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 421036c86a9eSQu Wenruo { 421136c86a9eSQu Wenruo struct btrfs_transaction *trans; 421236c86a9eSQu Wenruo struct btrfs_transaction *tmp; 421336c86a9eSQu Wenruo bool found = false; 421436c86a9eSQu Wenruo 421536c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 421636c86a9eSQu Wenruo return; 421736c86a9eSQu Wenruo 421836c86a9eSQu Wenruo /* 421936c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 422036c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 422136c86a9eSQu Wenruo */ 422236c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 422336c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 422436c86a9eSQu Wenruo struct extent_state *cached = NULL; 422536c86a9eSQu Wenruo u64 dirty_bytes = 0; 422636c86a9eSQu Wenruo u64 cur = 0; 422736c86a9eSQu Wenruo u64 found_start; 422836c86a9eSQu Wenruo u64 found_end; 422936c86a9eSQu Wenruo 423036c86a9eSQu Wenruo found = true; 4231e5860f82SFilipe Manana while (find_first_extent_bit(&trans->dirty_pages, cur, 423236c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 423336c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 423436c86a9eSQu Wenruo cur = found_end + 1; 423536c86a9eSQu Wenruo } 423636c86a9eSQu Wenruo btrfs_warn(fs_info, 423736c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 423836c86a9eSQu Wenruo trans->transid, dirty_bytes); 423936c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 424036c86a9eSQu Wenruo 424136c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 424236c86a9eSQu Wenruo fs_info->running_transaction = NULL; 424336c86a9eSQu Wenruo list_del_init(&trans->list); 424436c86a9eSQu Wenruo 424536c86a9eSQu Wenruo btrfs_put_transaction(trans); 424636c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 424736c86a9eSQu Wenruo } 424836c86a9eSQu Wenruo ASSERT(!found); 424936c86a9eSQu Wenruo } 425036c86a9eSQu Wenruo 4251b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4252eb60ceacSChris Mason { 4253c146afadSYan Zheng int ret; 4254e089f05cSChris Mason 4255afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 425631e70e52SFilipe Manana 425731e70e52SFilipe Manana /* 42588a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 42598a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 42608a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 42618a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 42628a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 42638a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 42648a1f1e3dSFilipe Manana * return 1. 42658a1f1e3dSFilipe Manana */ 42668a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 42678a1f1e3dSFilipe Manana 42688a1f1e3dSFilipe Manana /* 426931e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 427031e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 427131e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 427231e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 427331e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 427431e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 427531e70e52SFilipe Manana * trying to stop the async reclaim task. 427631e70e52SFilipe Manana */ 427731e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4278d6fd0ae2SOmar Sandoval /* 4279d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4280d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4281d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4282d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4283d6fd0ae2SOmar Sandoval */ 4284d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4285a2135011SChris Mason 42867343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4287d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 42887343dd61SJustin Maggard 4289803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4290803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4291803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4292803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4293803b2f54SStefan Behrens 4294837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4295aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4296837d5b6eSIlya Dryomov 42978dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 42988dabb742SStefan Behrens 4299aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 43004cb5300bSChris Mason 43014cb5300bSChris Mason /* wait for any defraggers to finish */ 43024cb5300bSChris Mason wait_event(fs_info->transaction_wait, 43034cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 43044cb5300bSChris Mason 43054cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 430626176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 43074cb5300bSChris Mason 4308a362bb86SFilipe Manana /* 4309a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4310a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4311a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4312a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4313a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4314a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4315a362bb86SFilipe Manana * no one else to run iputs. 4316a362bb86SFilipe Manana * 4317a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4318a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4319a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4320a362bb86SFilipe Manana * through some other means. 4321a362bb86SFilipe Manana * 4322a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4323a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4324a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4325a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4326a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4327a362bb86SFilipe Manana */ 4328a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4329a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4330a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4331a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4332a362bb86SFilipe Manana 433321c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 433457056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4335576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 433621c7e756SMiao Xie 4337b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4338b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4339b0643e59SDennis Zhou 4340bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4341e44163e1SJeff Mahoney /* 4342d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4343d6fd0ae2SOmar Sandoval * unused block groups. 4344e44163e1SJeff Mahoney */ 43450b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4346e44163e1SJeff Mahoney 4347f0cc2cd7SFilipe Manana /* 4348f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4349f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4350f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4351f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4352f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4353f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4354f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4355f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4356f0cc2cd7SFilipe Manana * This is a very rare case. 4357f0cc2cd7SFilipe Manana */ 4358f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4359f0cc2cd7SFilipe Manana 43606bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4361d397712bSChris Mason if (ret) 436204892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4363c146afadSYan Zheng } 4364ed2ff2cbSChris Mason 436584961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 43662ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4367acce952bSliubo 4368e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4369e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 43708929ecfaSYan, Zheng 4371e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4372afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4373f25784b3SYan Zheng 43745958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 43755958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 43765958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 43775958253cSQu Wenruo } 43785958253cSQu Wenruo 437904892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4380fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4381bcef60f2SArne Jansen 4382963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 438304892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4384963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4385b0c68f8bSChris Mason } 438631153d81SYan Zheng 43875deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 43884297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 43895deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 43904297ff84SJosef Bacik 43916618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4392b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 43935ac1d209SJeff Mahoney 43941a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 43951a4319ccSLiu Bo 4396de348ee0SWang Shilong /* 4397de348ee0SWang Shilong * we must make sure there is not any read request to 4398de348ee0SWang Shilong * submit after we stopping all workers. 4399de348ee0SWang Shilong */ 4400de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 440196192499SJosef Bacik btrfs_stop_all_workers(fs_info); 440296192499SJosef Bacik 44030a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 440436c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 44050a31daa4SFilipe Manana 4406afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 44074273eaffSAnand Jain free_root_pointers(fs_info, true); 44088c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 44099ad6b7bcSChris Mason 44104e19443dSJosef Bacik /* 44114e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 44124e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 44134e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 44144e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 44154e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 44164e19443dSJosef Bacik */ 44174e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 44184e19443dSJosef Bacik 441913e6c37bSJosef Bacik iput(fs_info->btree_inode); 4420d6bfde87SChris Mason 442121adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 44220b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 44232ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 442421adbd5cSStefan Behrens #endif 442521adbd5cSStefan Behrens 44260b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 442768c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4428eb60ceacSChris Mason } 4429eb60ceacSChris Mason 44305f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 44315f39d397SChris Mason { 44322f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 44335f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4434b4ce94deSChris Mason 443506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 443606ea65a3SJosef Bacik /* 443706ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 443852042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 443906ea65a3SJosef Bacik * outside of the sanity tests. 444006ea65a3SJosef Bacik */ 4441b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 444206ea65a3SJosef Bacik return; 444306ea65a3SJosef Bacik #endif 444449d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 44450b246afaSJeff Mahoney if (transid != fs_info->generation) 44465d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 44470b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 4448f18cc978SChristoph Hellwig set_extent_buffer_dirty(buf); 44491f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 445069fc6cbbSQu Wenruo /* 445185d8a826SJosef Bacik * btrfs_check_leaf() won't check item data if we don't have WRITTEN 445285d8a826SJosef Bacik * set, so this will only validate the basic structure of the items. 445369fc6cbbSQu Wenruo */ 445485d8a826SJosef Bacik if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(buf)) { 4455a4f78750SDavid Sterba btrfs_print_leaf(buf); 44561f21ef0aSFilipe Manana ASSERT(0); 44571f21ef0aSFilipe Manana } 44581f21ef0aSFilipe Manana #endif 4459eb60ceacSChris Mason } 4460eb60ceacSChris Mason 44612ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4462b53d3f5dSLiu Bo int flush_delayed) 446335b7e476SChris Mason { 4464188de649SChris Mason /* 4465188de649SChris Mason * looks as though older kernels can get into trouble with 4466188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4467188de649SChris Mason */ 4468e2d84521SMiao Xie int ret; 4469d6bfde87SChris Mason 44706933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4471d6bfde87SChris Mason return; 4472d6bfde87SChris Mason 4473b53d3f5dSLiu Bo if (flush_delayed) 44742ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 447516cdcec7SMiao Xie 4476d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4477d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4478d814a491SEthan Lien fs_info->dirty_metadata_batch); 4479e2d84521SMiao Xie if (ret > 0) { 44800b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 448116cdcec7SMiao Xie } 448216cdcec7SMiao Xie } 448316cdcec7SMiao Xie 44842ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 448516cdcec7SMiao Xie { 44862ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 448735b7e476SChris Mason } 4488b53d3f5dSLiu Bo 44892ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4490b53d3f5dSLiu Bo { 44912ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4492d6bfde87SChris Mason } 44936b80053dSChris Mason 44942ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4495acce952bSliubo { 4496fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4497fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4498fe816d0fSNikolay Borisov 44990b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45002ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45010b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4502acce952bSliubo 45030b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45040b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4505acce952bSliubo } 4506acce952bSliubo 4507ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4508ef67963dSJosef Bacik { 4509fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4510fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4511fc7cbcd4SDavid Sterba int ret; 4512ef67963dSJosef Bacik 4513fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4514fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4515fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4516fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4517fc7cbcd4SDavid Sterba int i; 4518ef67963dSJosef Bacik 4519fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4520fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4521fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4522fc7cbcd4SDavid Sterba 4523fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4524fc7cbcd4SDavid Sterba if (!gang[i]) 4525ef67963dSJosef Bacik continue; 4526fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4527fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4528fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4529ef67963dSJosef Bacik } 4530fc7cbcd4SDavid Sterba root_objectid++; 4531fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4532ef67963dSJosef Bacik } 4533fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4534ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4535ef67963dSJosef Bacik } 4536ef67963dSJosef Bacik 4537143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4538acce952bSliubo { 4539acce952bSliubo struct btrfs_ordered_extent *ordered; 4540acce952bSliubo 4541199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4542779880efSJosef Bacik /* 4543779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4544779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4545779880efSJosef Bacik */ 4546199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4547779880efSJosef Bacik root_extent_list) 4548779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4549199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4550199c2a9cSMiao Xie } 4551199c2a9cSMiao Xie 4552199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4553199c2a9cSMiao Xie { 4554199c2a9cSMiao Xie struct btrfs_root *root; 4555199c2a9cSMiao Xie struct list_head splice; 4556199c2a9cSMiao Xie 4557199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4558199c2a9cSMiao Xie 4559199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4560199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4561199c2a9cSMiao Xie while (!list_empty(&splice)) { 4562199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4563199c2a9cSMiao Xie ordered_root); 45641de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 45651de2cfdeSJosef Bacik &fs_info->ordered_roots); 4566199c2a9cSMiao Xie 45672a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4568199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4569199c2a9cSMiao Xie 45702a85d9caSLiu Bo cond_resched(); 45712a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4572199c2a9cSMiao Xie } 4573199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 457474d5d229SJosef Bacik 457574d5d229SJosef Bacik /* 457674d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 457774d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 457874d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 457974d5d229SJosef Bacik * actually get run and error out properly. 458074d5d229SJosef Bacik */ 458174d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4582acce952bSliubo } 4583acce952bSliubo 458499f09ce3SFilipe Manana static void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 45852ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4586acce952bSliubo { 4587acce952bSliubo struct rb_node *node; 4588acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4589acce952bSliubo struct btrfs_delayed_ref_node *ref; 4590acce952bSliubo 4591acce952bSliubo delayed_refs = &trans->delayed_refs; 4592acce952bSliubo 4593acce952bSliubo spin_lock(&delayed_refs->lock); 4594d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4595cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4596b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 459799f09ce3SFilipe Manana return; 4598acce952bSliubo } 4599acce952bSliubo 46005c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4601d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 46020e0adbcfSJosef Bacik struct rb_node *n; 4603e78417d1SJosef Bacik bool pin_bytes = false; 4604acce952bSliubo 4605d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4606d7df2c79SJosef Bacik href_node); 46073069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4608b939d1abSJosef Bacik continue; 46093069bd26SJosef Bacik 4610d7df2c79SJosef Bacik spin_lock(&head->lock); 4611e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 46120e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 46130e0adbcfSJosef Bacik ref_node); 4614e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 46150e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 46161d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 46171d57ee94SWang Xiaoguang list_del(&ref->add_list); 4618d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4619d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4620d7df2c79SJosef Bacik } 462154067ae9SJosef Bacik if (head->must_insert_reserved) 4622e78417d1SJosef Bacik pin_bytes = true; 462378a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4624fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4625d7df2c79SJosef Bacik spin_unlock(&head->lock); 4626acce952bSliubo spin_unlock(&delayed_refs->lock); 4627e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4628acce952bSliubo 4629f603bb94SNikolay Borisov if (pin_bytes) { 4630f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4631f603bb94SNikolay Borisov 4632f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4633f603bb94SNikolay Borisov BUG_ON(!cache); 4634f603bb94SNikolay Borisov 4635f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4636f603bb94SNikolay Borisov spin_lock(&cache->lock); 4637f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4638f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4639f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4640f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4641f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4642f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4643f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4644f603bb94SNikolay Borisov 4645f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4646f603bb94SNikolay Borisov 4647f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4648f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4649f603bb94SNikolay Borisov } 465031890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4651d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4652acce952bSliubo cond_resched(); 4653acce952bSliubo spin_lock(&delayed_refs->lock); 4654acce952bSliubo } 465581f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4656acce952bSliubo 4657acce952bSliubo spin_unlock(&delayed_refs->lock); 4658acce952bSliubo } 4659acce952bSliubo 4660143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4661acce952bSliubo { 4662acce952bSliubo struct btrfs_inode *btrfs_inode; 4663acce952bSliubo struct list_head splice; 4664acce952bSliubo 4665acce952bSliubo INIT_LIST_HEAD(&splice); 4666acce952bSliubo 4667eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4668eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4669acce952bSliubo 4670acce952bSliubo while (!list_empty(&splice)) { 4671fe816d0fSNikolay Borisov struct inode *inode = NULL; 4672eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4673acce952bSliubo delalloc_inodes); 4674fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4675eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4676acce952bSliubo 4677fe816d0fSNikolay Borisov /* 4678fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4679fe816d0fSNikolay Borisov * meanwhile. 4680fe816d0fSNikolay Borisov */ 4681fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4682fe816d0fSNikolay Borisov if (inode) { 4683597441b3SJosef Bacik unsigned int nofs_flag; 4684597441b3SJosef Bacik 4685597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4686fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4687597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4688fe816d0fSNikolay Borisov iput(inode); 4689fe816d0fSNikolay Borisov } 4690eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4691acce952bSliubo } 4692eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4693eb73c1b7SMiao Xie } 4694eb73c1b7SMiao Xie 4695eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4696eb73c1b7SMiao Xie { 4697eb73c1b7SMiao Xie struct btrfs_root *root; 4698eb73c1b7SMiao Xie struct list_head splice; 4699eb73c1b7SMiao Xie 4700eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4701eb73c1b7SMiao Xie 4702eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4703eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4704eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4705eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4706eb73c1b7SMiao Xie delalloc_root); 470700246528SJosef Bacik root = btrfs_grab_root(root); 4708eb73c1b7SMiao Xie BUG_ON(!root); 4709eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4710eb73c1b7SMiao Xie 4711eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 471200246528SJosef Bacik btrfs_put_root(root); 4713eb73c1b7SMiao Xie 4714eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4715eb73c1b7SMiao Xie } 4716eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4717acce952bSliubo } 4718acce952bSliubo 4719aec5716cSFilipe Manana static void btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4720acce952bSliubo struct extent_io_tree *dirty_pages, 4721acce952bSliubo int mark) 4722acce952bSliubo { 4723acce952bSliubo struct extent_buffer *eb; 4724acce952bSliubo u64 start = 0; 4725acce952bSliubo u64 end; 4726acce952bSliubo 4727e5860f82SFilipe Manana while (find_first_extent_bit(dirty_pages, start, &start, &end, 4728aec5716cSFilipe Manana mark, NULL)) { 472991166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4730acce952bSliubo while (start <= end) { 47310b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 47320b246afaSJeff Mahoney start += fs_info->nodesize; 4733fd8b2b61SJosef Bacik if (!eb) 4734acce952bSliubo continue; 4735acce952bSliubo 4736c4e54a65SJosef Bacik btrfs_tree_lock(eb); 4737c4e54a65SJosef Bacik wait_on_extent_buffer_writeback(eb); 4738190a8339SJosef Bacik btrfs_clear_buffer_dirty(NULL, eb); 4739c4e54a65SJosef Bacik btrfs_tree_unlock(eb); 4740c4e54a65SJosef Bacik 4741fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4742acce952bSliubo } 4743acce952bSliubo } 4744acce952bSliubo } 4745acce952bSliubo 474646d81ebdSFilipe Manana static void btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4747fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4748acce952bSliubo { 4749acce952bSliubo u64 start; 4750acce952bSliubo u64 end; 4751acce952bSliubo 4752acce952bSliubo while (1) { 47530e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 47540e6ec385SFilipe Manana 4755fcd5e742SLu Fengqi /* 4756fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4757fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4758fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4759fcd5e742SLu Fengqi * the same extent range. 4760fcd5e742SLu Fengqi */ 4761fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4762e5860f82SFilipe Manana if (!find_first_extent_bit(unpin, 0, &start, &end, 476346d81ebdSFilipe Manana EXTENT_DIRTY, &cached_state)) { 4764fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4765acce952bSliubo break; 4766fcd5e742SLu Fengqi } 4767acce952bSliubo 47680e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 47690e6ec385SFilipe Manana free_extent_state(cached_state); 47702ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4771fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4772acce952bSliubo cond_resched(); 4773acce952bSliubo } 4774acce952bSliubo } 4775acce952bSliubo 477632da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4777c79a1751SLiu Bo { 4778c79a1751SLiu Bo struct inode *inode; 4779c79a1751SLiu Bo 4780c79a1751SLiu Bo inode = cache->io_ctl.inode; 4781c79a1751SLiu Bo if (inode) { 4782597441b3SJosef Bacik unsigned int nofs_flag; 4783597441b3SJosef Bacik 4784597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4785c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4786597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4787597441b3SJosef Bacik 4788c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4789c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4790c79a1751SLiu Bo iput(inode); 4791c79a1751SLiu Bo } 4792bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4793c79a1751SLiu Bo btrfs_put_block_group(cache); 4794c79a1751SLiu Bo } 4795c79a1751SLiu Bo 4796c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 47972ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4798c79a1751SLiu Bo { 479932da5386SDavid Sterba struct btrfs_block_group *cache; 4800c79a1751SLiu Bo 4801c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4802c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4803c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 480432da5386SDavid Sterba struct btrfs_block_group, 4805c79a1751SLiu Bo dirty_list); 4806c79a1751SLiu Bo 4807c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4808c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4809c79a1751SLiu Bo list_del_init(&cache->io_list); 4810c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4811c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4812c79a1751SLiu Bo } 4813c79a1751SLiu Bo 4814c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4815c79a1751SLiu Bo spin_lock(&cache->lock); 4816c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4817c79a1751SLiu Bo spin_unlock(&cache->lock); 4818c79a1751SLiu Bo 4819c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4820c79a1751SLiu Bo btrfs_put_block_group(cache); 4821ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4822c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4823c79a1751SLiu Bo } 4824c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4825c79a1751SLiu Bo 482645ae2c18SNikolay Borisov /* 482745ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 482845ae2c18SNikolay Borisov * to use it without any locking 482945ae2c18SNikolay Borisov */ 4830c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4831c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 483232da5386SDavid Sterba struct btrfs_block_group, 4833c79a1751SLiu Bo io_list); 4834c79a1751SLiu Bo 4835c79a1751SLiu Bo list_del_init(&cache->io_list); 4836c79a1751SLiu Bo spin_lock(&cache->lock); 4837c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4838c79a1751SLiu Bo spin_unlock(&cache->lock); 4839c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4840c79a1751SLiu Bo } 4841c79a1751SLiu Bo } 4842c79a1751SLiu Bo 484349b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 48442ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 484549b25e05SJeff Mahoney { 4846bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4847bbbf7243SNikolay Borisov 48482ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4849c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4850c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4851c79a1751SLiu Bo 4852bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4853bbbf7243SNikolay Borisov post_commit_list) { 4854bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4855bbbf7243SNikolay Borisov } 4856bbbf7243SNikolay Borisov 48572ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 485849b25e05SJeff Mahoney 48594a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 48600b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 486149b25e05SJeff Mahoney 48624a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 48630b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 486449b25e05SJeff Mahoney 4865ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 486667cde344SMiao Xie 48672ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 486849b25e05SJeff Mahoney EXTENT_DIRTY); 4869fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 487049b25e05SJeff Mahoney 48714a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 48724a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 487349b25e05SJeff Mahoney } 487449b25e05SJeff Mahoney 48752ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4876acce952bSliubo { 4877acce952bSliubo struct btrfs_transaction *t; 4878acce952bSliubo 48790b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4880acce952bSliubo 48810b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 48820b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 48830b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4884724e2315SJosef Bacik struct btrfs_transaction, list); 4885724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 48869b64f57dSElena Reshetova refcount_inc(&t->use_count); 48870b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 48882ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4889724e2315SJosef Bacik btrfs_put_transaction(t); 48900b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4891724e2315SJosef Bacik continue; 4892724e2315SJosef Bacik } 48930b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4894724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 48950b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4896724e2315SJosef Bacik /* 4897724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4898724e2315SJosef Bacik * handle open currently for this transaction. 4899724e2315SJosef Bacik */ 4900724e2315SJosef Bacik wait_event(t->writer_wait, 4901724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4902724e2315SJosef Bacik } else { 49030b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4904724e2315SJosef Bacik } 49052ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4906724e2315SJosef Bacik 49070b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49080b246afaSJeff Mahoney if (t == fs_info->running_transaction) 49090b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4910724e2315SJosef Bacik list_del_init(&t->list); 49110b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4912a4abeea4SJosef Bacik 4913724e2315SJosef Bacik btrfs_put_transaction(t); 49142e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 49150b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4916724e2315SJosef Bacik } 49170b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49180b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4919ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4920ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 49210b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4922ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 49230b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4924acce952bSliubo 4925acce952bSliubo return 0; 4926acce952bSliubo } 4927ec7d6dfdSNikolay Borisov 4928453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 4929ec7d6dfdSNikolay Borisov { 4930ec7d6dfdSNikolay Borisov struct btrfs_path *path; 4931ec7d6dfdSNikolay Borisov int ret; 4932ec7d6dfdSNikolay Borisov struct extent_buffer *l; 4933ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 4934ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 4935ec7d6dfdSNikolay Borisov int slot; 4936ec7d6dfdSNikolay Borisov 4937ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 4938ec7d6dfdSNikolay Borisov if (!path) 4939ec7d6dfdSNikolay Borisov return -ENOMEM; 4940ec7d6dfdSNikolay Borisov 4941ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 4942ec7d6dfdSNikolay Borisov search_key.type = -1; 4943ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 4944ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 4945ec7d6dfdSNikolay Borisov if (ret < 0) 4946ec7d6dfdSNikolay Borisov goto error; 4947ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 4948ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 4949ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 4950ec7d6dfdSNikolay Borisov l = path->nodes[0]; 4951ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 495223125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 495323125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 4954ec7d6dfdSNikolay Borisov } else { 495523125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 4956ec7d6dfdSNikolay Borisov } 4957ec7d6dfdSNikolay Borisov ret = 0; 4958ec7d6dfdSNikolay Borisov error: 4959ec7d6dfdSNikolay Borisov btrfs_free_path(path); 4960ec7d6dfdSNikolay Borisov return ret; 4961ec7d6dfdSNikolay Borisov } 4962ec7d6dfdSNikolay Borisov 4963543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 4964ec7d6dfdSNikolay Borisov { 4965ec7d6dfdSNikolay Borisov int ret; 4966ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 4967ec7d6dfdSNikolay Borisov 49686b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 4969ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 4970ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 4971ec7d6dfdSNikolay Borisov root->root_key.objectid); 4972ec7d6dfdSNikolay Borisov ret = -ENOSPC; 4973ec7d6dfdSNikolay Borisov goto out; 4974ec7d6dfdSNikolay Borisov } 4975ec7d6dfdSNikolay Borisov 497623125104SNikolay Borisov *objectid = root->free_objectid++; 4977ec7d6dfdSNikolay Borisov ret = 0; 4978ec7d6dfdSNikolay Borisov out: 4979ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 4980ec7d6dfdSNikolay Borisov return ret; 4981ec7d6dfdSNikolay Borisov } 4982