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 && 110637f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 110792a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1108f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1109f39e4571SJosef Bacik } 1110f39e4571SJosef Bacik 1111851fd730SQu Wenruo /* 1112851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1113851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1114851fd730SQu Wenruo */ 1115851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1116851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 11172dfb1e43SQu Wenruo if (!anon_dev) { 1118cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1119cb517eabSMiao Xie if (ret) 1120876d2cf1SLiu Bo goto fail; 11212dfb1e43SQu Wenruo } else { 11222dfb1e43SQu Wenruo root->anon_dev = anon_dev; 11232dfb1e43SQu Wenruo } 1124851fd730SQu Wenruo } 1125f32e48e9SChandan Rajendra 1126f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1127453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1128f32e48e9SChandan Rajendra if (ret) { 1129f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1130876d2cf1SLiu Bo goto fail; 1131f32e48e9SChandan Rajendra } 1132f32e48e9SChandan Rajendra 11336b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1134f32e48e9SChandan Rajendra 1135f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1136f32e48e9SChandan Rajendra 1137cb517eabSMiao Xie return 0; 1138cb517eabSMiao Xie fail: 113984db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1140cb517eabSMiao Xie return ret; 1141cb517eabSMiao Xie } 1142cb517eabSMiao Xie 1143a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1144cb517eabSMiao Xie u64 root_id) 1145cb517eabSMiao Xie { 1146cb517eabSMiao Xie struct btrfs_root *root; 1147cb517eabSMiao Xie 1148fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1149fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1150fc7cbcd4SDavid Sterba (unsigned long)root_id); 115100246528SJosef Bacik root = btrfs_grab_root(root); 1152fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1153cb517eabSMiao Xie return root; 1154cb517eabSMiao Xie } 1155cb517eabSMiao Xie 115649d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 115749d11beaSJosef Bacik u64 objectid) 115849d11beaSJosef Bacik { 1159abed4aaaSJosef Bacik struct btrfs_key key = { 1160abed4aaaSJosef Bacik .objectid = objectid, 1161abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1162abed4aaaSJosef Bacik .offset = 0, 1163abed4aaaSJosef Bacik }; 1164abed4aaaSJosef Bacik 1165e91909aaSChristoph Hellwig switch (objectid) { 1166e91909aaSChristoph Hellwig case BTRFS_ROOT_TREE_OBJECTID: 116749d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 1168e91909aaSChristoph Hellwig case BTRFS_EXTENT_TREE_OBJECTID: 1169abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1170e91909aaSChristoph Hellwig case BTRFS_CHUNK_TREE_OBJECTID: 117149d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 1172e91909aaSChristoph Hellwig case BTRFS_DEV_TREE_OBJECTID: 117349d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 1174e91909aaSChristoph Hellwig case BTRFS_CSUM_TREE_OBJECTID: 1175abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1176e91909aaSChristoph Hellwig case BTRFS_QUOTA_TREE_OBJECTID: 117785724171SChristoph Hellwig return btrfs_grab_root(fs_info->quota_root); 1178e91909aaSChristoph Hellwig case BTRFS_UUID_TREE_OBJECTID: 117985724171SChristoph Hellwig return btrfs_grab_root(fs_info->uuid_root); 1180e91909aaSChristoph Hellwig case BTRFS_BLOCK_GROUP_TREE_OBJECTID: 118185724171SChristoph Hellwig return btrfs_grab_root(fs_info->block_group_root); 1182e91909aaSChristoph Hellwig case BTRFS_FREE_SPACE_TREE_OBJECTID: 118385724171SChristoph Hellwig return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1184e91909aaSChristoph Hellwig default: 118549d11beaSJosef Bacik return NULL; 118649d11beaSJosef Bacik } 1187e91909aaSChristoph Hellwig } 118849d11beaSJosef Bacik 1189cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1190cb517eabSMiao Xie struct btrfs_root *root) 1191cb517eabSMiao Xie { 1192cb517eabSMiao Xie int ret; 1193cb517eabSMiao Xie 1194fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1195fc7cbcd4SDavid Sterba if (ret) 1196fc7cbcd4SDavid Sterba return ret; 1197fc7cbcd4SDavid Sterba 1198fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1199fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1200fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1201fc7cbcd4SDavid Sterba root); 1202af01d2e5SJosef Bacik if (ret == 0) { 120300246528SJosef Bacik btrfs_grab_root(root); 1204fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1205af01d2e5SJosef Bacik } 1206fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1207fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1208cb517eabSMiao Xie 1209cb517eabSMiao Xie return ret; 1210cb517eabSMiao Xie } 1211cb517eabSMiao Xie 1212bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1213bd647ce3SJosef Bacik { 1214bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1215bd647ce3SJosef Bacik struct btrfs_root *root; 1216bd647ce3SJosef Bacik 1217bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1218457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1219457f1864SJosef Bacik 1220bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1221bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1222457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 122371008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1224bd647ce3SJosef Bacik refcount_read(&root->refs)); 1225bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 122600246528SJosef Bacik btrfs_put_root(root); 122700246528SJosef Bacik btrfs_put_root(root); 1228bd647ce3SJosef Bacik } 1229bd647ce3SJosef Bacik #endif 1230bd647ce3SJosef Bacik } 1231bd647ce3SJosef Bacik 1232abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1233abed4aaaSJosef Bacik { 1234abed4aaaSJosef Bacik struct btrfs_root *root; 1235abed4aaaSJosef Bacik struct rb_node *node; 1236abed4aaaSJosef Bacik 1237abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1238abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1239abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1240abed4aaaSJosef Bacik btrfs_put_root(root); 1241abed4aaaSJosef Bacik } 1242abed4aaaSJosef Bacik } 1243abed4aaaSJosef Bacik 12440d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 12450d4b0463SJosef Bacik { 1246141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1247141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 12485deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1249141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1250141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1251141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1252141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 12530d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 12540d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1255abed4aaaSJosef Bacik free_global_roots(fs_info); 125600246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 125700246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 125800246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 125900246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 126000246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 126100246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1262aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 12639c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1264bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 12653fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 12660d4b0463SJosef Bacik kfree(fs_info->super_copy); 12670d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 12688481dd80SQu Wenruo kfree(fs_info->subpage_info); 12690d4b0463SJosef Bacik kvfree(fs_info); 12700d4b0463SJosef Bacik } 12710d4b0463SJosef Bacik 12720d4b0463SJosef Bacik 12732dfb1e43SQu Wenruo /* 12742dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 12752dfb1e43SQu Wenruo * 12762dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 12772dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 12782dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 12792dfb1e43SQu Wenruo * 12802dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 12812dfb1e43SQu Wenruo * 12822dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 12832dfb1e43SQu Wenruo * same root objectid. 12842dfb1e43SQu Wenruo * 12852dfb1e43SQu Wenruo * @objectid: root id 12862dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 12872dfb1e43SQu Wenruo * pass 0 for new allocation. 12882dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 12892dfb1e43SQu Wenruo * for orphan roots 12902dfb1e43SQu Wenruo */ 12912dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 12922dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 12932dfb1e43SQu Wenruo bool check_ref) 12945eda7b5eSChris Mason { 12955eda7b5eSChris Mason struct btrfs_root *root; 1296381cf658SDavid Sterba struct btrfs_path *path; 12971d4c08e0SDavid Sterba struct btrfs_key key; 12985eda7b5eSChris Mason int ret; 12995eda7b5eSChris Mason 130049d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 130149d11beaSJosef Bacik if (root) 130249d11beaSJosef Bacik return root; 13034df27c4dSYan, Zheng again: 130456e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 130548475471SStefan Behrens if (root) { 13062dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 13072dfb1e43SQu Wenruo ASSERT(!anon_dev); 1308bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 130900246528SJosef Bacik btrfs_put_root(root); 131048475471SStefan Behrens return ERR_PTR(-ENOENT); 1311bc44d7c4SJosef Bacik } 13125eda7b5eSChris Mason return root; 131348475471SStefan Behrens } 13145eda7b5eSChris Mason 131556e9357aSDavid Sterba key.objectid = objectid; 131656e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 131756e9357aSDavid Sterba key.offset = (u64)-1; 131856e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 13195eda7b5eSChris Mason if (IS_ERR(root)) 13205eda7b5eSChris Mason return root; 13213394e160SChris Mason 1322c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1323d68fc57bSYan, Zheng ret = -ENOENT; 1324d68fc57bSYan, Zheng goto fail; 1325d68fc57bSYan, Zheng } 1326d68fc57bSYan, Zheng 13272dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1328cb517eabSMiao Xie if (ret) 1329cb517eabSMiao Xie goto fail; 1330cb517eabSMiao Xie 1331381cf658SDavid Sterba path = btrfs_alloc_path(); 1332381cf658SDavid Sterba if (!path) { 1333381cf658SDavid Sterba ret = -ENOMEM; 1334381cf658SDavid Sterba goto fail; 1335381cf658SDavid Sterba } 13361d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 13371d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 133856e9357aSDavid Sterba key.offset = objectid; 13391d4c08e0SDavid Sterba 13401d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1341381cf658SDavid Sterba btrfs_free_path(path); 1342d68fc57bSYan, Zheng if (ret < 0) 1343d68fc57bSYan, Zheng goto fail; 1344d68fc57bSYan, Zheng if (ret == 0) 134527cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1346d68fc57bSYan, Zheng 1347cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 13480f7d52f4SChris Mason if (ret) { 1349168a2f77SJia-Ju Bai if (ret == -EEXIST) { 135000246528SJosef Bacik btrfs_put_root(root); 13514df27c4dSYan, Zheng goto again; 1352168a2f77SJia-Ju Bai } 13534df27c4dSYan, Zheng goto fail; 13544df27c4dSYan, Zheng } 1355edbd8d4eSChris Mason return root; 13564df27c4dSYan, Zheng fail: 135733fab972SFilipe Manana /* 135833fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1359143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 136033fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 136133fab972SFilipe Manana * and once again by our caller. 136233fab972SFilipe Manana */ 136333fab972SFilipe Manana if (anon_dev) 136433fab972SFilipe Manana root->anon_dev = 0; 13658c38938cSJosef Bacik btrfs_put_root(root); 13664df27c4dSYan, Zheng return ERR_PTR(ret); 1367edbd8d4eSChris Mason } 1368edbd8d4eSChris Mason 13692dfb1e43SQu Wenruo /* 13702dfb1e43SQu Wenruo * Get in-memory reference of a root structure 13712dfb1e43SQu Wenruo * 13722dfb1e43SQu Wenruo * @objectid: tree objectid 13732dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 13742dfb1e43SQu Wenruo * one reference 13752dfb1e43SQu Wenruo */ 13762dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 13772dfb1e43SQu Wenruo u64 objectid, bool check_ref) 13782dfb1e43SQu Wenruo { 13792dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 13802dfb1e43SQu Wenruo } 13812dfb1e43SQu Wenruo 13822dfb1e43SQu Wenruo /* 13832dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 13842dfb1e43SQu Wenruo * the anonymous block device id 13852dfb1e43SQu Wenruo * 13862dfb1e43SQu Wenruo * @objectid: tree objectid 13872dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 13882dfb1e43SQu Wenruo * parameter value 13892dfb1e43SQu Wenruo */ 13902dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 13912dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 13922dfb1e43SQu Wenruo { 13932dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 13942dfb1e43SQu Wenruo } 13952dfb1e43SQu Wenruo 13968b712842SChris Mason /* 139749d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 139849d11beaSJosef Bacik * @fs_info: the fs_info 139949d11beaSJosef Bacik * @objectid: the objectid we need to lookup 140049d11beaSJosef Bacik * 140149d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 140249d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 140349d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 140449d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 140549d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 140649d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 140749d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 140849d11beaSJosef Bacik * backref code is finished using the root. 140949d11beaSJosef Bacik */ 141049d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 141149d11beaSJosef Bacik struct btrfs_path *path, 141249d11beaSJosef Bacik u64 objectid) 141349d11beaSJosef Bacik { 141449d11beaSJosef Bacik struct btrfs_root *root; 141549d11beaSJosef Bacik struct btrfs_key key; 141649d11beaSJosef Bacik 141749d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 141849d11beaSJosef Bacik 141949d11beaSJosef Bacik /* 142049d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 142149d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 142249d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 142349d11beaSJosef Bacik * != NULL just return it. 142449d11beaSJosef Bacik */ 142549d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 142649d11beaSJosef Bacik if (root) 142749d11beaSJosef Bacik return root; 142849d11beaSJosef Bacik 142949d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 143049d11beaSJosef Bacik if (root) 143149d11beaSJosef Bacik return root; 143249d11beaSJosef Bacik 143349d11beaSJosef Bacik key.objectid = objectid; 143449d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 143549d11beaSJosef Bacik key.offset = (u64)-1; 143649d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 143749d11beaSJosef Bacik btrfs_release_path(path); 143849d11beaSJosef Bacik 143949d11beaSJosef Bacik return root; 144049d11beaSJosef Bacik } 144149d11beaSJosef Bacik 1442a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1443a74a4b97SChris Mason { 14440d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1445d0278245SMiao Xie int again; 1446a74a4b97SChris Mason 1447d6fd0ae2SOmar Sandoval while (1) { 1448d0278245SMiao Xie again = 0; 14499d1a2a3aSDavid Sterba 1450fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1451fd340d0fSJosef Bacik 1452d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 14532ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1454d0278245SMiao Xie goto sleep; 1455d0278245SMiao Xie 145690c711abSZygo Blaxell /* 145790c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 145890c711abSZygo Blaxell * before we have finished mounting the filesystem. 145990c711abSZygo Blaxell */ 14600b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 146190c711abSZygo Blaxell goto sleep; 146290c711abSZygo Blaxell 14630b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1464d0278245SMiao Xie goto sleep; 1465d0278245SMiao Xie 1466dc7f370cSMiao Xie /* 1467dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1468dc7f370cSMiao Xie * during the above check and trylock. 1469dc7f370cSMiao Xie */ 14702ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 14710b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1472dc7f370cSMiao Xie goto sleep; 1473dc7f370cSMiao Xie } 1474dc7f370cSMiao Xie 1475b7625f46SQu Wenruo if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) 1476b7625f46SQu Wenruo btrfs_sysfs_feature_update(fs_info); 1477b7625f46SQu Wenruo 14782ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1479c2d6cb16SFilipe Manana 148033c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 14810b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1482a74a4b97SChris Mason 1483d0278245SMiao Xie /* 148405323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 148505323cd1SMiao Xie * needn't do anything special here. 1486d0278245SMiao Xie */ 14870b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 148867c5e7d4SFilipe Manana 148967c5e7d4SFilipe Manana /* 1490f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 149167c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 149267c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1493f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 149467c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 149567c5e7d4SFilipe Manana * unused block groups. 149667c5e7d4SFilipe Manana */ 14970b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 149818bb8bbfSJohannes Thumshirn 149918bb8bbfSJohannes Thumshirn /* 150018bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 150118bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 150218bb8bbfSJohannes Thumshirn * space. 150318bb8bbfSJohannes Thumshirn */ 150418bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1505d0278245SMiao Xie sleep: 1506a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1507d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1508d6fd0ae2SOmar Sandoval kthread_parkme(); 1509d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1510d6fd0ae2SOmar Sandoval return 0; 1511838fe188SJiri Kosina if (!again) { 1512a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1513a74a4b97SChris Mason schedule(); 1514a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1515a74a4b97SChris Mason } 1516da288d28SFilipe Manana } 1517a74a4b97SChris Mason } 1518a74a4b97SChris Mason 1519a74a4b97SChris Mason static int transaction_kthread(void *arg) 1520a74a4b97SChris Mason { 1521a74a4b97SChris Mason struct btrfs_root *root = arg; 15220b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1523a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1524a74a4b97SChris Mason struct btrfs_transaction *cur; 15258929ecfaSYan, Zheng u64 transid; 1526643900beSNikolay Borisov time64_t delta; 1527a74a4b97SChris Mason unsigned long delay; 1528914b2007SJan Kara bool cannot_commit; 1529a74a4b97SChris Mason 1530a74a4b97SChris Mason do { 1531914b2007SJan Kara cannot_commit = false; 1532ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 15330b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1534a74a4b97SChris Mason 15350b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 15360b246afaSJeff Mahoney cur = fs_info->running_transaction; 1537a74a4b97SChris Mason if (!cur) { 15380b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1539a74a4b97SChris Mason goto sleep; 1540a74a4b97SChris Mason } 154131153d81SYan Zheng 1542643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1543fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1544fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1545643900beSNikolay Borisov delta < fs_info->commit_interval) { 15460b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1547fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1548fb8a7e94SNikolay Borisov delay = min(delay, 1549fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1550a74a4b97SChris Mason goto sleep; 1551a74a4b97SChris Mason } 15528929ecfaSYan, Zheng transid = cur->transid; 15530b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 155456bec294SChris Mason 155579787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1556354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1557914b2007SJan Kara if (IS_ERR(trans)) { 1558354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1559914b2007SJan Kara cannot_commit = true; 156079787eaaSJeff Mahoney goto sleep; 1561914b2007SJan Kara } 15628929ecfaSYan, Zheng if (transid == trans->transid) { 15633a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 15648929ecfaSYan, Zheng } else { 15653a45bb20SJeff Mahoney btrfs_end_transaction(trans); 15668929ecfaSYan, Zheng } 1567a74a4b97SChris Mason sleep: 15680b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 15690b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1570a74a4b97SChris Mason 157184961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 15722ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 15738929ecfaSYan, Zheng if (!kthread_should_stop() && 15740b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1575914b2007SJan Kara cannot_commit)) 1576bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1577a74a4b97SChris Mason } while (!kthread_should_stop()); 1578a74a4b97SChris Mason return 0; 1579a74a4b97SChris Mason } 1580a74a4b97SChris Mason 1581af31f5e5SChris Mason /* 158201f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 158301f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 158401f0f9daSNikolay Borisov * anything. 1585af31f5e5SChris Mason * 1586af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1587af31f5e5SChris Mason * generation of the latest root in the array with the generation 1588af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1589af31f5e5SChris Mason */ 159001f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1591af31f5e5SChris Mason { 159201f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1593af31f5e5SChris Mason u64 cur; 1594af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1595af31f5e5SChris Mason int i; 1596af31f5e5SChris Mason 1597af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1598af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1599af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1600af31f5e5SChris Mason if (cur == newest_gen) 160101f0f9daSNikolay Borisov return i; 1602af31f5e5SChris Mason } 1603af31f5e5SChris Mason 160401f0f9daSNikolay Borisov return -EINVAL; 1605af31f5e5SChris Mason } 1606af31f5e5SChris Mason 1607af31f5e5SChris Mason /* 1608af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1609af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1610af31f5e5SChris Mason * done 1611af31f5e5SChris Mason */ 1612af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1613af31f5e5SChris Mason { 16146ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1615af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1616af31f5e5SChris Mason 1617af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1618af31f5e5SChris Mason 1619af31f5e5SChris Mason /* 1620af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1621af31f5e5SChris Mason * regardless of which ones we use today 1622af31f5e5SChris Mason */ 1623af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1624af31f5e5SChris Mason 1625af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1626af31f5e5SChris Mason 1627af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1628af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1629af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1630af31f5e5SChris Mason 1631af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1632af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1633af31f5e5SChris Mason 1634af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1635af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1636af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1637af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1638af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1639af31f5e5SChris Mason 16401c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 16419c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1642f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 16439c54e80dSJosef Bacik 16449c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 16459c54e80dSJosef Bacik extent_root->node->start); 1646af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 164729cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 1648af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 164929cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 1650af31f5e5SChris Mason 1651f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 1652f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 1653f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 1654f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 1655f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 16569c54e80dSJosef Bacik } 1657af31f5e5SChris Mason 16587c7e82a7SChris Mason /* 16597c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 16607c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 16617c7e82a7SChris Mason */ 16627c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 16637c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 16647c7e82a7SChris Mason info->fs_root->node->start); 1665af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1666af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1667af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1668af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 16697c7e82a7SChris Mason } 1670af31f5e5SChris Mason 1671af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1672af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1673af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1674af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1675af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1676af31f5e5SChris Mason 1677af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1678af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1679af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1680af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1681af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1682af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1683af31f5e5SChris Mason 1684af31f5e5SChris Mason /* 1685af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1686af31f5e5SChris Mason * for the next commit 1687af31f5e5SChris Mason */ 1688af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1689af31f5e5SChris Mason &info->super_for_commit->super_roots, 1690af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1691af31f5e5SChris Mason } 1692af31f5e5SChris Mason 1693af31f5e5SChris Mason /* 1694bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1695bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1696bd2336b2SNikolay Borisov * 1697bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1698bd2336b2SNikolay Borisov * priority - priority of backup root required 1699bd2336b2SNikolay Borisov * 1700bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1701bd2336b2SNikolay Borisov */ 1702bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1703bd2336b2SNikolay Borisov { 1704bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1705bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1706bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1707bd2336b2SNikolay Borisov 1708bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1709bd2336b2SNikolay Borisov if (priority == 0) 1710bd2336b2SNikolay Borisov return backup_index; 1711bd2336b2SNikolay Borisov 1712bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1713bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1714bd2336b2SNikolay Borisov } else { 1715bd2336b2SNikolay Borisov return -EINVAL; 1716bd2336b2SNikolay Borisov } 1717bd2336b2SNikolay Borisov 1718bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1719bd2336b2SNikolay Borisov 1720bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1721bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1722bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1723bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1724bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1725bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1726bd2336b2SNikolay Borisov 1727bd2336b2SNikolay Borisov /* 1728bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1729bd2336b2SNikolay Borisov * need a fsck 1730bd2336b2SNikolay Borisov */ 1731bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1732bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1733bd2336b2SNikolay Borisov 1734bd2336b2SNikolay Borisov return backup_index; 1735bd2336b2SNikolay Borisov } 1736bd2336b2SNikolay Borisov 17377abadb64SLiu Bo /* helper to cleanup workers */ 17387abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 17397abadb64SLiu Bo { 1740dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1741afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 17425cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1743d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 1744d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 1745385de0efSChristoph Hellwig if (fs_info->rmw_workers) 1746385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 1747fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 1748fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 1749fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1750fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 17515b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1752e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1753a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1754fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1755b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1756b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1757a9b9477dSFilipe Manana /* 1758a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 1759a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 1760a9b9477dSFilipe Manana * queues can do metadata I/O operations. 1761a9b9477dSFilipe Manana */ 1762d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 1763d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 17647abadb64SLiu Bo } 17657abadb64SLiu Bo 17662e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 17672e9f5954SRashika { 17682e9f5954SRashika if (root) { 17692e9f5954SRashika free_extent_buffer(root->node); 17702e9f5954SRashika free_extent_buffer(root->commit_root); 17712e9f5954SRashika root->node = NULL; 17722e9f5954SRashika root->commit_root = NULL; 17732e9f5954SRashika } 17742e9f5954SRashika } 17752e9f5954SRashika 1776abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 1777abed4aaaSJosef Bacik { 1778abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 1779abed4aaaSJosef Bacik 1780abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 1781abed4aaaSJosef Bacik &fs_info->global_root_tree, 1782abed4aaaSJosef Bacik rb_node) 1783abed4aaaSJosef Bacik free_root_extent_buffers(root); 1784abed4aaaSJosef Bacik } 1785abed4aaaSJosef Bacik 1786af31f5e5SChris Mason /* helper to cleanup tree roots */ 17874273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 1788af31f5e5SChris Mason { 17892e9f5954SRashika free_root_extent_buffers(info->tree_root); 1790655b09feSJosef Bacik 1791abed4aaaSJosef Bacik free_global_root_pointers(info); 17922e9f5954SRashika free_root_extent_buffers(info->dev_root); 17932e9f5954SRashika free_root_extent_buffers(info->quota_root); 17942e9f5954SRashika free_root_extent_buffers(info->uuid_root); 17958c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 1796aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 17979c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 17984273eaffSAnand Jain if (free_chunk_root) 17992e9f5954SRashika free_root_extent_buffers(info->chunk_root); 1800af31f5e5SChris Mason } 1801af31f5e5SChris Mason 18028c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 18038c38938cSJosef Bacik { 18048c38938cSJosef Bacik if (!root) 18058c38938cSJosef Bacik return; 18068c38938cSJosef Bacik 18078c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 18088c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 18091dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 18108c38938cSJosef Bacik if (root->anon_dev) 18118c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 1812923eb523SJohannes Thumshirn free_root_extent_buffers(root); 18138c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1814fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 18158c38938cSJosef Bacik list_del_init(&root->leak_list); 1816fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 18178c38938cSJosef Bacik #endif 18188c38938cSJosef Bacik kfree(root); 18198c38938cSJosef Bacik } 18208c38938cSJosef Bacik } 18218c38938cSJosef Bacik 1822faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 1823171f6537SJosef Bacik { 1824fc7cbcd4SDavid Sterba int ret; 1825fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 1826fc7cbcd4SDavid Sterba int i; 1827171f6537SJosef Bacik 1828171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 1829fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 1830171f6537SJosef Bacik struct btrfs_root, root_list); 1831fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 1832171f6537SJosef Bacik 1833fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 1834fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 1835fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 1836171f6537SJosef Bacik } 1837171f6537SJosef Bacik 1838fc7cbcd4SDavid Sterba while (1) { 1839fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 1840fc7cbcd4SDavid Sterba (void **)gang, 0, 1841fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 1842fc7cbcd4SDavid Sterba if (!ret) 1843fc7cbcd4SDavid Sterba break; 1844fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 1845fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 1846171f6537SJosef Bacik } 1847171f6537SJosef Bacik } 1848af31f5e5SChris Mason 1849638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 1850638aa7edSEric Sandeen { 1851638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 1852638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 1853638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 1854638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 1855638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 1856638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 1857ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 1858638aa7edSEric Sandeen } 1859638aa7edSEric Sandeen 1860779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 1861779a65a4SEric Sandeen { 1862779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 1863779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 1864779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 1865779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 1866779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 1867779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 1868907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 1869779a65a4SEric Sandeen } 1870779a65a4SEric Sandeen 1871dcb2137cSChristoph Hellwig static int btrfs_init_btree_inode(struct super_block *sb) 1872f37938e0SEric Sandeen { 1873dcb2137cSChristoph Hellwig struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1874e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 1875e256927bSJosef Bacik fs_info->tree_root); 1876dcb2137cSChristoph Hellwig struct inode *inode; 1877dcb2137cSChristoph Hellwig 1878dcb2137cSChristoph Hellwig inode = new_inode(sb); 1879dcb2137cSChristoph Hellwig if (!inode) 1880dcb2137cSChristoph Hellwig return -ENOMEM; 18812ff7e61eSJeff Mahoney 18822ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 18832ff7e61eSJeff Mahoney set_nlink(inode, 1); 1884f37938e0SEric Sandeen /* 1885f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 1886f37938e0SEric Sandeen * the real end of the address space is determined by all of 1887f37938e0SEric Sandeen * the devices in the system 1888f37938e0SEric Sandeen */ 18892ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 18902ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 1891dcb2137cSChristoph Hellwig mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); 1892f37938e0SEric Sandeen 18932ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 189443eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 189535da5a7eSDavid Sterba IO_TREE_BTREE_INODE_IO); 18962ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 1897f37938e0SEric Sandeen 18985c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 1899adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 1900adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 1901adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 19022ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 1903e256927bSJosef Bacik __insert_inode_hash(inode, hash); 1904dcb2137cSChristoph Hellwig fs_info->btree_inode = inode; 1905dcb2137cSChristoph Hellwig 1906dcb2137cSChristoph Hellwig return 0; 1907f37938e0SEric Sandeen } 1908f37938e0SEric Sandeen 1909ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 1910ad618368SEric Sandeen { 1911ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 1912129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 19137f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 1914ad618368SEric Sandeen } 1915ad618368SEric Sandeen 1916f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 1917f9e92e40SEric Sandeen { 1918f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 1919f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 1920f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 1921f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 1922f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 1923f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 1924d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 1925011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 1926f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 1927f9e92e40SEric Sandeen } 1928f9e92e40SEric Sandeen 1929d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 19302a458198SEric Sandeen { 1931f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 19326f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 193358e814fcSTejun Heo unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; 19342a458198SEric Sandeen 19352a458198SEric Sandeen fs_info->workers = 1936a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 19372a458198SEric Sandeen 19382a458198SEric Sandeen fs_info->delalloc_workers = 1939cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 1940cb001095SJeff Mahoney flags, max_active, 2); 19412a458198SEric Sandeen 19422a458198SEric Sandeen fs_info->flush_workers = 1943cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 1944cb001095SJeff Mahoney flags, max_active, 0); 19452a458198SEric Sandeen 19462a458198SEric Sandeen fs_info->caching_workers = 1947cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 19482a458198SEric Sandeen 19492a458198SEric Sandeen fs_info->fixup_workers = 195058e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); 19512a458198SEric Sandeen 19522a458198SEric Sandeen fs_info->endio_workers = 1953d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 19542a458198SEric Sandeen fs_info->endio_meta_workers = 1955d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 1956385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 19572a458198SEric Sandeen fs_info->endio_write_workers = 1958cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 1959cb001095SJeff Mahoney max_active, 2); 1960fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 1961fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 19622a458198SEric Sandeen fs_info->endio_freespace_worker = 1963cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 1964cb001095SJeff Mahoney max_active, 0); 19652a458198SEric Sandeen fs_info->delayed_workers = 1966cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 1967cb001095SJeff Mahoney max_active, 0); 19682a458198SEric Sandeen fs_info->qgroup_rescan_workers = 196958e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", 197058e814fcSTejun Heo ordered_flags); 1971b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 197258e814fcSTejun Heo alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); 19732a458198SEric Sandeen 19748bfec2e4SChristoph Hellwig if (!(fs_info->workers && 1975a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 19762a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 1977fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 19781a1a2851SQu Wenruo fs_info->endio_write_workers && 19792a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 1980f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 1981f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 1982b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 19832a458198SEric Sandeen return -ENOMEM; 19842a458198SEric Sandeen } 19852a458198SEric Sandeen 19862a458198SEric Sandeen return 0; 19872a458198SEric Sandeen } 19882a458198SEric Sandeen 19896d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 19906d97c6e3SJohannes Thumshirn { 19916d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 1992b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 19936d97c6e3SJohannes Thumshirn 1994b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 19956d97c6e3SJohannes Thumshirn 19966d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 19976d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 1998b4e967beSDavid Sterba csum_driver); 19996d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 20006d97c6e3SJohannes Thumshirn } 20016d97c6e3SJohannes Thumshirn 20026d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 20036d97c6e3SJohannes Thumshirn 200468d99ab0SChristoph Hellwig /* 200568d99ab0SChristoph Hellwig * Check if the checksum implementation is a fast accelerated one. 200668d99ab0SChristoph Hellwig * As-is this is a bit of a hack and should be replaced once the csum 200768d99ab0SChristoph Hellwig * implementations provide that information themselves. 200868d99ab0SChristoph Hellwig */ 200968d99ab0SChristoph Hellwig switch (csum_type) { 201068d99ab0SChristoph Hellwig case BTRFS_CSUM_TYPE_CRC32: 201168d99ab0SChristoph Hellwig if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) 201268d99ab0SChristoph Hellwig set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 201368d99ab0SChristoph Hellwig break; 2014efcfcbc6SDavid Sterba case BTRFS_CSUM_TYPE_XXHASH: 2015efcfcbc6SDavid Sterba set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 2016efcfcbc6SDavid Sterba break; 201768d99ab0SChristoph Hellwig default: 201868d99ab0SChristoph Hellwig break; 201968d99ab0SChristoph Hellwig } 202068d99ab0SChristoph Hellwig 2021c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2022c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2023c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 20246d97c6e3SJohannes Thumshirn return 0; 20256d97c6e3SJohannes Thumshirn } 20266d97c6e3SJohannes Thumshirn 202763443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 202863443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 202963443bf5SEric Sandeen { 203063443bf5SEric Sandeen int ret; 2031789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 203263443bf5SEric Sandeen struct btrfs_root *log_tree_root; 203363443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 203463443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2035581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 203663443bf5SEric Sandeen 203763443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2038f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 203963443bf5SEric Sandeen return -EIO; 204063443bf5SEric Sandeen } 204163443bf5SEric Sandeen 204296dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 204396dfcb46SJosef Bacik GFP_KERNEL); 204463443bf5SEric Sandeen if (!log_tree_root) 204563443bf5SEric Sandeen return -ENOMEM; 204663443bf5SEric Sandeen 2047789d6a3aSQu Wenruo check.level = level; 2048789d6a3aSQu Wenruo check.transid = fs_info->generation + 1; 2049789d6a3aSQu Wenruo check.owner_root = BTRFS_TREE_LOG_OBJECTID; 2050789d6a3aSQu Wenruo log_tree_root->node = read_tree_block(fs_info, bytenr, &check); 205164c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2052f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 20530eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 20548c38938cSJosef Bacik log_tree_root->node = NULL; 205500246528SJosef Bacik btrfs_put_root(log_tree_root); 20560eeff236SLiu Bo return ret; 20574eb150d6SQu Wenruo } 20584eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2059f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 206000246528SJosef Bacik btrfs_put_root(log_tree_root); 206163443bf5SEric Sandeen return -EIO; 206263443bf5SEric Sandeen } 20634eb150d6SQu Wenruo 206463443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 206563443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 206663443bf5SEric Sandeen if (ret) { 20670b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 206863443bf5SEric Sandeen "Failed to recover log tree"); 206900246528SJosef Bacik btrfs_put_root(log_tree_root); 207063443bf5SEric Sandeen return ret; 207163443bf5SEric Sandeen } 207263443bf5SEric Sandeen 2073bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 20746bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 207563443bf5SEric Sandeen if (ret) 207663443bf5SEric Sandeen return ret; 207763443bf5SEric Sandeen } 207863443bf5SEric Sandeen 207963443bf5SEric Sandeen return 0; 208063443bf5SEric Sandeen } 208163443bf5SEric Sandeen 2082abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2083abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2084abed4aaaSJosef Bacik const char *name) 2085abed4aaaSJosef Bacik { 2086abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2087abed4aaaSJosef Bacik struct btrfs_root *root; 2088f7238e50SJosef Bacik u64 max_global_id = 0; 2089abed4aaaSJosef Bacik int ret; 2090abed4aaaSJosef Bacik struct btrfs_key key = { 2091abed4aaaSJosef Bacik .objectid = objectid, 2092abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2093abed4aaaSJosef Bacik .offset = 0, 2094abed4aaaSJosef Bacik }; 2095abed4aaaSJosef Bacik bool found = false; 2096abed4aaaSJosef Bacik 2097abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2098abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2099abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2100abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2101abed4aaaSJosef Bacik return 0; 2102abed4aaaSJosef Bacik } 2103abed4aaaSJosef Bacik 2104abed4aaaSJosef Bacik while (1) { 2105abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2106abed4aaaSJosef Bacik if (ret < 0) 2107abed4aaaSJosef Bacik break; 2108abed4aaaSJosef Bacik 2109abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2110abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2111abed4aaaSJosef Bacik if (ret) { 2112abed4aaaSJosef Bacik if (ret > 0) 2113abed4aaaSJosef Bacik ret = 0; 2114abed4aaaSJosef Bacik break; 2115abed4aaaSJosef Bacik } 2116abed4aaaSJosef Bacik } 2117abed4aaaSJosef Bacik ret = 0; 2118abed4aaaSJosef Bacik 2119abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2120abed4aaaSJosef Bacik if (key.objectid != objectid) 2121abed4aaaSJosef Bacik break; 2122abed4aaaSJosef Bacik btrfs_release_path(path); 2123abed4aaaSJosef Bacik 2124f7238e50SJosef Bacik /* 2125f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2126f7238e50SJosef Bacik * everybody. 2127f7238e50SJosef Bacik */ 2128f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2129f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2130f7238e50SJosef Bacik 2131abed4aaaSJosef Bacik found = true; 2132abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2133abed4aaaSJosef Bacik if (IS_ERR(root)) { 2134abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2135abed4aaaSJosef Bacik ret = PTR_ERR(root); 2136abed4aaaSJosef Bacik break; 2137abed4aaaSJosef Bacik } 2138abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2139abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2140abed4aaaSJosef Bacik if (ret) { 2141abed4aaaSJosef Bacik btrfs_put_root(root); 2142abed4aaaSJosef Bacik break; 2143abed4aaaSJosef Bacik } 2144abed4aaaSJosef Bacik key.offset++; 2145abed4aaaSJosef Bacik } 2146abed4aaaSJosef Bacik btrfs_release_path(path); 2147abed4aaaSJosef Bacik 2148f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2149f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2150f7238e50SJosef Bacik 2151abed4aaaSJosef Bacik if (!found || ret) { 2152abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2153abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2154abed4aaaSJosef Bacik 2155abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2156abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2157abed4aaaSJosef Bacik else 2158abed4aaaSJosef Bacik ret = 0; 2159abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2160abed4aaaSJosef Bacik } 2161abed4aaaSJosef Bacik return ret; 2162abed4aaaSJosef Bacik } 2163abed4aaaSJosef Bacik 2164abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2165abed4aaaSJosef Bacik { 2166abed4aaaSJosef Bacik struct btrfs_path *path; 2167abed4aaaSJosef Bacik int ret = 0; 2168abed4aaaSJosef Bacik 2169abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2170abed4aaaSJosef Bacik if (!path) 2171abed4aaaSJosef Bacik return -ENOMEM; 2172abed4aaaSJosef Bacik 2173abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2174abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2175abed4aaaSJosef Bacik if (ret) 2176abed4aaaSJosef Bacik goto out; 2177abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2178abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2179abed4aaaSJosef Bacik if (ret) 2180abed4aaaSJosef Bacik goto out; 2181abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2182abed4aaaSJosef Bacik goto out; 2183abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2184abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2185abed4aaaSJosef Bacik "free space"); 2186abed4aaaSJosef Bacik out: 2187abed4aaaSJosef Bacik btrfs_free_path(path); 2188abed4aaaSJosef Bacik return ret; 2189abed4aaaSJosef Bacik } 2190abed4aaaSJosef Bacik 21916bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 21924bbcaa64SEric Sandeen { 21936bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2194a4f3d2c4SDavid Sterba struct btrfs_root *root; 21954bbcaa64SEric Sandeen struct btrfs_key location; 21964bbcaa64SEric Sandeen int ret; 21974bbcaa64SEric Sandeen 21986bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 21996bccf3abSJeff Mahoney 2200abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2201abed4aaaSJosef Bacik if (ret) 2202abed4aaaSJosef Bacik return ret; 2203abed4aaaSJosef Bacik 22044bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 22054bbcaa64SEric Sandeen location.offset = 0; 22064bbcaa64SEric Sandeen 22071c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 220814033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 220914033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 221014033b08SQu Wenruo if (IS_ERR(root)) { 221114033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 221214033b08SQu Wenruo ret = PTR_ERR(root); 221314033b08SQu Wenruo goto out; 221414033b08SQu Wenruo } 221514033b08SQu Wenruo } else { 221614033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 221714033b08SQu Wenruo fs_info->block_group_root = root; 221814033b08SQu Wenruo } 221914033b08SQu Wenruo } 222014033b08SQu Wenruo 222114033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2222a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2223f50f4353SLiu Bo if (IS_ERR(root)) { 222442437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2225f50f4353SLiu Bo ret = PTR_ERR(root); 2226f50f4353SLiu Bo goto out; 2227f50f4353SLiu Bo } 222842437a63SJosef Bacik } else { 2229a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2230a4f3d2c4SDavid Sterba fs_info->dev_root = root; 223142437a63SJosef Bacik } 2232820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2233a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2234a8d1b164SJohannes Thumshirn if (ret) 2235a8d1b164SJohannes Thumshirn goto out; 22364bbcaa64SEric Sandeen 2237aeb935a4SQu Wenruo /* 2238aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2239aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2240aeb935a4SQu Wenruo */ 224156e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 224256e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2243aeb935a4SQu Wenruo if (IS_ERR(root)) { 224442437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2245aeb935a4SQu Wenruo ret = PTR_ERR(root); 2246aeb935a4SQu Wenruo goto out; 2247aeb935a4SQu Wenruo } 224842437a63SJosef Bacik } else { 2249aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2250aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 225142437a63SJosef Bacik } 2252aeb935a4SQu Wenruo 22534bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2254a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2255a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2256a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2257afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2258a4f3d2c4SDavid Sterba fs_info->quota_root = root; 22594bbcaa64SEric Sandeen } 22604bbcaa64SEric Sandeen 22614bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2262a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2263a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 226442437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2265a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 22664bbcaa64SEric Sandeen if (ret != -ENOENT) 2267f50f4353SLiu Bo goto out; 226842437a63SJosef Bacik } 22694bbcaa64SEric Sandeen } else { 2270a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2271a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 22724bbcaa64SEric Sandeen } 22734bbcaa64SEric Sandeen 22744bbcaa64SEric Sandeen return 0; 2275f50f4353SLiu Bo out: 2276f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2277f50f4353SLiu Bo location.objectid, ret); 2278f50f4353SLiu Bo return ret; 22794bbcaa64SEric Sandeen } 22804bbcaa64SEric Sandeen 2281069ec957SQu Wenruo /* 2282069ec957SQu Wenruo * Real super block validation 2283069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2284069ec957SQu Wenruo * 2285069ec957SQu Wenruo * @sb: super block to check 2286069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2287069ec957SQu Wenruo * 0 the primary (1st) sb 2288069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2289069ec957SQu Wenruo * -1 skip bytenr check 2290069ec957SQu Wenruo */ 2291a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2292069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 229321a852b0SQu Wenruo { 229421a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 229521a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 229621a852b0SQu Wenruo int ret = 0; 229721a852b0SQu Wenruo 229821a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 229921a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 230021a852b0SQu Wenruo ret = -EINVAL; 230121a852b0SQu Wenruo } 230221a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 230321a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 230421a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 230521a852b0SQu Wenruo ret = -EINVAL; 230621a852b0SQu Wenruo } 230721a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 230821a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 230921a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 231021a852b0SQu Wenruo ret = -EINVAL; 231121a852b0SQu Wenruo } 231221a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 231321a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 231421a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 231521a852b0SQu Wenruo ret = -EINVAL; 231621a852b0SQu Wenruo } 231721a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 231821a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 231921a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 232021a852b0SQu Wenruo ret = -EINVAL; 232121a852b0SQu Wenruo } 232221a852b0SQu Wenruo 232321a852b0SQu Wenruo /* 232421a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 232521a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 232621a852b0SQu Wenruo */ 232721a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 232821a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 232921a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 233021a852b0SQu Wenruo ret = -EINVAL; 233121a852b0SQu Wenruo } 23320bb3eb3eSQu Wenruo 23330bb3eb3eSQu Wenruo /* 23341a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 23351a42daabSQu Wenruo * 23361a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 23371a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 23381a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 23391a42daabSQu Wenruo * beginning. 23400bb3eb3eSQu Wenruo */ 23411a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 234221a852b0SQu Wenruo btrfs_err(fs_info, 23430bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 234421a852b0SQu Wenruo sectorsize, PAGE_SIZE); 234521a852b0SQu Wenruo ret = -EINVAL; 234621a852b0SQu Wenruo } 23470bb3eb3eSQu Wenruo 234821a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 234921a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 235021a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 235121a852b0SQu Wenruo ret = -EINVAL; 235221a852b0SQu Wenruo } 235321a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 235421a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 235521a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 235621a852b0SQu Wenruo ret = -EINVAL; 235721a852b0SQu Wenruo } 235821a852b0SQu Wenruo 235921a852b0SQu Wenruo /* Root alignment check */ 236021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 236121a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 236221a852b0SQu Wenruo btrfs_super_root(sb)); 236321a852b0SQu Wenruo ret = -EINVAL; 236421a852b0SQu Wenruo } 236521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 236621a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 236721a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 236821a852b0SQu Wenruo ret = -EINVAL; 236921a852b0SQu Wenruo } 237021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 237121a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 237221a852b0SQu Wenruo btrfs_super_log_root(sb)); 237321a852b0SQu Wenruo ret = -EINVAL; 237421a852b0SQu Wenruo } 237521a852b0SQu Wenruo 2376aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2377aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2378aefd7f70SNikolay Borisov btrfs_err(fs_info, 2379aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2380aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2381aefd7f70SNikolay Borisov ret = -EINVAL; 2382aefd7f70SNikolay Borisov } 2383aefd7f70SNikolay Borisov 2384aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2385aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2386aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2387aefd7f70SNikolay Borisov btrfs_err(fs_info, 2388aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2389aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2390aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2391aefd7f70SNikolay Borisov ret = -EINVAL; 2392aefd7f70SNikolay Borisov } 2393aefd7f70SNikolay Borisov 239425984a5aSAnand Jain if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 239525984a5aSAnand Jain BTRFS_FSID_SIZE) != 0) { 239625984a5aSAnand Jain btrfs_err(fs_info, 239725984a5aSAnand Jain "dev_item UUID does not match metadata fsid: %pU != %pU", 239825984a5aSAnand Jain fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 239925984a5aSAnand Jain ret = -EINVAL; 240025984a5aSAnand Jain } 240125984a5aSAnand Jain 24021c56ab99SQu Wenruo /* 24031c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 24041c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 24051c56ab99SQu Wenruo */ 24061c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 24071c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 24081c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 24091c56ab99SQu Wenruo btrfs_err(fs_info, 24101c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 24111c56ab99SQu Wenruo ret = -EINVAL; 24121c56ab99SQu Wenruo } 24131c56ab99SQu Wenruo 241421a852b0SQu Wenruo /* 241521a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 241621a852b0SQu Wenruo * done later 241721a852b0SQu Wenruo */ 241821a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 241921a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 242021a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 242121a852b0SQu Wenruo ret = -EINVAL; 242221a852b0SQu Wenruo } 242321a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 242421a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 242521a852b0SQu Wenruo btrfs_super_stripesize(sb)); 242621a852b0SQu Wenruo ret = -EINVAL; 242721a852b0SQu Wenruo } 242821a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 242921a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 243021a852b0SQu Wenruo btrfs_super_num_devices(sb)); 243121a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 243221a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 243321a852b0SQu Wenruo ret = -EINVAL; 243421a852b0SQu Wenruo } 243521a852b0SQu Wenruo 2436069ec957SQu Wenruo if (mirror_num >= 0 && 2437069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 243821a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 243921a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 244021a852b0SQu Wenruo ret = -EINVAL; 244121a852b0SQu Wenruo } 244221a852b0SQu Wenruo 244321a852b0SQu Wenruo /* 244421a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 244521a852b0SQu Wenruo * and one chunk 244621a852b0SQu Wenruo */ 244721a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 244821a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 244921a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 245021a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 245121a852b0SQu Wenruo ret = -EINVAL; 245221a852b0SQu Wenruo } 245321a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 245421a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 245521a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 245621a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 245721a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 245821a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 245921a852b0SQu Wenruo ret = -EINVAL; 246021a852b0SQu Wenruo } 246121a852b0SQu Wenruo 246221a852b0SQu Wenruo /* 246321a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 246421a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 246521a852b0SQu Wenruo */ 246621a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 246721a852b0SQu Wenruo btrfs_warn(fs_info, 246821a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 246921a852b0SQu Wenruo btrfs_super_generation(sb), 247021a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 247121a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 247221a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 247321a852b0SQu Wenruo btrfs_warn(fs_info, 247421a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 247521a852b0SQu Wenruo btrfs_super_generation(sb), 247621a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 247721a852b0SQu Wenruo 247821a852b0SQu Wenruo return ret; 247921a852b0SQu Wenruo } 248021a852b0SQu Wenruo 2481069ec957SQu Wenruo /* 2482069ec957SQu Wenruo * Validation of super block at mount time. 2483069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2484069ec957SQu Wenruo * flags will be skipped. 2485069ec957SQu Wenruo */ 2486069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2487069ec957SQu Wenruo { 2488a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2489069ec957SQu Wenruo } 2490069ec957SQu Wenruo 249175cb857dSQu Wenruo /* 249275cb857dSQu Wenruo * Validation of super block at write time. 249375cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 249475cb857dSQu Wenruo * overwritten soon. 249575cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 249675cb857dSQu Wenruo */ 249775cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 249875cb857dSQu Wenruo struct btrfs_super_block *sb) 249975cb857dSQu Wenruo { 250075cb857dSQu Wenruo int ret; 250175cb857dSQu Wenruo 2502a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 250375cb857dSQu Wenruo if (ret < 0) 250475cb857dSQu Wenruo goto out; 2505e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 250675cb857dSQu Wenruo ret = -EUCLEAN; 250775cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 250875cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 250975cb857dSQu Wenruo goto out; 251075cb857dSQu Wenruo } 251175cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 251275cb857dSQu Wenruo ret = -EUCLEAN; 251375cb857dSQu Wenruo btrfs_err(fs_info, 251475cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 251575cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 251675cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 251775cb857dSQu Wenruo goto out; 251875cb857dSQu Wenruo } 251975cb857dSQu Wenruo out: 252075cb857dSQu Wenruo if (ret < 0) 252175cb857dSQu Wenruo btrfs_err(fs_info, 252275cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 252375cb857dSQu Wenruo return ret; 252475cb857dSQu Wenruo } 252575cb857dSQu Wenruo 2526bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2527bd676446SJosef Bacik { 2528789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 2529789d6a3aSQu Wenruo .level = level, 2530789d6a3aSQu Wenruo .transid = gen, 2531789d6a3aSQu Wenruo .owner_root = root->root_key.objectid 2532789d6a3aSQu Wenruo }; 2533bd676446SJosef Bacik int ret = 0; 2534bd676446SJosef Bacik 2535789d6a3aSQu Wenruo root->node = read_tree_block(root->fs_info, bytenr, &check); 2536bd676446SJosef Bacik if (IS_ERR(root->node)) { 2537bd676446SJosef Bacik ret = PTR_ERR(root->node); 2538bd676446SJosef Bacik root->node = NULL; 25394eb150d6SQu Wenruo return ret; 25404eb150d6SQu Wenruo } 25414eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2542bd676446SJosef Bacik free_extent_buffer(root->node); 2543bd676446SJosef Bacik root->node = NULL; 25444eb150d6SQu Wenruo return -EIO; 2545bd676446SJosef Bacik } 2546bd676446SJosef Bacik 2547bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2548bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2549bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2550bd676446SJosef Bacik return ret; 2551bd676446SJosef Bacik } 2552bd676446SJosef Bacik 2553bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2554bd676446SJosef Bacik { 2555bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2556bd676446SJosef Bacik u64 gen, bytenr; 2557bd676446SJosef Bacik int level, ret; 2558bd676446SJosef Bacik 2559bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2560bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2561bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2562bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 25639c54e80dSJosef Bacik if (ret) { 2564bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2565bd676446SJosef Bacik return ret; 2566bd676446SJosef Bacik } 25679c54e80dSJosef Bacik return 0; 25689c54e80dSJosef Bacik } 25699c54e80dSJosef Bacik 25706ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2571b8522a1eSNikolay Borisov { 25726ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2573b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2574b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2575b8522a1eSNikolay Borisov bool handle_error = false; 2576b8522a1eSNikolay Borisov int ret = 0; 2577b8522a1eSNikolay Borisov int i; 2578b8522a1eSNikolay Borisov 2579b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2580b8522a1eSNikolay Borisov if (handle_error) { 2581b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2582b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2583b8522a1eSNikolay Borisov tree_root->node = NULL; 2584b8522a1eSNikolay Borisov 2585b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2586b8522a1eSNikolay Borisov break; 2587b8522a1eSNikolay Borisov 2588b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2589b8522a1eSNikolay Borisov 2590b8522a1eSNikolay Borisov /* 2591b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2592b8522a1eSNikolay Borisov * valid 2593b8522a1eSNikolay Borisov */ 2594b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2595b8522a1eSNikolay Borisov 2596b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2597b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2598b8522a1eSNikolay Borisov 2599745806fbSQu Wenruo btrfs_warn(fs_info, "try to load backup roots slot %d", i); 2600b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26016ef108ddSNikolay Borisov backup_index = ret; 2602b8522a1eSNikolay Borisov if (ret < 0) 2603b8522a1eSNikolay Borisov return ret; 2604b8522a1eSNikolay Borisov } 2605b8522a1eSNikolay Borisov 2606bd676446SJosef Bacik ret = load_important_roots(fs_info); 2607bd676446SJosef Bacik if (ret) { 2608217f5004SNikolay Borisov handle_error = true; 2609b8522a1eSNikolay Borisov continue; 2610b8522a1eSNikolay Borisov } 2611b8522a1eSNikolay Borisov 2612336a0d8dSNikolay Borisov /* 2613336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2614336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2615336a0d8dSNikolay Borisov */ 2616453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2617b8522a1eSNikolay Borisov if (ret < 0) { 2618b8522a1eSNikolay Borisov handle_error = true; 2619b8522a1eSNikolay Borisov continue; 2620b8522a1eSNikolay Borisov } 2621b8522a1eSNikolay Borisov 26226b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2623b8522a1eSNikolay Borisov 2624b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2625b8522a1eSNikolay Borisov if (ret < 0) { 2626b8522a1eSNikolay Borisov handle_error = true; 2627b8522a1eSNikolay Borisov continue; 2628b8522a1eSNikolay Borisov } 2629b8522a1eSNikolay Borisov 2630b8522a1eSNikolay Borisov /* All successful */ 2631bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2632bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2633d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 26346ef108ddSNikolay Borisov 26356ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26366ef108ddSNikolay Borisov if (backup_index < 0) { 26376ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26386ef108ddSNikolay Borisov } else { 26396ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26406ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26416ef108ddSNikolay Borisov } 2642b8522a1eSNikolay Borisov break; 2643b8522a1eSNikolay Borisov } 2644b8522a1eSNikolay Borisov 2645b8522a1eSNikolay Borisov return ret; 2646b8522a1eSNikolay Borisov } 2647b8522a1eSNikolay Borisov 26488260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2649eb60ceacSChris Mason { 2650fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 265101cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26528fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2653facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 265424bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2655eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 265611833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2657eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2658a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2659fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 266024bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26614cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2662ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2663f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 266447ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 266540ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2666afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2667c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2668f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2669abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2670d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2671f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 26727585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2673573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 26740bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 26755f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 2676de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 267719c00ddcSChris Mason 2678e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 26795a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 26808b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 26815f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 26823e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 26833e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 26843e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 26853e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 26863e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 26873e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 26883e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 26893e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 2690e1489b4fSIoannis Angelakopoulos 26910b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 26926324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2693f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 269447ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 269518bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2696afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2697bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2698bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 26993fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27003fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2701bd647ce3SJosef Bacik #endif 2702c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 270366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 270466d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 270566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 270666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 270766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 270866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 270966d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2710ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2711ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2712ba2c4d4eSJosef Bacik 2713771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27144cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2715034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2716fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2717abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 271895ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27199ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27204cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2721a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2722f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27238b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2724fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2725c8b97818SChris Mason 2726b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2727b34b086cSChris Mason num_online_cpus() + 2, 8); 27280afbaf8cSChris Mason 2729199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2730199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 273169fe2d75SJosef Bacik 2732638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 273321adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 273421adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 273521adbd5cSStefan Behrens #endif 2736779a65a4SEric Sandeen btrfs_init_balance(fs_info); 273757056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2738a2de733cSArne Jansen 273916b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 274008dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 27410f9dd46cSJosef Bacik 2742fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 274335da5a7eSDavid Sterba IO_TREE_FS_EXCLUDED_EXTENTS); 274439279cc3SChris Mason 27455a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2746e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2747925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2748a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2749a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27501bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27519e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2752c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 275376dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2754803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2755fa9c0d79SChris Mason 2756ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2757f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2758b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2759416ac51dSArne Jansen 2760fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2761fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2762fa9c0d79SChris Mason 2763e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2764f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2765bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27664854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2767034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 27683768f368SChris Mason 2769da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2770da17066cSJeff Mahoney fs_info->nodesize = 4096; 2771da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2772ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2773da17066cSJeff Mahoney fs_info->stripesize = 4096; 2774da17066cSJeff Mahoney 2775f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 2776f7b12a62SNaohiro Aota 2777eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2778eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2779eede2bf3SOmar Sandoval 278018bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 278118bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 27828260edbaSJosef Bacik } 27838260edbaSJosef Bacik 27848260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 27858260edbaSJosef Bacik { 27868260edbaSJosef Bacik int ret; 27878260edbaSJosef Bacik 27888260edbaSJosef Bacik fs_info->sb = sb; 27898260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 27908260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 27919e967495SFilipe Manana 27925deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 2793ae18c37aSJosef Bacik if (ret) 2794ae18c37aSJosef Bacik return ret; 2795ae18c37aSJosef Bacik 2796ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2797ae18c37aSJosef Bacik if (ret) 2798c75e8394SJosef Bacik return ret; 2799ae18c37aSJosef Bacik 2800ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2801ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2802ae18c37aSJosef Bacik 2803ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2804ae18c37aSJosef Bacik if (ret) 2805c75e8394SJosef Bacik return ret; 2806ae18c37aSJosef Bacik 2807ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2808ae18c37aSJosef Bacik GFP_KERNEL); 2809ae18c37aSJosef Bacik if (ret) 2810c75e8394SJosef Bacik return ret; 2811ae18c37aSJosef Bacik 2812ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2813ae18c37aSJosef Bacik GFP_KERNEL); 2814c75e8394SJosef Bacik if (!fs_info->delayed_root) 2815c75e8394SJosef Bacik return -ENOMEM; 2816ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2817ae18c37aSJosef Bacik 2818a0a1db70SFilipe Manana if (sb_rdonly(sb)) 2819a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 2820a0a1db70SFilipe Manana 2821c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2822ae18c37aSJosef Bacik } 2823ae18c37aSJosef Bacik 282497f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 282597f4dd09SNikolay Borisov { 28260d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 282797f4dd09SNikolay Borisov int ret; 282897f4dd09SNikolay Borisov 282997f4dd09SNikolay Borisov /* 283097f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 283197f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 283297f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 283397f4dd09SNikolay Borisov */ 283497f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 283597f4dd09SNikolay Borisov if (ret < 0) { 2836c94bec2cSJosef Bacik if (ret != -EINTR) 2837c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2838c94bec2cSJosef Bacik ret); 283997f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 284097f4dd09SNikolay Borisov return ret; 284197f4dd09SNikolay Borisov } 284297f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 284397f4dd09SNikolay Borisov } 284497f4dd09SNikolay Borisov 284597f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 284697f4dd09SNikolay Borisov { 284797f4dd09SNikolay Borisov struct task_struct *task; 284897f4dd09SNikolay Borisov 284997f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 285097f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 285197f4dd09SNikolay Borisov if (IS_ERR(task)) { 285297f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 285397f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 285497f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 285597f4dd09SNikolay Borisov return PTR_ERR(task); 285697f4dd09SNikolay Borisov } 285797f4dd09SNikolay Borisov 285897f4dd09SNikolay Borisov return 0; 285997f4dd09SNikolay Borisov } 286097f4dd09SNikolay Borisov 286144c0ca21SBoris Burkov /* 28628cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 28638cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 28648cd29088SBoris Burkov * code paths. 28658cd29088SBoris Burkov */ 28668cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 28678cd29088SBoris Burkov { 28688cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 28698b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 28708cd29088SBoris Burkov } 28718cd29088SBoris Burkov 28728cd29088SBoris Burkov /* 287344c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 287444c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 287544c0ca21SBoris Burkov */ 287644c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 287744c0ca21SBoris Burkov { 287844c0ca21SBoris Burkov int ret; 287994846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 28801d6a4fc8SQu Wenruo bool rebuild_free_space_tree = false; 28818b228324SBoris Burkov 28828b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 28838b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 28841d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 28858b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 28868b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 28878b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 28881d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 28898b228324SBoris Burkov } 28908b228324SBoris Burkov 28911d6a4fc8SQu Wenruo if (rebuild_free_space_tree) { 28921d6a4fc8SQu Wenruo btrfs_info(fs_info, "rebuilding free space tree"); 28931d6a4fc8SQu Wenruo ret = btrfs_rebuild_free_space_tree(fs_info); 28948b228324SBoris Burkov if (ret) { 28958b228324SBoris Burkov btrfs_warn(fs_info, 28961d6a4fc8SQu Wenruo "failed to rebuild free space tree: %d", ret); 28971d6a4fc8SQu Wenruo goto out; 28981d6a4fc8SQu Wenruo } 28991d6a4fc8SQu Wenruo } 29001d6a4fc8SQu Wenruo 29011d6a4fc8SQu Wenruo if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29021d6a4fc8SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { 29031d6a4fc8SQu Wenruo btrfs_info(fs_info, "disabling free space tree"); 29041d6a4fc8SQu Wenruo ret = btrfs_delete_free_space_tree(fs_info); 29051d6a4fc8SQu Wenruo if (ret) { 29061d6a4fc8SQu Wenruo btrfs_warn(fs_info, 29071d6a4fc8SQu Wenruo "failed to disable free space tree: %d", ret); 29088b228324SBoris Burkov goto out; 29098b228324SBoris Burkov } 29108b228324SBoris Burkov } 291144c0ca21SBoris Burkov 29128d488a8cSFilipe Manana /* 29138d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 29148d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 2915fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 29168d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 29178d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 29188d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 29198d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 29208d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 29218d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 29228d488a8cSFilipe Manana */ 29238d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 29248d488a8cSFilipe Manana if (ret) 29258d488a8cSFilipe Manana goto out; 29268d488a8cSFilipe Manana 292744c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 292844c0ca21SBoris Burkov if (ret) 292944c0ca21SBoris Burkov goto out; 293044c0ca21SBoris Burkov 29318f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 29328f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29338f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29348f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29358f1c21d7SBoris Burkov goto out; 29368f1c21d7SBoris Burkov } 29378f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29388f1c21d7SBoris Burkov 293944c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 29407eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 294144c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 294244c0ca21SBoris Burkov if (ret < 0) { 294344c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 294444c0ca21SBoris Burkov goto out; 294544c0ca21SBoris Burkov } 294644c0ca21SBoris Burkov 29475011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 29485011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 29495011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 29505011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 29515011139aSBoris Burkov if (ret) { 29525011139aSBoris Burkov btrfs_warn(fs_info, 29535011139aSBoris Burkov "failed to create free space tree: %d", ret); 29545011139aSBoris Burkov goto out; 29555011139aSBoris Burkov } 29565011139aSBoris Burkov } 29575011139aSBoris Burkov 295894846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 295994846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 296094846229SBoris Burkov if (ret) 296194846229SBoris Burkov goto out; 296294846229SBoris Burkov } 296394846229SBoris Burkov 296444c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 296544c0ca21SBoris Burkov if (ret) 296644c0ca21SBoris Burkov goto out; 296744c0ca21SBoris Burkov 296844c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 296944c0ca21SBoris Burkov if (ret) { 297044c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 297144c0ca21SBoris Burkov goto out; 297244c0ca21SBoris Burkov } 297344c0ca21SBoris Burkov 297444c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 297544c0ca21SBoris Burkov 297644c0ca21SBoris Burkov if (!fs_info->uuid_root) { 297744c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 297844c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 297944c0ca21SBoris Burkov if (ret) { 298044c0ca21SBoris Burkov btrfs_warn(fs_info, 298144c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 298244c0ca21SBoris Burkov goto out; 298344c0ca21SBoris Burkov } 298444c0ca21SBoris Burkov } 298544c0ca21SBoris Burkov 298644c0ca21SBoris Burkov out: 298744c0ca21SBoris Burkov return ret; 298844c0ca21SBoris Burkov } 298944c0ca21SBoris Burkov 2990d7f67ac9SQu Wenruo /* 2991d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 2992d7f67ac9SQu Wenruo * 29932ba48b20SQu Wenruo * @is_rw_mount: If the mount is read-write. 29942ba48b20SQu Wenruo * 2995d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 2996d7f67ac9SQu Wenruo * feature dependencies). 2997d7f67ac9SQu Wenruo * 2998d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 2999d7f67ac9SQu Wenruo * btrfs_validate_super(). 3000d7f67ac9SQu Wenruo * 3001d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3002d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3003d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3004d7f67ac9SQu Wenruo */ 30052ba48b20SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) 3006d7f67ac9SQu Wenruo { 3007d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3008d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3009d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3010d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3011d7f67ac9SQu Wenruo 3012d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3013d7f67ac9SQu Wenruo btrfs_err(fs_info, 3014d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3015d7f67ac9SQu Wenruo incompat); 3016d7f67ac9SQu Wenruo return -EINVAL; 3017d7f67ac9SQu Wenruo } 3018d7f67ac9SQu Wenruo 3019d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3020d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3021d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3022d7f67ac9SQu Wenruo btrfs_err(fs_info, 3023d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3024d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3025d7f67ac9SQu Wenruo return -EINVAL; 3026d7f67ac9SQu Wenruo } 3027d7f67ac9SQu Wenruo 3028d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3029d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3030d7f67ac9SQu Wenruo 3031d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3032d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3033d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3034d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3035d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3036d7f67ac9SQu Wenruo 3037d7f67ac9SQu Wenruo /* 3038d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3039d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3040d7f67ac9SQu Wenruo */ 3041d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3042d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3043d7f67ac9SQu Wenruo 30442ba48b20SQu Wenruo if (compat_ro_unsupp && is_rw_mount) { 3045d7f67ac9SQu Wenruo btrfs_err(fs_info, 3046d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3047d7f67ac9SQu Wenruo compat_ro); 3048d7f67ac9SQu Wenruo return -EINVAL; 3049d7f67ac9SQu Wenruo } 3050d7f67ac9SQu Wenruo 3051d7f67ac9SQu Wenruo /* 3052d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3053d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3054d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3055d7f67ac9SQu Wenruo */ 3056d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3057d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3058d7f67ac9SQu Wenruo btrfs_err(fs_info, 3059d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3060d7f67ac9SQu Wenruo compat_ro); 3061d7f67ac9SQu Wenruo return -EINVAL; 3062d7f67ac9SQu Wenruo } 3063d7f67ac9SQu Wenruo 3064d7f67ac9SQu Wenruo /* 3065d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3066d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3067d7f67ac9SQu Wenruo */ 3068d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3069d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3070d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3071d7f67ac9SQu Wenruo btrfs_err(fs_info, 3072d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3073d7f67ac9SQu Wenruo return -EINVAL; 3074d7f67ac9SQu Wenruo } 3075d7f67ac9SQu Wenruo 3076d7f67ac9SQu Wenruo /* 3077d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3078d7f67ac9SQu Wenruo * 3079d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3080d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3081d7f67ac9SQu Wenruo * going to be deprecated anyway. 3082d7f67ac9SQu Wenruo */ 3083d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3084d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3085d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3086d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3087d7f67ac9SQu Wenruo return -EINVAL; 3088d7f67ac9SQu Wenruo } 3089d7f67ac9SQu Wenruo 3090d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3091d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3092d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3093d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3094d7f67ac9SQu Wenruo 3095d7f67ac9SQu Wenruo return 0; 3096d7f67ac9SQu Wenruo } 3097d7f67ac9SQu Wenruo 3098ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3099ae18c37aSJosef Bacik char *options) 3100ae18c37aSJosef Bacik { 3101ae18c37aSJosef Bacik u32 sectorsize; 3102ae18c37aSJosef Bacik u32 nodesize; 3103ae18c37aSJosef Bacik u32 stripesize; 3104ae18c37aSJosef Bacik u64 generation; 3105ae18c37aSJosef Bacik u64 features; 3106ae18c37aSJosef Bacik u16 csum_type; 3107ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3108ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3109ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3110ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3111ae18c37aSJosef Bacik int ret; 3112ae18c37aSJosef Bacik int level; 3113ae18c37aSJosef Bacik 31148260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 31154871c33bSQu Wenruo if (ret) 3116ae18c37aSJosef Bacik goto fail; 311753b381b3SDavid Woodhouse 3118ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3119ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3120ae18c37aSJosef Bacik GFP_KERNEL); 3121ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3122ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3123ae18c37aSJosef Bacik GFP_KERNEL); 3124ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3125ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 31264871c33bSQu Wenruo ret = -ENOMEM; 3127c75e8394SJosef Bacik goto fail; 3128ae18c37aSJosef Bacik } 3129ae18c37aSJosef Bacik 3130dcb2137cSChristoph Hellwig ret = btrfs_init_btree_inode(sb); 31314871c33bSQu Wenruo if (ret) 3132c75e8394SJosef Bacik goto fail; 3133ae18c37aSJosef Bacik 3134d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 31351104a885SDavid Sterba 31361104a885SDavid Sterba /* 31371104a885SDavid Sterba * Read super block and check the signature bytes only 31381104a885SDavid Sterba */ 3139d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 31408f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 31414871c33bSQu Wenruo ret = PTR_ERR(disk_super); 314216cdcec7SMiao Xie goto fail_alloc; 314320b45077SDave Young } 314439279cc3SChris Mason 31451104a885SDavid Sterba /* 3146260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 31478dc3f22cSJohannes Thumshirn * corrupted, we'll find out 31488dc3f22cSJohannes Thumshirn */ 31498f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 315051bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 31518dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 315251bce6c9SJohannes Thumshirn csum_type); 31534871c33bSQu Wenruo ret = -EINVAL; 31548f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31558dc3f22cSJohannes Thumshirn goto fail_alloc; 31568dc3f22cSJohannes Thumshirn } 31578dc3f22cSJohannes Thumshirn 315883c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 315983c68bbcSSu Yue 31606d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 31616d97c6e3SJohannes Thumshirn if (ret) { 31628f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31636d97c6e3SJohannes Thumshirn goto fail_alloc; 31646d97c6e3SJohannes Thumshirn } 31656d97c6e3SJohannes Thumshirn 31668dc3f22cSJohannes Thumshirn /* 31671104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 31681104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 31691104a885SDavid Sterba */ 31703d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 317105135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 31724871c33bSQu Wenruo ret = -EINVAL; 31738f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3174141386e1SJosef Bacik goto fail_alloc; 31751104a885SDavid Sterba } 31761104a885SDavid Sterba 31771104a885SDavid Sterba /* 31781104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 31791104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 31801104a885SDavid Sterba * the whole block of INFO_SIZE 31811104a885SDavid Sterba */ 31828f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 31838f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31845f39d397SChris Mason 3185fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3186fbc6feaeSNikolay Borisov 31870b86a832SChris Mason 3188fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3189fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3190fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3191fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3192fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3193fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3194fbc6feaeSNikolay Borisov } 3195fbc6feaeSNikolay Borisov 3196fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3197fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3198de37aa51SNikolay Borisov 3199069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 32001104a885SDavid Sterba if (ret) { 320105135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 32024871c33bSQu Wenruo ret = -EINVAL; 3203141386e1SJosef Bacik goto fail_alloc; 32041104a885SDavid Sterba } 32051104a885SDavid Sterba 32064871c33bSQu Wenruo if (!btrfs_super_root(disk_super)) { 32074871c33bSQu Wenruo btrfs_err(fs_info, "invalid superblock tree root bytenr"); 32084871c33bSQu Wenruo ret = -EINVAL; 3209141386e1SJosef Bacik goto fail_alloc; 32104871c33bSQu Wenruo } 32110f7d52f4SChris Mason 3212acce952bSliubo /* check FS state, whether FS is broken. */ 321387533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 321487533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3215acce952bSliubo 321675e7cb7fSLiu Bo /* 321775e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 321875e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 321975e7cb7fSLiu Bo */ 322075e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 322175e7cb7fSLiu Bo 32226f93e834SAnand Jain 32236f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 32246f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 32256f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 32266f93e834SAnand Jain stripesize = sectorsize; 32276f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 32286f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 32296f93e834SAnand Jain 32306f93e834SAnand Jain fs_info->nodesize = nodesize; 32316f93e834SAnand Jain fs_info->sectorsize = sectorsize; 32326f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 32336f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 32346f93e834SAnand Jain fs_info->stripesize = stripesize; 32356f93e834SAnand Jain 32362ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 32374871c33bSQu Wenruo if (ret) 3238141386e1SJosef Bacik goto fail_alloc; 3239dfe25020SChris Mason 32402ba48b20SQu Wenruo ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); 32414871c33bSQu Wenruo if (ret < 0) 3242141386e1SJosef Bacik goto fail_alloc; 3243f2b636e8SJosef Bacik 32448481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 32458481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 32468481dd80SQu Wenruo 32479f73f1aeSQu Wenruo /* 32489f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 32499f73f1aeSQu Wenruo * going to be deprecated. 32509f73f1aeSQu Wenruo * 32519f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 32529f73f1aeSQu Wenruo */ 32539f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 32549f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 32559f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 32569f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 32579f73f1aeSQu Wenruo 325895ea0486SQu Wenruo btrfs_warn(fs_info, 325995ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 32600bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 32618481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 32624871c33bSQu Wenruo if (!subpage_info) { 32634871c33bSQu Wenruo ret = -ENOMEM; 32648481dd80SQu Wenruo goto fail_alloc; 32654871c33bSQu Wenruo } 32668481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 32678481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3268c8050b3bSQu Wenruo } 32690bb3eb3eSQu Wenruo 3270d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 32714871c33bSQu Wenruo if (ret) 32720dc3b84aSJosef Bacik goto fail_sb_buffer; 32734543df7eSChris Mason 32749e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 32759e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 32764575c9ccSChris Mason 3277a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3278a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3279de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3280db94535dSChris Mason 3281925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 32826bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3283925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 328484eed90fSChris Mason if (ret) { 328505135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 32865d4f98a2SYan Zheng goto fail_sb_buffer; 328784eed90fSChris Mason } 32880b86a832SChris Mason 328984234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3290581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3291bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3292bd676446SJosef Bacik generation, level); 3293bd676446SJosef Bacik if (ret) { 329405135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3295af31f5e5SChris Mason goto fail_tree_roots; 329683121942SDavid Woodhouse } 32970b86a832SChris Mason 3298e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3299c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3300c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3301e17cade2SChris Mason 33025b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 33032b82032cSYan Zheng if (ret) { 330405135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3305af31f5e5SChris Mason goto fail_tree_roots; 33062b82032cSYan Zheng } 33070b86a832SChris Mason 33088dabb742SStefan Behrens /* 3309bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3310bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3311bacce86aSAnand Jain * target is required. So free devices that are not part of this 33121a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3313bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 33148dabb742SStefan Behrens */ 3315bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3316d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 331705135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 33184871c33bSQu Wenruo ret = -EIO; 3319a6b0d5c8SChris Mason goto fail_tree_roots; 3320a6b0d5c8SChris Mason } 3321a6b0d5c8SChris Mason 3322b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 33234bbcaa64SEric Sandeen if (ret) 3324b8522a1eSNikolay Borisov goto fail_tree_roots; 33258929ecfaSYan, Zheng 332675ec1db8SJosef Bacik /* 332773651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 332873651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 332973651042SNaohiro Aota * zoned incompat feature flag set. 333073651042SNaohiro Aota */ 333173651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 333273651042SNaohiro Aota if (ret) { 333373651042SNaohiro Aota btrfs_err(fs_info, 33344871c33bSQu Wenruo "zoned: failed to read device zone info: %d", ret); 333573651042SNaohiro Aota goto fail_block_groups; 333673651042SNaohiro Aota } 333773651042SNaohiro Aota 333873651042SNaohiro Aota /* 333975ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 334075ec1db8SJosef Bacik * check the generation here so we can set the 334175ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 334275ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 334375ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 334475ec1db8SJosef Bacik * even though it was perfectly fine. 334575ec1db8SJosef Bacik */ 334675ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 334775ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 334875ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 334975ec1db8SJosef Bacik 3350cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3351cf90d884SQu Wenruo if (ret) { 3352cf90d884SQu Wenruo btrfs_err(fs_info, 3353cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3354cf90d884SQu Wenruo ret); 3355cf90d884SQu Wenruo goto fail_block_groups; 3356cf90d884SQu Wenruo } 335768310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 335868310a5eSIlya Dryomov if (ret) { 335905135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 336068310a5eSIlya Dryomov goto fail_block_groups; 336168310a5eSIlya Dryomov } 336268310a5eSIlya Dryomov 3363733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3364733f4fbbSStefan Behrens if (ret) { 336505135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3366733f4fbbSStefan Behrens goto fail_block_groups; 3367733f4fbbSStefan Behrens } 3368733f4fbbSStefan Behrens 33698dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 33708dabb742SStefan Behrens if (ret) { 337105135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 33728dabb742SStefan Behrens goto fail_block_groups; 33738dabb742SStefan Behrens } 33748dabb742SStefan Behrens 3375b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3376b70f5097SNaohiro Aota if (ret) { 3377b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3378b70f5097SNaohiro Aota ret); 3379b70f5097SNaohiro Aota goto fail_block_groups; 3380b70f5097SNaohiro Aota } 3381b70f5097SNaohiro Aota 3382c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3383b7c35e81SAnand Jain if (ret) { 338405135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 338505135f59SDavid Sterba ret); 3386b7c35e81SAnand Jain goto fail_block_groups; 3387b7c35e81SAnand Jain } 3388b7c35e81SAnand Jain 338996f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 33905ac1d209SJeff Mahoney if (ret) { 339105135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3392b7c35e81SAnand Jain goto fail_fsdev_sysfs; 33935ac1d209SJeff Mahoney } 33945ac1d209SJeff Mahoney 3395c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3396c59021f8Sliubo if (ret) { 339705135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 33982365dd3cSAnand Jain goto fail_sysfs; 3399c59021f8Sliubo } 3400c59021f8Sliubo 34015b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 34021b1d1f66SJosef Bacik if (ret) { 340305135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 34042365dd3cSAnand Jain goto fail_sysfs; 34051b1d1f66SJosef Bacik } 34064330e183SQu Wenruo 340716beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 340816beac87SNaohiro Aota 34095c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 34105c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 341105135f59SDavid Sterba btrfs_warn(fs_info, 341252042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 34134871c33bSQu Wenruo ret = -EINVAL; 34142365dd3cSAnand Jain goto fail_sysfs; 3415292fd7fcSStefan Behrens } 34169078a3e1SChris Mason 341733c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3418a74a4b97SChris Mason "btrfs-cleaner"); 34194871c33bSQu Wenruo if (IS_ERR(fs_info->cleaner_kthread)) { 34204871c33bSQu Wenruo ret = PTR_ERR(fs_info->cleaner_kthread); 34212365dd3cSAnand Jain goto fail_sysfs; 34224871c33bSQu Wenruo } 3423a74a4b97SChris Mason 3424a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3425a74a4b97SChris Mason tree_root, 3426a74a4b97SChris Mason "btrfs-transaction"); 34274871c33bSQu Wenruo if (IS_ERR(fs_info->transaction_kthread)) { 34284871c33bSQu Wenruo ret = PTR_ERR(fs_info->transaction_kthread); 34293f157a2fSChris Mason goto fail_cleaner; 34304871c33bSQu Wenruo } 3431a74a4b97SChris Mason 3432583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3433c289811cSChris Mason !fs_info->fs_devices->rotating) { 3434583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3435c289811cSChris Mason } 3436c289811cSChris Mason 3437572d9ab7SDavid Sterba /* 343863a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 343963a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 344063a7cb13SDavid Sterba * nodiscard for the same mount. 3441*95ca6599SNaohiro Aota * 3442*95ca6599SNaohiro Aota * The zoned mode piggy backs on the discard functionality for 3443*95ca6599SNaohiro Aota * resetting a zone. There is no reason to delay the zone reset as it is 3444*95ca6599SNaohiro Aota * fast enough. So, do not enable async discard for zoned mode. 344563a7cb13SDavid Sterba */ 344663a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 344763a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 344863a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 3449*95ca6599SNaohiro Aota fs_info->fs_devices->discardable && 3450*95ca6599SNaohiro Aota !btrfs_is_zoned(fs_info)) { 345163a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 345263a7cb13SDavid Sterba "auto enabling async discard"); 345363a7cb13SDavid Sterba } 345463a7cb13SDavid Sterba 345521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 34560b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 34572ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 34580b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3459cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 346021adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 346121adbd5cSStefan Behrens if (ret) 346205135f59SDavid Sterba btrfs_warn(fs_info, 346305135f59SDavid Sterba "failed to initialize integrity check module: %d", 346405135f59SDavid Sterba ret); 346521adbd5cSStefan Behrens } 346621adbd5cSStefan Behrens #endif 3467bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3468bcef60f2SArne Jansen if (ret) 3469bcef60f2SArne Jansen goto fail_trans_kthread; 347021adbd5cSStefan Behrens 3471fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3472fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3473fd708b81SJosef Bacik 347496da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 347596da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 34760b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3477e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 347863443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 34794871c33bSQu Wenruo if (ret) 348028c16cbbSWang Shilong goto fail_qgroup; 3481e556ce2cSYan Zheng } 34821a40e23bSZheng Yan 348356e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 34843140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 34854871c33bSQu Wenruo ret = PTR_ERR(fs_info->fs_root); 34864871c33bSQu Wenruo btrfs_warn(fs_info, "failed to read fs tree: %d", ret); 3487315bf8efSJosef Bacik fs_info->fs_root = NULL; 3488bcef60f2SArne Jansen goto fail_qgroup; 34893140c9a3SDan Carpenter } 3490c289811cSChris Mason 3491bc98a42cSDavid Howells if (sb_rdonly(sb)) 34928cd29088SBoris Burkov goto clear_oneshot; 34932b6ba629SIlya Dryomov 349444c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 34952b6ba629SIlya Dryomov if (ret) { 34966bccf3abSJeff Mahoney close_ctree(fs_info); 34972b6ba629SIlya Dryomov return ret; 3498e3acc2a6SJosef Bacik } 3499b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3500b382a324SJan Schmidt 350144c0ca21SBoris Burkov if (fs_info->uuid_root && 350244c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 350344c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 350405135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 350570f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 350670f80175SStefan Behrens if (ret) { 350705135f59SDavid Sterba btrfs_warn(fs_info, 350805135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 35096bccf3abSJeff Mahoney close_ctree(fs_info); 351070f80175SStefan Behrens return ret; 351170f80175SStefan Behrens } 3512f7a81ea4SStefan Behrens } 351394846229SBoris Burkov 3514afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 351547ab2a6cSJosef Bacik 3516b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3517b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3518b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3519b4be6aefSJosef Bacik 35208cd29088SBoris Burkov clear_oneshot: 35218cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3522ad2b2c80SAl Viro return 0; 352339279cc3SChris Mason 3524bcef60f2SArne Jansen fail_qgroup: 3525bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 35267c2ca468SChris Mason fail_trans_kthread: 35277c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 35282ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3529faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 35303f157a2fSChris Mason fail_cleaner: 3531a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 35327c2ca468SChris Mason 35337c2ca468SChris Mason /* 35347c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 35357c2ca468SChris Mason * kthreads 35367c2ca468SChris Mason */ 35377c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 35387c2ca468SChris Mason 35392365dd3cSAnand Jain fail_sysfs: 35406618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 35412365dd3cSAnand Jain 3542b7c35e81SAnand Jain fail_fsdev_sysfs: 3543b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3544b7c35e81SAnand Jain 35451b1d1f66SJosef Bacik fail_block_groups: 354654067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3547af31f5e5SChris Mason 3548af31f5e5SChris Mason fail_tree_roots: 35499e3aa805SJosef Bacik if (fs_info->data_reloc_root) 35509e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 35514273eaffSAnand Jain free_root_pointers(fs_info, true); 35522b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3553af31f5e5SChris Mason 355439279cc3SChris Mason fail_sb_buffer: 35557abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 35565cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 355716cdcec7SMiao Xie fail_alloc: 3558586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3559586e46e2SIlya Dryomov 35604543df7eSChris Mason iput(fs_info->btree_inode); 35617e662854SQinghuang Feng fail: 3562586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 35634871c33bSQu Wenruo ASSERT(ret < 0); 35644871c33bSQu Wenruo return ret; 3565eb60ceacSChris Mason } 3566663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3567eb60ceacSChris Mason 3568314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3569f2984462SChris Mason { 3570314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3571314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3572314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3573314b6dd0SJohannes Thumshirn struct page *page; 3574442a4f63SStefan Behrens 3575314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3576314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3577314b6dd0SJohannes Thumshirn 3578314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3579fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3580314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3581cb3e217bSQu Wenruo btrfs_dev_name(device), 3582314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3583314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3584314b6dd0SJohannes Thumshirn SetPageError(page); 3585314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3586314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3587314b6dd0SJohannes Thumshirn } else { 3588314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3589f2984462SChris Mason } 3590314b6dd0SJohannes Thumshirn 3591314b6dd0SJohannes Thumshirn put_page(page); 3592314b6dd0SJohannes Thumshirn unlock_page(page); 3593314b6dd0SJohannes Thumshirn } 3594314b6dd0SJohannes Thumshirn 3595314b6dd0SJohannes Thumshirn bio_put(bio); 3596f2984462SChris Mason } 3597f2984462SChris Mason 35988f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3599a05d3c91SQu Wenruo int copy_num, bool drop_cache) 360029c36d72SAnand Jain { 360129c36d72SAnand Jain struct btrfs_super_block *super; 36028f32380dSJohannes Thumshirn struct page *page; 360312659251SNaohiro Aota u64 bytenr, bytenr_orig; 36048f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 360512659251SNaohiro Aota int ret; 360629c36d72SAnand Jain 360712659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 360812659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 360912659251SNaohiro Aota if (ret == -ENOENT) 361012659251SNaohiro Aota return ERR_PTR(-EINVAL); 361112659251SNaohiro Aota else if (ret) 361212659251SNaohiro Aota return ERR_PTR(ret); 361312659251SNaohiro Aota 3614cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 36158f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 361629c36d72SAnand Jain 3617a05d3c91SQu Wenruo if (drop_cache) { 3618a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3619a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3620a05d3c91SQu Wenruo 3621a05d3c91SQu Wenruo /* 3622a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3623a05d3c91SQu Wenruo * always read from the device. 3624a05d3c91SQu Wenruo */ 3625a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3626a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3627a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3628a05d3c91SQu Wenruo } 3629a05d3c91SQu Wenruo 36308f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 36318f32380dSJohannes Thumshirn if (IS_ERR(page)) 36328f32380dSJohannes Thumshirn return ERR_CAST(page); 363329c36d72SAnand Jain 36348f32380dSJohannes Thumshirn super = page_address(page); 363596c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 363696c2e067SAnand Jain btrfs_release_disk_super(super); 363796c2e067SAnand Jain return ERR_PTR(-ENODATA); 363896c2e067SAnand Jain } 363996c2e067SAnand Jain 364012659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 36418f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36428f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 364329c36d72SAnand Jain } 364429c36d72SAnand Jain 36458f32380dSJohannes Thumshirn return super; 364629c36d72SAnand Jain } 364729c36d72SAnand Jain 364829c36d72SAnand Jain 36498f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3650a512bbf8SYan Zheng { 36518f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3652a512bbf8SYan Zheng int i; 3653a512bbf8SYan Zheng u64 transid = 0; 3654a512bbf8SYan Zheng 3655a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3656a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3657a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3658a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3659a512bbf8SYan Zheng */ 3660a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3661a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 36628f32380dSJohannes Thumshirn if (IS_ERR(super)) 3663a512bbf8SYan Zheng continue; 3664a512bbf8SYan Zheng 3665a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 36668f32380dSJohannes Thumshirn if (latest) 36678f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36688f32380dSJohannes Thumshirn 36698f32380dSJohannes Thumshirn latest = super; 3670a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3671a512bbf8SYan Zheng } 3672a512bbf8SYan Zheng } 367392fc03fbSAnand Jain 36748f32380dSJohannes Thumshirn return super; 3675a512bbf8SYan Zheng } 3676a512bbf8SYan Zheng 36774eedeb75SHisashi Hifumi /* 3678abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3679314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 36804eedeb75SHisashi Hifumi * 3681abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3682abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3683abbb3b8eSDavid Sterba * same for write and wait phases. 36844eedeb75SHisashi Hifumi * 3685314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 36864eedeb75SHisashi Hifumi */ 3687a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3688abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3689a512bbf8SYan Zheng { 3690d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3691314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3692d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3693a512bbf8SYan Zheng int i; 3694a512bbf8SYan Zheng int errors = 0; 369512659251SNaohiro Aota int ret; 369612659251SNaohiro Aota u64 bytenr, bytenr_orig; 3697a512bbf8SYan Zheng 3698a512bbf8SYan Zheng if (max_mirrors == 0) 3699a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3700a512bbf8SYan Zheng 3701d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3702d5178578SJohannes Thumshirn 3703a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3704314b6dd0SJohannes Thumshirn struct page *page; 3705314b6dd0SJohannes Thumshirn struct bio *bio; 3706314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3707314b6dd0SJohannes Thumshirn 370812659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 370912659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 371012659251SNaohiro Aota if (ret == -ENOENT) { 371112659251SNaohiro Aota continue; 371212659251SNaohiro Aota } else if (ret < 0) { 371312659251SNaohiro Aota btrfs_err(device->fs_info, 371412659251SNaohiro Aota "couldn't get super block location for mirror %d", 371512659251SNaohiro Aota i); 371612659251SNaohiro Aota errors++; 371712659251SNaohiro Aota continue; 371812659251SNaohiro Aota } 3719935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3720935e5cc9SMiao Xie device->commit_total_bytes) 3721a512bbf8SYan Zheng break; 3722a512bbf8SYan Zheng 372312659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3724a512bbf8SYan Zheng 3725fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3726fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3727fd08001fSEric Biggers sb->csum); 3728a512bbf8SYan Zheng 3729314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3730314b6dd0SJohannes Thumshirn GFP_NOFS); 3731314b6dd0SJohannes Thumshirn if (!page) { 3732fb456252SJeff Mahoney btrfs_err(device->fs_info, 3733314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3734f14d104dSDavid Sterba bytenr); 3735634554dcSJosef Bacik errors++; 3736634554dcSJosef Bacik continue; 3737634554dcSJosef Bacik } 3738634554dcSJosef Bacik 3739314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3740314b6dd0SJohannes Thumshirn get_page(page); 3741a512bbf8SYan Zheng 3742314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3743314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3744a512bbf8SYan Zheng 3745387125fcSChris Mason /* 3746314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3747314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3748314b6dd0SJohannes Thumshirn * checking. 3749387125fcSChris Mason */ 375007888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 375107888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 375207888c66SChristoph Hellwig GFP_NOFS); 3753314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3754314b6dd0SJohannes Thumshirn bio->bi_private = device; 3755314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3756314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3757314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3758314b6dd0SJohannes Thumshirn 3759314b6dd0SJohannes Thumshirn /* 3760314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3761314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3762314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3763314b6dd0SJohannes Thumshirn */ 37641b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3765314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3766314b6dd0SJohannes Thumshirn 376758ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 376858ff51f1SChristoph Hellwig submit_bio(bio); 37698376d9e1SNaohiro Aota 37708376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 37718376d9e1SNaohiro Aota errors++; 3772a512bbf8SYan Zheng } 3773a512bbf8SYan Zheng return errors < i ? 0 : -1; 3774a512bbf8SYan Zheng } 3775a512bbf8SYan Zheng 3776387125fcSChris Mason /* 3777abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3778abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3779abbb3b8eSDavid Sterba * 3780314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3781abbb3b8eSDavid Sterba * date. 3782abbb3b8eSDavid Sterba */ 3783abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3784abbb3b8eSDavid Sterba { 3785abbb3b8eSDavid Sterba int i; 3786abbb3b8eSDavid Sterba int errors = 0; 3787b6a535faSHoward McLauchlan bool primary_failed = false; 378812659251SNaohiro Aota int ret; 3789abbb3b8eSDavid Sterba u64 bytenr; 3790abbb3b8eSDavid Sterba 3791abbb3b8eSDavid Sterba if (max_mirrors == 0) 3792abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3793abbb3b8eSDavid Sterba 3794abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3795314b6dd0SJohannes Thumshirn struct page *page; 3796314b6dd0SJohannes Thumshirn 379712659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 379812659251SNaohiro Aota if (ret == -ENOENT) { 379912659251SNaohiro Aota break; 380012659251SNaohiro Aota } else if (ret < 0) { 380112659251SNaohiro Aota errors++; 380212659251SNaohiro Aota if (i == 0) 380312659251SNaohiro Aota primary_failed = true; 380412659251SNaohiro Aota continue; 380512659251SNaohiro Aota } 3806abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3807abbb3b8eSDavid Sterba device->commit_total_bytes) 3808abbb3b8eSDavid Sterba break; 3809abbb3b8eSDavid Sterba 3810314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3811314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3812314b6dd0SJohannes Thumshirn if (!page) { 3813abbb3b8eSDavid Sterba errors++; 3814b6a535faSHoward McLauchlan if (i == 0) 3815b6a535faSHoward McLauchlan primary_failed = true; 3816abbb3b8eSDavid Sterba continue; 3817abbb3b8eSDavid Sterba } 3818314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3819314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3820314b6dd0SJohannes Thumshirn if (PageError(page)) { 3821abbb3b8eSDavid Sterba errors++; 3822b6a535faSHoward McLauchlan if (i == 0) 3823b6a535faSHoward McLauchlan primary_failed = true; 3824b6a535faSHoward McLauchlan } 3825abbb3b8eSDavid Sterba 3826314b6dd0SJohannes Thumshirn /* Drop our reference */ 3827314b6dd0SJohannes Thumshirn put_page(page); 3828abbb3b8eSDavid Sterba 3829314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3830314b6dd0SJohannes Thumshirn put_page(page); 3831abbb3b8eSDavid Sterba } 3832abbb3b8eSDavid Sterba 3833b6a535faSHoward McLauchlan /* log error, force error return */ 3834b6a535faSHoward McLauchlan if (primary_failed) { 3835b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3836b6a535faSHoward McLauchlan device->devid); 3837b6a535faSHoward McLauchlan return -1; 3838b6a535faSHoward McLauchlan } 3839b6a535faSHoward McLauchlan 3840abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3841abbb3b8eSDavid Sterba } 3842abbb3b8eSDavid Sterba 3843abbb3b8eSDavid Sterba /* 3844387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3845387125fcSChris Mason * for the barrier when it is done 3846387125fcSChris Mason */ 38474246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3848387125fcSChris Mason { 3849f9e69aa9SChristoph Hellwig bio_uninit(bio); 3850387125fcSChris Mason complete(bio->bi_private); 3851387125fcSChris Mason } 3852387125fcSChris Mason 3853387125fcSChris Mason /* 38544fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 38554fc6441aSAnand Jain * done in the waiting counterpart. 3856387125fcSChris Mason */ 38574fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3858387125fcSChris Mason { 3859f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 3860387125fcSChris Mason 3861bfd3ea94SAnand Jain device->last_flush_error = BLK_STS_OK; 3862bfd3ea94SAnand Jain 3863a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3864a91cf0ffSWang Yugui /* 3865a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 3866a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 3867a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 3868a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 3869a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 3870a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 3871a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 3872a91cf0ffSWang Yugui * non-debug builds. 3873a91cf0ffSWang Yugui */ 387408e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 38754fc6441aSAnand Jain return; 3876a91cf0ffSWang Yugui #endif 3877387125fcSChris Mason 3878f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 3879f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 3880387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3881387125fcSChris Mason init_completion(&device->flush_wait); 3882387125fcSChris Mason bio->bi_private = &device->flush_wait; 3883387125fcSChris Mason 388458ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 388558ff51f1SChristoph Hellwig submit_bio(bio); 38861c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 38874fc6441aSAnand Jain } 3888387125fcSChris Mason 38894fc6441aSAnand Jain /* 38904fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 38911b465784SAnand Jain * Return true for any error, and false otherwise. 38924fc6441aSAnand Jain */ 38931b465784SAnand Jain static bool wait_dev_flush(struct btrfs_device *device) 38944fc6441aSAnand Jain { 3895f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 38964fc6441aSAnand Jain 38977e812f20SAnand Jain if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 38981b465784SAnand Jain return false; 38994fc6441aSAnand Jain 39002980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 39014fc6441aSAnand Jain 3902bfd3ea94SAnand Jain if (bio->bi_status) { 3903bfd3ea94SAnand Jain device->last_flush_error = bio->bi_status; 3904bfd3ea94SAnand Jain btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); 39051b465784SAnand Jain return true; 3906bfd3ea94SAnand Jain } 3907bfd3ea94SAnand Jain 39081b465784SAnand Jain return false; 3909387125fcSChris Mason } 3910387125fcSChris Mason 3911387125fcSChris Mason /* 3912387125fcSChris Mason * send an empty flush down to each device in parallel, 3913387125fcSChris Mason * then wait for them 3914387125fcSChris Mason */ 3915387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3916387125fcSChris Mason { 3917387125fcSChris Mason struct list_head *head; 3918387125fcSChris Mason struct btrfs_device *dev; 39195af3e8ccSStefan Behrens int errors_wait = 0; 3920387125fcSChris Mason 39211538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3922387125fcSChris Mason /* send down all the barriers */ 3923387125fcSChris Mason head = &info->fs_devices->devices; 39241538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3925e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3926f88ba6a2SHidetoshi Seto continue; 3927cea7c8bfSAnand Jain if (!dev->bdev) 3928387125fcSChris Mason continue; 3929e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3930ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3931387125fcSChris Mason continue; 3932387125fcSChris Mason 39334fc6441aSAnand Jain write_dev_flush(dev); 3934387125fcSChris Mason } 3935387125fcSChris Mason 3936387125fcSChris Mason /* wait for all the barriers */ 39371538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3938e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3939f88ba6a2SHidetoshi Seto continue; 3940387125fcSChris Mason if (!dev->bdev) { 39415af3e8ccSStefan Behrens errors_wait++; 3942387125fcSChris Mason continue; 3943387125fcSChris Mason } 3944e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3945ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3946387125fcSChris Mason continue; 3947387125fcSChris Mason 39481b465784SAnand Jain if (wait_dev_flush(dev)) 39495af3e8ccSStefan Behrens errors_wait++; 3950387125fcSChris Mason } 3951401b41e5SAnand Jain 3952401b41e5SAnand Jain /* 3953de38a206SAnand Jain * Checks last_flush_error of disks in order to determine the device 3954de38a206SAnand Jain * state. 3955401b41e5SAnand Jain */ 3956de38a206SAnand Jain if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) 3957de38a206SAnand Jain return -EIO; 3958de38a206SAnand Jain 3959387125fcSChris Mason return 0; 3960387125fcSChris Mason } 3961387125fcSChris Mason 3962943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3963943c6e99SZhao Lei { 39648789f4feSZhao Lei int raid_type; 39658789f4feSZhao Lei int min_tolerated = INT_MAX; 3966943c6e99SZhao Lei 39678789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 39688789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 39698c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 39708789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 39718789f4feSZhao Lei tolerated_failures); 3972943c6e99SZhao Lei 39738789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 39748789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 39758789f4feSZhao Lei continue; 397641a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 39778789f4feSZhao Lei continue; 39788c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 39798789f4feSZhao Lei btrfs_raid_array[raid_type]. 39808789f4feSZhao Lei tolerated_failures); 39818789f4feSZhao Lei } 3982943c6e99SZhao Lei 39838789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3984ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 39858789f4feSZhao Lei min_tolerated = 0; 39868789f4feSZhao Lei } 39878789f4feSZhao Lei 39888789f4feSZhao Lei return min_tolerated; 3989943c6e99SZhao Lei } 3990943c6e99SZhao Lei 3991eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3992f2984462SChris Mason { 3993e5e9a520SChris Mason struct list_head *head; 3994f2984462SChris Mason struct btrfs_device *dev; 3995a061fc8dSChris Mason struct btrfs_super_block *sb; 3996f2984462SChris Mason struct btrfs_dev_item *dev_item; 3997f2984462SChris Mason int ret; 3998f2984462SChris Mason int do_barriers; 3999a236aed1SChris Mason int max_errors; 4000a236aed1SChris Mason int total_errors = 0; 4001a061fc8dSChris Mason u64 flags; 4002f2984462SChris Mason 40030b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4004fed3b381SLiu Bo 4005fed3b381SLiu Bo /* 4006fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4007fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4008fed3b381SLiu Bo * been consistent on disk. 4009fed3b381SLiu Bo */ 4010fed3b381SLiu Bo if (max_mirrors == 0) 40110b246afaSJeff Mahoney backup_super_roots(fs_info); 4012f2984462SChris Mason 40130b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4014a061fc8dSChris Mason dev_item = &sb->dev_item; 4015e5e9a520SChris Mason 40160b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 40170b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 40180b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4019387125fcSChris Mason 40205af3e8ccSStefan Behrens if (do_barriers) { 40210b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 40225af3e8ccSStefan Behrens if (ret) { 40235af3e8ccSStefan Behrens mutex_unlock( 40240b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 40250b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 40265af3e8ccSStefan Behrens "errors while submitting device barriers."); 40275af3e8ccSStefan Behrens return ret; 40285af3e8ccSStefan Behrens } 40295af3e8ccSStefan Behrens } 4030387125fcSChris Mason 40311538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4032dfe25020SChris Mason if (!dev->bdev) { 4033dfe25020SChris Mason total_errors++; 4034dfe25020SChris Mason continue; 4035dfe25020SChris Mason } 4036e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4037ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4038dfe25020SChris Mason continue; 4039dfe25020SChris Mason 40402b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4041a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4042a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 40437df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4044935e5cc9SMiao Xie dev->commit_total_bytes); 4045ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4046ce7213c7SMiao Xie dev->commit_bytes_used); 4047a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4048a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4049a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4050a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 40517239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 40527239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4053a512bbf8SYan Zheng 4054a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4055a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4056f2984462SChris Mason 405775cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 405875cb857dSQu Wenruo if (ret < 0) { 405975cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 406075cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 406175cb857dSQu Wenruo "unexpected superblock corruption detected"); 406275cb857dSQu Wenruo return -EUCLEAN; 406375cb857dSQu Wenruo } 406475cb857dSQu Wenruo 4065abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4066a236aed1SChris Mason if (ret) 4067a236aed1SChris Mason total_errors++; 4068f2984462SChris Mason } 4069a236aed1SChris Mason if (total_errors > max_errors) { 40700b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4071d397712bSChris Mason total_errors); 40720b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 407379787eaaSJeff Mahoney 40749d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 40750b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 40760b246afaSJeff Mahoney "%d errors while writing supers", 40770b246afaSJeff Mahoney total_errors); 40789d565ba4SStefan Behrens return -EIO; 4079a236aed1SChris Mason } 4080f2984462SChris Mason 4081a512bbf8SYan Zheng total_errors = 0; 40821538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4083dfe25020SChris Mason if (!dev->bdev) 4084dfe25020SChris Mason continue; 4085e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4086ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4087dfe25020SChris Mason continue; 4088dfe25020SChris Mason 4089abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4090a512bbf8SYan Zheng if (ret) 40911259ab75SChris Mason total_errors++; 4092f2984462SChris Mason } 40930b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4094a236aed1SChris Mason if (total_errors > max_errors) { 40950b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 40960b246afaSJeff Mahoney "%d errors while writing supers", 40970b246afaSJeff Mahoney total_errors); 409879787eaaSJeff Mahoney return -EIO; 4099a236aed1SChris Mason } 4100f2984462SChris Mason return 0; 4101f2984462SChris Mason } 4102f2984462SChris Mason 4103cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4104cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4105cb517eabSMiao Xie struct btrfs_root *root) 41062619ba1fSChris Mason { 41074785e24fSJosef Bacik bool drop_ref = false; 41084785e24fSJosef Bacik 4109fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4110fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4111fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4112fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 41134785e24fSJosef Bacik drop_ref = true; 4114fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 411576dda93cSYan, Zheng 411684961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4117ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 41181c1ea4f7SLiu Bo if (root->reloc_root) { 411900246528SJosef Bacik btrfs_put_root(root->reloc_root); 41201c1ea4f7SLiu Bo root->reloc_root = NULL; 41211c1ea4f7SLiu Bo } 41221c1ea4f7SLiu Bo } 41233321719eSLiu Bo 41244785e24fSJosef Bacik if (drop_ref) 412500246528SJosef Bacik btrfs_put_root(root); 41262619ba1fSChris Mason } 41272619ba1fSChris Mason 4128c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4129c146afadSYan Zheng { 4130fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4131fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4132fc7cbcd4SDavid Sterba int i = 0; 413365d33fd7SQu Wenruo int err = 0; 4134fc7cbcd4SDavid Sterba unsigned int ret = 0; 4135c146afadSYan Zheng 4136c146afadSYan Zheng while (1) { 4137fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4138fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4139fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4140fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4141fc7cbcd4SDavid Sterba if (!ret) { 4142fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4143c146afadSYan Zheng break; 414465d33fd7SQu Wenruo } 4145fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 414666b4ffd1SJosef Bacik 4147fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4148fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4149fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4150fc7cbcd4SDavid Sterba gang[i] = NULL; 415165d33fd7SQu Wenruo continue; 4152c146afadSYan Zheng } 4153fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4154fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4155fc7cbcd4SDavid Sterba } 4156fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4157fc7cbcd4SDavid Sterba 4158fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4159fc7cbcd4SDavid Sterba if (!gang[i]) 4160fc7cbcd4SDavid Sterba continue; 4161fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4162fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4163fc7cbcd4SDavid Sterba if (err) 41646989627dSJosef Bacik goto out; 4165fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4166fc7cbcd4SDavid Sterba } 4167fc7cbcd4SDavid Sterba root_objectid++; 4168c146afadSYan Zheng } 41696989627dSJosef Bacik out: 4170fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4171fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4172fc7cbcd4SDavid Sterba if (gang[i]) 4173fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 417465d33fd7SQu Wenruo } 417565d33fd7SQu Wenruo return err; 4176c146afadSYan Zheng } 4177c146afadSYan Zheng 41786bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4179c146afadSYan Zheng { 41806bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4181c146afadSYan Zheng struct btrfs_trans_handle *trans; 4182c146afadSYan Zheng 41830b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 41842ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 41850b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 41860b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4187c71bf099SYan, Zheng 4188c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 41890b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 41900b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4191c71bf099SYan, Zheng 41927a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 41933612b495STsutomu Itoh if (IS_ERR(trans)) 41943612b495STsutomu Itoh return PTR_ERR(trans); 41953a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4196c146afadSYan Zheng } 4197c146afadSYan Zheng 419836c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 419936c86a9eSQu Wenruo { 420036c86a9eSQu Wenruo struct btrfs_transaction *trans; 420136c86a9eSQu Wenruo struct btrfs_transaction *tmp; 420236c86a9eSQu Wenruo bool found = false; 420336c86a9eSQu Wenruo 420436c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 420536c86a9eSQu Wenruo return; 420636c86a9eSQu Wenruo 420736c86a9eSQu Wenruo /* 420836c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 420936c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 421036c86a9eSQu Wenruo */ 421136c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 421236c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 421336c86a9eSQu Wenruo struct extent_state *cached = NULL; 421436c86a9eSQu Wenruo u64 dirty_bytes = 0; 421536c86a9eSQu Wenruo u64 cur = 0; 421636c86a9eSQu Wenruo u64 found_start; 421736c86a9eSQu Wenruo u64 found_end; 421836c86a9eSQu Wenruo 421936c86a9eSQu Wenruo found = true; 422036c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 422136c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 422236c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 422336c86a9eSQu Wenruo cur = found_end + 1; 422436c86a9eSQu Wenruo } 422536c86a9eSQu Wenruo btrfs_warn(fs_info, 422636c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 422736c86a9eSQu Wenruo trans->transid, dirty_bytes); 422836c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 422936c86a9eSQu Wenruo 423036c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 423136c86a9eSQu Wenruo fs_info->running_transaction = NULL; 423236c86a9eSQu Wenruo list_del_init(&trans->list); 423336c86a9eSQu Wenruo 423436c86a9eSQu Wenruo btrfs_put_transaction(trans); 423536c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 423636c86a9eSQu Wenruo } 423736c86a9eSQu Wenruo ASSERT(!found); 423836c86a9eSQu Wenruo } 423936c86a9eSQu Wenruo 4240b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4241eb60ceacSChris Mason { 4242c146afadSYan Zheng int ret; 4243e089f05cSChris Mason 4244afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 424531e70e52SFilipe Manana 424631e70e52SFilipe Manana /* 42478a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 42488a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 42498a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 42508a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 42518a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 42528a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 42538a1f1e3dSFilipe Manana * return 1. 42548a1f1e3dSFilipe Manana */ 42558a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 42568a1f1e3dSFilipe Manana 42578a1f1e3dSFilipe Manana /* 425831e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 425931e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 426031e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 426131e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 426231e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 426331e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 426431e70e52SFilipe Manana * trying to stop the async reclaim task. 426531e70e52SFilipe Manana */ 426631e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4267d6fd0ae2SOmar Sandoval /* 4268d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4269d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4270d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4271d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4272d6fd0ae2SOmar Sandoval */ 4273d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4274a2135011SChris Mason 42757343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4276d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 42777343dd61SJustin Maggard 4278803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4279803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4280803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4281803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4282803b2f54SStefan Behrens 4283837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4284aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4285837d5b6eSIlya Dryomov 42868dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 42878dabb742SStefan Behrens 4288aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 42894cb5300bSChris Mason 42904cb5300bSChris Mason /* wait for any defraggers to finish */ 42914cb5300bSChris Mason wait_event(fs_info->transaction_wait, 42924cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 42934cb5300bSChris Mason 42944cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 429526176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 42964cb5300bSChris Mason 4297a362bb86SFilipe Manana /* 4298a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4299a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4300a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4301a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4302a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4303a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4304a362bb86SFilipe Manana * no one else to run iputs. 4305a362bb86SFilipe Manana * 4306a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4307a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4308a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4309a362bb86SFilipe Manana * through some other means. 4310a362bb86SFilipe Manana * 4311a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4312a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4313a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4314a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4315a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4316a362bb86SFilipe Manana */ 4317a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4318a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4319a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4320a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4321a362bb86SFilipe Manana 432221c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 432357056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4324576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 432521c7e756SMiao Xie 4326b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4327b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4328b0643e59SDennis Zhou 4329bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4330e44163e1SJeff Mahoney /* 4331d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4332d6fd0ae2SOmar Sandoval * unused block groups. 4333e44163e1SJeff Mahoney */ 43340b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4335e44163e1SJeff Mahoney 4336f0cc2cd7SFilipe Manana /* 4337f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4338f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4339f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4340f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4341f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4342f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4343f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4344f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4345f0cc2cd7SFilipe Manana * This is a very rare case. 4346f0cc2cd7SFilipe Manana */ 4347f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4348f0cc2cd7SFilipe Manana 43496bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4350d397712bSChris Mason if (ret) 435104892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4352c146afadSYan Zheng } 4353ed2ff2cbSChris Mason 435484961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 43552ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4356acce952bSliubo 4357e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4358e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 43598929ecfaSYan, Zheng 4360e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4361afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4362f25784b3SYan Zheng 43635958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 43645958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 43655958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 43665958253cSQu Wenruo } 43675958253cSQu Wenruo 436804892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4369fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4370bcef60f2SArne Jansen 4371963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 437204892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4373963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4374b0c68f8bSChris Mason } 437531153d81SYan Zheng 43765deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 43774297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 43785deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 43794297ff84SJosef Bacik 43806618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4381b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 43825ac1d209SJeff Mahoney 43831a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 43841a4319ccSLiu Bo 4385de348ee0SWang Shilong /* 4386de348ee0SWang Shilong * we must make sure there is not any read request to 4387de348ee0SWang Shilong * submit after we stopping all workers. 4388de348ee0SWang Shilong */ 4389de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 439096192499SJosef Bacik btrfs_stop_all_workers(fs_info); 439196192499SJosef Bacik 43920a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 439336c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 43940a31daa4SFilipe Manana 4395afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 43964273eaffSAnand Jain free_root_pointers(fs_info, true); 43978c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 43989ad6b7bcSChris Mason 43994e19443dSJosef Bacik /* 44004e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 44014e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 44024e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 44034e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 44044e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 44054e19443dSJosef Bacik */ 44064e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 44074e19443dSJosef Bacik 440813e6c37bSJosef Bacik iput(fs_info->btree_inode); 4409d6bfde87SChris Mason 441021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 44110b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 44122ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 441321adbd5cSStefan Behrens #endif 441421adbd5cSStefan Behrens 44150b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 441668c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4417eb60ceacSChris Mason } 4418eb60ceacSChris Mason 44195f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 44205f39d397SChris Mason { 44212f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 44225f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4423b4ce94deSChris Mason 442406ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 442506ea65a3SJosef Bacik /* 442606ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 442752042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 442806ea65a3SJosef Bacik * outside of the sanity tests. 442906ea65a3SJosef Bacik */ 4430b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 443106ea65a3SJosef Bacik return; 443206ea65a3SJosef Bacik #endif 443349d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 44340b246afaSJeff Mahoney if (transid != fs_info->generation) 44355d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 44360b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 4437f18cc978SChristoph Hellwig set_extent_buffer_dirty(buf); 44381f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 443969fc6cbbSQu Wenruo /* 444085d8a826SJosef Bacik * btrfs_check_leaf() won't check item data if we don't have WRITTEN 444185d8a826SJosef Bacik * set, so this will only validate the basic structure of the items. 444269fc6cbbSQu Wenruo */ 444385d8a826SJosef Bacik if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(buf)) { 4444a4f78750SDavid Sterba btrfs_print_leaf(buf); 44451f21ef0aSFilipe Manana ASSERT(0); 44461f21ef0aSFilipe Manana } 44471f21ef0aSFilipe Manana #endif 4448eb60ceacSChris Mason } 4449eb60ceacSChris Mason 44502ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4451b53d3f5dSLiu Bo int flush_delayed) 445235b7e476SChris Mason { 4453188de649SChris Mason /* 4454188de649SChris Mason * looks as though older kernels can get into trouble with 4455188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4456188de649SChris Mason */ 4457e2d84521SMiao Xie int ret; 4458d6bfde87SChris Mason 44596933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4460d6bfde87SChris Mason return; 4461d6bfde87SChris Mason 4462b53d3f5dSLiu Bo if (flush_delayed) 44632ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 446416cdcec7SMiao Xie 4465d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4466d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4467d814a491SEthan Lien fs_info->dirty_metadata_batch); 4468e2d84521SMiao Xie if (ret > 0) { 44690b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 447016cdcec7SMiao Xie } 447116cdcec7SMiao Xie } 447216cdcec7SMiao Xie 44732ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 447416cdcec7SMiao Xie { 44752ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 447635b7e476SChris Mason } 4477b53d3f5dSLiu Bo 44782ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4479b53d3f5dSLiu Bo { 44802ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4481d6bfde87SChris Mason } 44826b80053dSChris Mason 44832ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4484acce952bSliubo { 4485fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4486fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4487fe816d0fSNikolay Borisov 44880b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 44892ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 44900b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4491acce952bSliubo 44920b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 44930b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4494acce952bSliubo } 4495acce952bSliubo 4496ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4497ef67963dSJosef Bacik { 4498fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4499fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4500fc7cbcd4SDavid Sterba int ret; 4501ef67963dSJosef Bacik 4502fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4503fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4504fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4505fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4506fc7cbcd4SDavid Sterba int i; 4507ef67963dSJosef Bacik 4508fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4509fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4510fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4511fc7cbcd4SDavid Sterba 4512fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4513fc7cbcd4SDavid Sterba if (!gang[i]) 4514ef67963dSJosef Bacik continue; 4515fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4516fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4517fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4518ef67963dSJosef Bacik } 4519fc7cbcd4SDavid Sterba root_objectid++; 4520fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4521ef67963dSJosef Bacik } 4522fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4523ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4524ef67963dSJosef Bacik } 4525ef67963dSJosef Bacik 4526143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4527acce952bSliubo { 4528acce952bSliubo struct btrfs_ordered_extent *ordered; 4529acce952bSliubo 4530199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4531779880efSJosef Bacik /* 4532779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4533779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4534779880efSJosef Bacik */ 4535199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4536779880efSJosef Bacik root_extent_list) 4537779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4538199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4539199c2a9cSMiao Xie } 4540199c2a9cSMiao Xie 4541199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4542199c2a9cSMiao Xie { 4543199c2a9cSMiao Xie struct btrfs_root *root; 4544199c2a9cSMiao Xie struct list_head splice; 4545199c2a9cSMiao Xie 4546199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4547199c2a9cSMiao Xie 4548199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4549199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4550199c2a9cSMiao Xie while (!list_empty(&splice)) { 4551199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4552199c2a9cSMiao Xie ordered_root); 45531de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 45541de2cfdeSJosef Bacik &fs_info->ordered_roots); 4555199c2a9cSMiao Xie 45562a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4557199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4558199c2a9cSMiao Xie 45592a85d9caSLiu Bo cond_resched(); 45602a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4561199c2a9cSMiao Xie } 4562199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 456374d5d229SJosef Bacik 456474d5d229SJosef Bacik /* 456574d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 456674d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 456774d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 456874d5d229SJosef Bacik * actually get run and error out properly. 456974d5d229SJosef Bacik */ 457074d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4571acce952bSliubo } 4572acce952bSliubo 457399f09ce3SFilipe Manana static void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 45742ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4575acce952bSliubo { 4576acce952bSliubo struct rb_node *node; 4577acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4578acce952bSliubo struct btrfs_delayed_ref_node *ref; 4579acce952bSliubo 4580acce952bSliubo delayed_refs = &trans->delayed_refs; 4581acce952bSliubo 4582acce952bSliubo spin_lock(&delayed_refs->lock); 4583d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4584cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4585b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 458699f09ce3SFilipe Manana return; 4587acce952bSliubo } 4588acce952bSliubo 45895c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4590d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 45910e0adbcfSJosef Bacik struct rb_node *n; 4592e78417d1SJosef Bacik bool pin_bytes = false; 4593acce952bSliubo 4594d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4595d7df2c79SJosef Bacik href_node); 45963069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4597b939d1abSJosef Bacik continue; 45983069bd26SJosef Bacik 4599d7df2c79SJosef Bacik spin_lock(&head->lock); 4600e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 46010e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 46020e0adbcfSJosef Bacik ref_node); 4603e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 46040e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 46051d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 46061d57ee94SWang Xiaoguang list_del(&ref->add_list); 4607d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4608d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4609d7df2c79SJosef Bacik } 461054067ae9SJosef Bacik if (head->must_insert_reserved) 4611e78417d1SJosef Bacik pin_bytes = true; 461278a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4613fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4614d7df2c79SJosef Bacik spin_unlock(&head->lock); 4615acce952bSliubo spin_unlock(&delayed_refs->lock); 4616e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4617acce952bSliubo 4618f603bb94SNikolay Borisov if (pin_bytes) { 4619f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4620f603bb94SNikolay Borisov 4621f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4622f603bb94SNikolay Borisov BUG_ON(!cache); 4623f603bb94SNikolay Borisov 4624f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4625f603bb94SNikolay Borisov spin_lock(&cache->lock); 4626f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4627f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4628f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4629f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4630f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4631f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4632f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4633f603bb94SNikolay Borisov 4634f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4635f603bb94SNikolay Borisov 4636f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4637f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4638f603bb94SNikolay Borisov } 463931890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4640d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4641acce952bSliubo cond_resched(); 4642acce952bSliubo spin_lock(&delayed_refs->lock); 4643acce952bSliubo } 464481f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4645acce952bSliubo 4646acce952bSliubo spin_unlock(&delayed_refs->lock); 4647acce952bSliubo } 4648acce952bSliubo 4649143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4650acce952bSliubo { 4651acce952bSliubo struct btrfs_inode *btrfs_inode; 4652acce952bSliubo struct list_head splice; 4653acce952bSliubo 4654acce952bSliubo INIT_LIST_HEAD(&splice); 4655acce952bSliubo 4656eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4657eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4658acce952bSliubo 4659acce952bSliubo while (!list_empty(&splice)) { 4660fe816d0fSNikolay Borisov struct inode *inode = NULL; 4661eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4662acce952bSliubo delalloc_inodes); 4663fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4664eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4665acce952bSliubo 4666fe816d0fSNikolay Borisov /* 4667fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4668fe816d0fSNikolay Borisov * meanwhile. 4669fe816d0fSNikolay Borisov */ 4670fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4671fe816d0fSNikolay Borisov if (inode) { 4672597441b3SJosef Bacik unsigned int nofs_flag; 4673597441b3SJosef Bacik 4674597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4675fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4676597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4677fe816d0fSNikolay Borisov iput(inode); 4678fe816d0fSNikolay Borisov } 4679eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4680acce952bSliubo } 4681eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4682eb73c1b7SMiao Xie } 4683eb73c1b7SMiao Xie 4684eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4685eb73c1b7SMiao Xie { 4686eb73c1b7SMiao Xie struct btrfs_root *root; 4687eb73c1b7SMiao Xie struct list_head splice; 4688eb73c1b7SMiao Xie 4689eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4690eb73c1b7SMiao Xie 4691eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4692eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4693eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4694eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4695eb73c1b7SMiao Xie delalloc_root); 469600246528SJosef Bacik root = btrfs_grab_root(root); 4697eb73c1b7SMiao Xie BUG_ON(!root); 4698eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4699eb73c1b7SMiao Xie 4700eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 470100246528SJosef Bacik btrfs_put_root(root); 4702eb73c1b7SMiao Xie 4703eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4704eb73c1b7SMiao Xie } 4705eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4706acce952bSliubo } 4707acce952bSliubo 47082ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4709acce952bSliubo struct extent_io_tree *dirty_pages, 4710acce952bSliubo int mark) 4711acce952bSliubo { 4712acce952bSliubo int ret; 4713acce952bSliubo struct extent_buffer *eb; 4714acce952bSliubo u64 start = 0; 4715acce952bSliubo u64 end; 4716acce952bSliubo 4717acce952bSliubo while (1) { 4718acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4719e6138876SJosef Bacik mark, NULL); 4720acce952bSliubo if (ret) 4721acce952bSliubo break; 4722acce952bSliubo 472391166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4724acce952bSliubo while (start <= end) { 47250b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 47260b246afaSJeff Mahoney start += fs_info->nodesize; 4727fd8b2b61SJosef Bacik if (!eb) 4728acce952bSliubo continue; 4729acce952bSliubo 4730c4e54a65SJosef Bacik btrfs_tree_lock(eb); 4731c4e54a65SJosef Bacik wait_on_extent_buffer_writeback(eb); 4732190a8339SJosef Bacik btrfs_clear_buffer_dirty(NULL, eb); 4733c4e54a65SJosef Bacik btrfs_tree_unlock(eb); 4734c4e54a65SJosef Bacik 4735fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4736acce952bSliubo } 4737acce952bSliubo } 4738acce952bSliubo 4739acce952bSliubo return ret; 4740acce952bSliubo } 4741acce952bSliubo 47422ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4743fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4744acce952bSliubo { 4745acce952bSliubo u64 start; 4746acce952bSliubo u64 end; 4747acce952bSliubo int ret; 4748acce952bSliubo 4749acce952bSliubo while (1) { 47500e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 47510e6ec385SFilipe Manana 4752fcd5e742SLu Fengqi /* 4753fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4754fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4755fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4756fcd5e742SLu Fengqi * the same extent range. 4757fcd5e742SLu Fengqi */ 4758fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4759acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 47600e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4761fcd5e742SLu Fengqi if (ret) { 4762fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4763acce952bSliubo break; 4764fcd5e742SLu Fengqi } 4765acce952bSliubo 47660e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 47670e6ec385SFilipe Manana free_extent_state(cached_state); 47682ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4769fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4770acce952bSliubo cond_resched(); 4771acce952bSliubo } 4772acce952bSliubo 4773acce952bSliubo return 0; 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