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 63143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 64acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 652ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 66143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 672ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 68acce952bSliubo struct extent_io_tree *dirty_pages, 69acce952bSliubo int mark); 702ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 71acce952bSliubo struct extent_io_tree *pinned_extents); 722ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 732ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 74ce9adaa5SChris Mason 75141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 76141386e1SJosef Bacik { 77141386e1SJosef Bacik if (fs_info->csum_shash) 78141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 79141386e1SJosef Bacik } 80141386e1SJosef Bacik 81d352ac68SChris Mason /* 822996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 83d352ac68SChris Mason */ 84c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 8519c00ddcSChris Mason { 86d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 877280305eSDavid Sterba const int num_pages = num_extent_pages(buf); 88a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 89d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 9019c00ddcSChris Mason char *kaddr; 91e9be5a30SDavid Sterba int i; 92d5178578SJohannes Thumshirn 93d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 94d5178578SJohannes Thumshirn crypto_shash_init(shash); 95a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 96e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 97a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 9819c00ddcSChris Mason 995ad9b471Spengfuyuan for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) { 100e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 101e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 10219c00ddcSChris Mason } 10371a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 104d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 10519c00ddcSChris Mason } 10619c00ddcSChris Mason 107d352ac68SChris Mason /* 108d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 109d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 110d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 111d352ac68SChris Mason * in the wrong place. 112d352ac68SChris Mason */ 113d87e6575SChristoph Hellwig int btrfs_buffer_uptodate(struct extent_buffer *eb, u64 parent_transid, int atomic) 1141259ab75SChris Mason { 115d87e6575SChristoph Hellwig if (!extent_buffer_uptodate(eb)) 116d87e6575SChristoph Hellwig return 0; 1171259ab75SChris Mason 1181259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 119d87e6575SChristoph Hellwig return 1; 1201259ab75SChris Mason 121b9fab919SChris Mason if (atomic) 122b9fab919SChris Mason return -EAGAIN; 123b9fab919SChris Mason 124d87e6575SChristoph Hellwig if (!extent_buffer_uptodate(eb) || 125d87e6575SChristoph Hellwig btrfs_header_generation(eb) != parent_transid) { 12694647322SDavid Sterba btrfs_err_rl(eb->fs_info, 1278f0ed7d4SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 1288f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 12929549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 1300b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 1319e2aff90SChristoph Hellwig return 0; 132d87e6575SChristoph Hellwig } 1339e2aff90SChristoph Hellwig return 1; 1341259ab75SChris Mason } 1351259ab75SChris Mason 136e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 137e7e16f48SJohannes Thumshirn { 138e7e16f48SJohannes Thumshirn switch (csum_type) { 139e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 1403951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 1413831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 142352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 143e7e16f48SJohannes Thumshirn return true; 144e7e16f48SJohannes Thumshirn default: 145e7e16f48SJohannes Thumshirn return false; 146e7e16f48SJohannes Thumshirn } 147e7e16f48SJohannes Thumshirn } 148e7e16f48SJohannes Thumshirn 149d352ac68SChris Mason /* 1501104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 1511104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 1521104a885SDavid Sterba */ 1533d17adeaSQu Wenruo int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 1543d17adeaSQu Wenruo const struct btrfs_super_block *disk_sb) 1551104a885SDavid Sterba { 15651bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 157d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 158d5178578SJohannes Thumshirn 159d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 1601104a885SDavid Sterba 1611104a885SDavid Sterba /* 1621104a885SDavid Sterba * The super_block structure does not span the whole 16351bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 16451bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 1651104a885SDavid Sterba */ 1663d17adeaSQu Wenruo crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, 167fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 1681104a885SDavid Sterba 16955fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 170e7e16f48SJohannes Thumshirn return 1; 1711104a885SDavid Sterba 172e7e16f48SJohannes Thumshirn return 0; 1731104a885SDavid Sterba } 1741104a885SDavid Sterba 175bacf60e5SChristoph Hellwig static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, 176bacf60e5SChristoph Hellwig int mirror_num) 177bacf60e5SChristoph Hellwig { 178bacf60e5SChristoph Hellwig struct btrfs_fs_info *fs_info = eb->fs_info; 179bacf60e5SChristoph Hellwig int i, num_pages = num_extent_pages(eb); 180bacf60e5SChristoph Hellwig int ret = 0; 181bacf60e5SChristoph Hellwig 182bacf60e5SChristoph Hellwig if (sb_rdonly(fs_info->sb)) 183bacf60e5SChristoph Hellwig return -EROFS; 184bacf60e5SChristoph Hellwig 185bacf60e5SChristoph Hellwig for (i = 0; i < num_pages; i++) { 186bacf60e5SChristoph Hellwig struct page *p = eb->pages[i]; 187917ac778SQu Wenruo u64 start = max_t(u64, eb->start, page_offset(p)); 188917ac778SQu Wenruo u64 end = min_t(u64, eb->start + eb->len, page_offset(p) + PAGE_SIZE); 189917ac778SQu Wenruo u32 len = end - start; 190bacf60e5SChristoph Hellwig 191917ac778SQu Wenruo ret = btrfs_repair_io_failure(fs_info, 0, start, len, 192917ac778SQu Wenruo start, p, offset_in_page(start), mirror_num); 193bacf60e5SChristoph Hellwig if (ret) 194bacf60e5SChristoph Hellwig break; 195bacf60e5SChristoph Hellwig } 196bacf60e5SChristoph Hellwig 197bacf60e5SChristoph Hellwig return ret; 198bacf60e5SChristoph Hellwig } 199bacf60e5SChristoph Hellwig 2001104a885SDavid Sterba /* 201d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 202d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 203581c1760SQu Wenruo * 204789d6a3aSQu Wenruo * @check: expected tree parentness check, see the comments of the 205789d6a3aSQu Wenruo * structure for details. 206d352ac68SChris Mason */ 2076a2e9dc4SFilipe Manana int btrfs_read_extent_buffer(struct extent_buffer *eb, 208789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 209f188591eSChris Mason { 2105ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 211ea466794SJosef Bacik int failed = 0; 212f188591eSChris Mason int ret; 213f188591eSChris Mason int num_copies = 0; 214f188591eSChris Mason int mirror_num = 0; 215ea466794SJosef Bacik int failed_mirror = 0; 216f188591eSChris Mason 217789d6a3aSQu Wenruo ASSERT(check); 218789d6a3aSQu Wenruo 219f188591eSChris Mason while (1) { 220f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 221947a6299SQu Wenruo ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num, check); 222947a6299SQu Wenruo if (!ret) 223581c1760SQu Wenruo break; 224d397712bSChris Mason 2250b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 226f188591eSChris Mason eb->start, eb->len); 2274235298eSChris Mason if (num_copies == 1) 228ea466794SJosef Bacik break; 2294235298eSChris Mason 2305cf1ab56SJosef Bacik if (!failed_mirror) { 2315cf1ab56SJosef Bacik failed = 1; 2325cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 2335cf1ab56SJosef Bacik } 2345cf1ab56SJosef Bacik 235f188591eSChris Mason mirror_num++; 236ea466794SJosef Bacik if (mirror_num == failed_mirror) 237ea466794SJosef Bacik mirror_num++; 238ea466794SJosef Bacik 2394235298eSChris Mason if (mirror_num > num_copies) 240ea466794SJosef Bacik break; 241f188591eSChris Mason } 242ea466794SJosef Bacik 243c0901581SStefan Behrens if (failed && !ret && failed_mirror) 24420a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 245ea466794SJosef Bacik 246ea466794SJosef Bacik return ret; 247f188591eSChris Mason } 24819c00ddcSChris Mason 24931d89399SChristoph Hellwig /* 25031d89399SChristoph Hellwig * Checksum a dirty tree block before IO. 25131d89399SChristoph Hellwig */ 25231d89399SChristoph Hellwig blk_status_t btree_csum_one_bio(struct btrfs_bio *bbio) 253eca0f6f6SQu Wenruo { 25431d89399SChristoph Hellwig struct extent_buffer *eb = bbio->private; 255eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 25631d89399SChristoph Hellwig u64 found_start = btrfs_header_bytenr(eb); 257eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 258eca0f6f6SQu Wenruo int ret; 259eca0f6f6SQu Wenruo 26031d89399SChristoph Hellwig /* Btree blocks are always contiguous on disk. */ 26131d89399SChristoph Hellwig if (WARN_ON_ONCE(bbio->file_offset != eb->start)) 26231d89399SChristoph Hellwig return BLK_STS_IOERR; 26331d89399SChristoph Hellwig if (WARN_ON_ONCE(bbio->bio.bi_iter.bi_size != eb->len)) 26431d89399SChristoph Hellwig return BLK_STS_IOERR; 26531d89399SChristoph Hellwig 26631d89399SChristoph Hellwig if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 26731d89399SChristoph Hellwig WARN_ON_ONCE(found_start != 0); 26831d89399SChristoph Hellwig return BLK_STS_OK; 26931d89399SChristoph Hellwig } 27031d89399SChristoph Hellwig 27131d89399SChristoph Hellwig if (WARN_ON_ONCE(found_start != eb->start)) 27231d89399SChristoph Hellwig return BLK_STS_IOERR; 27331d89399SChristoph Hellwig if (WARN_ON(!btrfs_page_test_uptodate(fs_info, eb->pages[0], eb->start, 27431d89399SChristoph Hellwig eb->len))) 27531d89399SChristoph Hellwig return BLK_STS_IOERR; 27631d89399SChristoph Hellwig 277eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 278eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 279eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 280eca0f6f6SQu Wenruo csum_tree_block(eb, result); 281eca0f6f6SQu Wenruo 282eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 283eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 284eca0f6f6SQu Wenruo else 28585d8a826SJosef Bacik ret = btrfs_check_leaf(eb); 286eca0f6f6SQu Wenruo 2873777369fSQu Wenruo if (ret < 0) 2883777369fSQu Wenruo goto error; 2893777369fSQu Wenruo 2903777369fSQu Wenruo /* 2913777369fSQu Wenruo * Also check the generation, the eb reached here must be newer than 2923777369fSQu Wenruo * last committed. Or something seriously wrong happened. 2933777369fSQu Wenruo */ 2943777369fSQu Wenruo if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { 2953777369fSQu Wenruo ret = -EUCLEAN; 296eca0f6f6SQu Wenruo btrfs_err(fs_info, 2973777369fSQu Wenruo "block=%llu bad generation, have %llu expect > %llu", 2983777369fSQu Wenruo eb->start, btrfs_header_generation(eb), 2993777369fSQu Wenruo fs_info->last_trans_committed); 3003777369fSQu Wenruo goto error; 301eca0f6f6SQu Wenruo } 302eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 30331d89399SChristoph Hellwig return BLK_STS_OK; 3043777369fSQu Wenruo 3053777369fSQu Wenruo error: 3063777369fSQu Wenruo btrfs_print_tree(eb, 0); 3073777369fSQu Wenruo btrfs_err(fs_info, "block=%llu write time tree block corruption detected", 3083777369fSQu Wenruo eb->start); 30916199ad9SFilipe Manana /* 31016199ad9SFilipe Manana * Be noisy if this is an extent buffer from a log tree. We don't abort 31116199ad9SFilipe Manana * a transaction in case there's a bad log tree extent buffer, we just 31216199ad9SFilipe Manana * fallback to a transaction commit. Still we want to know when there is 31316199ad9SFilipe Manana * a bad log tree extent buffer, as that may signal a bug somewhere. 31416199ad9SFilipe Manana */ 31516199ad9SFilipe Manana WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG) || 31616199ad9SFilipe Manana btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID); 317deb6216fSChristoph Hellwig return errno_to_blk_status(ret); 318deb6216fSChristoph Hellwig } 319deb6216fSChristoph Hellwig 320413fb1bcSAnand Jain static bool check_tree_block_fsid(struct extent_buffer *eb) 3212b82032cSYan Zheng { 322b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 323944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 32444880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 325944d3f9fSNikolay Borisov u8 *metadata_uuid; 3262b82032cSYan Zheng 3279a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 3289a8658e3SDavid Sterba BTRFS_FSID_SIZE); 3297239ff4bSNikolay Borisov /* 330944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 331944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 332944d3f9fSNikolay Borisov * ->fsid in this case is fine 3337239ff4bSNikolay Borisov */ 334944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 3357239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 3367239ff4bSNikolay Borisov else 3377239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 3387239ff4bSNikolay Borisov 339944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 340413fb1bcSAnand Jain return false; 341944d3f9fSNikolay Borisov 342944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 343944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 344413fb1bcSAnand Jain return false; 345944d3f9fSNikolay Borisov 346413fb1bcSAnand Jain return true; 3472b82032cSYan Zheng } 3482b82032cSYan Zheng 34977bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 350046b562bSChristoph Hellwig int btrfs_validate_extent_buffer(struct extent_buffer *eb, 351947a6299SQu Wenruo struct btrfs_tree_parent_check *check) 352ce9adaa5SChris Mason { 35377bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 354ce9adaa5SChris Mason u64 found_start; 35577bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 35677bf40a2SQu Wenruo u8 found_level; 3572996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 358dfd29eedSDavid Sterba const u8 *header_csum; 35977bf40a2SQu Wenruo int ret = 0; 360ea466794SJosef Bacik 361947a6299SQu Wenruo ASSERT(check); 362947a6299SQu Wenruo 363ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 364727011e0SChris Mason if (found_start != eb->start) { 3658f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, 3668f0ed7d4SQu Wenruo "bad tree block start, mirror %u want %llu have %llu", 3678f0ed7d4SQu Wenruo eb->read_mirror, eb->start, found_start); 368f188591eSChris Mason ret = -EIO; 36977bf40a2SQu Wenruo goto out; 370ce9adaa5SChris Mason } 371b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 3728f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", 3738f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 3741259ab75SChris Mason ret = -EIO; 37577bf40a2SQu Wenruo goto out; 3761259ab75SChris Mason } 377ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 3781c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 3798f0ed7d4SQu Wenruo btrfs_err(fs_info, 3808f0ed7d4SQu Wenruo "bad tree block level, mirror %u level %d on logical %llu", 3818f0ed7d4SQu Wenruo eb->read_mirror, btrfs_header_level(eb), eb->start); 3821c24c3ceSJosef Bacik ret = -EIO; 38377bf40a2SQu Wenruo goto out; 3841c24c3ceSJosef Bacik } 385ce9adaa5SChris Mason 386c67b3892SDavid Sterba csum_tree_block(eb, result); 387dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 388dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 389a826d6dcSJosef Bacik 390dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 3912996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 3928f0ed7d4SQu Wenruo "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", 3938f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 394dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 39535be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 39635be8851SJohannes Thumshirn btrfs_header_level(eb)); 3972996e1f8SJohannes Thumshirn ret = -EUCLEAN; 39877bf40a2SQu Wenruo goto out; 3992996e1f8SJohannes Thumshirn } 4002996e1f8SJohannes Thumshirn 401947a6299SQu Wenruo if (found_level != check->level) { 40277177ed1SQu Wenruo btrfs_err(fs_info, 40377177ed1SQu Wenruo "level verify failed on logical %llu mirror %u wanted %u found %u", 40477177ed1SQu Wenruo eb->start, eb->read_mirror, check->level, found_level); 405947a6299SQu Wenruo ret = -EIO; 406947a6299SQu Wenruo goto out; 407947a6299SQu Wenruo } 408947a6299SQu Wenruo if (unlikely(check->transid && 409947a6299SQu Wenruo btrfs_header_generation(eb) != check->transid)) { 410947a6299SQu Wenruo btrfs_err_rl(eb->fs_info, 411947a6299SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 412947a6299SQu Wenruo eb->start, eb->read_mirror, check->transid, 413947a6299SQu Wenruo btrfs_header_generation(eb)); 414947a6299SQu Wenruo ret = -EIO; 415947a6299SQu Wenruo goto out; 416947a6299SQu Wenruo } 417947a6299SQu Wenruo if (check->has_first_key) { 418947a6299SQu Wenruo struct btrfs_key *expect_key = &check->first_key; 419947a6299SQu Wenruo struct btrfs_key found_key; 420947a6299SQu Wenruo 421947a6299SQu Wenruo if (found_level) 422947a6299SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 423947a6299SQu Wenruo else 424947a6299SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 425947a6299SQu Wenruo if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { 426947a6299SQu Wenruo btrfs_err(fs_info, 427947a6299SQu Wenruo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 428947a6299SQu Wenruo eb->start, check->transid, 429947a6299SQu Wenruo expect_key->objectid, 430947a6299SQu Wenruo expect_key->type, expect_key->offset, 431947a6299SQu Wenruo found_key.objectid, found_key.type, 432947a6299SQu Wenruo found_key.offset); 433947a6299SQu Wenruo ret = -EUCLEAN; 434947a6299SQu Wenruo goto out; 435947a6299SQu Wenruo } 436947a6299SQu Wenruo } 437947a6299SQu Wenruo if (check->owner_root) { 438947a6299SQu Wenruo ret = btrfs_check_eb_owner(eb, check->owner_root); 439947a6299SQu Wenruo if (ret < 0) 440947a6299SQu Wenruo goto out; 441947a6299SQu Wenruo } 442947a6299SQu Wenruo 443a826d6dcSJosef Bacik /* 444a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 445a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 446a826d6dcSJosef Bacik * return -EIO. 447a826d6dcSJosef Bacik */ 44885d8a826SJosef Bacik if (found_level == 0 && btrfs_check_leaf(eb)) { 449a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 450a826d6dcSJosef Bacik ret = -EIO; 451a826d6dcSJosef Bacik } 452ce9adaa5SChris Mason 453813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 454053ab70fSLiu Bo ret = -EIO; 455053ab70fSLiu Bo 456aebcc159SChristoph Hellwig if (ret) 45775391f0dSQu Wenruo btrfs_err(fs_info, 4588f0ed7d4SQu Wenruo "read time tree block corruption detected on logical %llu mirror %u", 4598f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 46077bf40a2SQu Wenruo out: 46177bf40a2SQu Wenruo return ret; 46277bf40a2SQu Wenruo } 46377bf40a2SQu Wenruo 4643dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 4658958b551SMatthew Wilcox (Oracle) static int btree_migrate_folio(struct address_space *mapping, 4668958b551SMatthew Wilcox (Oracle) struct folio *dst, struct folio *src, enum migrate_mode mode) 467784b4e29SChris Mason { 468784b4e29SChris Mason /* 469784b4e29SChris Mason * we can't safely write a btree page from here, 470784b4e29SChris Mason * we haven't done the locking hook 471784b4e29SChris Mason */ 4728958b551SMatthew Wilcox (Oracle) if (folio_test_dirty(src)) 473784b4e29SChris Mason return -EAGAIN; 474784b4e29SChris Mason /* 475784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 476784b4e29SChris Mason * We must have no buffers or drop them. 477784b4e29SChris Mason */ 4788958b551SMatthew Wilcox (Oracle) if (folio_get_private(src) && 4798958b551SMatthew Wilcox (Oracle) !filemap_release_folio(src, GFP_KERNEL)) 480784b4e29SChris Mason return -EAGAIN; 48154184650SMatthew Wilcox (Oracle) return migrate_folio(mapping, dst, src, mode); 482784b4e29SChris Mason } 4838958b551SMatthew Wilcox (Oracle) #else 4848958b551SMatthew Wilcox (Oracle) #define btree_migrate_folio NULL 4853dd1462eSJan Beulich #endif 486784b4e29SChris Mason 4870da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 4880da5468fSChris Mason struct writeback_control *wbc) 4890da5468fSChris Mason { 490e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 491e2d84521SMiao Xie int ret; 492e2d84521SMiao Xie 493d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 494448d640bSChris Mason 495448d640bSChris Mason if (wbc->for_kupdate) 496448d640bSChris Mason return 0; 497448d640bSChris Mason 498e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 499b9473439SChris Mason /* this is a bit racy, but that's ok */ 500d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 501d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 502d814a491SEthan Lien fs_info->dirty_metadata_batch); 503e2d84521SMiao Xie if (ret < 0) 504793955bcSChris Mason return 0; 505793955bcSChris Mason } 5060b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 5070da5468fSChris Mason } 5080da5468fSChris Mason 509f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 5105f39d397SChris Mason { 511f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 512f913cff3SMatthew Wilcox (Oracle) return false; 5130c4e538bSDavid Sterba 514f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 515d98237b3SChris Mason } 516d98237b3SChris Mason 517895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 518895586ebSMatthew Wilcox (Oracle) size_t length) 519d98237b3SChris Mason { 520d1310b2eSChris Mason struct extent_io_tree *tree; 521895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 522895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 523f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 524895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 525895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 526895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 527895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 528895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 5299ad6b7bcSChris Mason } 530d98237b3SChris Mason } 531d98237b3SChris Mason 532bb146eb2SJosef Bacik #ifdef DEBUG 5330079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 5340079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 5350079c3b1SMatthew Wilcox (Oracle) { 5360079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 537139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 5380b32f4bbSJosef Bacik struct extent_buffer *eb; 539139e8cd3SQu Wenruo int cur_bit = 0; 5400079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 5410b32f4bbSJosef Bacik 542139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 5430079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 5440b32f4bbSJosef Bacik BUG_ON(!eb); 5450b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 5460b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 54749d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 5480079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 549139e8cd3SQu Wenruo } 5500079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 551139e8cd3SQu Wenruo 552139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 553139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 554139e8cd3SQu Wenruo unsigned long flags; 555139e8cd3SQu Wenruo u64 cur; 556139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 557139e8cd3SQu Wenruo 558139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 559139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 560139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 561139e8cd3SQu Wenruo cur_bit++; 562139e8cd3SQu Wenruo continue; 563139e8cd3SQu Wenruo } 564139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 565139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 566139e8cd3SQu Wenruo 567139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 568139e8cd3SQu Wenruo ASSERT(eb); 569139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 570139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 57149d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 572139e8cd3SQu Wenruo free_extent_buffer(eb); 573139e8cd3SQu Wenruo 574139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 575139e8cd3SQu Wenruo } 5760079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 5770b32f4bbSJosef Bacik } 5780079c3b1SMatthew Wilcox (Oracle) #else 5790079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 5800079c3b1SMatthew Wilcox (Oracle) #endif 5810b32f4bbSJosef Bacik 5827f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 5830da5468fSChris Mason .writepages = btree_writepages, 584f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 585895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 5868958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 5870079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 588d98237b3SChris Mason }; 589123abc88SChris Mason 5902ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 5912ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 5923fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 5933fbaf258SJosef Bacik int level) 5940999df54SChris Mason { 5950b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 5960b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 5973fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 5980999df54SChris Mason } 5990999df54SChris Mason 600581c1760SQu Wenruo /* 601581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 602581c1760SQu Wenruo * verification. 603581c1760SQu Wenruo * 604789d6a3aSQu Wenruo * @check: expected tree parentness check, see comments of the 605789d6a3aSQu Wenruo * structure for details. 606581c1760SQu Wenruo */ 6072ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 608789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 609e20d96d6SChris Mason { 6105f39d397SChris Mason struct extent_buffer *buf = NULL; 61119c00ddcSChris Mason int ret; 61219c00ddcSChris Mason 613789d6a3aSQu Wenruo ASSERT(check); 614789d6a3aSQu Wenruo 615789d6a3aSQu Wenruo buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, 616789d6a3aSQu Wenruo check->level); 617c871b0f2SLiu Bo if (IS_ERR(buf)) 618c871b0f2SLiu Bo return buf; 619e4204dedSChris Mason 620789d6a3aSQu Wenruo ret = btrfs_read_extent_buffer(buf, check); 6210f0fe8f7SFilipe David Borba Manana if (ret) { 622537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 62364c043deSLiu Bo return ERR_PTR(ret); 6240f0fe8f7SFilipe David Borba Manana } 625789d6a3aSQu Wenruo if (btrfs_check_eb_owner(buf, check->owner_root)) { 62688c602abSQu Wenruo free_extent_buffer_stale(buf); 62788c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 62888c602abSQu Wenruo } 6295f39d397SChris Mason return buf; 630ce9adaa5SChris Mason 631eb60ceacSChris Mason } 632eb60ceacSChris Mason 633da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 634e20d96d6SChris Mason u64 objectid) 635d97e63b6SChris Mason { 6367c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 6372e608bd1SJosef Bacik 6382e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 6392e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 6402e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 64196dfcb46SJosef Bacik root->fs_info = fs_info; 6422e608bd1SJosef Bacik root->root_key.objectid = objectid; 643cfaa7295SChris Mason root->node = NULL; 644a28ec197SChris Mason root->commit_root = NULL; 64527cdeb70SMiao Xie root->state = 0; 646abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 6470b86a832SChris Mason 6480f7d52f4SChris Mason root->last_trans = 0; 6496b8fad57SNikolay Borisov root->free_objectid = 0; 650eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 651199c2a9cSMiao Xie root->nr_ordered_extents = 0; 6526bef4d31SEric Paris root->inode_tree = RB_ROOT; 653088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 6542e608bd1SJosef Bacik 6552e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 6560b86a832SChris Mason 6570b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 6585d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 659eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 660eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 661199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 662199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 663d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 6642ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 6652ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 6665d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 667eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 668199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 669f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 6702ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 6712ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 6728287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 673a2135011SChris Mason mutex_init(&root->objectid_mutex); 674e02119d5SChris Mason mutex_init(&root->log_mutex); 67531f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 676573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 677c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 6787237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 6797237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 6807237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 6818b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 6828b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 6837237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 6847237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 6857237f183SYan Zheng atomic_set(&root->log_writers, 0); 6862ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 6870700cea7SElena Reshetova refcount_set(&root->refs, 1); 6888ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 689eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 6907237f183SYan Zheng root->log_transid = 0; 691d1433debSMiao Xie root->log_transid_committed = -1; 692257c62e1SChris Mason root->last_log_commit = 0; 6932e608bd1SJosef Bacik root->anon_dev = 0; 694e289f03eSFilipe Manana if (!dummy) { 69543eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 69635da5a7eSDavid Sterba IO_TREE_ROOT_DIRTY_LOG_PAGES); 697e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 69835da5a7eSDavid Sterba IO_TREE_LOG_CSUM_RANGE); 699e289f03eSFilipe Manana } 700017e5369SChris Mason 7015f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 702370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 703bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 704bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 705fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 706bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 707fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 708bd647ce3SJosef Bacik #endif 7093768f368SChris Mason } 7103768f368SChris Mason 71174e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 71296dfcb46SJosef Bacik u64 objectid, gfp_t flags) 7136f07e42eSAl Viro { 71474e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 7156f07e42eSAl Viro if (root) 71696dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 7176f07e42eSAl Viro return root; 7186f07e42eSAl Viro } 7196f07e42eSAl Viro 72006ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 72106ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 722da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 72306ea65a3SJosef Bacik { 72406ea65a3SJosef Bacik struct btrfs_root *root; 72506ea65a3SJosef Bacik 7267c0260eeSJeff Mahoney if (!fs_info) 7277c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 7287c0260eeSJeff Mahoney 72996dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 73006ea65a3SJosef Bacik if (!root) 73106ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 732da17066cSJeff Mahoney 733b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 734faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 73506ea65a3SJosef Bacik 73606ea65a3SJosef Bacik return root; 73706ea65a3SJosef Bacik } 73806ea65a3SJosef Bacik #endif 73906ea65a3SJosef Bacik 740abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 741abed4aaaSJosef Bacik { 742abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 743abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 744abed4aaaSJosef Bacik 745abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 746abed4aaaSJosef Bacik } 747abed4aaaSJosef Bacik 748abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 749abed4aaaSJosef Bacik { 750abed4aaaSJosef Bacik const struct btrfs_key *key = k; 751abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 752abed4aaaSJosef Bacik 753abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 754abed4aaaSJosef Bacik } 755abed4aaaSJosef Bacik 756abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 757abed4aaaSJosef Bacik { 758abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 759abed4aaaSJosef Bacik struct rb_node *tmp; 760745806fbSQu Wenruo int ret = 0; 761abed4aaaSJosef Bacik 762abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 763abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 764abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 765abed4aaaSJosef Bacik 766745806fbSQu Wenruo if (tmp) { 767745806fbSQu Wenruo ret = -EEXIST; 768745806fbSQu Wenruo btrfs_warn(fs_info, "global root %llu %llu already exists", 769745806fbSQu Wenruo root->root_key.objectid, root->root_key.offset); 770745806fbSQu Wenruo } 771745806fbSQu Wenruo return ret; 772abed4aaaSJosef Bacik } 773abed4aaaSJosef Bacik 774abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 775abed4aaaSJosef Bacik { 776abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 777abed4aaaSJosef Bacik 778abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 779abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 780abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 781abed4aaaSJosef Bacik } 782abed4aaaSJosef Bacik 783abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 784abed4aaaSJosef Bacik struct btrfs_key *key) 785abed4aaaSJosef Bacik { 786abed4aaaSJosef Bacik struct rb_node *node; 787abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 788abed4aaaSJosef Bacik 789abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 790abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 791abed4aaaSJosef Bacik if (node) 792abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 793abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 794abed4aaaSJosef Bacik 795abed4aaaSJosef Bacik return root; 796abed4aaaSJosef Bacik } 797abed4aaaSJosef Bacik 798f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 799f7238e50SJosef Bacik { 800f7238e50SJosef Bacik struct btrfs_block_group *block_group; 801f7238e50SJosef Bacik u64 ret; 802f7238e50SJosef Bacik 803f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 804f7238e50SJosef Bacik return 0; 805f7238e50SJosef Bacik 806f7238e50SJosef Bacik if (bytenr) 807f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 808f7238e50SJosef Bacik else 809f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 810f7238e50SJosef Bacik ASSERT(block_group); 811f7238e50SJosef Bacik if (!block_group) 812f7238e50SJosef Bacik return 0; 813f7238e50SJosef Bacik ret = block_group->global_root_id; 814f7238e50SJosef Bacik btrfs_put_block_group(block_group); 815f7238e50SJosef Bacik 816f7238e50SJosef Bacik return ret; 817f7238e50SJosef Bacik } 818f7238e50SJosef Bacik 819abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 820abed4aaaSJosef Bacik { 821abed4aaaSJosef Bacik struct btrfs_key key = { 822abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 823abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 824f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 825abed4aaaSJosef Bacik }; 826abed4aaaSJosef Bacik 827abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 828abed4aaaSJosef Bacik } 829abed4aaaSJosef Bacik 830abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 831abed4aaaSJosef Bacik { 832abed4aaaSJosef Bacik struct btrfs_key key = { 833abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 834abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 835f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 836abed4aaaSJosef Bacik }; 837abed4aaaSJosef Bacik 838abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 839abed4aaaSJosef Bacik } 840abed4aaaSJosef Bacik 84151129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 84251129b33SJosef Bacik { 84351129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 84451129b33SJosef Bacik return fs_info->block_group_root; 84551129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 84651129b33SJosef Bacik } 84751129b33SJosef Bacik 84820897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 84920897f5cSArne Jansen u64 objectid) 85020897f5cSArne Jansen { 8519b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 85220897f5cSArne Jansen struct extent_buffer *leaf; 85320897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 85420897f5cSArne Jansen struct btrfs_root *root; 85520897f5cSArne Jansen struct btrfs_key key; 856b89f6d1fSFilipe Manana unsigned int nofs_flag; 85720897f5cSArne Jansen int ret = 0; 85820897f5cSArne Jansen 859b89f6d1fSFilipe Manana /* 860b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 861b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 862b89f6d1fSFilipe Manana */ 863b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 86496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 865b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 86620897f5cSArne Jansen if (!root) 86720897f5cSArne Jansen return ERR_PTR(-ENOMEM); 86820897f5cSArne Jansen 86920897f5cSArne Jansen root->root_key.objectid = objectid; 87020897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 87120897f5cSArne Jansen root->root_key.offset = 0; 87220897f5cSArne Jansen 8739631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 8749631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 87520897f5cSArne Jansen if (IS_ERR(leaf)) { 87620897f5cSArne Jansen ret = PTR_ERR(leaf); 8771dd05682STsutomu Itoh leaf = NULL; 878c1b07854SPeng Hao goto fail; 87920897f5cSArne Jansen } 88020897f5cSArne Jansen 88120897f5cSArne Jansen root->node = leaf; 88220897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 88320897f5cSArne Jansen 88420897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 88527cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 88620897f5cSArne Jansen 887f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 888f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 88920897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 89020897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 89120897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 89220897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 89320897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 89420897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 89520897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 89633d85fdaSQu Wenruo if (is_fstree(objectid)) 897807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 898807fc790SAndy Shevchenko else 899807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 900c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 90120897f5cSArne Jansen 9028a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 9038a6a87cdSBoris Burkov 90420897f5cSArne Jansen key.objectid = objectid; 90520897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 90620897f5cSArne Jansen key.offset = 0; 90720897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 90820897f5cSArne Jansen if (ret) 90920897f5cSArne Jansen goto fail; 91020897f5cSArne Jansen 91120897f5cSArne Jansen return root; 9121dd05682STsutomu Itoh 9138a6a87cdSBoris Burkov fail: 91400246528SJosef Bacik btrfs_put_root(root); 9151dd05682STsutomu Itoh 9161dd05682STsutomu Itoh return ERR_PTR(ret); 91720897f5cSArne Jansen } 91820897f5cSArne Jansen 9197237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 920e02119d5SChris Mason struct btrfs_fs_info *fs_info) 9210f7d52f4SChris Mason { 9220f7d52f4SChris Mason struct btrfs_root *root; 923e02119d5SChris Mason 92496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 925e02119d5SChris Mason if (!root) 9267237f183SYan Zheng return ERR_PTR(-ENOMEM); 927e02119d5SChris Mason 928e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 929e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 930e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 93127cdeb70SMiao Xie 9326ab6ebb7SNaohiro Aota return root; 9336ab6ebb7SNaohiro Aota } 9346ab6ebb7SNaohiro Aota 9356ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 9366ab6ebb7SNaohiro Aota struct btrfs_root *root) 9376ab6ebb7SNaohiro Aota { 9386ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 9396ab6ebb7SNaohiro Aota 9407237f183SYan Zheng /* 94192a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 94227cdeb70SMiao Xie * 94392a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 94492a7cc42SQu Wenruo * and they go away before a real commit is actually done. 94592a7cc42SQu Wenruo * 94692a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 94792a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 9487237f183SYan Zheng */ 949e02119d5SChris Mason 9504d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 9519631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 9526ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 9536ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 954e02119d5SChris Mason 9557237f183SYan Zheng root->node = leaf; 956e02119d5SChris Mason 957e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 958e02119d5SChris Mason btrfs_tree_unlock(root->node); 9596ab6ebb7SNaohiro Aota 9606ab6ebb7SNaohiro Aota return 0; 9617237f183SYan Zheng } 9627237f183SYan Zheng 9637237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 9647237f183SYan Zheng struct btrfs_fs_info *fs_info) 9657237f183SYan Zheng { 9667237f183SYan Zheng struct btrfs_root *log_root; 9677237f183SYan Zheng 9687237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 9697237f183SYan Zheng if (IS_ERR(log_root)) 9707237f183SYan Zheng return PTR_ERR(log_root); 9716ab6ebb7SNaohiro Aota 9723ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 9733ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 9743ddebf27SNaohiro Aota 9756ab6ebb7SNaohiro Aota if (ret) { 9766ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 9776ab6ebb7SNaohiro Aota return ret; 9786ab6ebb7SNaohiro Aota } 9793ddebf27SNaohiro Aota } 9806ab6ebb7SNaohiro Aota 9817237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 9827237f183SYan Zheng fs_info->log_root_tree = log_root; 9837237f183SYan Zheng return 0; 9847237f183SYan Zheng } 9857237f183SYan Zheng 9867237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 9877237f183SYan Zheng struct btrfs_root *root) 9887237f183SYan Zheng { 9890b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 9907237f183SYan Zheng struct btrfs_root *log_root; 9917237f183SYan Zheng struct btrfs_inode_item *inode_item; 9926ab6ebb7SNaohiro Aota int ret; 9937237f183SYan Zheng 9940b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 9957237f183SYan Zheng if (IS_ERR(log_root)) 9967237f183SYan Zheng return PTR_ERR(log_root); 9977237f183SYan Zheng 9986ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 9996ab6ebb7SNaohiro Aota if (ret) { 10006ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 10016ab6ebb7SNaohiro Aota return ret; 10026ab6ebb7SNaohiro Aota } 10036ab6ebb7SNaohiro Aota 10047237f183SYan Zheng log_root->last_trans = trans->transid; 10057237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 10067237f183SYan Zheng 10077237f183SYan Zheng inode_item = &log_root->root_item.inode; 10083cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 10093cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 10103cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1011da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 10120b246afaSJeff Mahoney fs_info->nodesize); 10133cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 10147237f183SYan Zheng 10155d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 10167237f183SYan Zheng 10177237f183SYan Zheng WARN_ON(root->log_root); 10187237f183SYan Zheng root->log_root = log_root; 10197237f183SYan Zheng root->log_transid = 0; 1020d1433debSMiao Xie root->log_transid_committed = -1; 1021257c62e1SChris Mason root->last_log_commit = 0; 1022e02119d5SChris Mason return 0; 1023e02119d5SChris Mason } 1024e02119d5SChris Mason 102549d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 102649d11beaSJosef Bacik struct btrfs_path *path, 1027cb517eabSMiao Xie struct btrfs_key *key) 1028e02119d5SChris Mason { 1029e02119d5SChris Mason struct btrfs_root *root; 1030789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 1031e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 103284234f3aSYan Zheng u64 generation; 1033cb517eabSMiao Xie int ret; 1034581c1760SQu Wenruo int level; 1035cb517eabSMiao Xie 103696dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 103749d11beaSJosef Bacik if (!root) 103849d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 10390f7d52f4SChris Mason 1040cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1041cb517eabSMiao Xie &root->root_item, &root->root_key); 10420f7d52f4SChris Mason if (ret) { 104313a8a7c8SYan, Zheng if (ret > 0) 104413a8a7c8SYan, Zheng ret = -ENOENT; 104549d11beaSJosef Bacik goto fail; 10460f7d52f4SChris Mason } 104713a8a7c8SYan, Zheng 104884234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1049581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 1050789d6a3aSQu Wenruo check.level = level; 1051789d6a3aSQu Wenruo check.transid = generation; 1052789d6a3aSQu Wenruo check.owner_root = key->objectid; 1053789d6a3aSQu Wenruo root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), 1054789d6a3aSQu Wenruo &check); 105564c043deSLiu Bo if (IS_ERR(root->node)) { 105664c043deSLiu Bo ret = PTR_ERR(root->node); 10578c38938cSJosef Bacik root->node = NULL; 105849d11beaSJosef Bacik goto fail; 10594eb150d6SQu Wenruo } 10604eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1061cb517eabSMiao Xie ret = -EIO; 106249d11beaSJosef Bacik goto fail; 1063416bc658SJosef Bacik } 106488c602abSQu Wenruo 106588c602abSQu Wenruo /* 106688c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 106788c602abSQu Wenruo * match its root node owner 106888c602abSQu Wenruo */ 106988c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 107088c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 107188c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 107288c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 107388c602abSQu Wenruo btrfs_crit(fs_info, 107488c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 107588c602abSQu Wenruo root->root_key.objectid, root->node->start, 107688c602abSQu Wenruo btrfs_header_owner(root->node), 107788c602abSQu Wenruo root->root_key.objectid); 107888c602abSQu Wenruo ret = -EUCLEAN; 107988c602abSQu Wenruo goto fail; 108088c602abSQu Wenruo } 10815d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1082cb517eabSMiao Xie return root; 108349d11beaSJosef Bacik fail: 108400246528SJosef Bacik btrfs_put_root(root); 108549d11beaSJosef Bacik return ERR_PTR(ret); 108649d11beaSJosef Bacik } 108749d11beaSJosef Bacik 108849d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 108949d11beaSJosef Bacik struct btrfs_key *key) 109049d11beaSJosef Bacik { 109149d11beaSJosef Bacik struct btrfs_root *root; 109249d11beaSJosef Bacik struct btrfs_path *path; 109349d11beaSJosef Bacik 109449d11beaSJosef Bacik path = btrfs_alloc_path(); 109549d11beaSJosef Bacik if (!path) 109649d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 109749d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 109849d11beaSJosef Bacik btrfs_free_path(path); 109949d11beaSJosef Bacik 110049d11beaSJosef Bacik return root; 1101cb517eabSMiao Xie } 1102cb517eabSMiao Xie 11032dfb1e43SQu Wenruo /* 11042dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 11052dfb1e43SQu Wenruo * 11062dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 11072dfb1e43SQu Wenruo */ 11082dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1109cb517eabSMiao Xie { 1110cb517eabSMiao Xie int ret; 1111cb517eabSMiao Xie 11120b548539SDavid Sterba btrfs_drew_lock_init(&root->snapshot_lock); 11138257b2dcSMiao Xie 1114aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 111537f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 111692a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1117f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1118f39e4571SJosef Bacik } 1119f39e4571SJosef Bacik 1120851fd730SQu Wenruo /* 1121851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1122851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1123851fd730SQu Wenruo */ 1124851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1125851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 11262dfb1e43SQu Wenruo if (!anon_dev) { 1127cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1128cb517eabSMiao Xie if (ret) 1129876d2cf1SLiu Bo goto fail; 11302dfb1e43SQu Wenruo } else { 11312dfb1e43SQu Wenruo root->anon_dev = anon_dev; 11322dfb1e43SQu Wenruo } 1133851fd730SQu Wenruo } 1134f32e48e9SChandan Rajendra 1135f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1136453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1137f32e48e9SChandan Rajendra if (ret) { 1138f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1139876d2cf1SLiu Bo goto fail; 1140f32e48e9SChandan Rajendra } 1141f32e48e9SChandan Rajendra 11426b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1143f32e48e9SChandan Rajendra 1144f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1145f32e48e9SChandan Rajendra 1146cb517eabSMiao Xie return 0; 1147cb517eabSMiao Xie fail: 114884db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1149cb517eabSMiao Xie return ret; 1150cb517eabSMiao Xie } 1151cb517eabSMiao Xie 1152a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1153cb517eabSMiao Xie u64 root_id) 1154cb517eabSMiao Xie { 1155cb517eabSMiao Xie struct btrfs_root *root; 1156cb517eabSMiao Xie 1157fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1158fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1159fc7cbcd4SDavid Sterba (unsigned long)root_id); 116000246528SJosef Bacik root = btrfs_grab_root(root); 1161fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1162cb517eabSMiao Xie return root; 1163cb517eabSMiao Xie } 1164cb517eabSMiao Xie 116549d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 116649d11beaSJosef Bacik u64 objectid) 116749d11beaSJosef Bacik { 1168abed4aaaSJosef Bacik struct btrfs_key key = { 1169abed4aaaSJosef Bacik .objectid = objectid, 1170abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1171abed4aaaSJosef Bacik .offset = 0, 1172abed4aaaSJosef Bacik }; 1173abed4aaaSJosef Bacik 1174e91909aaSChristoph Hellwig switch (objectid) { 1175e91909aaSChristoph Hellwig case BTRFS_ROOT_TREE_OBJECTID: 117649d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 1177e91909aaSChristoph Hellwig case BTRFS_EXTENT_TREE_OBJECTID: 1178abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1179e91909aaSChristoph Hellwig case BTRFS_CHUNK_TREE_OBJECTID: 118049d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 1181e91909aaSChristoph Hellwig case BTRFS_DEV_TREE_OBJECTID: 118249d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 1183e91909aaSChristoph Hellwig case BTRFS_CSUM_TREE_OBJECTID: 1184abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1185e91909aaSChristoph Hellwig case BTRFS_QUOTA_TREE_OBJECTID: 118685724171SChristoph Hellwig return btrfs_grab_root(fs_info->quota_root); 1187e91909aaSChristoph Hellwig case BTRFS_UUID_TREE_OBJECTID: 118885724171SChristoph Hellwig return btrfs_grab_root(fs_info->uuid_root); 1189e91909aaSChristoph Hellwig case BTRFS_BLOCK_GROUP_TREE_OBJECTID: 119085724171SChristoph Hellwig return btrfs_grab_root(fs_info->block_group_root); 1191e91909aaSChristoph Hellwig case BTRFS_FREE_SPACE_TREE_OBJECTID: 119285724171SChristoph Hellwig return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 1193e91909aaSChristoph Hellwig default: 119449d11beaSJosef Bacik return NULL; 119549d11beaSJosef Bacik } 1196e91909aaSChristoph Hellwig } 119749d11beaSJosef Bacik 1198cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1199cb517eabSMiao Xie struct btrfs_root *root) 1200cb517eabSMiao Xie { 1201cb517eabSMiao Xie int ret; 1202cb517eabSMiao Xie 1203fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1204fc7cbcd4SDavid Sterba if (ret) 1205fc7cbcd4SDavid Sterba return ret; 1206fc7cbcd4SDavid Sterba 1207fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1208fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1209fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1210fc7cbcd4SDavid Sterba root); 1211af01d2e5SJosef Bacik if (ret == 0) { 121200246528SJosef Bacik btrfs_grab_root(root); 1213fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1214af01d2e5SJosef Bacik } 1215fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1216fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1217cb517eabSMiao Xie 1218cb517eabSMiao Xie return ret; 1219cb517eabSMiao Xie } 1220cb517eabSMiao Xie 1221bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1222bd647ce3SJosef Bacik { 1223bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1224bd647ce3SJosef Bacik struct btrfs_root *root; 1225bd647ce3SJosef Bacik 1226bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1227457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1228457f1864SJosef Bacik 1229bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1230bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1231457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 123271008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1233bd647ce3SJosef Bacik refcount_read(&root->refs)); 1234bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 123500246528SJosef Bacik btrfs_put_root(root); 123600246528SJosef Bacik btrfs_put_root(root); 1237bd647ce3SJosef Bacik } 1238bd647ce3SJosef Bacik #endif 1239bd647ce3SJosef Bacik } 1240bd647ce3SJosef Bacik 1241abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1242abed4aaaSJosef Bacik { 1243abed4aaaSJosef Bacik struct btrfs_root *root; 1244abed4aaaSJosef Bacik struct rb_node *node; 1245abed4aaaSJosef Bacik 1246abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1247abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1248abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1249abed4aaaSJosef Bacik btrfs_put_root(root); 1250abed4aaaSJosef Bacik } 1251abed4aaaSJosef Bacik } 1252abed4aaaSJosef Bacik 12530d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 12540d4b0463SJosef Bacik { 1255141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1256141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 12575deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1258141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1259141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1260141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1261141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 12620d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 12630d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1264abed4aaaSJosef Bacik free_global_roots(fs_info); 126500246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 126600246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 126700246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 126800246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 126900246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 127000246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1271aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 12729c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1273bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 12743fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 12750d4b0463SJosef Bacik kfree(fs_info->super_copy); 12760d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 12778481dd80SQu Wenruo kfree(fs_info->subpage_info); 12780d4b0463SJosef Bacik kvfree(fs_info); 12790d4b0463SJosef Bacik } 12800d4b0463SJosef Bacik 12810d4b0463SJosef Bacik 12822dfb1e43SQu Wenruo /* 12832dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 12842dfb1e43SQu Wenruo * 12852dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 12862dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 12872dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 12882dfb1e43SQu Wenruo * 12892dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 12902dfb1e43SQu Wenruo * 12912dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 12922dfb1e43SQu Wenruo * same root objectid. 12932dfb1e43SQu Wenruo * 12942dfb1e43SQu Wenruo * @objectid: root id 12952dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 12962dfb1e43SQu Wenruo * pass 0 for new allocation. 12972dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 12982dfb1e43SQu Wenruo * for orphan roots 12992dfb1e43SQu Wenruo */ 13002dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 13012dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 13022dfb1e43SQu Wenruo bool check_ref) 13035eda7b5eSChris Mason { 13045eda7b5eSChris Mason struct btrfs_root *root; 1305381cf658SDavid Sterba struct btrfs_path *path; 13061d4c08e0SDavid Sterba struct btrfs_key key; 13075eda7b5eSChris Mason int ret; 13085eda7b5eSChris Mason 130949d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 131049d11beaSJosef Bacik if (root) 131149d11beaSJosef Bacik return root; 13124df27c4dSYan, Zheng again: 131356e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 131448475471SStefan Behrens if (root) { 13152dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 13162dfb1e43SQu Wenruo ASSERT(!anon_dev); 1317bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 131800246528SJosef Bacik btrfs_put_root(root); 131948475471SStefan Behrens return ERR_PTR(-ENOENT); 1320bc44d7c4SJosef Bacik } 13215eda7b5eSChris Mason return root; 132248475471SStefan Behrens } 13235eda7b5eSChris Mason 132456e9357aSDavid Sterba key.objectid = objectid; 132556e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 132656e9357aSDavid Sterba key.offset = (u64)-1; 132756e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 13285eda7b5eSChris Mason if (IS_ERR(root)) 13295eda7b5eSChris Mason return root; 13303394e160SChris Mason 1331c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1332d68fc57bSYan, Zheng ret = -ENOENT; 1333d68fc57bSYan, Zheng goto fail; 1334d68fc57bSYan, Zheng } 1335d68fc57bSYan, Zheng 13362dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1337cb517eabSMiao Xie if (ret) 1338cb517eabSMiao Xie goto fail; 1339cb517eabSMiao Xie 1340381cf658SDavid Sterba path = btrfs_alloc_path(); 1341381cf658SDavid Sterba if (!path) { 1342381cf658SDavid Sterba ret = -ENOMEM; 1343381cf658SDavid Sterba goto fail; 1344381cf658SDavid Sterba } 13451d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 13461d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 134756e9357aSDavid Sterba key.offset = objectid; 13481d4c08e0SDavid Sterba 13491d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1350381cf658SDavid Sterba btrfs_free_path(path); 1351d68fc57bSYan, Zheng if (ret < 0) 1352d68fc57bSYan, Zheng goto fail; 1353d68fc57bSYan, Zheng if (ret == 0) 135427cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1355d68fc57bSYan, Zheng 1356cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 13570f7d52f4SChris Mason if (ret) { 1358168a2f77SJia-Ju Bai if (ret == -EEXIST) { 135900246528SJosef Bacik btrfs_put_root(root); 13604df27c4dSYan, Zheng goto again; 1361168a2f77SJia-Ju Bai } 13624df27c4dSYan, Zheng goto fail; 13634df27c4dSYan, Zheng } 1364edbd8d4eSChris Mason return root; 13654df27c4dSYan, Zheng fail: 136633fab972SFilipe Manana /* 136733fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1368143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 136933fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 137033fab972SFilipe Manana * and once again by our caller. 137133fab972SFilipe Manana */ 137233fab972SFilipe Manana if (anon_dev) 137333fab972SFilipe Manana root->anon_dev = 0; 13748c38938cSJosef Bacik btrfs_put_root(root); 13754df27c4dSYan, Zheng return ERR_PTR(ret); 1376edbd8d4eSChris Mason } 1377edbd8d4eSChris Mason 13782dfb1e43SQu Wenruo /* 13792dfb1e43SQu Wenruo * Get in-memory reference of a root structure 13802dfb1e43SQu Wenruo * 13812dfb1e43SQu Wenruo * @objectid: tree objectid 13822dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 13832dfb1e43SQu Wenruo * one reference 13842dfb1e43SQu Wenruo */ 13852dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 13862dfb1e43SQu Wenruo u64 objectid, bool check_ref) 13872dfb1e43SQu Wenruo { 13882dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 13892dfb1e43SQu Wenruo } 13902dfb1e43SQu Wenruo 13912dfb1e43SQu Wenruo /* 13922dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 13932dfb1e43SQu Wenruo * the anonymous block device id 13942dfb1e43SQu Wenruo * 13952dfb1e43SQu Wenruo * @objectid: tree objectid 13962dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 13972dfb1e43SQu Wenruo * parameter value 13982dfb1e43SQu Wenruo */ 13992dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 14002dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 14012dfb1e43SQu Wenruo { 14022dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 14032dfb1e43SQu Wenruo } 14042dfb1e43SQu Wenruo 14058b712842SChris Mason /* 140649d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 140749d11beaSJosef Bacik * @fs_info: the fs_info 140849d11beaSJosef Bacik * @objectid: the objectid we need to lookup 140949d11beaSJosef Bacik * 141049d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 141149d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 141249d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 141349d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 141449d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 141549d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 141649d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 141749d11beaSJosef Bacik * backref code is finished using the root. 141849d11beaSJosef Bacik */ 141949d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 142049d11beaSJosef Bacik struct btrfs_path *path, 142149d11beaSJosef Bacik u64 objectid) 142249d11beaSJosef Bacik { 142349d11beaSJosef Bacik struct btrfs_root *root; 142449d11beaSJosef Bacik struct btrfs_key key; 142549d11beaSJosef Bacik 142649d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 142749d11beaSJosef Bacik 142849d11beaSJosef Bacik /* 142949d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 143049d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 143149d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 143249d11beaSJosef Bacik * != NULL just return it. 143349d11beaSJosef Bacik */ 143449d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 143549d11beaSJosef Bacik if (root) 143649d11beaSJosef Bacik return root; 143749d11beaSJosef Bacik 143849d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 143949d11beaSJosef Bacik if (root) 144049d11beaSJosef Bacik return root; 144149d11beaSJosef Bacik 144249d11beaSJosef Bacik key.objectid = objectid; 144349d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 144449d11beaSJosef Bacik key.offset = (u64)-1; 144549d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 144649d11beaSJosef Bacik btrfs_release_path(path); 144749d11beaSJosef Bacik 144849d11beaSJosef Bacik return root; 144949d11beaSJosef Bacik } 145049d11beaSJosef Bacik 1451a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1452a74a4b97SChris Mason { 14530d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1454d0278245SMiao Xie int again; 1455a74a4b97SChris Mason 1456d6fd0ae2SOmar Sandoval while (1) { 1457d0278245SMiao Xie again = 0; 14589d1a2a3aSDavid Sterba 1459fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1460fd340d0fSJosef Bacik 1461d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 14622ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1463d0278245SMiao Xie goto sleep; 1464d0278245SMiao Xie 146590c711abSZygo Blaxell /* 146690c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 146790c711abSZygo Blaxell * before we have finished mounting the filesystem. 146890c711abSZygo Blaxell */ 14690b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 147090c711abSZygo Blaxell goto sleep; 147190c711abSZygo Blaxell 14720b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1473d0278245SMiao Xie goto sleep; 1474d0278245SMiao Xie 1475dc7f370cSMiao Xie /* 1476dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1477dc7f370cSMiao Xie * during the above check and trylock. 1478dc7f370cSMiao Xie */ 14792ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 14800b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1481dc7f370cSMiao Xie goto sleep; 1482dc7f370cSMiao Xie } 1483dc7f370cSMiao Xie 1484b7625f46SQu Wenruo if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) 1485b7625f46SQu Wenruo btrfs_sysfs_feature_update(fs_info); 1486b7625f46SQu Wenruo 14872ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1488c2d6cb16SFilipe Manana 148933c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 14900b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1491a74a4b97SChris Mason 1492d0278245SMiao Xie /* 149305323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 149405323cd1SMiao Xie * needn't do anything special here. 1495d0278245SMiao Xie */ 14960b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 149767c5e7d4SFilipe Manana 149867c5e7d4SFilipe Manana /* 1499f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 150067c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 150167c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1502f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 150367c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 150467c5e7d4SFilipe Manana * unused block groups. 150567c5e7d4SFilipe Manana */ 15060b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 150718bb8bbfSJohannes Thumshirn 150818bb8bbfSJohannes Thumshirn /* 150918bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 151018bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 151118bb8bbfSJohannes Thumshirn * space. 151218bb8bbfSJohannes Thumshirn */ 151318bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1514d0278245SMiao Xie sleep: 1515a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1516d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1517d6fd0ae2SOmar Sandoval kthread_parkme(); 1518d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1519d6fd0ae2SOmar Sandoval return 0; 1520838fe188SJiri Kosina if (!again) { 1521a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1522a74a4b97SChris Mason schedule(); 1523a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1524a74a4b97SChris Mason } 1525da288d28SFilipe Manana } 1526a74a4b97SChris Mason } 1527a74a4b97SChris Mason 1528a74a4b97SChris Mason static int transaction_kthread(void *arg) 1529a74a4b97SChris Mason { 1530a74a4b97SChris Mason struct btrfs_root *root = arg; 15310b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1532a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1533a74a4b97SChris Mason struct btrfs_transaction *cur; 15348929ecfaSYan, Zheng u64 transid; 1535643900beSNikolay Borisov time64_t delta; 1536a74a4b97SChris Mason unsigned long delay; 1537914b2007SJan Kara bool cannot_commit; 1538a74a4b97SChris Mason 1539a74a4b97SChris Mason do { 1540914b2007SJan Kara cannot_commit = false; 1541ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 15420b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1543a74a4b97SChris Mason 15440b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 15450b246afaSJeff Mahoney cur = fs_info->running_transaction; 1546a74a4b97SChris Mason if (!cur) { 15470b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1548a74a4b97SChris Mason goto sleep; 1549a74a4b97SChris Mason } 155031153d81SYan Zheng 1551643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1552fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1553fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1554643900beSNikolay Borisov delta < fs_info->commit_interval) { 15550b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1556fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1557fb8a7e94SNikolay Borisov delay = min(delay, 1558fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1559a74a4b97SChris Mason goto sleep; 1560a74a4b97SChris Mason } 15618929ecfaSYan, Zheng transid = cur->transid; 15620b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 156356bec294SChris Mason 156479787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1565354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1566914b2007SJan Kara if (IS_ERR(trans)) { 1567354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1568914b2007SJan Kara cannot_commit = true; 156979787eaaSJeff Mahoney goto sleep; 1570914b2007SJan Kara } 15718929ecfaSYan, Zheng if (transid == trans->transid) { 15723a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 15738929ecfaSYan, Zheng } else { 15743a45bb20SJeff Mahoney btrfs_end_transaction(trans); 15758929ecfaSYan, Zheng } 1576a74a4b97SChris Mason sleep: 15770b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 15780b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1579a74a4b97SChris Mason 158084961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 15812ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 15828929ecfaSYan, Zheng if (!kthread_should_stop() && 15830b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1584914b2007SJan Kara cannot_commit)) 1585bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1586a74a4b97SChris Mason } while (!kthread_should_stop()); 1587a74a4b97SChris Mason return 0; 1588a74a4b97SChris Mason } 1589a74a4b97SChris Mason 1590af31f5e5SChris Mason /* 159101f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 159201f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 159301f0f9daSNikolay Borisov * anything. 1594af31f5e5SChris Mason * 1595af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1596af31f5e5SChris Mason * generation of the latest root in the array with the generation 1597af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1598af31f5e5SChris Mason */ 159901f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1600af31f5e5SChris Mason { 160101f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1602af31f5e5SChris Mason u64 cur; 1603af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1604af31f5e5SChris Mason int i; 1605af31f5e5SChris Mason 1606af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1607af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1608af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1609af31f5e5SChris Mason if (cur == newest_gen) 161001f0f9daSNikolay Borisov return i; 1611af31f5e5SChris Mason } 1612af31f5e5SChris Mason 161301f0f9daSNikolay Borisov return -EINVAL; 1614af31f5e5SChris Mason } 1615af31f5e5SChris Mason 1616af31f5e5SChris Mason /* 1617af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1618af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1619af31f5e5SChris Mason * done 1620af31f5e5SChris Mason */ 1621af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1622af31f5e5SChris Mason { 16236ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1624af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1625af31f5e5SChris Mason 1626af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1627af31f5e5SChris Mason 1628af31f5e5SChris Mason /* 1629af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1630af31f5e5SChris Mason * regardless of which ones we use today 1631af31f5e5SChris Mason */ 1632af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1633af31f5e5SChris Mason 1634af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1635af31f5e5SChris Mason 1636af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1637af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1638af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1639af31f5e5SChris Mason 1640af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1641af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1642af31f5e5SChris Mason 1643af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1644af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1645af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1646af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1647af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1648af31f5e5SChris Mason 16491c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 16509c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1651f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 16529c54e80dSJosef Bacik 16539c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 16549c54e80dSJosef Bacik extent_root->node->start); 1655af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 165629cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 1657af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 165829cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 1659af31f5e5SChris Mason 1660f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 1661f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 1662f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 1663f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 1664f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 16659c54e80dSJosef Bacik } 1666af31f5e5SChris Mason 16677c7e82a7SChris Mason /* 16687c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 16697c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 16707c7e82a7SChris Mason */ 16717c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 16727c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 16737c7e82a7SChris Mason info->fs_root->node->start); 1674af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1675af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1676af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1677af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 16787c7e82a7SChris Mason } 1679af31f5e5SChris Mason 1680af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1681af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1682af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1683af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1684af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1685af31f5e5SChris Mason 1686af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1687af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1688af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1689af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1690af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1691af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1692af31f5e5SChris Mason 1693af31f5e5SChris Mason /* 1694af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1695af31f5e5SChris Mason * for the next commit 1696af31f5e5SChris Mason */ 1697af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1698af31f5e5SChris Mason &info->super_for_commit->super_roots, 1699af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1700af31f5e5SChris Mason } 1701af31f5e5SChris Mason 1702af31f5e5SChris Mason /* 1703bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1704bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1705bd2336b2SNikolay Borisov * 1706bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1707bd2336b2SNikolay Borisov * priority - priority of backup root required 1708bd2336b2SNikolay Borisov * 1709bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1710bd2336b2SNikolay Borisov */ 1711bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1712bd2336b2SNikolay Borisov { 1713bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1714bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1715bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1716bd2336b2SNikolay Borisov 1717bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1718bd2336b2SNikolay Borisov if (priority == 0) 1719bd2336b2SNikolay Borisov return backup_index; 1720bd2336b2SNikolay Borisov 1721bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1722bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1723bd2336b2SNikolay Borisov } else { 1724bd2336b2SNikolay Borisov return -EINVAL; 1725bd2336b2SNikolay Borisov } 1726bd2336b2SNikolay Borisov 1727bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1728bd2336b2SNikolay Borisov 1729bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1730bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1731bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1732bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1733bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1734bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1735bd2336b2SNikolay Borisov 1736bd2336b2SNikolay Borisov /* 1737bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1738bd2336b2SNikolay Borisov * need a fsck 1739bd2336b2SNikolay Borisov */ 1740bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1741bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1742bd2336b2SNikolay Borisov 1743bd2336b2SNikolay Borisov return backup_index; 1744bd2336b2SNikolay Borisov } 1745bd2336b2SNikolay Borisov 17467abadb64SLiu Bo /* helper to cleanup workers */ 17477abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 17487abadb64SLiu Bo { 1749dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1750afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 17515cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1752d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 1753d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 1754385de0efSChristoph Hellwig if (fs_info->rmw_workers) 1755385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 1756fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 1757fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 1758fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1759fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 17605b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1761e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1762a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1763fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1764b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1765b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1766a9b9477dSFilipe Manana /* 1767a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 1768a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 1769a9b9477dSFilipe Manana * queues can do metadata I/O operations. 1770a9b9477dSFilipe Manana */ 1771d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 1772d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 17737abadb64SLiu Bo } 17747abadb64SLiu Bo 17752e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 17762e9f5954SRashika { 17772e9f5954SRashika if (root) { 17782e9f5954SRashika free_extent_buffer(root->node); 17792e9f5954SRashika free_extent_buffer(root->commit_root); 17802e9f5954SRashika root->node = NULL; 17812e9f5954SRashika root->commit_root = NULL; 17822e9f5954SRashika } 17832e9f5954SRashika } 17842e9f5954SRashika 1785abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 1786abed4aaaSJosef Bacik { 1787abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 1788abed4aaaSJosef Bacik 1789abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 1790abed4aaaSJosef Bacik &fs_info->global_root_tree, 1791abed4aaaSJosef Bacik rb_node) 1792abed4aaaSJosef Bacik free_root_extent_buffers(root); 1793abed4aaaSJosef Bacik } 1794abed4aaaSJosef Bacik 1795af31f5e5SChris Mason /* helper to cleanup tree roots */ 17964273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 1797af31f5e5SChris Mason { 17982e9f5954SRashika free_root_extent_buffers(info->tree_root); 1799655b09feSJosef Bacik 1800abed4aaaSJosef Bacik free_global_root_pointers(info); 18012e9f5954SRashika free_root_extent_buffers(info->dev_root); 18022e9f5954SRashika free_root_extent_buffers(info->quota_root); 18032e9f5954SRashika free_root_extent_buffers(info->uuid_root); 18048c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 1805aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 18069c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 18074273eaffSAnand Jain if (free_chunk_root) 18082e9f5954SRashika free_root_extent_buffers(info->chunk_root); 1809af31f5e5SChris Mason } 1810af31f5e5SChris Mason 18118c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 18128c38938cSJosef Bacik { 18138c38938cSJosef Bacik if (!root) 18148c38938cSJosef Bacik return; 18158c38938cSJosef Bacik 18168c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 18178c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 18181dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 18198c38938cSJosef Bacik if (root->anon_dev) 18208c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 1821923eb523SJohannes Thumshirn free_root_extent_buffers(root); 18228c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1823fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 18248c38938cSJosef Bacik list_del_init(&root->leak_list); 1825fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 18268c38938cSJosef Bacik #endif 18278c38938cSJosef Bacik kfree(root); 18288c38938cSJosef Bacik } 18298c38938cSJosef Bacik } 18308c38938cSJosef Bacik 1831faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 1832171f6537SJosef Bacik { 1833fc7cbcd4SDavid Sterba int ret; 1834fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 1835fc7cbcd4SDavid Sterba int i; 1836171f6537SJosef Bacik 1837171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 1838fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 1839171f6537SJosef Bacik struct btrfs_root, root_list); 1840fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 1841171f6537SJosef Bacik 1842fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 1843fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 1844fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 1845171f6537SJosef Bacik } 1846171f6537SJosef Bacik 1847fc7cbcd4SDavid Sterba while (1) { 1848fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 1849fc7cbcd4SDavid Sterba (void **)gang, 0, 1850fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 1851fc7cbcd4SDavid Sterba if (!ret) 1852fc7cbcd4SDavid Sterba break; 1853fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 1854fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 1855171f6537SJosef Bacik } 1856171f6537SJosef Bacik } 1857af31f5e5SChris Mason 1858638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 1859638aa7edSEric Sandeen { 1860638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 1861638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 1862638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 1863638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 1864638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 1865638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 1866ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 1867638aa7edSEric Sandeen } 1868638aa7edSEric Sandeen 1869779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 1870779a65a4SEric Sandeen { 1871779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 1872779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 1873779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 1874779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 1875779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 1876779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 1877907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 1878779a65a4SEric Sandeen } 1879779a65a4SEric Sandeen 1880dcb2137cSChristoph Hellwig static int btrfs_init_btree_inode(struct super_block *sb) 1881f37938e0SEric Sandeen { 1882dcb2137cSChristoph Hellwig struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1883e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 1884e256927bSJosef Bacik fs_info->tree_root); 1885dcb2137cSChristoph Hellwig struct inode *inode; 1886dcb2137cSChristoph Hellwig 1887dcb2137cSChristoph Hellwig inode = new_inode(sb); 1888dcb2137cSChristoph Hellwig if (!inode) 1889dcb2137cSChristoph Hellwig return -ENOMEM; 18902ff7e61eSJeff Mahoney 18912ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 18922ff7e61eSJeff Mahoney set_nlink(inode, 1); 1893f37938e0SEric Sandeen /* 1894f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 1895f37938e0SEric Sandeen * the real end of the address space is determined by all of 1896f37938e0SEric Sandeen * the devices in the system 1897f37938e0SEric Sandeen */ 18982ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 18992ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 1900dcb2137cSChristoph Hellwig mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); 1901f37938e0SEric Sandeen 19022ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 190343eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 190435da5a7eSDavid Sterba IO_TREE_BTREE_INODE_IO); 19052ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 1906f37938e0SEric Sandeen 19075c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 1908adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 1909adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 1910adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 19112ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 1912e256927bSJosef Bacik __insert_inode_hash(inode, hash); 1913dcb2137cSChristoph Hellwig fs_info->btree_inode = inode; 1914dcb2137cSChristoph Hellwig 1915dcb2137cSChristoph Hellwig return 0; 1916f37938e0SEric Sandeen } 1917f37938e0SEric Sandeen 1918ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 1919ad618368SEric Sandeen { 1920ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 1921129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 19227f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 1923ad618368SEric Sandeen } 1924ad618368SEric Sandeen 1925f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 1926f9e92e40SEric Sandeen { 1927f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 1928f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 1929f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 1930f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 1931f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 1932f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 1933d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 1934011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 1935f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 1936f9e92e40SEric Sandeen } 1937f9e92e40SEric Sandeen 1938d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 19392a458198SEric Sandeen { 1940f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 19416f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 1942*58e814fcSTejun Heo unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; 19432a458198SEric Sandeen 19442a458198SEric Sandeen fs_info->workers = 1945a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 19462a458198SEric Sandeen 19472a458198SEric Sandeen fs_info->delalloc_workers = 1948cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 1949cb001095SJeff Mahoney flags, max_active, 2); 19502a458198SEric Sandeen 19512a458198SEric Sandeen fs_info->flush_workers = 1952cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 1953cb001095SJeff Mahoney flags, max_active, 0); 19542a458198SEric Sandeen 19552a458198SEric Sandeen fs_info->caching_workers = 1956cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 19572a458198SEric Sandeen 19582a458198SEric Sandeen fs_info->fixup_workers = 1959*58e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); 19602a458198SEric Sandeen 19612a458198SEric Sandeen fs_info->endio_workers = 1962d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 19632a458198SEric Sandeen fs_info->endio_meta_workers = 1964d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 1965385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 19662a458198SEric Sandeen fs_info->endio_write_workers = 1967cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 1968cb001095SJeff Mahoney max_active, 2); 1969fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 1970fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 19712a458198SEric Sandeen fs_info->endio_freespace_worker = 1972cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 1973cb001095SJeff Mahoney max_active, 0); 19742a458198SEric Sandeen fs_info->delayed_workers = 1975cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 1976cb001095SJeff Mahoney max_active, 0); 19772a458198SEric Sandeen fs_info->qgroup_rescan_workers = 1978*58e814fcSTejun Heo btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", 1979*58e814fcSTejun Heo ordered_flags); 1980b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 1981*58e814fcSTejun Heo alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); 19822a458198SEric Sandeen 19838bfec2e4SChristoph Hellwig if (!(fs_info->workers && 1984a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 19852a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 1986fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 19871a1a2851SQu Wenruo fs_info->endio_write_workers && 19882a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 1989f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 1990f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 1991b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 19922a458198SEric Sandeen return -ENOMEM; 19932a458198SEric Sandeen } 19942a458198SEric Sandeen 19952a458198SEric Sandeen return 0; 19962a458198SEric Sandeen } 19972a458198SEric Sandeen 19986d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 19996d97c6e3SJohannes Thumshirn { 20006d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2001b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 20026d97c6e3SJohannes Thumshirn 2003b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 20046d97c6e3SJohannes Thumshirn 20056d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 20066d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2007b4e967beSDavid Sterba csum_driver); 20086d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 20096d97c6e3SJohannes Thumshirn } 20106d97c6e3SJohannes Thumshirn 20116d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 20126d97c6e3SJohannes Thumshirn 201368d99ab0SChristoph Hellwig /* 201468d99ab0SChristoph Hellwig * Check if the checksum implementation is a fast accelerated one. 201568d99ab0SChristoph Hellwig * As-is this is a bit of a hack and should be replaced once the csum 201668d99ab0SChristoph Hellwig * implementations provide that information themselves. 201768d99ab0SChristoph Hellwig */ 201868d99ab0SChristoph Hellwig switch (csum_type) { 201968d99ab0SChristoph Hellwig case BTRFS_CSUM_TYPE_CRC32: 202068d99ab0SChristoph Hellwig if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) 202168d99ab0SChristoph Hellwig set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 202268d99ab0SChristoph Hellwig break; 202368d99ab0SChristoph Hellwig default: 202468d99ab0SChristoph Hellwig break; 202568d99ab0SChristoph Hellwig } 202668d99ab0SChristoph Hellwig 2027c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2028c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2029c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 20306d97c6e3SJohannes Thumshirn return 0; 20316d97c6e3SJohannes Thumshirn } 20326d97c6e3SJohannes Thumshirn 203363443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 203463443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 203563443bf5SEric Sandeen { 203663443bf5SEric Sandeen int ret; 2037789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 203863443bf5SEric Sandeen struct btrfs_root *log_tree_root; 203963443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 204063443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2041581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 204263443bf5SEric Sandeen 204363443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2044f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 204563443bf5SEric Sandeen return -EIO; 204663443bf5SEric Sandeen } 204763443bf5SEric Sandeen 204896dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 204996dfcb46SJosef Bacik GFP_KERNEL); 205063443bf5SEric Sandeen if (!log_tree_root) 205163443bf5SEric Sandeen return -ENOMEM; 205263443bf5SEric Sandeen 2053789d6a3aSQu Wenruo check.level = level; 2054789d6a3aSQu Wenruo check.transid = fs_info->generation + 1; 2055789d6a3aSQu Wenruo check.owner_root = BTRFS_TREE_LOG_OBJECTID; 2056789d6a3aSQu Wenruo log_tree_root->node = read_tree_block(fs_info, bytenr, &check); 205764c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2058f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 20590eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 20608c38938cSJosef Bacik log_tree_root->node = NULL; 206100246528SJosef Bacik btrfs_put_root(log_tree_root); 20620eeff236SLiu Bo return ret; 20634eb150d6SQu Wenruo } 20644eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2065f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 206600246528SJosef Bacik btrfs_put_root(log_tree_root); 206763443bf5SEric Sandeen return -EIO; 206863443bf5SEric Sandeen } 20694eb150d6SQu Wenruo 207063443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 207163443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 207263443bf5SEric Sandeen if (ret) { 20730b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 207463443bf5SEric Sandeen "Failed to recover log tree"); 207500246528SJosef Bacik btrfs_put_root(log_tree_root); 207663443bf5SEric Sandeen return ret; 207763443bf5SEric Sandeen } 207863443bf5SEric Sandeen 2079bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 20806bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 208163443bf5SEric Sandeen if (ret) 208263443bf5SEric Sandeen return ret; 208363443bf5SEric Sandeen } 208463443bf5SEric Sandeen 208563443bf5SEric Sandeen return 0; 208663443bf5SEric Sandeen } 208763443bf5SEric Sandeen 2088abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2089abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2090abed4aaaSJosef Bacik const char *name) 2091abed4aaaSJosef Bacik { 2092abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2093abed4aaaSJosef Bacik struct btrfs_root *root; 2094f7238e50SJosef Bacik u64 max_global_id = 0; 2095abed4aaaSJosef Bacik int ret; 2096abed4aaaSJosef Bacik struct btrfs_key key = { 2097abed4aaaSJosef Bacik .objectid = objectid, 2098abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2099abed4aaaSJosef Bacik .offset = 0, 2100abed4aaaSJosef Bacik }; 2101abed4aaaSJosef Bacik bool found = false; 2102abed4aaaSJosef Bacik 2103abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2104abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2105abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2106abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2107abed4aaaSJosef Bacik return 0; 2108abed4aaaSJosef Bacik } 2109abed4aaaSJosef Bacik 2110abed4aaaSJosef Bacik while (1) { 2111abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2112abed4aaaSJosef Bacik if (ret < 0) 2113abed4aaaSJosef Bacik break; 2114abed4aaaSJosef Bacik 2115abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2116abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2117abed4aaaSJosef Bacik if (ret) { 2118abed4aaaSJosef Bacik if (ret > 0) 2119abed4aaaSJosef Bacik ret = 0; 2120abed4aaaSJosef Bacik break; 2121abed4aaaSJosef Bacik } 2122abed4aaaSJosef Bacik } 2123abed4aaaSJosef Bacik ret = 0; 2124abed4aaaSJosef Bacik 2125abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2126abed4aaaSJosef Bacik if (key.objectid != objectid) 2127abed4aaaSJosef Bacik break; 2128abed4aaaSJosef Bacik btrfs_release_path(path); 2129abed4aaaSJosef Bacik 2130f7238e50SJosef Bacik /* 2131f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2132f7238e50SJosef Bacik * everybody. 2133f7238e50SJosef Bacik */ 2134f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2135f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2136f7238e50SJosef Bacik 2137abed4aaaSJosef Bacik found = true; 2138abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2139abed4aaaSJosef Bacik if (IS_ERR(root)) { 2140abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2141abed4aaaSJosef Bacik ret = PTR_ERR(root); 2142abed4aaaSJosef Bacik break; 2143abed4aaaSJosef Bacik } 2144abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2145abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2146abed4aaaSJosef Bacik if (ret) { 2147abed4aaaSJosef Bacik btrfs_put_root(root); 2148abed4aaaSJosef Bacik break; 2149abed4aaaSJosef Bacik } 2150abed4aaaSJosef Bacik key.offset++; 2151abed4aaaSJosef Bacik } 2152abed4aaaSJosef Bacik btrfs_release_path(path); 2153abed4aaaSJosef Bacik 2154f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2155f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2156f7238e50SJosef Bacik 2157abed4aaaSJosef Bacik if (!found || ret) { 2158abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2159abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2160abed4aaaSJosef Bacik 2161abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2162abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2163abed4aaaSJosef Bacik else 2164abed4aaaSJosef Bacik ret = 0; 2165abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2166abed4aaaSJosef Bacik } 2167abed4aaaSJosef Bacik return ret; 2168abed4aaaSJosef Bacik } 2169abed4aaaSJosef Bacik 2170abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2171abed4aaaSJosef Bacik { 2172abed4aaaSJosef Bacik struct btrfs_path *path; 2173abed4aaaSJosef Bacik int ret = 0; 2174abed4aaaSJosef Bacik 2175abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2176abed4aaaSJosef Bacik if (!path) 2177abed4aaaSJosef Bacik return -ENOMEM; 2178abed4aaaSJosef Bacik 2179abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2180abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2181abed4aaaSJosef Bacik if (ret) 2182abed4aaaSJosef Bacik goto out; 2183abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2184abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2185abed4aaaSJosef Bacik if (ret) 2186abed4aaaSJosef Bacik goto out; 2187abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2188abed4aaaSJosef Bacik goto out; 2189abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2190abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2191abed4aaaSJosef Bacik "free space"); 2192abed4aaaSJosef Bacik out: 2193abed4aaaSJosef Bacik btrfs_free_path(path); 2194abed4aaaSJosef Bacik return ret; 2195abed4aaaSJosef Bacik } 2196abed4aaaSJosef Bacik 21976bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 21984bbcaa64SEric Sandeen { 21996bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2200a4f3d2c4SDavid Sterba struct btrfs_root *root; 22014bbcaa64SEric Sandeen struct btrfs_key location; 22024bbcaa64SEric Sandeen int ret; 22034bbcaa64SEric Sandeen 22046bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 22056bccf3abSJeff Mahoney 2206abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2207abed4aaaSJosef Bacik if (ret) 2208abed4aaaSJosef Bacik return ret; 2209abed4aaaSJosef Bacik 22104bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 22114bbcaa64SEric Sandeen location.offset = 0; 22124bbcaa64SEric Sandeen 22131c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 221414033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 221514033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 221614033b08SQu Wenruo if (IS_ERR(root)) { 221714033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 221814033b08SQu Wenruo ret = PTR_ERR(root); 221914033b08SQu Wenruo goto out; 222014033b08SQu Wenruo } 222114033b08SQu Wenruo } else { 222214033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 222314033b08SQu Wenruo fs_info->block_group_root = root; 222414033b08SQu Wenruo } 222514033b08SQu Wenruo } 222614033b08SQu Wenruo 222714033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2228a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2229f50f4353SLiu Bo if (IS_ERR(root)) { 223042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2231f50f4353SLiu Bo ret = PTR_ERR(root); 2232f50f4353SLiu Bo goto out; 2233f50f4353SLiu Bo } 223442437a63SJosef Bacik } else { 2235a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2236a4f3d2c4SDavid Sterba fs_info->dev_root = root; 223742437a63SJosef Bacik } 2238820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2239a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2240a8d1b164SJohannes Thumshirn if (ret) 2241a8d1b164SJohannes Thumshirn goto out; 22424bbcaa64SEric Sandeen 2243aeb935a4SQu Wenruo /* 2244aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2245aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2246aeb935a4SQu Wenruo */ 224756e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 224856e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2249aeb935a4SQu Wenruo if (IS_ERR(root)) { 225042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2251aeb935a4SQu Wenruo ret = PTR_ERR(root); 2252aeb935a4SQu Wenruo goto out; 2253aeb935a4SQu Wenruo } 225442437a63SJosef Bacik } else { 2255aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2256aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 225742437a63SJosef Bacik } 2258aeb935a4SQu Wenruo 22594bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2260a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2261a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2262a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2263afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2264a4f3d2c4SDavid Sterba fs_info->quota_root = root; 22654bbcaa64SEric Sandeen } 22664bbcaa64SEric Sandeen 22674bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2268a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2269a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 227042437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2271a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 22724bbcaa64SEric Sandeen if (ret != -ENOENT) 2273f50f4353SLiu Bo goto out; 227442437a63SJosef Bacik } 22754bbcaa64SEric Sandeen } else { 2276a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2277a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 22784bbcaa64SEric Sandeen } 22794bbcaa64SEric Sandeen 22804bbcaa64SEric Sandeen return 0; 2281f50f4353SLiu Bo out: 2282f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2283f50f4353SLiu Bo location.objectid, ret); 2284f50f4353SLiu Bo return ret; 22854bbcaa64SEric Sandeen } 22864bbcaa64SEric Sandeen 2287069ec957SQu Wenruo /* 2288069ec957SQu Wenruo * Real super block validation 2289069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2290069ec957SQu Wenruo * 2291069ec957SQu Wenruo * @sb: super block to check 2292069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2293069ec957SQu Wenruo * 0 the primary (1st) sb 2294069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2295069ec957SQu Wenruo * -1 skip bytenr check 2296069ec957SQu Wenruo */ 2297a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2298069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 229921a852b0SQu Wenruo { 230021a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 230121a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 230221a852b0SQu Wenruo int ret = 0; 230321a852b0SQu Wenruo 230421a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 230521a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 230621a852b0SQu Wenruo ret = -EINVAL; 230721a852b0SQu Wenruo } 230821a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 230921a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 231021a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 231121a852b0SQu Wenruo ret = -EINVAL; 231221a852b0SQu Wenruo } 231321a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 231421a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 231521a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 231621a852b0SQu Wenruo ret = -EINVAL; 231721a852b0SQu Wenruo } 231821a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 231921a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 232021a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 232121a852b0SQu Wenruo ret = -EINVAL; 232221a852b0SQu Wenruo } 232321a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 232421a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 232521a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 232621a852b0SQu Wenruo ret = -EINVAL; 232721a852b0SQu Wenruo } 232821a852b0SQu Wenruo 232921a852b0SQu Wenruo /* 233021a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 233121a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 233221a852b0SQu Wenruo */ 233321a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 233421a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 233521a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 233621a852b0SQu Wenruo ret = -EINVAL; 233721a852b0SQu Wenruo } 23380bb3eb3eSQu Wenruo 23390bb3eb3eSQu Wenruo /* 23401a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 23411a42daabSQu Wenruo * 23421a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 23431a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 23441a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 23451a42daabSQu Wenruo * beginning. 23460bb3eb3eSQu Wenruo */ 23471a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 234821a852b0SQu Wenruo btrfs_err(fs_info, 23490bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 235021a852b0SQu Wenruo sectorsize, PAGE_SIZE); 235121a852b0SQu Wenruo ret = -EINVAL; 235221a852b0SQu Wenruo } 23530bb3eb3eSQu Wenruo 235421a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 235521a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 235621a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 235721a852b0SQu Wenruo ret = -EINVAL; 235821a852b0SQu Wenruo } 235921a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 236021a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 236121a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 236221a852b0SQu Wenruo ret = -EINVAL; 236321a852b0SQu Wenruo } 236421a852b0SQu Wenruo 236521a852b0SQu Wenruo /* Root alignment check */ 236621a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 236721a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 236821a852b0SQu Wenruo btrfs_super_root(sb)); 236921a852b0SQu Wenruo ret = -EINVAL; 237021a852b0SQu Wenruo } 237121a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 237221a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 237321a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 237421a852b0SQu Wenruo ret = -EINVAL; 237521a852b0SQu Wenruo } 237621a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 237721a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 237821a852b0SQu Wenruo btrfs_super_log_root(sb)); 237921a852b0SQu Wenruo ret = -EINVAL; 238021a852b0SQu Wenruo } 238121a852b0SQu Wenruo 2382aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2383aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2384aefd7f70SNikolay Borisov btrfs_err(fs_info, 2385aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2386aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2387aefd7f70SNikolay Borisov ret = -EINVAL; 2388aefd7f70SNikolay Borisov } 2389aefd7f70SNikolay Borisov 2390aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2391aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2392aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2393aefd7f70SNikolay Borisov btrfs_err(fs_info, 2394aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2395aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2396aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2397aefd7f70SNikolay Borisov ret = -EINVAL; 2398aefd7f70SNikolay Borisov } 2399aefd7f70SNikolay Borisov 240025984a5aSAnand Jain if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 240125984a5aSAnand Jain BTRFS_FSID_SIZE) != 0) { 240225984a5aSAnand Jain btrfs_err(fs_info, 240325984a5aSAnand Jain "dev_item UUID does not match metadata fsid: %pU != %pU", 240425984a5aSAnand Jain fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 240525984a5aSAnand Jain ret = -EINVAL; 240625984a5aSAnand Jain } 240725984a5aSAnand Jain 24081c56ab99SQu Wenruo /* 24091c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 24101c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 24111c56ab99SQu Wenruo */ 24121c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 24131c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 24141c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 24151c56ab99SQu Wenruo btrfs_err(fs_info, 24161c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 24171c56ab99SQu Wenruo ret = -EINVAL; 24181c56ab99SQu Wenruo } 24191c56ab99SQu Wenruo 242021a852b0SQu Wenruo /* 242121a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 242221a852b0SQu Wenruo * done later 242321a852b0SQu Wenruo */ 242421a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 242521a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 242621a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 242721a852b0SQu Wenruo ret = -EINVAL; 242821a852b0SQu Wenruo } 242921a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 243021a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 243121a852b0SQu Wenruo btrfs_super_stripesize(sb)); 243221a852b0SQu Wenruo ret = -EINVAL; 243321a852b0SQu Wenruo } 243421a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 243521a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 243621a852b0SQu Wenruo btrfs_super_num_devices(sb)); 243721a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 243821a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 243921a852b0SQu Wenruo ret = -EINVAL; 244021a852b0SQu Wenruo } 244121a852b0SQu Wenruo 2442069ec957SQu Wenruo if (mirror_num >= 0 && 2443069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 244421a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 244521a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 244621a852b0SQu Wenruo ret = -EINVAL; 244721a852b0SQu Wenruo } 244821a852b0SQu Wenruo 244921a852b0SQu Wenruo /* 245021a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 245121a852b0SQu Wenruo * and one chunk 245221a852b0SQu Wenruo */ 245321a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 245421a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 245521a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 245621a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 245721a852b0SQu Wenruo ret = -EINVAL; 245821a852b0SQu Wenruo } 245921a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 246021a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 246121a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 246221a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 246321a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 246421a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 246521a852b0SQu Wenruo ret = -EINVAL; 246621a852b0SQu Wenruo } 246721a852b0SQu Wenruo 246821a852b0SQu Wenruo /* 246921a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 247021a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 247121a852b0SQu Wenruo */ 247221a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 247321a852b0SQu Wenruo btrfs_warn(fs_info, 247421a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 247521a852b0SQu Wenruo btrfs_super_generation(sb), 247621a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 247721a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 247821a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 247921a852b0SQu Wenruo btrfs_warn(fs_info, 248021a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 248121a852b0SQu Wenruo btrfs_super_generation(sb), 248221a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 248321a852b0SQu Wenruo 248421a852b0SQu Wenruo return ret; 248521a852b0SQu Wenruo } 248621a852b0SQu Wenruo 2487069ec957SQu Wenruo /* 2488069ec957SQu Wenruo * Validation of super block at mount time. 2489069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2490069ec957SQu Wenruo * flags will be skipped. 2491069ec957SQu Wenruo */ 2492069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2493069ec957SQu Wenruo { 2494a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2495069ec957SQu Wenruo } 2496069ec957SQu Wenruo 249775cb857dSQu Wenruo /* 249875cb857dSQu Wenruo * Validation of super block at write time. 249975cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 250075cb857dSQu Wenruo * overwritten soon. 250175cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 250275cb857dSQu Wenruo */ 250375cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 250475cb857dSQu Wenruo struct btrfs_super_block *sb) 250575cb857dSQu Wenruo { 250675cb857dSQu Wenruo int ret; 250775cb857dSQu Wenruo 2508a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 250975cb857dSQu Wenruo if (ret < 0) 251075cb857dSQu Wenruo goto out; 2511e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 251275cb857dSQu Wenruo ret = -EUCLEAN; 251375cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 251475cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 251575cb857dSQu Wenruo goto out; 251675cb857dSQu Wenruo } 251775cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 251875cb857dSQu Wenruo ret = -EUCLEAN; 251975cb857dSQu Wenruo btrfs_err(fs_info, 252075cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 252175cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 252275cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 252375cb857dSQu Wenruo goto out; 252475cb857dSQu Wenruo } 252575cb857dSQu Wenruo out: 252675cb857dSQu Wenruo if (ret < 0) 252775cb857dSQu Wenruo btrfs_err(fs_info, 252875cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 252975cb857dSQu Wenruo return ret; 253075cb857dSQu Wenruo } 253175cb857dSQu Wenruo 2532bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2533bd676446SJosef Bacik { 2534789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 2535789d6a3aSQu Wenruo .level = level, 2536789d6a3aSQu Wenruo .transid = gen, 2537789d6a3aSQu Wenruo .owner_root = root->root_key.objectid 2538789d6a3aSQu Wenruo }; 2539bd676446SJosef Bacik int ret = 0; 2540bd676446SJosef Bacik 2541789d6a3aSQu Wenruo root->node = read_tree_block(root->fs_info, bytenr, &check); 2542bd676446SJosef Bacik if (IS_ERR(root->node)) { 2543bd676446SJosef Bacik ret = PTR_ERR(root->node); 2544bd676446SJosef Bacik root->node = NULL; 25454eb150d6SQu Wenruo return ret; 25464eb150d6SQu Wenruo } 25474eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2548bd676446SJosef Bacik free_extent_buffer(root->node); 2549bd676446SJosef Bacik root->node = NULL; 25504eb150d6SQu Wenruo return -EIO; 2551bd676446SJosef Bacik } 2552bd676446SJosef Bacik 2553bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2554bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2555bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2556bd676446SJosef Bacik return ret; 2557bd676446SJosef Bacik } 2558bd676446SJosef Bacik 2559bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2560bd676446SJosef Bacik { 2561bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2562bd676446SJosef Bacik u64 gen, bytenr; 2563bd676446SJosef Bacik int level, ret; 2564bd676446SJosef Bacik 2565bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2566bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2567bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2568bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 25699c54e80dSJosef Bacik if (ret) { 2570bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2571bd676446SJosef Bacik return ret; 2572bd676446SJosef Bacik } 25739c54e80dSJosef Bacik return 0; 25749c54e80dSJosef Bacik } 25759c54e80dSJosef Bacik 25766ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2577b8522a1eSNikolay Borisov { 25786ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2579b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2580b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2581b8522a1eSNikolay Borisov bool handle_error = false; 2582b8522a1eSNikolay Borisov int ret = 0; 2583b8522a1eSNikolay Borisov int i; 2584b8522a1eSNikolay Borisov 2585b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2586b8522a1eSNikolay Borisov if (handle_error) { 2587b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2588b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2589b8522a1eSNikolay Borisov tree_root->node = NULL; 2590b8522a1eSNikolay Borisov 2591b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2592b8522a1eSNikolay Borisov break; 2593b8522a1eSNikolay Borisov 2594b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2595b8522a1eSNikolay Borisov 2596b8522a1eSNikolay Borisov /* 2597b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2598b8522a1eSNikolay Borisov * valid 2599b8522a1eSNikolay Borisov */ 2600b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2601b8522a1eSNikolay Borisov 2602b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2603b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2604b8522a1eSNikolay Borisov 2605745806fbSQu Wenruo btrfs_warn(fs_info, "try to load backup roots slot %d", i); 2606b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26076ef108ddSNikolay Borisov backup_index = ret; 2608b8522a1eSNikolay Borisov if (ret < 0) 2609b8522a1eSNikolay Borisov return ret; 2610b8522a1eSNikolay Borisov } 2611b8522a1eSNikolay Borisov 2612bd676446SJosef Bacik ret = load_important_roots(fs_info); 2613bd676446SJosef Bacik if (ret) { 2614217f5004SNikolay Borisov handle_error = true; 2615b8522a1eSNikolay Borisov continue; 2616b8522a1eSNikolay Borisov } 2617b8522a1eSNikolay Borisov 2618336a0d8dSNikolay Borisov /* 2619336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2620336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2621336a0d8dSNikolay Borisov */ 2622453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2623b8522a1eSNikolay Borisov if (ret < 0) { 2624b8522a1eSNikolay Borisov handle_error = true; 2625b8522a1eSNikolay Borisov continue; 2626b8522a1eSNikolay Borisov } 2627b8522a1eSNikolay Borisov 26286b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2629b8522a1eSNikolay Borisov 2630b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2631b8522a1eSNikolay Borisov if (ret < 0) { 2632b8522a1eSNikolay Borisov handle_error = true; 2633b8522a1eSNikolay Borisov continue; 2634b8522a1eSNikolay Borisov } 2635b8522a1eSNikolay Borisov 2636b8522a1eSNikolay Borisov /* All successful */ 2637bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2638bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2639d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 26406ef108ddSNikolay Borisov 26416ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26426ef108ddSNikolay Borisov if (backup_index < 0) { 26436ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26446ef108ddSNikolay Borisov } else { 26456ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26466ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26476ef108ddSNikolay Borisov } 2648b8522a1eSNikolay Borisov break; 2649b8522a1eSNikolay Borisov } 2650b8522a1eSNikolay Borisov 2651b8522a1eSNikolay Borisov return ret; 2652b8522a1eSNikolay Borisov } 2653b8522a1eSNikolay Borisov 26548260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2655eb60ceacSChris Mason { 2656fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 265701cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26588fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2659facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 266024bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2661eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 266211833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2663eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2664a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2665fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 266624bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26674cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2668ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2669f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 267047ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 267140ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2672afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2673c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2674f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2675abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2676d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2677f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 26787585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2679573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 26800bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 26815f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 2682de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 268319c00ddcSChris Mason 2684e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 26855a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 26868b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 26875f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 26883e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 26893e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 26903e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 26913e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 26923e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 26933e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 26943e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 26953e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 2696e1489b4fSIoannis Angelakopoulos 26970b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 26986324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2699f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 270047ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 270118bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2702afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2703bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2704bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 27053fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27063fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2707bd647ce3SJosef Bacik #endif 2708c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 270966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 271066d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 271166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 271266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 271366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 271466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 271566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2716ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2717ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2718ba2c4d4eSJosef Bacik 2719771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27204cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2721034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2722fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2723abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 272495ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27259ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27264cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2727a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2728f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27298b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2730fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2731c8b97818SChris Mason 2732b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2733b34b086cSChris Mason num_online_cpus() + 2, 8); 27340afbaf8cSChris Mason 2735199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2736199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 273769fe2d75SJosef Bacik 2738638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 273921adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 274021adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 274121adbd5cSStefan Behrens #endif 2742779a65a4SEric Sandeen btrfs_init_balance(fs_info); 274357056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2744a2de733cSArne Jansen 274516b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 274608dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 27470f9dd46cSJosef Bacik 2748fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 274935da5a7eSDavid Sterba IO_TREE_FS_EXCLUDED_EXTENTS); 275039279cc3SChris Mason 27515a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2752e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2753925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2754a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2755a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27561bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27579e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2758c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 275976dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2760803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2761fa9c0d79SChris Mason 2762ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2763f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2764b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2765416ac51dSArne Jansen 2766fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2767fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2768fa9c0d79SChris Mason 2769e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2770f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2771bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27724854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2773034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 27743768f368SChris Mason 2775da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2776da17066cSJeff Mahoney fs_info->nodesize = 4096; 2777da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2778ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2779da17066cSJeff Mahoney fs_info->stripesize = 4096; 2780da17066cSJeff Mahoney 2781f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 2782f7b12a62SNaohiro Aota 2783eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2784eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2785eede2bf3SOmar Sandoval 278618bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 278718bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 27888260edbaSJosef Bacik } 27898260edbaSJosef Bacik 27908260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 27918260edbaSJosef Bacik { 27928260edbaSJosef Bacik int ret; 27938260edbaSJosef Bacik 27948260edbaSJosef Bacik fs_info->sb = sb; 27958260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 27968260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 27979e967495SFilipe Manana 27985deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 2799ae18c37aSJosef Bacik if (ret) 2800ae18c37aSJosef Bacik return ret; 2801ae18c37aSJosef Bacik 2802ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2803ae18c37aSJosef Bacik if (ret) 2804c75e8394SJosef Bacik return ret; 2805ae18c37aSJosef Bacik 2806ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2807ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2808ae18c37aSJosef Bacik 2809ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2810ae18c37aSJosef Bacik if (ret) 2811c75e8394SJosef Bacik return ret; 2812ae18c37aSJosef Bacik 2813ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2814ae18c37aSJosef Bacik GFP_KERNEL); 2815ae18c37aSJosef Bacik if (ret) 2816c75e8394SJosef Bacik return ret; 2817ae18c37aSJosef Bacik 2818ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2819ae18c37aSJosef Bacik GFP_KERNEL); 2820c75e8394SJosef Bacik if (!fs_info->delayed_root) 2821c75e8394SJosef Bacik return -ENOMEM; 2822ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2823ae18c37aSJosef Bacik 2824a0a1db70SFilipe Manana if (sb_rdonly(sb)) 2825a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 2826a0a1db70SFilipe Manana 2827c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2828ae18c37aSJosef Bacik } 2829ae18c37aSJosef Bacik 283097f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 283197f4dd09SNikolay Borisov { 28320d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 283397f4dd09SNikolay Borisov int ret; 283497f4dd09SNikolay Borisov 283597f4dd09SNikolay Borisov /* 283697f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 283797f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 283897f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 283997f4dd09SNikolay Borisov */ 284097f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 284197f4dd09SNikolay Borisov if (ret < 0) { 2842c94bec2cSJosef Bacik if (ret != -EINTR) 2843c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2844c94bec2cSJosef Bacik ret); 284597f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 284697f4dd09SNikolay Borisov return ret; 284797f4dd09SNikolay Borisov } 284897f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 284997f4dd09SNikolay Borisov } 285097f4dd09SNikolay Borisov 285197f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 285297f4dd09SNikolay Borisov { 285397f4dd09SNikolay Borisov struct task_struct *task; 285497f4dd09SNikolay Borisov 285597f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 285697f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 285797f4dd09SNikolay Borisov if (IS_ERR(task)) { 285897f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 285997f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 286097f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 286197f4dd09SNikolay Borisov return PTR_ERR(task); 286297f4dd09SNikolay Borisov } 286397f4dd09SNikolay Borisov 286497f4dd09SNikolay Borisov return 0; 286597f4dd09SNikolay Borisov } 286697f4dd09SNikolay Borisov 286744c0ca21SBoris Burkov /* 28688cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 28698cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 28708cd29088SBoris Burkov * code paths. 28718cd29088SBoris Burkov */ 28728cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 28738cd29088SBoris Burkov { 28748cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 28758b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 28768cd29088SBoris Burkov } 28778cd29088SBoris Burkov 28788cd29088SBoris Burkov /* 287944c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 288044c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 288144c0ca21SBoris Burkov */ 288244c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 288344c0ca21SBoris Burkov { 288444c0ca21SBoris Burkov int ret; 288594846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 28861d6a4fc8SQu Wenruo bool rebuild_free_space_tree = false; 28878b228324SBoris Burkov 28888b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 28898b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 28901d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 28918b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 28928b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 28938b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 28941d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 28958b228324SBoris Burkov } 28968b228324SBoris Burkov 28971d6a4fc8SQu Wenruo if (rebuild_free_space_tree) { 28981d6a4fc8SQu Wenruo btrfs_info(fs_info, "rebuilding free space tree"); 28991d6a4fc8SQu Wenruo ret = btrfs_rebuild_free_space_tree(fs_info); 29008b228324SBoris Burkov if (ret) { 29018b228324SBoris Burkov btrfs_warn(fs_info, 29021d6a4fc8SQu Wenruo "failed to rebuild free space tree: %d", ret); 29031d6a4fc8SQu Wenruo goto out; 29041d6a4fc8SQu Wenruo } 29051d6a4fc8SQu Wenruo } 29061d6a4fc8SQu Wenruo 29071d6a4fc8SQu Wenruo if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29081d6a4fc8SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { 29091d6a4fc8SQu Wenruo btrfs_info(fs_info, "disabling free space tree"); 29101d6a4fc8SQu Wenruo ret = btrfs_delete_free_space_tree(fs_info); 29111d6a4fc8SQu Wenruo if (ret) { 29121d6a4fc8SQu Wenruo btrfs_warn(fs_info, 29131d6a4fc8SQu Wenruo "failed to disable free space tree: %d", ret); 29148b228324SBoris Burkov goto out; 29158b228324SBoris Burkov } 29168b228324SBoris Burkov } 291744c0ca21SBoris Burkov 29188d488a8cSFilipe Manana /* 29198d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 29208d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 2921fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 29228d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 29238d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 29248d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 29258d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 29268d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 29278d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 29288d488a8cSFilipe Manana */ 29298d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 29308d488a8cSFilipe Manana if (ret) 29318d488a8cSFilipe Manana goto out; 29328d488a8cSFilipe Manana 293344c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 293444c0ca21SBoris Burkov if (ret) 293544c0ca21SBoris Burkov goto out; 293644c0ca21SBoris Burkov 29378f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 29388f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29398f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29408f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29418f1c21d7SBoris Burkov goto out; 29428f1c21d7SBoris Burkov } 29438f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29448f1c21d7SBoris Burkov 294544c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 29467eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 294744c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 294844c0ca21SBoris Burkov if (ret < 0) { 294944c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 295044c0ca21SBoris Burkov goto out; 295144c0ca21SBoris Burkov } 295244c0ca21SBoris Burkov 29535011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 29545011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 29555011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 29565011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 29575011139aSBoris Burkov if (ret) { 29585011139aSBoris Burkov btrfs_warn(fs_info, 29595011139aSBoris Burkov "failed to create free space tree: %d", ret); 29605011139aSBoris Burkov goto out; 29615011139aSBoris Burkov } 29625011139aSBoris Burkov } 29635011139aSBoris Burkov 296494846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 296594846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 296694846229SBoris Burkov if (ret) 296794846229SBoris Burkov goto out; 296894846229SBoris Burkov } 296994846229SBoris Burkov 297044c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 297144c0ca21SBoris Burkov if (ret) 297244c0ca21SBoris Burkov goto out; 297344c0ca21SBoris Burkov 297444c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 297544c0ca21SBoris Burkov if (ret) { 297644c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 297744c0ca21SBoris Burkov goto out; 297844c0ca21SBoris Burkov } 297944c0ca21SBoris Burkov 298044c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 298144c0ca21SBoris Burkov 298244c0ca21SBoris Burkov if (!fs_info->uuid_root) { 298344c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 298444c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 298544c0ca21SBoris Burkov if (ret) { 298644c0ca21SBoris Burkov btrfs_warn(fs_info, 298744c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 298844c0ca21SBoris Burkov goto out; 298944c0ca21SBoris Burkov } 299044c0ca21SBoris Burkov } 299144c0ca21SBoris Burkov 299244c0ca21SBoris Burkov out: 299344c0ca21SBoris Burkov return ret; 299444c0ca21SBoris Burkov } 299544c0ca21SBoris Burkov 2996d7f67ac9SQu Wenruo /* 2997d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 2998d7f67ac9SQu Wenruo * 29992ba48b20SQu Wenruo * @is_rw_mount: If the mount is read-write. 30002ba48b20SQu Wenruo * 3001d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3002d7f67ac9SQu Wenruo * feature dependencies). 3003d7f67ac9SQu Wenruo * 3004d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3005d7f67ac9SQu Wenruo * btrfs_validate_super(). 3006d7f67ac9SQu Wenruo * 3007d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3008d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3009d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3010d7f67ac9SQu Wenruo */ 30112ba48b20SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) 3012d7f67ac9SQu Wenruo { 3013d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3014d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3015d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3016d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3017d7f67ac9SQu Wenruo 3018d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3019d7f67ac9SQu Wenruo btrfs_err(fs_info, 3020d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3021d7f67ac9SQu Wenruo incompat); 3022d7f67ac9SQu Wenruo return -EINVAL; 3023d7f67ac9SQu Wenruo } 3024d7f67ac9SQu Wenruo 3025d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3026d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3027d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3028d7f67ac9SQu Wenruo btrfs_err(fs_info, 3029d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3030d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3031d7f67ac9SQu Wenruo return -EINVAL; 3032d7f67ac9SQu Wenruo } 3033d7f67ac9SQu Wenruo 3034d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3035d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3036d7f67ac9SQu Wenruo 3037d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3038d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3039d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3040d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3041d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3042d7f67ac9SQu Wenruo 3043d7f67ac9SQu Wenruo /* 3044d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3045d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3046d7f67ac9SQu Wenruo */ 3047d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3048d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3049d7f67ac9SQu Wenruo 30502ba48b20SQu Wenruo if (compat_ro_unsupp && is_rw_mount) { 3051d7f67ac9SQu Wenruo btrfs_err(fs_info, 3052d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3053d7f67ac9SQu Wenruo compat_ro); 3054d7f67ac9SQu Wenruo return -EINVAL; 3055d7f67ac9SQu Wenruo } 3056d7f67ac9SQu Wenruo 3057d7f67ac9SQu Wenruo /* 3058d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3059d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3060d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3061d7f67ac9SQu Wenruo */ 3062d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3063d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3064d7f67ac9SQu Wenruo btrfs_err(fs_info, 3065d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3066d7f67ac9SQu Wenruo compat_ro); 3067d7f67ac9SQu Wenruo return -EINVAL; 3068d7f67ac9SQu Wenruo } 3069d7f67ac9SQu Wenruo 3070d7f67ac9SQu Wenruo /* 3071d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3072d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3073d7f67ac9SQu Wenruo */ 3074d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3075d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3076d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3077d7f67ac9SQu Wenruo btrfs_err(fs_info, 3078d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3079d7f67ac9SQu Wenruo return -EINVAL; 3080d7f67ac9SQu Wenruo } 3081d7f67ac9SQu Wenruo 3082d7f67ac9SQu Wenruo /* 3083d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3084d7f67ac9SQu Wenruo * 3085d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3086d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3087d7f67ac9SQu Wenruo * going to be deprecated anyway. 3088d7f67ac9SQu Wenruo */ 3089d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3090d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3091d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3092d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3093d7f67ac9SQu Wenruo return -EINVAL; 3094d7f67ac9SQu Wenruo } 3095d7f67ac9SQu Wenruo 3096d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3097d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3098d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3099d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3100d7f67ac9SQu Wenruo 3101d7f67ac9SQu Wenruo return 0; 3102d7f67ac9SQu Wenruo } 3103d7f67ac9SQu Wenruo 3104ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3105ae18c37aSJosef Bacik char *options) 3106ae18c37aSJosef Bacik { 3107ae18c37aSJosef Bacik u32 sectorsize; 3108ae18c37aSJosef Bacik u32 nodesize; 3109ae18c37aSJosef Bacik u32 stripesize; 3110ae18c37aSJosef Bacik u64 generation; 3111ae18c37aSJosef Bacik u64 features; 3112ae18c37aSJosef Bacik u16 csum_type; 3113ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3114ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3115ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3116ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3117ae18c37aSJosef Bacik int ret; 3118ae18c37aSJosef Bacik int level; 3119ae18c37aSJosef Bacik 31208260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 31214871c33bSQu Wenruo if (ret) 3122ae18c37aSJosef Bacik goto fail; 312353b381b3SDavid Woodhouse 3124ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3125ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3126ae18c37aSJosef Bacik GFP_KERNEL); 3127ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3128ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3129ae18c37aSJosef Bacik GFP_KERNEL); 3130ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3131ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 31324871c33bSQu Wenruo ret = -ENOMEM; 3133c75e8394SJosef Bacik goto fail; 3134ae18c37aSJosef Bacik } 3135ae18c37aSJosef Bacik 3136dcb2137cSChristoph Hellwig ret = btrfs_init_btree_inode(sb); 31374871c33bSQu Wenruo if (ret) 3138c75e8394SJosef Bacik goto fail; 3139ae18c37aSJosef Bacik 3140d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 31411104a885SDavid Sterba 31421104a885SDavid Sterba /* 31431104a885SDavid Sterba * Read super block and check the signature bytes only 31441104a885SDavid Sterba */ 3145d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 31468f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 31474871c33bSQu Wenruo ret = PTR_ERR(disk_super); 314816cdcec7SMiao Xie goto fail_alloc; 314920b45077SDave Young } 315039279cc3SChris Mason 31511104a885SDavid Sterba /* 3152260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 31538dc3f22cSJohannes Thumshirn * corrupted, we'll find out 31548dc3f22cSJohannes Thumshirn */ 31558f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 315651bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 31578dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 315851bce6c9SJohannes Thumshirn csum_type); 31594871c33bSQu Wenruo ret = -EINVAL; 31608f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31618dc3f22cSJohannes Thumshirn goto fail_alloc; 31628dc3f22cSJohannes Thumshirn } 31638dc3f22cSJohannes Thumshirn 316483c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 316583c68bbcSSu Yue 31666d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 31676d97c6e3SJohannes Thumshirn if (ret) { 31688f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31696d97c6e3SJohannes Thumshirn goto fail_alloc; 31706d97c6e3SJohannes Thumshirn } 31716d97c6e3SJohannes Thumshirn 31728dc3f22cSJohannes Thumshirn /* 31731104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 31741104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 31751104a885SDavid Sterba */ 31763d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 317705135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 31784871c33bSQu Wenruo ret = -EINVAL; 31798f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3180141386e1SJosef Bacik goto fail_alloc; 31811104a885SDavid Sterba } 31821104a885SDavid Sterba 31831104a885SDavid Sterba /* 31841104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 31851104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 31861104a885SDavid Sterba * the whole block of INFO_SIZE 31871104a885SDavid Sterba */ 31888f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 31898f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31905f39d397SChris Mason 3191fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3192fbc6feaeSNikolay Borisov 31930b86a832SChris Mason 3194fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3195fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3196fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3197fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3198fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3199fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3200fbc6feaeSNikolay Borisov } 3201fbc6feaeSNikolay Borisov 3202fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3203fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3204de37aa51SNikolay Borisov 3205069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 32061104a885SDavid Sterba if (ret) { 320705135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 32084871c33bSQu Wenruo ret = -EINVAL; 3209141386e1SJosef Bacik goto fail_alloc; 32101104a885SDavid Sterba } 32111104a885SDavid Sterba 32124871c33bSQu Wenruo if (!btrfs_super_root(disk_super)) { 32134871c33bSQu Wenruo btrfs_err(fs_info, "invalid superblock tree root bytenr"); 32144871c33bSQu Wenruo ret = -EINVAL; 3215141386e1SJosef Bacik goto fail_alloc; 32164871c33bSQu Wenruo } 32170f7d52f4SChris Mason 3218acce952bSliubo /* check FS state, whether FS is broken. */ 321987533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 322087533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3221acce952bSliubo 322275e7cb7fSLiu Bo /* 322375e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 322475e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 322575e7cb7fSLiu Bo */ 322675e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 322775e7cb7fSLiu Bo 32286f93e834SAnand Jain 32296f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 32306f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 32316f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 32326f93e834SAnand Jain stripesize = sectorsize; 32336f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 32346f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 32356f93e834SAnand Jain 32366f93e834SAnand Jain fs_info->nodesize = nodesize; 32376f93e834SAnand Jain fs_info->sectorsize = sectorsize; 32386f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 32396f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 32406f93e834SAnand Jain fs_info->stripesize = stripesize; 32416f93e834SAnand Jain 32422ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 32434871c33bSQu Wenruo if (ret) 3244141386e1SJosef Bacik goto fail_alloc; 3245dfe25020SChris Mason 32462ba48b20SQu Wenruo ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); 32474871c33bSQu Wenruo if (ret < 0) 3248141386e1SJosef Bacik goto fail_alloc; 3249f2b636e8SJosef Bacik 32508481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 32518481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 32528481dd80SQu Wenruo 32539f73f1aeSQu Wenruo /* 32549f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 32559f73f1aeSQu Wenruo * going to be deprecated. 32569f73f1aeSQu Wenruo * 32579f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 32589f73f1aeSQu Wenruo */ 32599f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 32609f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 32619f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 32629f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 32639f73f1aeSQu Wenruo 326495ea0486SQu Wenruo btrfs_warn(fs_info, 326595ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 32660bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 32678481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 32684871c33bSQu Wenruo if (!subpage_info) { 32694871c33bSQu Wenruo ret = -ENOMEM; 32708481dd80SQu Wenruo goto fail_alloc; 32714871c33bSQu Wenruo } 32728481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 32738481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3274c8050b3bSQu Wenruo } 32750bb3eb3eSQu Wenruo 3276d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 32774871c33bSQu Wenruo if (ret) 32780dc3b84aSJosef Bacik goto fail_sb_buffer; 32794543df7eSChris Mason 32809e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 32819e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 32824575c9ccSChris Mason 3283a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3284a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3285de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3286db94535dSChris Mason 3287925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 32886bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3289925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 329084eed90fSChris Mason if (ret) { 329105135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 32925d4f98a2SYan Zheng goto fail_sb_buffer; 329384eed90fSChris Mason } 32940b86a832SChris Mason 329584234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3296581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3297bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3298bd676446SJosef Bacik generation, level); 3299bd676446SJosef Bacik if (ret) { 330005135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3301af31f5e5SChris Mason goto fail_tree_roots; 330283121942SDavid Woodhouse } 33030b86a832SChris Mason 3304e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3305c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3306c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3307e17cade2SChris Mason 33085b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 33092b82032cSYan Zheng if (ret) { 331005135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3311af31f5e5SChris Mason goto fail_tree_roots; 33122b82032cSYan Zheng } 33130b86a832SChris Mason 33148dabb742SStefan Behrens /* 3315bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3316bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3317bacce86aSAnand Jain * target is required. So free devices that are not part of this 33181a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3319bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 33208dabb742SStefan Behrens */ 3321bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3322d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 332305135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 33244871c33bSQu Wenruo ret = -EIO; 3325a6b0d5c8SChris Mason goto fail_tree_roots; 3326a6b0d5c8SChris Mason } 3327a6b0d5c8SChris Mason 3328b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 33294bbcaa64SEric Sandeen if (ret) 3330b8522a1eSNikolay Borisov goto fail_tree_roots; 33318929ecfaSYan, Zheng 333275ec1db8SJosef Bacik /* 333373651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 333473651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 333573651042SNaohiro Aota * zoned incompat feature flag set. 333673651042SNaohiro Aota */ 333773651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 333873651042SNaohiro Aota if (ret) { 333973651042SNaohiro Aota btrfs_err(fs_info, 33404871c33bSQu Wenruo "zoned: failed to read device zone info: %d", ret); 334173651042SNaohiro Aota goto fail_block_groups; 334273651042SNaohiro Aota } 334373651042SNaohiro Aota 334473651042SNaohiro Aota /* 334575ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 334675ec1db8SJosef Bacik * check the generation here so we can set the 334775ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 334875ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 334975ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 335075ec1db8SJosef Bacik * even though it was perfectly fine. 335175ec1db8SJosef Bacik */ 335275ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 335375ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 335475ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 335575ec1db8SJosef Bacik 3356cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3357cf90d884SQu Wenruo if (ret) { 3358cf90d884SQu Wenruo btrfs_err(fs_info, 3359cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3360cf90d884SQu Wenruo ret); 3361cf90d884SQu Wenruo goto fail_block_groups; 3362cf90d884SQu Wenruo } 336368310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 336468310a5eSIlya Dryomov if (ret) { 336505135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 336668310a5eSIlya Dryomov goto fail_block_groups; 336768310a5eSIlya Dryomov } 336868310a5eSIlya Dryomov 3369733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3370733f4fbbSStefan Behrens if (ret) { 337105135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3372733f4fbbSStefan Behrens goto fail_block_groups; 3373733f4fbbSStefan Behrens } 3374733f4fbbSStefan Behrens 33758dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 33768dabb742SStefan Behrens if (ret) { 337705135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 33788dabb742SStefan Behrens goto fail_block_groups; 33798dabb742SStefan Behrens } 33808dabb742SStefan Behrens 3381b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3382b70f5097SNaohiro Aota if (ret) { 3383b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3384b70f5097SNaohiro Aota ret); 3385b70f5097SNaohiro Aota goto fail_block_groups; 3386b70f5097SNaohiro Aota } 3387b70f5097SNaohiro Aota 3388c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3389b7c35e81SAnand Jain if (ret) { 339005135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 339105135f59SDavid Sterba ret); 3392b7c35e81SAnand Jain goto fail_block_groups; 3393b7c35e81SAnand Jain } 3394b7c35e81SAnand Jain 339596f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 33965ac1d209SJeff Mahoney if (ret) { 339705135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3398b7c35e81SAnand Jain goto fail_fsdev_sysfs; 33995ac1d209SJeff Mahoney } 34005ac1d209SJeff Mahoney 3401c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3402c59021f8Sliubo if (ret) { 340305135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 34042365dd3cSAnand Jain goto fail_sysfs; 3405c59021f8Sliubo } 3406c59021f8Sliubo 34075b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 34081b1d1f66SJosef Bacik if (ret) { 340905135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 34102365dd3cSAnand Jain goto fail_sysfs; 34111b1d1f66SJosef Bacik } 34124330e183SQu Wenruo 341316beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 341416beac87SNaohiro Aota 34155c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 34165c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 341705135f59SDavid Sterba btrfs_warn(fs_info, 341852042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 34194871c33bSQu Wenruo ret = -EINVAL; 34202365dd3cSAnand Jain goto fail_sysfs; 3421292fd7fcSStefan Behrens } 34229078a3e1SChris Mason 342333c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3424a74a4b97SChris Mason "btrfs-cleaner"); 34254871c33bSQu Wenruo if (IS_ERR(fs_info->cleaner_kthread)) { 34264871c33bSQu Wenruo ret = PTR_ERR(fs_info->cleaner_kthread); 34272365dd3cSAnand Jain goto fail_sysfs; 34284871c33bSQu Wenruo } 3429a74a4b97SChris Mason 3430a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3431a74a4b97SChris Mason tree_root, 3432a74a4b97SChris Mason "btrfs-transaction"); 34334871c33bSQu Wenruo if (IS_ERR(fs_info->transaction_kthread)) { 34344871c33bSQu Wenruo ret = PTR_ERR(fs_info->transaction_kthread); 34353f157a2fSChris Mason goto fail_cleaner; 34364871c33bSQu Wenruo } 3437a74a4b97SChris Mason 3438583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3439c289811cSChris Mason !fs_info->fs_devices->rotating) { 3440583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3441c289811cSChris Mason } 3442c289811cSChris Mason 3443572d9ab7SDavid Sterba /* 344463a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 344563a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 344663a7cb13SDavid Sterba * nodiscard for the same mount. 344763a7cb13SDavid Sterba */ 344863a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 344963a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 345063a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 345163a7cb13SDavid Sterba fs_info->fs_devices->discardable) { 345263a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 345363a7cb13SDavid Sterba "auto enabling async discard"); 345463a7cb13SDavid Sterba } 345563a7cb13SDavid Sterba 345621adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 34570b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 34582ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 34590b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3460cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 346121adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 346221adbd5cSStefan Behrens if (ret) 346305135f59SDavid Sterba btrfs_warn(fs_info, 346405135f59SDavid Sterba "failed to initialize integrity check module: %d", 346505135f59SDavid Sterba ret); 346621adbd5cSStefan Behrens } 346721adbd5cSStefan Behrens #endif 3468bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3469bcef60f2SArne Jansen if (ret) 3470bcef60f2SArne Jansen goto fail_trans_kthread; 347121adbd5cSStefan Behrens 3472fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3473fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3474fd708b81SJosef Bacik 347596da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 347696da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 34770b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3478e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 347963443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 34804871c33bSQu Wenruo if (ret) 348128c16cbbSWang Shilong goto fail_qgroup; 3482e556ce2cSYan Zheng } 34831a40e23bSZheng Yan 348456e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 34853140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 34864871c33bSQu Wenruo ret = PTR_ERR(fs_info->fs_root); 34874871c33bSQu Wenruo btrfs_warn(fs_info, "failed to read fs tree: %d", ret); 3488315bf8efSJosef Bacik fs_info->fs_root = NULL; 3489bcef60f2SArne Jansen goto fail_qgroup; 34903140c9a3SDan Carpenter } 3491c289811cSChris Mason 3492bc98a42cSDavid Howells if (sb_rdonly(sb)) 34938cd29088SBoris Burkov goto clear_oneshot; 34942b6ba629SIlya Dryomov 349544c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 34962b6ba629SIlya Dryomov if (ret) { 34976bccf3abSJeff Mahoney close_ctree(fs_info); 34982b6ba629SIlya Dryomov return ret; 3499e3acc2a6SJosef Bacik } 3500b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3501b382a324SJan Schmidt 350244c0ca21SBoris Burkov if (fs_info->uuid_root && 350344c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 350444c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 350505135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 350670f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 350770f80175SStefan Behrens if (ret) { 350805135f59SDavid Sterba btrfs_warn(fs_info, 350905135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 35106bccf3abSJeff Mahoney close_ctree(fs_info); 351170f80175SStefan Behrens return ret; 351270f80175SStefan Behrens } 3513f7a81ea4SStefan Behrens } 351494846229SBoris Burkov 3515afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 351647ab2a6cSJosef Bacik 3517b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3518b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3519b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3520b4be6aefSJosef Bacik 35218cd29088SBoris Burkov clear_oneshot: 35228cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3523ad2b2c80SAl Viro return 0; 352439279cc3SChris Mason 3525bcef60f2SArne Jansen fail_qgroup: 3526bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 35277c2ca468SChris Mason fail_trans_kthread: 35287c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 35292ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3530faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 35313f157a2fSChris Mason fail_cleaner: 3532a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 35337c2ca468SChris Mason 35347c2ca468SChris Mason /* 35357c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 35367c2ca468SChris Mason * kthreads 35377c2ca468SChris Mason */ 35387c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 35397c2ca468SChris Mason 35402365dd3cSAnand Jain fail_sysfs: 35416618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 35422365dd3cSAnand Jain 3543b7c35e81SAnand Jain fail_fsdev_sysfs: 3544b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3545b7c35e81SAnand Jain 35461b1d1f66SJosef Bacik fail_block_groups: 354754067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3548af31f5e5SChris Mason 3549af31f5e5SChris Mason fail_tree_roots: 35509e3aa805SJosef Bacik if (fs_info->data_reloc_root) 35519e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 35524273eaffSAnand Jain free_root_pointers(fs_info, true); 35532b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3554af31f5e5SChris Mason 355539279cc3SChris Mason fail_sb_buffer: 35567abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 35575cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 355816cdcec7SMiao Xie fail_alloc: 3559586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3560586e46e2SIlya Dryomov 35614543df7eSChris Mason iput(fs_info->btree_inode); 35627e662854SQinghuang Feng fail: 3563586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 35644871c33bSQu Wenruo ASSERT(ret < 0); 35654871c33bSQu Wenruo return ret; 3566eb60ceacSChris Mason } 3567663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3568eb60ceacSChris Mason 3569314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3570f2984462SChris Mason { 3571314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3572314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3573314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3574314b6dd0SJohannes Thumshirn struct page *page; 3575442a4f63SStefan Behrens 3576314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3577314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3578314b6dd0SJohannes Thumshirn 3579314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3580fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3581314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3582cb3e217bSQu Wenruo btrfs_dev_name(device), 3583314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3584314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3585314b6dd0SJohannes Thumshirn SetPageError(page); 3586314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3587314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3588314b6dd0SJohannes Thumshirn } else { 3589314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3590f2984462SChris Mason } 3591314b6dd0SJohannes Thumshirn 3592314b6dd0SJohannes Thumshirn put_page(page); 3593314b6dd0SJohannes Thumshirn unlock_page(page); 3594314b6dd0SJohannes Thumshirn } 3595314b6dd0SJohannes Thumshirn 3596314b6dd0SJohannes Thumshirn bio_put(bio); 3597f2984462SChris Mason } 3598f2984462SChris Mason 35998f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3600a05d3c91SQu Wenruo int copy_num, bool drop_cache) 360129c36d72SAnand Jain { 360229c36d72SAnand Jain struct btrfs_super_block *super; 36038f32380dSJohannes Thumshirn struct page *page; 360412659251SNaohiro Aota u64 bytenr, bytenr_orig; 36058f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 360612659251SNaohiro Aota int ret; 360729c36d72SAnand Jain 360812659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 360912659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 361012659251SNaohiro Aota if (ret == -ENOENT) 361112659251SNaohiro Aota return ERR_PTR(-EINVAL); 361212659251SNaohiro Aota else if (ret) 361312659251SNaohiro Aota return ERR_PTR(ret); 361412659251SNaohiro Aota 3615cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 36168f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 361729c36d72SAnand Jain 3618a05d3c91SQu Wenruo if (drop_cache) { 3619a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3620a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3621a05d3c91SQu Wenruo 3622a05d3c91SQu Wenruo /* 3623a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3624a05d3c91SQu Wenruo * always read from the device. 3625a05d3c91SQu Wenruo */ 3626a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3627a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3628a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3629a05d3c91SQu Wenruo } 3630a05d3c91SQu Wenruo 36318f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 36328f32380dSJohannes Thumshirn if (IS_ERR(page)) 36338f32380dSJohannes Thumshirn return ERR_CAST(page); 363429c36d72SAnand Jain 36358f32380dSJohannes Thumshirn super = page_address(page); 363696c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 363796c2e067SAnand Jain btrfs_release_disk_super(super); 363896c2e067SAnand Jain return ERR_PTR(-ENODATA); 363996c2e067SAnand Jain } 364096c2e067SAnand Jain 364112659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 36428f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36438f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 364429c36d72SAnand Jain } 364529c36d72SAnand Jain 36468f32380dSJohannes Thumshirn return super; 364729c36d72SAnand Jain } 364829c36d72SAnand Jain 364929c36d72SAnand Jain 36508f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3651a512bbf8SYan Zheng { 36528f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3653a512bbf8SYan Zheng int i; 3654a512bbf8SYan Zheng u64 transid = 0; 3655a512bbf8SYan Zheng 3656a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3657a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3658a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3659a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3660a512bbf8SYan Zheng */ 3661a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3662a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 36638f32380dSJohannes Thumshirn if (IS_ERR(super)) 3664a512bbf8SYan Zheng continue; 3665a512bbf8SYan Zheng 3666a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 36678f32380dSJohannes Thumshirn if (latest) 36688f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36698f32380dSJohannes Thumshirn 36708f32380dSJohannes Thumshirn latest = super; 3671a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3672a512bbf8SYan Zheng } 3673a512bbf8SYan Zheng } 367492fc03fbSAnand Jain 36758f32380dSJohannes Thumshirn return super; 3676a512bbf8SYan Zheng } 3677a512bbf8SYan Zheng 36784eedeb75SHisashi Hifumi /* 3679abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3680314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 36814eedeb75SHisashi Hifumi * 3682abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3683abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3684abbb3b8eSDavid Sterba * same for write and wait phases. 36854eedeb75SHisashi Hifumi * 3686314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 36874eedeb75SHisashi Hifumi */ 3688a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3689abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3690a512bbf8SYan Zheng { 3691d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3692314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3693d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3694a512bbf8SYan Zheng int i; 3695a512bbf8SYan Zheng int errors = 0; 369612659251SNaohiro Aota int ret; 369712659251SNaohiro Aota u64 bytenr, bytenr_orig; 3698a512bbf8SYan Zheng 3699a512bbf8SYan Zheng if (max_mirrors == 0) 3700a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3701a512bbf8SYan Zheng 3702d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3703d5178578SJohannes Thumshirn 3704a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3705314b6dd0SJohannes Thumshirn struct page *page; 3706314b6dd0SJohannes Thumshirn struct bio *bio; 3707314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3708314b6dd0SJohannes Thumshirn 370912659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 371012659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 371112659251SNaohiro Aota if (ret == -ENOENT) { 371212659251SNaohiro Aota continue; 371312659251SNaohiro Aota } else if (ret < 0) { 371412659251SNaohiro Aota btrfs_err(device->fs_info, 371512659251SNaohiro Aota "couldn't get super block location for mirror %d", 371612659251SNaohiro Aota i); 371712659251SNaohiro Aota errors++; 371812659251SNaohiro Aota continue; 371912659251SNaohiro Aota } 3720935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3721935e5cc9SMiao Xie device->commit_total_bytes) 3722a512bbf8SYan Zheng break; 3723a512bbf8SYan Zheng 372412659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3725a512bbf8SYan Zheng 3726fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3727fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3728fd08001fSEric Biggers sb->csum); 3729a512bbf8SYan Zheng 3730314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3731314b6dd0SJohannes Thumshirn GFP_NOFS); 3732314b6dd0SJohannes Thumshirn if (!page) { 3733fb456252SJeff Mahoney btrfs_err(device->fs_info, 3734314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3735f14d104dSDavid Sterba bytenr); 3736634554dcSJosef Bacik errors++; 3737634554dcSJosef Bacik continue; 3738634554dcSJosef Bacik } 3739634554dcSJosef Bacik 3740314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3741314b6dd0SJohannes Thumshirn get_page(page); 3742a512bbf8SYan Zheng 3743314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3744314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3745a512bbf8SYan Zheng 3746387125fcSChris Mason /* 3747314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3748314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3749314b6dd0SJohannes Thumshirn * checking. 3750387125fcSChris Mason */ 375107888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 375207888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 375307888c66SChristoph Hellwig GFP_NOFS); 3754314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3755314b6dd0SJohannes Thumshirn bio->bi_private = device; 3756314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3757314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3758314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3759314b6dd0SJohannes Thumshirn 3760314b6dd0SJohannes Thumshirn /* 3761314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3762314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3763314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3764314b6dd0SJohannes Thumshirn */ 37651b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3766314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3767314b6dd0SJohannes Thumshirn 376858ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 376958ff51f1SChristoph Hellwig submit_bio(bio); 37708376d9e1SNaohiro Aota 37718376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 37728376d9e1SNaohiro Aota errors++; 3773a512bbf8SYan Zheng } 3774a512bbf8SYan Zheng return errors < i ? 0 : -1; 3775a512bbf8SYan Zheng } 3776a512bbf8SYan Zheng 3777387125fcSChris Mason /* 3778abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3779abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3780abbb3b8eSDavid Sterba * 3781314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3782abbb3b8eSDavid Sterba * date. 3783abbb3b8eSDavid Sterba */ 3784abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3785abbb3b8eSDavid Sterba { 3786abbb3b8eSDavid Sterba int i; 3787abbb3b8eSDavid Sterba int errors = 0; 3788b6a535faSHoward McLauchlan bool primary_failed = false; 378912659251SNaohiro Aota int ret; 3790abbb3b8eSDavid Sterba u64 bytenr; 3791abbb3b8eSDavid Sterba 3792abbb3b8eSDavid Sterba if (max_mirrors == 0) 3793abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3794abbb3b8eSDavid Sterba 3795abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3796314b6dd0SJohannes Thumshirn struct page *page; 3797314b6dd0SJohannes Thumshirn 379812659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 379912659251SNaohiro Aota if (ret == -ENOENT) { 380012659251SNaohiro Aota break; 380112659251SNaohiro Aota } else if (ret < 0) { 380212659251SNaohiro Aota errors++; 380312659251SNaohiro Aota if (i == 0) 380412659251SNaohiro Aota primary_failed = true; 380512659251SNaohiro Aota continue; 380612659251SNaohiro Aota } 3807abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3808abbb3b8eSDavid Sterba device->commit_total_bytes) 3809abbb3b8eSDavid Sterba break; 3810abbb3b8eSDavid Sterba 3811314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3812314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3813314b6dd0SJohannes Thumshirn if (!page) { 3814abbb3b8eSDavid Sterba errors++; 3815b6a535faSHoward McLauchlan if (i == 0) 3816b6a535faSHoward McLauchlan primary_failed = true; 3817abbb3b8eSDavid Sterba continue; 3818abbb3b8eSDavid Sterba } 3819314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3820314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3821314b6dd0SJohannes Thumshirn if (PageError(page)) { 3822abbb3b8eSDavid Sterba errors++; 3823b6a535faSHoward McLauchlan if (i == 0) 3824b6a535faSHoward McLauchlan primary_failed = true; 3825b6a535faSHoward McLauchlan } 3826abbb3b8eSDavid Sterba 3827314b6dd0SJohannes Thumshirn /* Drop our reference */ 3828314b6dd0SJohannes Thumshirn put_page(page); 3829abbb3b8eSDavid Sterba 3830314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3831314b6dd0SJohannes Thumshirn put_page(page); 3832abbb3b8eSDavid Sterba } 3833abbb3b8eSDavid Sterba 3834b6a535faSHoward McLauchlan /* log error, force error return */ 3835b6a535faSHoward McLauchlan if (primary_failed) { 3836b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3837b6a535faSHoward McLauchlan device->devid); 3838b6a535faSHoward McLauchlan return -1; 3839b6a535faSHoward McLauchlan } 3840b6a535faSHoward McLauchlan 3841abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3842abbb3b8eSDavid Sterba } 3843abbb3b8eSDavid Sterba 3844abbb3b8eSDavid Sterba /* 3845387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3846387125fcSChris Mason * for the barrier when it is done 3847387125fcSChris Mason */ 38484246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3849387125fcSChris Mason { 3850f9e69aa9SChristoph Hellwig bio_uninit(bio); 3851387125fcSChris Mason complete(bio->bi_private); 3852387125fcSChris Mason } 3853387125fcSChris Mason 3854387125fcSChris Mason /* 38554fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 38564fc6441aSAnand Jain * done in the waiting counterpart. 3857387125fcSChris Mason */ 38584fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3859387125fcSChris Mason { 3860f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 3861387125fcSChris Mason 3862bfd3ea94SAnand Jain device->last_flush_error = BLK_STS_OK; 3863bfd3ea94SAnand Jain 3864a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3865a91cf0ffSWang Yugui /* 3866a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 3867a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 3868a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 3869a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 3870a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 3871a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 3872a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 3873a91cf0ffSWang Yugui * non-debug builds. 3874a91cf0ffSWang Yugui */ 387508e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 38764fc6441aSAnand Jain return; 3877a91cf0ffSWang Yugui #endif 3878387125fcSChris Mason 3879f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 3880f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 3881387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3882387125fcSChris Mason init_completion(&device->flush_wait); 3883387125fcSChris Mason bio->bi_private = &device->flush_wait; 3884387125fcSChris Mason 388558ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 388658ff51f1SChristoph Hellwig submit_bio(bio); 38871c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 38884fc6441aSAnand Jain } 3889387125fcSChris Mason 38904fc6441aSAnand Jain /* 38914fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 38921b465784SAnand Jain * Return true for any error, and false otherwise. 38934fc6441aSAnand Jain */ 38941b465784SAnand Jain static bool wait_dev_flush(struct btrfs_device *device) 38954fc6441aSAnand Jain { 3896f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 38974fc6441aSAnand Jain 38987e812f20SAnand Jain if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 38991b465784SAnand Jain return false; 39004fc6441aSAnand Jain 39012980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 39024fc6441aSAnand Jain 3903bfd3ea94SAnand Jain if (bio->bi_status) { 3904bfd3ea94SAnand Jain device->last_flush_error = bio->bi_status; 3905bfd3ea94SAnand Jain btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); 39061b465784SAnand Jain return true; 3907bfd3ea94SAnand Jain } 3908bfd3ea94SAnand Jain 39091b465784SAnand Jain return false; 3910387125fcSChris Mason } 3911387125fcSChris Mason 3912387125fcSChris Mason /* 3913387125fcSChris Mason * send an empty flush down to each device in parallel, 3914387125fcSChris Mason * then wait for them 3915387125fcSChris Mason */ 3916387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3917387125fcSChris Mason { 3918387125fcSChris Mason struct list_head *head; 3919387125fcSChris Mason struct btrfs_device *dev; 39205af3e8ccSStefan Behrens int errors_wait = 0; 3921387125fcSChris Mason 39221538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3923387125fcSChris Mason /* send down all the barriers */ 3924387125fcSChris Mason head = &info->fs_devices->devices; 39251538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3926e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3927f88ba6a2SHidetoshi Seto continue; 3928cea7c8bfSAnand Jain if (!dev->bdev) 3929387125fcSChris Mason continue; 3930e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3931ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3932387125fcSChris Mason continue; 3933387125fcSChris Mason 39344fc6441aSAnand Jain write_dev_flush(dev); 3935387125fcSChris Mason } 3936387125fcSChris Mason 3937387125fcSChris Mason /* wait for all the barriers */ 39381538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3939e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3940f88ba6a2SHidetoshi Seto continue; 3941387125fcSChris Mason if (!dev->bdev) { 39425af3e8ccSStefan Behrens errors_wait++; 3943387125fcSChris Mason continue; 3944387125fcSChris Mason } 3945e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3946ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3947387125fcSChris Mason continue; 3948387125fcSChris Mason 39491b465784SAnand Jain if (wait_dev_flush(dev)) 39505af3e8ccSStefan Behrens errors_wait++; 3951387125fcSChris Mason } 3952401b41e5SAnand Jain 3953401b41e5SAnand Jain /* 3954de38a206SAnand Jain * Checks last_flush_error of disks in order to determine the device 3955de38a206SAnand Jain * state. 3956401b41e5SAnand Jain */ 3957de38a206SAnand Jain if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) 3958de38a206SAnand Jain return -EIO; 3959de38a206SAnand Jain 3960387125fcSChris Mason return 0; 3961387125fcSChris Mason } 3962387125fcSChris Mason 3963943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3964943c6e99SZhao Lei { 39658789f4feSZhao Lei int raid_type; 39668789f4feSZhao Lei int min_tolerated = INT_MAX; 3967943c6e99SZhao Lei 39688789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 39698789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 39708c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 39718789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 39728789f4feSZhao Lei tolerated_failures); 3973943c6e99SZhao Lei 39748789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 39758789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 39768789f4feSZhao Lei continue; 397741a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 39788789f4feSZhao Lei continue; 39798c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 39808789f4feSZhao Lei btrfs_raid_array[raid_type]. 39818789f4feSZhao Lei tolerated_failures); 39828789f4feSZhao Lei } 3983943c6e99SZhao Lei 39848789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3985ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 39868789f4feSZhao Lei min_tolerated = 0; 39878789f4feSZhao Lei } 39888789f4feSZhao Lei 39898789f4feSZhao Lei return min_tolerated; 3990943c6e99SZhao Lei } 3991943c6e99SZhao Lei 3992eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3993f2984462SChris Mason { 3994e5e9a520SChris Mason struct list_head *head; 3995f2984462SChris Mason struct btrfs_device *dev; 3996a061fc8dSChris Mason struct btrfs_super_block *sb; 3997f2984462SChris Mason struct btrfs_dev_item *dev_item; 3998f2984462SChris Mason int ret; 3999f2984462SChris Mason int do_barriers; 4000a236aed1SChris Mason int max_errors; 4001a236aed1SChris Mason int total_errors = 0; 4002a061fc8dSChris Mason u64 flags; 4003f2984462SChris Mason 40040b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4005fed3b381SLiu Bo 4006fed3b381SLiu Bo /* 4007fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4008fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4009fed3b381SLiu Bo * been consistent on disk. 4010fed3b381SLiu Bo */ 4011fed3b381SLiu Bo if (max_mirrors == 0) 40120b246afaSJeff Mahoney backup_super_roots(fs_info); 4013f2984462SChris Mason 40140b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4015a061fc8dSChris Mason dev_item = &sb->dev_item; 4016e5e9a520SChris Mason 40170b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 40180b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 40190b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4020387125fcSChris Mason 40215af3e8ccSStefan Behrens if (do_barriers) { 40220b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 40235af3e8ccSStefan Behrens if (ret) { 40245af3e8ccSStefan Behrens mutex_unlock( 40250b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 40260b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 40275af3e8ccSStefan Behrens "errors while submitting device barriers."); 40285af3e8ccSStefan Behrens return ret; 40295af3e8ccSStefan Behrens } 40305af3e8ccSStefan Behrens } 4031387125fcSChris Mason 40321538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4033dfe25020SChris Mason if (!dev->bdev) { 4034dfe25020SChris Mason total_errors++; 4035dfe25020SChris Mason continue; 4036dfe25020SChris Mason } 4037e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4038ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4039dfe25020SChris Mason continue; 4040dfe25020SChris Mason 40412b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4042a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4043a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 40447df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4045935e5cc9SMiao Xie dev->commit_total_bytes); 4046ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4047ce7213c7SMiao Xie dev->commit_bytes_used); 4048a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4049a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4050a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4051a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 40527239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 40537239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4054a512bbf8SYan Zheng 4055a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4056a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4057f2984462SChris Mason 405875cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 405975cb857dSQu Wenruo if (ret < 0) { 406075cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 406175cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 406275cb857dSQu Wenruo "unexpected superblock corruption detected"); 406375cb857dSQu Wenruo return -EUCLEAN; 406475cb857dSQu Wenruo } 406575cb857dSQu Wenruo 4066abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4067a236aed1SChris Mason if (ret) 4068a236aed1SChris Mason total_errors++; 4069f2984462SChris Mason } 4070a236aed1SChris Mason if (total_errors > max_errors) { 40710b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4072d397712bSChris Mason total_errors); 40730b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 407479787eaaSJeff Mahoney 40759d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 40760b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 40770b246afaSJeff Mahoney "%d errors while writing supers", 40780b246afaSJeff Mahoney total_errors); 40799d565ba4SStefan Behrens return -EIO; 4080a236aed1SChris Mason } 4081f2984462SChris Mason 4082a512bbf8SYan Zheng total_errors = 0; 40831538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4084dfe25020SChris Mason if (!dev->bdev) 4085dfe25020SChris Mason continue; 4086e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4087ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4088dfe25020SChris Mason continue; 4089dfe25020SChris Mason 4090abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4091a512bbf8SYan Zheng if (ret) 40921259ab75SChris Mason total_errors++; 4093f2984462SChris Mason } 40940b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4095a236aed1SChris Mason if (total_errors > max_errors) { 40960b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 40970b246afaSJeff Mahoney "%d errors while writing supers", 40980b246afaSJeff Mahoney total_errors); 409979787eaaSJeff Mahoney return -EIO; 4100a236aed1SChris Mason } 4101f2984462SChris Mason return 0; 4102f2984462SChris Mason } 4103f2984462SChris Mason 4104cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4105cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4106cb517eabSMiao Xie struct btrfs_root *root) 41072619ba1fSChris Mason { 41084785e24fSJosef Bacik bool drop_ref = false; 41094785e24fSJosef Bacik 4110fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4111fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4112fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4113fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 41144785e24fSJosef Bacik drop_ref = true; 4115fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 411676dda93cSYan, Zheng 411784961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4118ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 41191c1ea4f7SLiu Bo if (root->reloc_root) { 412000246528SJosef Bacik btrfs_put_root(root->reloc_root); 41211c1ea4f7SLiu Bo root->reloc_root = NULL; 41221c1ea4f7SLiu Bo } 41231c1ea4f7SLiu Bo } 41243321719eSLiu Bo 41254785e24fSJosef Bacik if (drop_ref) 412600246528SJosef Bacik btrfs_put_root(root); 41272619ba1fSChris Mason } 41282619ba1fSChris Mason 4129c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4130c146afadSYan Zheng { 4131fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4132fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4133fc7cbcd4SDavid Sterba int i = 0; 413465d33fd7SQu Wenruo int err = 0; 4135fc7cbcd4SDavid Sterba unsigned int ret = 0; 4136c146afadSYan Zheng 4137c146afadSYan Zheng while (1) { 4138fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4139fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4140fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4141fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4142fc7cbcd4SDavid Sterba if (!ret) { 4143fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4144c146afadSYan Zheng break; 414565d33fd7SQu Wenruo } 4146fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 414766b4ffd1SJosef Bacik 4148fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4149fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4150fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4151fc7cbcd4SDavid Sterba gang[i] = NULL; 415265d33fd7SQu Wenruo continue; 4153c146afadSYan Zheng } 4154fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4155fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4156fc7cbcd4SDavid Sterba } 4157fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4158fc7cbcd4SDavid Sterba 4159fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4160fc7cbcd4SDavid Sterba if (!gang[i]) 4161fc7cbcd4SDavid Sterba continue; 4162fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4163fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4164fc7cbcd4SDavid Sterba if (err) 41656989627dSJosef Bacik goto out; 4166fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4167fc7cbcd4SDavid Sterba } 4168fc7cbcd4SDavid Sterba root_objectid++; 4169c146afadSYan Zheng } 41706989627dSJosef Bacik out: 4171fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4172fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4173fc7cbcd4SDavid Sterba if (gang[i]) 4174fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 417565d33fd7SQu Wenruo } 417665d33fd7SQu Wenruo return err; 4177c146afadSYan Zheng } 4178c146afadSYan Zheng 41796bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4180c146afadSYan Zheng { 41816bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4182c146afadSYan Zheng struct btrfs_trans_handle *trans; 4183c146afadSYan Zheng 41840b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 41852ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 41860b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 41870b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4188c71bf099SYan, Zheng 4189c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 41900b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 41910b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4192c71bf099SYan, Zheng 41937a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 41943612b495STsutomu Itoh if (IS_ERR(trans)) 41953612b495STsutomu Itoh return PTR_ERR(trans); 41963a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4197c146afadSYan Zheng } 4198c146afadSYan Zheng 419936c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 420036c86a9eSQu Wenruo { 420136c86a9eSQu Wenruo struct btrfs_transaction *trans; 420236c86a9eSQu Wenruo struct btrfs_transaction *tmp; 420336c86a9eSQu Wenruo bool found = false; 420436c86a9eSQu Wenruo 420536c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 420636c86a9eSQu Wenruo return; 420736c86a9eSQu Wenruo 420836c86a9eSQu Wenruo /* 420936c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 421036c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 421136c86a9eSQu Wenruo */ 421236c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 421336c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 421436c86a9eSQu Wenruo struct extent_state *cached = NULL; 421536c86a9eSQu Wenruo u64 dirty_bytes = 0; 421636c86a9eSQu Wenruo u64 cur = 0; 421736c86a9eSQu Wenruo u64 found_start; 421836c86a9eSQu Wenruo u64 found_end; 421936c86a9eSQu Wenruo 422036c86a9eSQu Wenruo found = true; 422136c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 422236c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 422336c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 422436c86a9eSQu Wenruo cur = found_end + 1; 422536c86a9eSQu Wenruo } 422636c86a9eSQu Wenruo btrfs_warn(fs_info, 422736c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 422836c86a9eSQu Wenruo trans->transid, dirty_bytes); 422936c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 423036c86a9eSQu Wenruo 423136c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 423236c86a9eSQu Wenruo fs_info->running_transaction = NULL; 423336c86a9eSQu Wenruo list_del_init(&trans->list); 423436c86a9eSQu Wenruo 423536c86a9eSQu Wenruo btrfs_put_transaction(trans); 423636c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 423736c86a9eSQu Wenruo } 423836c86a9eSQu Wenruo ASSERT(!found); 423936c86a9eSQu Wenruo } 424036c86a9eSQu Wenruo 4241b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4242eb60ceacSChris Mason { 4243c146afadSYan Zheng int ret; 4244e089f05cSChris Mason 4245afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 424631e70e52SFilipe Manana 424731e70e52SFilipe Manana /* 42488a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 42498a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 42508a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 42518a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 42528a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 42538a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 42548a1f1e3dSFilipe Manana * return 1. 42558a1f1e3dSFilipe Manana */ 42568a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 42578a1f1e3dSFilipe Manana 42588a1f1e3dSFilipe Manana /* 425931e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 426031e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 426131e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 426231e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 426331e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 426431e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 426531e70e52SFilipe Manana * trying to stop the async reclaim task. 426631e70e52SFilipe Manana */ 426731e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4268d6fd0ae2SOmar Sandoval /* 4269d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4270d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4271d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4272d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4273d6fd0ae2SOmar Sandoval */ 4274d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4275a2135011SChris Mason 42767343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4277d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 42787343dd61SJustin Maggard 4279803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4280803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4281803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4282803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4283803b2f54SStefan Behrens 4284837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4285aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4286837d5b6eSIlya Dryomov 42878dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 42888dabb742SStefan Behrens 4289aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 42904cb5300bSChris Mason 42914cb5300bSChris Mason /* wait for any defraggers to finish */ 42924cb5300bSChris Mason wait_event(fs_info->transaction_wait, 42934cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 42944cb5300bSChris Mason 42954cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 429626176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 42974cb5300bSChris Mason 4298a362bb86SFilipe Manana /* 4299a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4300a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4301a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4302a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4303a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4304a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4305a362bb86SFilipe Manana * no one else to run iputs. 4306a362bb86SFilipe Manana * 4307a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4308a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4309a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4310a362bb86SFilipe Manana * through some other means. 4311a362bb86SFilipe Manana * 4312a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4313a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4314a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4315a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4316a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4317a362bb86SFilipe Manana */ 4318a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4319a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4320a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4321a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4322a362bb86SFilipe Manana 432321c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 432457056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4325576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 432621c7e756SMiao Xie 4327b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4328b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4329b0643e59SDennis Zhou 4330bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4331e44163e1SJeff Mahoney /* 4332d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4333d6fd0ae2SOmar Sandoval * unused block groups. 4334e44163e1SJeff Mahoney */ 43350b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4336e44163e1SJeff Mahoney 4337f0cc2cd7SFilipe Manana /* 4338f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4339f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4340f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4341f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4342f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4343f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4344f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4345f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4346f0cc2cd7SFilipe Manana * This is a very rare case. 4347f0cc2cd7SFilipe Manana */ 4348f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4349f0cc2cd7SFilipe Manana 43506bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4351d397712bSChris Mason if (ret) 435204892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4353c146afadSYan Zheng } 4354ed2ff2cbSChris Mason 435584961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 43562ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4357acce952bSliubo 4358e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4359e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 43608929ecfaSYan, Zheng 4361e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4362afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4363f25784b3SYan Zheng 43645958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 43655958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 43665958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 43675958253cSQu Wenruo } 43685958253cSQu Wenruo 436904892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4370fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4371bcef60f2SArne Jansen 4372963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 437304892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4374963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4375b0c68f8bSChris Mason } 437631153d81SYan Zheng 43775deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 43784297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 43795deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 43804297ff84SJosef Bacik 43816618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4382b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 43835ac1d209SJeff Mahoney 43841a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 43851a4319ccSLiu Bo 4386de348ee0SWang Shilong /* 4387de348ee0SWang Shilong * we must make sure there is not any read request to 4388de348ee0SWang Shilong * submit after we stopping all workers. 4389de348ee0SWang Shilong */ 4390de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 439196192499SJosef Bacik btrfs_stop_all_workers(fs_info); 439296192499SJosef Bacik 43930a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 439436c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 43950a31daa4SFilipe Manana 4396afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 43974273eaffSAnand Jain free_root_pointers(fs_info, true); 43988c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 43999ad6b7bcSChris Mason 44004e19443dSJosef Bacik /* 44014e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 44024e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 44034e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 44044e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 44054e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 44064e19443dSJosef Bacik */ 44074e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 44084e19443dSJosef Bacik 440913e6c37bSJosef Bacik iput(fs_info->btree_inode); 4410d6bfde87SChris Mason 441121adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 44120b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 44132ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 441421adbd5cSStefan Behrens #endif 441521adbd5cSStefan Behrens 44160b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 441768c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4418eb60ceacSChris Mason } 4419eb60ceacSChris Mason 44205f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 44215f39d397SChris Mason { 44222f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 44235f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4424b4ce94deSChris Mason 442506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 442606ea65a3SJosef Bacik /* 442706ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 442852042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 442906ea65a3SJosef Bacik * outside of the sanity tests. 443006ea65a3SJosef Bacik */ 4431b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 443206ea65a3SJosef Bacik return; 443306ea65a3SJosef Bacik #endif 443449d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 44350b246afaSJeff Mahoney if (transid != fs_info->generation) 44365d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 44370b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 4438f18cc978SChristoph Hellwig set_extent_buffer_dirty(buf); 44391f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 444069fc6cbbSQu Wenruo /* 444185d8a826SJosef Bacik * btrfs_check_leaf() won't check item data if we don't have WRITTEN 444285d8a826SJosef Bacik * set, so this will only validate the basic structure of the items. 444369fc6cbbSQu Wenruo */ 444485d8a826SJosef Bacik if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(buf)) { 4445a4f78750SDavid Sterba btrfs_print_leaf(buf); 44461f21ef0aSFilipe Manana ASSERT(0); 44471f21ef0aSFilipe Manana } 44481f21ef0aSFilipe Manana #endif 4449eb60ceacSChris Mason } 4450eb60ceacSChris Mason 44512ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4452b53d3f5dSLiu Bo int flush_delayed) 445335b7e476SChris Mason { 4454188de649SChris Mason /* 4455188de649SChris Mason * looks as though older kernels can get into trouble with 4456188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4457188de649SChris Mason */ 4458e2d84521SMiao Xie int ret; 4459d6bfde87SChris Mason 44606933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4461d6bfde87SChris Mason return; 4462d6bfde87SChris Mason 4463b53d3f5dSLiu Bo if (flush_delayed) 44642ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 446516cdcec7SMiao Xie 4466d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4467d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4468d814a491SEthan Lien fs_info->dirty_metadata_batch); 4469e2d84521SMiao Xie if (ret > 0) { 44700b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 447116cdcec7SMiao Xie } 447216cdcec7SMiao Xie } 447316cdcec7SMiao Xie 44742ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 447516cdcec7SMiao Xie { 44762ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 447735b7e476SChris Mason } 4478b53d3f5dSLiu Bo 44792ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4480b53d3f5dSLiu Bo { 44812ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4482d6bfde87SChris Mason } 44836b80053dSChris Mason 44842ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4485acce952bSliubo { 4486fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4487fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4488fe816d0fSNikolay Borisov 44890b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 44902ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 44910b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4492acce952bSliubo 44930b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 44940b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4495acce952bSliubo } 4496acce952bSliubo 4497ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4498ef67963dSJosef Bacik { 4499fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4500fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4501fc7cbcd4SDavid Sterba int ret; 4502ef67963dSJosef Bacik 4503fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4504fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4505fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4506fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4507fc7cbcd4SDavid Sterba int i; 4508ef67963dSJosef Bacik 4509fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4510fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4511fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4512fc7cbcd4SDavid Sterba 4513fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4514fc7cbcd4SDavid Sterba if (!gang[i]) 4515ef67963dSJosef Bacik continue; 4516fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4517fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4518fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4519ef67963dSJosef Bacik } 4520fc7cbcd4SDavid Sterba root_objectid++; 4521fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4522ef67963dSJosef Bacik } 4523fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4524ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4525ef67963dSJosef Bacik } 4526ef67963dSJosef Bacik 4527143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4528acce952bSliubo { 4529acce952bSliubo struct btrfs_ordered_extent *ordered; 4530acce952bSliubo 4531199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4532779880efSJosef Bacik /* 4533779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4534779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4535779880efSJosef Bacik */ 4536199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4537779880efSJosef Bacik root_extent_list) 4538779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4539199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4540199c2a9cSMiao Xie } 4541199c2a9cSMiao Xie 4542199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4543199c2a9cSMiao Xie { 4544199c2a9cSMiao Xie struct btrfs_root *root; 4545199c2a9cSMiao Xie struct list_head splice; 4546199c2a9cSMiao Xie 4547199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4548199c2a9cSMiao Xie 4549199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4550199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4551199c2a9cSMiao Xie while (!list_empty(&splice)) { 4552199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4553199c2a9cSMiao Xie ordered_root); 45541de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 45551de2cfdeSJosef Bacik &fs_info->ordered_roots); 4556199c2a9cSMiao Xie 45572a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4558199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4559199c2a9cSMiao Xie 45602a85d9caSLiu Bo cond_resched(); 45612a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4562199c2a9cSMiao Xie } 4563199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 456474d5d229SJosef Bacik 456574d5d229SJosef Bacik /* 456674d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 456774d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 456874d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 456974d5d229SJosef Bacik * actually get run and error out properly. 457074d5d229SJosef Bacik */ 457174d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4572acce952bSliubo } 4573acce952bSliubo 457435a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 45752ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4576acce952bSliubo { 4577acce952bSliubo struct rb_node *node; 4578acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4579acce952bSliubo struct btrfs_delayed_ref_node *ref; 4580acce952bSliubo int ret = 0; 4581acce952bSliubo 4582acce952bSliubo delayed_refs = &trans->delayed_refs; 4583acce952bSliubo 4584acce952bSliubo spin_lock(&delayed_refs->lock); 4585d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4586cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4587b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4588acce952bSliubo return ret; 4589acce952bSliubo } 4590acce952bSliubo 45915c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4592d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 45930e0adbcfSJosef Bacik struct rb_node *n; 4594e78417d1SJosef Bacik bool pin_bytes = false; 4595acce952bSliubo 4596d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4597d7df2c79SJosef Bacik href_node); 45983069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4599b939d1abSJosef Bacik continue; 46003069bd26SJosef Bacik 4601d7df2c79SJosef Bacik spin_lock(&head->lock); 4602e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 46030e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 46040e0adbcfSJosef Bacik ref_node); 4605d7df2c79SJosef Bacik ref->in_tree = 0; 4606e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 46070e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 46081d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 46091d57ee94SWang Xiaoguang list_del(&ref->add_list); 4610d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4611d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4612d7df2c79SJosef Bacik } 461354067ae9SJosef Bacik if (head->must_insert_reserved) 4614e78417d1SJosef Bacik pin_bytes = true; 461578a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4616fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4617d7df2c79SJosef Bacik spin_unlock(&head->lock); 4618acce952bSliubo spin_unlock(&delayed_refs->lock); 4619e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4620acce952bSliubo 4621f603bb94SNikolay Borisov if (pin_bytes) { 4622f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4623f603bb94SNikolay Borisov 4624f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4625f603bb94SNikolay Borisov BUG_ON(!cache); 4626f603bb94SNikolay Borisov 4627f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4628f603bb94SNikolay Borisov spin_lock(&cache->lock); 4629f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4630f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4631f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4632f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4633f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4634f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4635f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4636f603bb94SNikolay Borisov 4637f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4638f603bb94SNikolay Borisov 4639f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4640f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4641f603bb94SNikolay Borisov } 464231890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4643d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4644acce952bSliubo cond_resched(); 4645acce952bSliubo spin_lock(&delayed_refs->lock); 4646acce952bSliubo } 464781f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4648acce952bSliubo 4649acce952bSliubo spin_unlock(&delayed_refs->lock); 4650acce952bSliubo 4651acce952bSliubo return ret; 4652acce952bSliubo } 4653acce952bSliubo 4654143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4655acce952bSliubo { 4656acce952bSliubo struct btrfs_inode *btrfs_inode; 4657acce952bSliubo struct list_head splice; 4658acce952bSliubo 4659acce952bSliubo INIT_LIST_HEAD(&splice); 4660acce952bSliubo 4661eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4662eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4663acce952bSliubo 4664acce952bSliubo while (!list_empty(&splice)) { 4665fe816d0fSNikolay Borisov struct inode *inode = NULL; 4666eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4667acce952bSliubo delalloc_inodes); 4668fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4669eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4670acce952bSliubo 4671fe816d0fSNikolay Borisov /* 4672fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4673fe816d0fSNikolay Borisov * meanwhile. 4674fe816d0fSNikolay Borisov */ 4675fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4676fe816d0fSNikolay Borisov if (inode) { 4677597441b3SJosef Bacik unsigned int nofs_flag; 4678597441b3SJosef Bacik 4679597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4680fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4681597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4682fe816d0fSNikolay Borisov iput(inode); 4683fe816d0fSNikolay Borisov } 4684eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4685acce952bSliubo } 4686eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4687eb73c1b7SMiao Xie } 4688eb73c1b7SMiao Xie 4689eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4690eb73c1b7SMiao Xie { 4691eb73c1b7SMiao Xie struct btrfs_root *root; 4692eb73c1b7SMiao Xie struct list_head splice; 4693eb73c1b7SMiao Xie 4694eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4695eb73c1b7SMiao Xie 4696eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4697eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4698eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4699eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4700eb73c1b7SMiao Xie delalloc_root); 470100246528SJosef Bacik root = btrfs_grab_root(root); 4702eb73c1b7SMiao Xie BUG_ON(!root); 4703eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4704eb73c1b7SMiao Xie 4705eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 470600246528SJosef Bacik btrfs_put_root(root); 4707eb73c1b7SMiao Xie 4708eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4709eb73c1b7SMiao Xie } 4710eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4711acce952bSliubo } 4712acce952bSliubo 47132ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4714acce952bSliubo struct extent_io_tree *dirty_pages, 4715acce952bSliubo int mark) 4716acce952bSliubo { 4717acce952bSliubo int ret; 4718acce952bSliubo struct extent_buffer *eb; 4719acce952bSliubo u64 start = 0; 4720acce952bSliubo u64 end; 4721acce952bSliubo 4722acce952bSliubo while (1) { 4723acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4724e6138876SJosef Bacik mark, NULL); 4725acce952bSliubo if (ret) 4726acce952bSliubo break; 4727acce952bSliubo 472891166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4729acce952bSliubo while (start <= end) { 47300b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 47310b246afaSJeff Mahoney start += fs_info->nodesize; 4732fd8b2b61SJosef Bacik if (!eb) 4733acce952bSliubo continue; 4734acce952bSliubo 4735c4e54a65SJosef Bacik btrfs_tree_lock(eb); 4736c4e54a65SJosef Bacik wait_on_extent_buffer_writeback(eb); 4737190a8339SJosef Bacik btrfs_clear_buffer_dirty(NULL, eb); 4738c4e54a65SJosef Bacik btrfs_tree_unlock(eb); 4739c4e54a65SJosef Bacik 4740fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4741acce952bSliubo } 4742acce952bSliubo } 4743acce952bSliubo 4744acce952bSliubo return ret; 4745acce952bSliubo } 4746acce952bSliubo 47472ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4748fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4749acce952bSliubo { 4750acce952bSliubo u64 start; 4751acce952bSliubo u64 end; 4752acce952bSliubo int ret; 4753acce952bSliubo 4754acce952bSliubo while (1) { 47550e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 47560e6ec385SFilipe Manana 4757fcd5e742SLu Fengqi /* 4758fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4759fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4760fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4761fcd5e742SLu Fengqi * the same extent range. 4762fcd5e742SLu Fengqi */ 4763fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4764acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 47650e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4766fcd5e742SLu Fengqi if (ret) { 4767fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4768acce952bSliubo break; 4769fcd5e742SLu Fengqi } 4770acce952bSliubo 47710e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 47720e6ec385SFilipe Manana free_extent_state(cached_state); 47732ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4774fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4775acce952bSliubo cond_resched(); 4776acce952bSliubo } 4777acce952bSliubo 4778acce952bSliubo return 0; 4779acce952bSliubo } 4780acce952bSliubo 478132da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4782c79a1751SLiu Bo { 4783c79a1751SLiu Bo struct inode *inode; 4784c79a1751SLiu Bo 4785c79a1751SLiu Bo inode = cache->io_ctl.inode; 4786c79a1751SLiu Bo if (inode) { 4787597441b3SJosef Bacik unsigned int nofs_flag; 4788597441b3SJosef Bacik 4789597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4790c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4791597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4792597441b3SJosef Bacik 4793c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4794c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4795c79a1751SLiu Bo iput(inode); 4796c79a1751SLiu Bo } 4797bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4798c79a1751SLiu Bo btrfs_put_block_group(cache); 4799c79a1751SLiu Bo } 4800c79a1751SLiu Bo 4801c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 48022ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4803c79a1751SLiu Bo { 480432da5386SDavid Sterba struct btrfs_block_group *cache; 4805c79a1751SLiu Bo 4806c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4807c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4808c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 480932da5386SDavid Sterba struct btrfs_block_group, 4810c79a1751SLiu Bo dirty_list); 4811c79a1751SLiu Bo 4812c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4813c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4814c79a1751SLiu Bo list_del_init(&cache->io_list); 4815c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4816c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4817c79a1751SLiu Bo } 4818c79a1751SLiu Bo 4819c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4820c79a1751SLiu Bo spin_lock(&cache->lock); 4821c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4822c79a1751SLiu Bo spin_unlock(&cache->lock); 4823c79a1751SLiu Bo 4824c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4825c79a1751SLiu Bo btrfs_put_block_group(cache); 4826ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4827c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4828c79a1751SLiu Bo } 4829c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4830c79a1751SLiu Bo 483145ae2c18SNikolay Borisov /* 483245ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 483345ae2c18SNikolay Borisov * to use it without any locking 483445ae2c18SNikolay Borisov */ 4835c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4836c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 483732da5386SDavid Sterba struct btrfs_block_group, 4838c79a1751SLiu Bo io_list); 4839c79a1751SLiu Bo 4840c79a1751SLiu Bo list_del_init(&cache->io_list); 4841c79a1751SLiu Bo spin_lock(&cache->lock); 4842c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4843c79a1751SLiu Bo spin_unlock(&cache->lock); 4844c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4845c79a1751SLiu Bo } 4846c79a1751SLiu Bo } 4847c79a1751SLiu Bo 484849b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 48492ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 485049b25e05SJeff Mahoney { 4851bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4852bbbf7243SNikolay Borisov 48532ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4854c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4855c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4856c79a1751SLiu Bo 4857bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4858bbbf7243SNikolay Borisov post_commit_list) { 4859bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4860bbbf7243SNikolay Borisov } 4861bbbf7243SNikolay Borisov 48622ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 486349b25e05SJeff Mahoney 48644a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 48650b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 486649b25e05SJeff Mahoney 48674a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 48680b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 486949b25e05SJeff Mahoney 4870ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 487167cde344SMiao Xie 48722ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 487349b25e05SJeff Mahoney EXTENT_DIRTY); 4874fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 487549b25e05SJeff Mahoney 48764a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 48774a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 487849b25e05SJeff Mahoney } 487949b25e05SJeff Mahoney 48802ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4881acce952bSliubo { 4882acce952bSliubo struct btrfs_transaction *t; 4883acce952bSliubo 48840b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4885acce952bSliubo 48860b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 48870b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 48880b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4889724e2315SJosef Bacik struct btrfs_transaction, list); 4890724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 48919b64f57dSElena Reshetova refcount_inc(&t->use_count); 48920b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 48932ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4894724e2315SJosef Bacik btrfs_put_transaction(t); 48950b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4896724e2315SJosef Bacik continue; 4897724e2315SJosef Bacik } 48980b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4899724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 49000b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4901724e2315SJosef Bacik /* 4902724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4903724e2315SJosef Bacik * handle open currently for this transaction. 4904724e2315SJosef Bacik */ 4905724e2315SJosef Bacik wait_event(t->writer_wait, 4906724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4907724e2315SJosef Bacik } else { 49080b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4909724e2315SJosef Bacik } 49102ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4911724e2315SJosef Bacik 49120b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49130b246afaSJeff Mahoney if (t == fs_info->running_transaction) 49140b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4915724e2315SJosef Bacik list_del_init(&t->list); 49160b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4917a4abeea4SJosef Bacik 4918724e2315SJosef Bacik btrfs_put_transaction(t); 49192e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 49200b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4921724e2315SJosef Bacik } 49220b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49230b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4924ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4925ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 49260b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4927ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 49280b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4929acce952bSliubo 4930acce952bSliubo return 0; 4931acce952bSliubo } 4932ec7d6dfdSNikolay Borisov 4933453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 4934ec7d6dfdSNikolay Borisov { 4935ec7d6dfdSNikolay Borisov struct btrfs_path *path; 4936ec7d6dfdSNikolay Borisov int ret; 4937ec7d6dfdSNikolay Borisov struct extent_buffer *l; 4938ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 4939ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 4940ec7d6dfdSNikolay Borisov int slot; 4941ec7d6dfdSNikolay Borisov 4942ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 4943ec7d6dfdSNikolay Borisov if (!path) 4944ec7d6dfdSNikolay Borisov return -ENOMEM; 4945ec7d6dfdSNikolay Borisov 4946ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 4947ec7d6dfdSNikolay Borisov search_key.type = -1; 4948ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 4949ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 4950ec7d6dfdSNikolay Borisov if (ret < 0) 4951ec7d6dfdSNikolay Borisov goto error; 4952ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 4953ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 4954ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 4955ec7d6dfdSNikolay Borisov l = path->nodes[0]; 4956ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 495723125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 495823125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 4959ec7d6dfdSNikolay Borisov } else { 496023125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 4961ec7d6dfdSNikolay Borisov } 4962ec7d6dfdSNikolay Borisov ret = 0; 4963ec7d6dfdSNikolay Borisov error: 4964ec7d6dfdSNikolay Borisov btrfs_free_path(path); 4965ec7d6dfdSNikolay Borisov return ret; 4966ec7d6dfdSNikolay Borisov } 4967ec7d6dfdSNikolay Borisov 4968543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 4969ec7d6dfdSNikolay Borisov { 4970ec7d6dfdSNikolay Borisov int ret; 4971ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 4972ec7d6dfdSNikolay Borisov 49736b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 4974ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 4975ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 4976ec7d6dfdSNikolay Borisov root->root_key.objectid); 4977ec7d6dfdSNikolay Borisov ret = -ENOSPC; 4978ec7d6dfdSNikolay Borisov goto out; 4979ec7d6dfdSNikolay Borisov } 4980ec7d6dfdSNikolay Borisov 498123125104SNikolay Borisov *objectid = root->free_objectid++; 4982ec7d6dfdSNikolay Borisov ret = 0; 4983ec7d6dfdSNikolay Borisov out: 4984ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 4985ec7d6dfdSNikolay Borisov return ret; 4986ec7d6dfdSNikolay Borisov } 4987