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 320*413fb1bcSAnand 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)) 340*413fb1bcSAnand 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)) 344*413fb1bcSAnand Jain return false; 345944d3f9fSNikolay Borisov 346*413fb1bcSAnand 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); 1160bc44d7c4SJosef Bacik if (root) 116100246528SJosef Bacik root = btrfs_grab_root(root); 1162fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1163cb517eabSMiao Xie return root; 1164cb517eabSMiao Xie } 1165cb517eabSMiao Xie 116649d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 116749d11beaSJosef Bacik u64 objectid) 116849d11beaSJosef Bacik { 1169abed4aaaSJosef Bacik struct btrfs_key key = { 1170abed4aaaSJosef Bacik .objectid = objectid, 1171abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1172abed4aaaSJosef Bacik .offset = 0, 1173abed4aaaSJosef Bacik }; 1174abed4aaaSJosef Bacik 117549d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 117649d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 117749d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 1178abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 117949d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 118049d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 118149d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 118249d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 118349d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 1184abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 118549d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 118649d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 118749d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 118849d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 118949d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 119049d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 119114033b08SQu Wenruo if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) 119214033b08SQu Wenruo return btrfs_grab_root(fs_info->block_group_root) ? 119314033b08SQu Wenruo fs_info->block_group_root : ERR_PTR(-ENOENT); 1194abed4aaaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { 1195abed4aaaSJosef Bacik struct btrfs_root *root = btrfs_global_root(fs_info, &key); 1196abed4aaaSJosef Bacik 1197abed4aaaSJosef Bacik return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); 1198abed4aaaSJosef Bacik } 119949d11beaSJosef Bacik return NULL; 120049d11beaSJosef Bacik } 120149d11beaSJosef Bacik 1202cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1203cb517eabSMiao Xie struct btrfs_root *root) 1204cb517eabSMiao Xie { 1205cb517eabSMiao Xie int ret; 1206cb517eabSMiao Xie 1207fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1208fc7cbcd4SDavid Sterba if (ret) 1209fc7cbcd4SDavid Sterba return ret; 1210fc7cbcd4SDavid Sterba 1211fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1212fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1213fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1214fc7cbcd4SDavid Sterba root); 1215af01d2e5SJosef Bacik if (ret == 0) { 121600246528SJosef Bacik btrfs_grab_root(root); 1217fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1218af01d2e5SJosef Bacik } 1219fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1220fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1221cb517eabSMiao Xie 1222cb517eabSMiao Xie return ret; 1223cb517eabSMiao Xie } 1224cb517eabSMiao Xie 1225bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1226bd647ce3SJosef Bacik { 1227bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1228bd647ce3SJosef Bacik struct btrfs_root *root; 1229bd647ce3SJosef Bacik 1230bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1231457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1232457f1864SJosef Bacik 1233bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1234bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1235457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 123671008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1237bd647ce3SJosef Bacik refcount_read(&root->refs)); 1238bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 123900246528SJosef Bacik btrfs_put_root(root); 124000246528SJosef Bacik btrfs_put_root(root); 1241bd647ce3SJosef Bacik } 1242bd647ce3SJosef Bacik #endif 1243bd647ce3SJosef Bacik } 1244bd647ce3SJosef Bacik 1245abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1246abed4aaaSJosef Bacik { 1247abed4aaaSJosef Bacik struct btrfs_root *root; 1248abed4aaaSJosef Bacik struct rb_node *node; 1249abed4aaaSJosef Bacik 1250abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1251abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1252abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1253abed4aaaSJosef Bacik btrfs_put_root(root); 1254abed4aaaSJosef Bacik } 1255abed4aaaSJosef Bacik } 1256abed4aaaSJosef Bacik 12570d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 12580d4b0463SJosef Bacik { 1259141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1260141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 12615deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1262141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1263141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1264141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1265141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 12660d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 12670d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1268abed4aaaSJosef Bacik free_global_roots(fs_info); 126900246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 127000246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 127100246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 127200246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 127300246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 127400246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1275aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 12769c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1277bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 12783fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 12790d4b0463SJosef Bacik kfree(fs_info->super_copy); 12800d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 12818481dd80SQu Wenruo kfree(fs_info->subpage_info); 12820d4b0463SJosef Bacik kvfree(fs_info); 12830d4b0463SJosef Bacik } 12840d4b0463SJosef Bacik 12850d4b0463SJosef Bacik 12862dfb1e43SQu Wenruo /* 12872dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 12882dfb1e43SQu Wenruo * 12892dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 12902dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 12912dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 12922dfb1e43SQu Wenruo * 12932dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 12942dfb1e43SQu Wenruo * 12952dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 12962dfb1e43SQu Wenruo * same root objectid. 12972dfb1e43SQu Wenruo * 12982dfb1e43SQu Wenruo * @objectid: root id 12992dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 13002dfb1e43SQu Wenruo * pass 0 for new allocation. 13012dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 13022dfb1e43SQu Wenruo * for orphan roots 13032dfb1e43SQu Wenruo */ 13042dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 13052dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 13062dfb1e43SQu Wenruo bool check_ref) 13075eda7b5eSChris Mason { 13085eda7b5eSChris Mason struct btrfs_root *root; 1309381cf658SDavid Sterba struct btrfs_path *path; 13101d4c08e0SDavid Sterba struct btrfs_key key; 13115eda7b5eSChris Mason int ret; 13125eda7b5eSChris Mason 131349d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 131449d11beaSJosef Bacik if (root) 131549d11beaSJosef Bacik return root; 13164df27c4dSYan, Zheng again: 131756e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 131848475471SStefan Behrens if (root) { 13192dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 13202dfb1e43SQu Wenruo ASSERT(!anon_dev); 1321bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 132200246528SJosef Bacik btrfs_put_root(root); 132348475471SStefan Behrens return ERR_PTR(-ENOENT); 1324bc44d7c4SJosef Bacik } 13255eda7b5eSChris Mason return root; 132648475471SStefan Behrens } 13275eda7b5eSChris Mason 132856e9357aSDavid Sterba key.objectid = objectid; 132956e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 133056e9357aSDavid Sterba key.offset = (u64)-1; 133156e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 13325eda7b5eSChris Mason if (IS_ERR(root)) 13335eda7b5eSChris Mason return root; 13343394e160SChris Mason 1335c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1336d68fc57bSYan, Zheng ret = -ENOENT; 1337d68fc57bSYan, Zheng goto fail; 1338d68fc57bSYan, Zheng } 1339d68fc57bSYan, Zheng 13402dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1341cb517eabSMiao Xie if (ret) 1342cb517eabSMiao Xie goto fail; 1343cb517eabSMiao Xie 1344381cf658SDavid Sterba path = btrfs_alloc_path(); 1345381cf658SDavid Sterba if (!path) { 1346381cf658SDavid Sterba ret = -ENOMEM; 1347381cf658SDavid Sterba goto fail; 1348381cf658SDavid Sterba } 13491d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 13501d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 135156e9357aSDavid Sterba key.offset = objectid; 13521d4c08e0SDavid Sterba 13531d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1354381cf658SDavid Sterba btrfs_free_path(path); 1355d68fc57bSYan, Zheng if (ret < 0) 1356d68fc57bSYan, Zheng goto fail; 1357d68fc57bSYan, Zheng if (ret == 0) 135827cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1359d68fc57bSYan, Zheng 1360cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 13610f7d52f4SChris Mason if (ret) { 1362168a2f77SJia-Ju Bai if (ret == -EEXIST) { 136300246528SJosef Bacik btrfs_put_root(root); 13644df27c4dSYan, Zheng goto again; 1365168a2f77SJia-Ju Bai } 13664df27c4dSYan, Zheng goto fail; 13674df27c4dSYan, Zheng } 1368edbd8d4eSChris Mason return root; 13694df27c4dSYan, Zheng fail: 137033fab972SFilipe Manana /* 137133fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1372143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 137333fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 137433fab972SFilipe Manana * and once again by our caller. 137533fab972SFilipe Manana */ 137633fab972SFilipe Manana if (anon_dev) 137733fab972SFilipe Manana root->anon_dev = 0; 13788c38938cSJosef Bacik btrfs_put_root(root); 13794df27c4dSYan, Zheng return ERR_PTR(ret); 1380edbd8d4eSChris Mason } 1381edbd8d4eSChris Mason 13822dfb1e43SQu Wenruo /* 13832dfb1e43SQu Wenruo * Get in-memory reference of a root structure 13842dfb1e43SQu Wenruo * 13852dfb1e43SQu Wenruo * @objectid: tree objectid 13862dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 13872dfb1e43SQu Wenruo * one reference 13882dfb1e43SQu Wenruo */ 13892dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 13902dfb1e43SQu Wenruo u64 objectid, bool check_ref) 13912dfb1e43SQu Wenruo { 13922dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 13932dfb1e43SQu Wenruo } 13942dfb1e43SQu Wenruo 13952dfb1e43SQu Wenruo /* 13962dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 13972dfb1e43SQu Wenruo * the anonymous block device id 13982dfb1e43SQu Wenruo * 13992dfb1e43SQu Wenruo * @objectid: tree objectid 14002dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 14012dfb1e43SQu Wenruo * parameter value 14022dfb1e43SQu Wenruo */ 14032dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 14042dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 14052dfb1e43SQu Wenruo { 14062dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 14072dfb1e43SQu Wenruo } 14082dfb1e43SQu Wenruo 14098b712842SChris Mason /* 141049d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 141149d11beaSJosef Bacik * @fs_info: the fs_info 141249d11beaSJosef Bacik * @objectid: the objectid we need to lookup 141349d11beaSJosef Bacik * 141449d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 141549d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 141649d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 141749d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 141849d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 141949d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 142049d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 142149d11beaSJosef Bacik * backref code is finished using the root. 142249d11beaSJosef Bacik */ 142349d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 142449d11beaSJosef Bacik struct btrfs_path *path, 142549d11beaSJosef Bacik u64 objectid) 142649d11beaSJosef Bacik { 142749d11beaSJosef Bacik struct btrfs_root *root; 142849d11beaSJosef Bacik struct btrfs_key key; 142949d11beaSJosef Bacik 143049d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 143149d11beaSJosef Bacik 143249d11beaSJosef Bacik /* 143349d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 143449d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 143549d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 143649d11beaSJosef Bacik * != NULL just return it. 143749d11beaSJosef Bacik */ 143849d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 143949d11beaSJosef Bacik if (root) 144049d11beaSJosef Bacik return root; 144149d11beaSJosef Bacik 144249d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 144349d11beaSJosef Bacik if (root) 144449d11beaSJosef Bacik return root; 144549d11beaSJosef Bacik 144649d11beaSJosef Bacik key.objectid = objectid; 144749d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 144849d11beaSJosef Bacik key.offset = (u64)-1; 144949d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 145049d11beaSJosef Bacik btrfs_release_path(path); 145149d11beaSJosef Bacik 145249d11beaSJosef Bacik return root; 145349d11beaSJosef Bacik } 145449d11beaSJosef Bacik 1455a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1456a74a4b97SChris Mason { 14570d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1458d0278245SMiao Xie int again; 1459a74a4b97SChris Mason 1460d6fd0ae2SOmar Sandoval while (1) { 1461d0278245SMiao Xie again = 0; 14629d1a2a3aSDavid Sterba 1463fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1464fd340d0fSJosef Bacik 1465d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 14662ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1467d0278245SMiao Xie goto sleep; 1468d0278245SMiao Xie 146990c711abSZygo Blaxell /* 147090c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 147190c711abSZygo Blaxell * before we have finished mounting the filesystem. 147290c711abSZygo Blaxell */ 14730b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 147490c711abSZygo Blaxell goto sleep; 147590c711abSZygo Blaxell 14760b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1477d0278245SMiao Xie goto sleep; 1478d0278245SMiao Xie 1479dc7f370cSMiao Xie /* 1480dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1481dc7f370cSMiao Xie * during the above check and trylock. 1482dc7f370cSMiao Xie */ 14832ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 14840b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1485dc7f370cSMiao Xie goto sleep; 1486dc7f370cSMiao Xie } 1487dc7f370cSMiao Xie 1488b7625f46SQu Wenruo if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) 1489b7625f46SQu Wenruo btrfs_sysfs_feature_update(fs_info); 1490b7625f46SQu Wenruo 14912ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1492c2d6cb16SFilipe Manana 149333c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 14940b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1495a74a4b97SChris Mason 1496d0278245SMiao Xie /* 149705323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 149805323cd1SMiao Xie * needn't do anything special here. 1499d0278245SMiao Xie */ 15000b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 150167c5e7d4SFilipe Manana 150267c5e7d4SFilipe Manana /* 1503f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 150467c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 150567c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1506f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 150767c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 150867c5e7d4SFilipe Manana * unused block groups. 150967c5e7d4SFilipe Manana */ 15100b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 151118bb8bbfSJohannes Thumshirn 151218bb8bbfSJohannes Thumshirn /* 151318bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 151418bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 151518bb8bbfSJohannes Thumshirn * space. 151618bb8bbfSJohannes Thumshirn */ 151718bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1518d0278245SMiao Xie sleep: 1519a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1520d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1521d6fd0ae2SOmar Sandoval kthread_parkme(); 1522d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1523d6fd0ae2SOmar Sandoval return 0; 1524838fe188SJiri Kosina if (!again) { 1525a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1526a74a4b97SChris Mason schedule(); 1527a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1528a74a4b97SChris Mason } 1529da288d28SFilipe Manana } 1530a74a4b97SChris Mason } 1531a74a4b97SChris Mason 1532a74a4b97SChris Mason static int transaction_kthread(void *arg) 1533a74a4b97SChris Mason { 1534a74a4b97SChris Mason struct btrfs_root *root = arg; 15350b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1536a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1537a74a4b97SChris Mason struct btrfs_transaction *cur; 15388929ecfaSYan, Zheng u64 transid; 1539643900beSNikolay Borisov time64_t delta; 1540a74a4b97SChris Mason unsigned long delay; 1541914b2007SJan Kara bool cannot_commit; 1542a74a4b97SChris Mason 1543a74a4b97SChris Mason do { 1544914b2007SJan Kara cannot_commit = false; 1545ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 15460b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1547a74a4b97SChris Mason 15480b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 15490b246afaSJeff Mahoney cur = fs_info->running_transaction; 1550a74a4b97SChris Mason if (!cur) { 15510b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1552a74a4b97SChris Mason goto sleep; 1553a74a4b97SChris Mason } 155431153d81SYan Zheng 1555643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1556fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1557fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1558643900beSNikolay Borisov delta < fs_info->commit_interval) { 15590b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1560fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1561fb8a7e94SNikolay Borisov delay = min(delay, 1562fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1563a74a4b97SChris Mason goto sleep; 1564a74a4b97SChris Mason } 15658929ecfaSYan, Zheng transid = cur->transid; 15660b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 156756bec294SChris Mason 156879787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1569354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1570914b2007SJan Kara if (IS_ERR(trans)) { 1571354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1572914b2007SJan Kara cannot_commit = true; 157379787eaaSJeff Mahoney goto sleep; 1574914b2007SJan Kara } 15758929ecfaSYan, Zheng if (transid == trans->transid) { 15763a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 15778929ecfaSYan, Zheng } else { 15783a45bb20SJeff Mahoney btrfs_end_transaction(trans); 15798929ecfaSYan, Zheng } 1580a74a4b97SChris Mason sleep: 15810b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 15820b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1583a74a4b97SChris Mason 158484961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 15852ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 15868929ecfaSYan, Zheng if (!kthread_should_stop() && 15870b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1588914b2007SJan Kara cannot_commit)) 1589bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1590a74a4b97SChris Mason } while (!kthread_should_stop()); 1591a74a4b97SChris Mason return 0; 1592a74a4b97SChris Mason } 1593a74a4b97SChris Mason 1594af31f5e5SChris Mason /* 159501f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 159601f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 159701f0f9daSNikolay Borisov * anything. 1598af31f5e5SChris Mason * 1599af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1600af31f5e5SChris Mason * generation of the latest root in the array with the generation 1601af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1602af31f5e5SChris Mason */ 160301f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1604af31f5e5SChris Mason { 160501f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1606af31f5e5SChris Mason u64 cur; 1607af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1608af31f5e5SChris Mason int i; 1609af31f5e5SChris Mason 1610af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1611af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1612af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1613af31f5e5SChris Mason if (cur == newest_gen) 161401f0f9daSNikolay Borisov return i; 1615af31f5e5SChris Mason } 1616af31f5e5SChris Mason 161701f0f9daSNikolay Borisov return -EINVAL; 1618af31f5e5SChris Mason } 1619af31f5e5SChris Mason 1620af31f5e5SChris Mason /* 1621af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1622af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1623af31f5e5SChris Mason * done 1624af31f5e5SChris Mason */ 1625af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1626af31f5e5SChris Mason { 16276ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1628af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1629af31f5e5SChris Mason 1630af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1631af31f5e5SChris Mason 1632af31f5e5SChris Mason /* 1633af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1634af31f5e5SChris Mason * regardless of which ones we use today 1635af31f5e5SChris Mason */ 1636af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1637af31f5e5SChris Mason 1638af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1639af31f5e5SChris Mason 1640af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1641af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1642af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1643af31f5e5SChris Mason 1644af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1645af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1646af31f5e5SChris Mason 1647af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1648af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1649af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1650af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1651af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1652af31f5e5SChris Mason 16531c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 16549c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1655f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 16569c54e80dSJosef Bacik 16579c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 16589c54e80dSJosef Bacik extent_root->node->start); 1659af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 166029cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 1661af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 166229cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 1663af31f5e5SChris Mason 1664f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 1665f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 1666f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 1667f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 1668f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 16699c54e80dSJosef Bacik } 1670af31f5e5SChris Mason 16717c7e82a7SChris Mason /* 16727c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 16737c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 16747c7e82a7SChris Mason */ 16757c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 16767c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 16777c7e82a7SChris Mason info->fs_root->node->start); 1678af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1679af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1680af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1681af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 16827c7e82a7SChris Mason } 1683af31f5e5SChris Mason 1684af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1685af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1686af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1687af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1688af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1689af31f5e5SChris Mason 1690af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1691af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1692af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1693af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1694af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1695af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1696af31f5e5SChris Mason 1697af31f5e5SChris Mason /* 1698af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1699af31f5e5SChris Mason * for the next commit 1700af31f5e5SChris Mason */ 1701af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1702af31f5e5SChris Mason &info->super_for_commit->super_roots, 1703af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1704af31f5e5SChris Mason } 1705af31f5e5SChris Mason 1706af31f5e5SChris Mason /* 1707bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1708bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1709bd2336b2SNikolay Borisov * 1710bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1711bd2336b2SNikolay Borisov * priority - priority of backup root required 1712bd2336b2SNikolay Borisov * 1713bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1714bd2336b2SNikolay Borisov */ 1715bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1716bd2336b2SNikolay Borisov { 1717bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1718bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1719bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1720bd2336b2SNikolay Borisov 1721bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1722bd2336b2SNikolay Borisov if (priority == 0) 1723bd2336b2SNikolay Borisov return backup_index; 1724bd2336b2SNikolay Borisov 1725bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1726bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1727bd2336b2SNikolay Borisov } else { 1728bd2336b2SNikolay Borisov return -EINVAL; 1729bd2336b2SNikolay Borisov } 1730bd2336b2SNikolay Borisov 1731bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1732bd2336b2SNikolay Borisov 1733bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1734bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1735bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1736bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1737bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1738bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1739bd2336b2SNikolay Borisov 1740bd2336b2SNikolay Borisov /* 1741bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1742bd2336b2SNikolay Borisov * need a fsck 1743bd2336b2SNikolay Borisov */ 1744bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1745bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1746bd2336b2SNikolay Borisov 1747bd2336b2SNikolay Borisov return backup_index; 1748bd2336b2SNikolay Borisov } 1749bd2336b2SNikolay Borisov 17507abadb64SLiu Bo /* helper to cleanup workers */ 17517abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 17527abadb64SLiu Bo { 1753dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1754afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 17555cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1756d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 1757d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 1758385de0efSChristoph Hellwig if (fs_info->rmw_workers) 1759385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 1760fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 1761fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 1762fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1763fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 17645b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1765e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1766a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1767fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1768b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1769b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1770a9b9477dSFilipe Manana /* 1771a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 1772a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 1773a9b9477dSFilipe Manana * queues can do metadata I/O operations. 1774a9b9477dSFilipe Manana */ 1775d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 1776d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 17777abadb64SLiu Bo } 17787abadb64SLiu Bo 17792e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 17802e9f5954SRashika { 17812e9f5954SRashika if (root) { 17822e9f5954SRashika free_extent_buffer(root->node); 17832e9f5954SRashika free_extent_buffer(root->commit_root); 17842e9f5954SRashika root->node = NULL; 17852e9f5954SRashika root->commit_root = NULL; 17862e9f5954SRashika } 17872e9f5954SRashika } 17882e9f5954SRashika 1789abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 1790abed4aaaSJosef Bacik { 1791abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 1792abed4aaaSJosef Bacik 1793abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 1794abed4aaaSJosef Bacik &fs_info->global_root_tree, 1795abed4aaaSJosef Bacik rb_node) 1796abed4aaaSJosef Bacik free_root_extent_buffers(root); 1797abed4aaaSJosef Bacik } 1798abed4aaaSJosef Bacik 1799af31f5e5SChris Mason /* helper to cleanup tree roots */ 18004273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 1801af31f5e5SChris Mason { 18022e9f5954SRashika free_root_extent_buffers(info->tree_root); 1803655b09feSJosef Bacik 1804abed4aaaSJosef Bacik free_global_root_pointers(info); 18052e9f5954SRashika free_root_extent_buffers(info->dev_root); 18062e9f5954SRashika free_root_extent_buffers(info->quota_root); 18072e9f5954SRashika free_root_extent_buffers(info->uuid_root); 18088c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 1809aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 18109c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 18114273eaffSAnand Jain if (free_chunk_root) 18122e9f5954SRashika free_root_extent_buffers(info->chunk_root); 1813af31f5e5SChris Mason } 1814af31f5e5SChris Mason 18158c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 18168c38938cSJosef Bacik { 18178c38938cSJosef Bacik if (!root) 18188c38938cSJosef Bacik return; 18198c38938cSJosef Bacik 18208c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 18218c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 18221dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 18238c38938cSJosef Bacik if (root->anon_dev) 18248c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 1825923eb523SJohannes Thumshirn free_root_extent_buffers(root); 18268c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1827fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 18288c38938cSJosef Bacik list_del_init(&root->leak_list); 1829fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 18308c38938cSJosef Bacik #endif 18318c38938cSJosef Bacik kfree(root); 18328c38938cSJosef Bacik } 18338c38938cSJosef Bacik } 18348c38938cSJosef Bacik 1835faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 1836171f6537SJosef Bacik { 1837fc7cbcd4SDavid Sterba int ret; 1838fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 1839fc7cbcd4SDavid Sterba int i; 1840171f6537SJosef Bacik 1841171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 1842fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 1843171f6537SJosef Bacik struct btrfs_root, root_list); 1844fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 1845171f6537SJosef Bacik 1846fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 1847fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 1848fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 1849171f6537SJosef Bacik } 1850171f6537SJosef Bacik 1851fc7cbcd4SDavid Sterba while (1) { 1852fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 1853fc7cbcd4SDavid Sterba (void **)gang, 0, 1854fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 1855fc7cbcd4SDavid Sterba if (!ret) 1856fc7cbcd4SDavid Sterba break; 1857fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 1858fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 1859171f6537SJosef Bacik } 1860171f6537SJosef Bacik } 1861af31f5e5SChris Mason 1862638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 1863638aa7edSEric Sandeen { 1864638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 1865638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 1866638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 1867638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 1868638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 1869638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 1870ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 1871638aa7edSEric Sandeen } 1872638aa7edSEric Sandeen 1873779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 1874779a65a4SEric Sandeen { 1875779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 1876779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 1877779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 1878779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 1879779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 1880779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 1881907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 1882779a65a4SEric Sandeen } 1883779a65a4SEric Sandeen 1884dcb2137cSChristoph Hellwig static int btrfs_init_btree_inode(struct super_block *sb) 1885f37938e0SEric Sandeen { 1886dcb2137cSChristoph Hellwig struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1887e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 1888e256927bSJosef Bacik fs_info->tree_root); 1889dcb2137cSChristoph Hellwig struct inode *inode; 1890dcb2137cSChristoph Hellwig 1891dcb2137cSChristoph Hellwig inode = new_inode(sb); 1892dcb2137cSChristoph Hellwig if (!inode) 1893dcb2137cSChristoph Hellwig return -ENOMEM; 18942ff7e61eSJeff Mahoney 18952ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 18962ff7e61eSJeff Mahoney set_nlink(inode, 1); 1897f37938e0SEric Sandeen /* 1898f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 1899f37938e0SEric Sandeen * the real end of the address space is determined by all of 1900f37938e0SEric Sandeen * the devices in the system 1901f37938e0SEric Sandeen */ 19022ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 19032ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 1904dcb2137cSChristoph Hellwig mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); 1905f37938e0SEric Sandeen 19062ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 190743eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 190835da5a7eSDavid Sterba IO_TREE_BTREE_INODE_IO); 19092ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 1910f37938e0SEric Sandeen 19115c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 1912adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 1913adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 1914adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 19152ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 1916e256927bSJosef Bacik __insert_inode_hash(inode, hash); 1917dcb2137cSChristoph Hellwig fs_info->btree_inode = inode; 1918dcb2137cSChristoph Hellwig 1919dcb2137cSChristoph Hellwig return 0; 1920f37938e0SEric Sandeen } 1921f37938e0SEric Sandeen 1922ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 1923ad618368SEric Sandeen { 1924ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 1925129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 19267f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 1927ad618368SEric Sandeen } 1928ad618368SEric Sandeen 1929f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 1930f9e92e40SEric Sandeen { 1931f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 1932f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 1933f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 1934f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 1935f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 1936f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 1937d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 1938011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 1939f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 1940f9e92e40SEric Sandeen } 1941f9e92e40SEric Sandeen 1942d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 19432a458198SEric Sandeen { 1944f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 19456f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 19462a458198SEric Sandeen 19472a458198SEric Sandeen fs_info->workers = 1948a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 19492a458198SEric Sandeen 19502a458198SEric Sandeen fs_info->delalloc_workers = 1951cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 1952cb001095SJeff Mahoney flags, max_active, 2); 19532a458198SEric Sandeen 19542a458198SEric Sandeen fs_info->flush_workers = 1955cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 1956cb001095SJeff Mahoney flags, max_active, 0); 19572a458198SEric Sandeen 19582a458198SEric Sandeen fs_info->caching_workers = 1959cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 19602a458198SEric Sandeen 19612a458198SEric Sandeen fs_info->fixup_workers = 1962cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 19632a458198SEric Sandeen 19642a458198SEric Sandeen fs_info->endio_workers = 1965d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 19662a458198SEric Sandeen fs_info->endio_meta_workers = 1967d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 1968385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 19692a458198SEric Sandeen fs_info->endio_write_workers = 1970cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 1971cb001095SJeff Mahoney max_active, 2); 1972fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 1973fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 19742a458198SEric Sandeen fs_info->endio_freespace_worker = 1975cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 1976cb001095SJeff Mahoney max_active, 0); 19772a458198SEric Sandeen fs_info->delayed_workers = 1978cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 1979cb001095SJeff Mahoney max_active, 0); 19802a458198SEric Sandeen fs_info->qgroup_rescan_workers = 1981cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 1982b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 1983b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 19842a458198SEric Sandeen 19858bfec2e4SChristoph Hellwig if (!(fs_info->workers && 1986a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 19872a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 1988fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 19891a1a2851SQu Wenruo fs_info->endio_write_workers && 19902a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 1991f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 1992f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 1993b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 19942a458198SEric Sandeen return -ENOMEM; 19952a458198SEric Sandeen } 19962a458198SEric Sandeen 19972a458198SEric Sandeen return 0; 19982a458198SEric Sandeen } 19992a458198SEric Sandeen 20006d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 20016d97c6e3SJohannes Thumshirn { 20026d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2003b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 20046d97c6e3SJohannes Thumshirn 2005b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 20066d97c6e3SJohannes Thumshirn 20076d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 20086d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2009b4e967beSDavid Sterba csum_driver); 20106d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 20116d97c6e3SJohannes Thumshirn } 20126d97c6e3SJohannes Thumshirn 20136d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 20146d97c6e3SJohannes Thumshirn 201568d99ab0SChristoph Hellwig /* 201668d99ab0SChristoph Hellwig * Check if the checksum implementation is a fast accelerated one. 201768d99ab0SChristoph Hellwig * As-is this is a bit of a hack and should be replaced once the csum 201868d99ab0SChristoph Hellwig * implementations provide that information themselves. 201968d99ab0SChristoph Hellwig */ 202068d99ab0SChristoph Hellwig switch (csum_type) { 202168d99ab0SChristoph Hellwig case BTRFS_CSUM_TYPE_CRC32: 202268d99ab0SChristoph Hellwig if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) 202368d99ab0SChristoph Hellwig set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 202468d99ab0SChristoph Hellwig break; 202568d99ab0SChristoph Hellwig default: 202668d99ab0SChristoph Hellwig break; 202768d99ab0SChristoph Hellwig } 202868d99ab0SChristoph Hellwig 2029c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2030c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2031c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 20326d97c6e3SJohannes Thumshirn return 0; 20336d97c6e3SJohannes Thumshirn } 20346d97c6e3SJohannes Thumshirn 203563443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 203663443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 203763443bf5SEric Sandeen { 203863443bf5SEric Sandeen int ret; 2039789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 204063443bf5SEric Sandeen struct btrfs_root *log_tree_root; 204163443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 204263443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2043581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 204463443bf5SEric Sandeen 204563443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2046f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 204763443bf5SEric Sandeen return -EIO; 204863443bf5SEric Sandeen } 204963443bf5SEric Sandeen 205096dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 205196dfcb46SJosef Bacik GFP_KERNEL); 205263443bf5SEric Sandeen if (!log_tree_root) 205363443bf5SEric Sandeen return -ENOMEM; 205463443bf5SEric Sandeen 2055789d6a3aSQu Wenruo check.level = level; 2056789d6a3aSQu Wenruo check.transid = fs_info->generation + 1; 2057789d6a3aSQu Wenruo check.owner_root = BTRFS_TREE_LOG_OBJECTID; 2058789d6a3aSQu Wenruo log_tree_root->node = read_tree_block(fs_info, bytenr, &check); 205964c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2060f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 20610eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 20628c38938cSJosef Bacik log_tree_root->node = NULL; 206300246528SJosef Bacik btrfs_put_root(log_tree_root); 20640eeff236SLiu Bo return ret; 20654eb150d6SQu Wenruo } 20664eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2067f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 206800246528SJosef Bacik btrfs_put_root(log_tree_root); 206963443bf5SEric Sandeen return -EIO; 207063443bf5SEric Sandeen } 20714eb150d6SQu Wenruo 207263443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 207363443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 207463443bf5SEric Sandeen if (ret) { 20750b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 207663443bf5SEric Sandeen "Failed to recover log tree"); 207700246528SJosef Bacik btrfs_put_root(log_tree_root); 207863443bf5SEric Sandeen return ret; 207963443bf5SEric Sandeen } 208063443bf5SEric Sandeen 2081bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 20826bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 208363443bf5SEric Sandeen if (ret) 208463443bf5SEric Sandeen return ret; 208563443bf5SEric Sandeen } 208663443bf5SEric Sandeen 208763443bf5SEric Sandeen return 0; 208863443bf5SEric Sandeen } 208963443bf5SEric Sandeen 2090abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2091abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2092abed4aaaSJosef Bacik const char *name) 2093abed4aaaSJosef Bacik { 2094abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2095abed4aaaSJosef Bacik struct btrfs_root *root; 2096f7238e50SJosef Bacik u64 max_global_id = 0; 2097abed4aaaSJosef Bacik int ret; 2098abed4aaaSJosef Bacik struct btrfs_key key = { 2099abed4aaaSJosef Bacik .objectid = objectid, 2100abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2101abed4aaaSJosef Bacik .offset = 0, 2102abed4aaaSJosef Bacik }; 2103abed4aaaSJosef Bacik bool found = false; 2104abed4aaaSJosef Bacik 2105abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2106abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2107abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2108abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2109abed4aaaSJosef Bacik return 0; 2110abed4aaaSJosef Bacik } 2111abed4aaaSJosef Bacik 2112abed4aaaSJosef Bacik while (1) { 2113abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2114abed4aaaSJosef Bacik if (ret < 0) 2115abed4aaaSJosef Bacik break; 2116abed4aaaSJosef Bacik 2117abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2118abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2119abed4aaaSJosef Bacik if (ret) { 2120abed4aaaSJosef Bacik if (ret > 0) 2121abed4aaaSJosef Bacik ret = 0; 2122abed4aaaSJosef Bacik break; 2123abed4aaaSJosef Bacik } 2124abed4aaaSJosef Bacik } 2125abed4aaaSJosef Bacik ret = 0; 2126abed4aaaSJosef Bacik 2127abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2128abed4aaaSJosef Bacik if (key.objectid != objectid) 2129abed4aaaSJosef Bacik break; 2130abed4aaaSJosef Bacik btrfs_release_path(path); 2131abed4aaaSJosef Bacik 2132f7238e50SJosef Bacik /* 2133f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2134f7238e50SJosef Bacik * everybody. 2135f7238e50SJosef Bacik */ 2136f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2137f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2138f7238e50SJosef Bacik 2139abed4aaaSJosef Bacik found = true; 2140abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2141abed4aaaSJosef Bacik if (IS_ERR(root)) { 2142abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2143abed4aaaSJosef Bacik ret = PTR_ERR(root); 2144abed4aaaSJosef Bacik break; 2145abed4aaaSJosef Bacik } 2146abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2147abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2148abed4aaaSJosef Bacik if (ret) { 2149abed4aaaSJosef Bacik btrfs_put_root(root); 2150abed4aaaSJosef Bacik break; 2151abed4aaaSJosef Bacik } 2152abed4aaaSJosef Bacik key.offset++; 2153abed4aaaSJosef Bacik } 2154abed4aaaSJosef Bacik btrfs_release_path(path); 2155abed4aaaSJosef Bacik 2156f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2157f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2158f7238e50SJosef Bacik 2159abed4aaaSJosef Bacik if (!found || ret) { 2160abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2161abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2162abed4aaaSJosef Bacik 2163abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2164abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2165abed4aaaSJosef Bacik else 2166abed4aaaSJosef Bacik ret = 0; 2167abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2168abed4aaaSJosef Bacik } 2169abed4aaaSJosef Bacik return ret; 2170abed4aaaSJosef Bacik } 2171abed4aaaSJosef Bacik 2172abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2173abed4aaaSJosef Bacik { 2174abed4aaaSJosef Bacik struct btrfs_path *path; 2175abed4aaaSJosef Bacik int ret = 0; 2176abed4aaaSJosef Bacik 2177abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2178abed4aaaSJosef Bacik if (!path) 2179abed4aaaSJosef Bacik return -ENOMEM; 2180abed4aaaSJosef Bacik 2181abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2182abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2183abed4aaaSJosef Bacik if (ret) 2184abed4aaaSJosef Bacik goto out; 2185abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2186abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2187abed4aaaSJosef Bacik if (ret) 2188abed4aaaSJosef Bacik goto out; 2189abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2190abed4aaaSJosef Bacik goto out; 2191abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2192abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2193abed4aaaSJosef Bacik "free space"); 2194abed4aaaSJosef Bacik out: 2195abed4aaaSJosef Bacik btrfs_free_path(path); 2196abed4aaaSJosef Bacik return ret; 2197abed4aaaSJosef Bacik } 2198abed4aaaSJosef Bacik 21996bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 22004bbcaa64SEric Sandeen { 22016bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2202a4f3d2c4SDavid Sterba struct btrfs_root *root; 22034bbcaa64SEric Sandeen struct btrfs_key location; 22044bbcaa64SEric Sandeen int ret; 22054bbcaa64SEric Sandeen 22066bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 22076bccf3abSJeff Mahoney 2208abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2209abed4aaaSJosef Bacik if (ret) 2210abed4aaaSJosef Bacik return ret; 2211abed4aaaSJosef Bacik 22124bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 22134bbcaa64SEric Sandeen location.offset = 0; 22144bbcaa64SEric Sandeen 22151c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 221614033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 221714033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 221814033b08SQu Wenruo if (IS_ERR(root)) { 221914033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 222014033b08SQu Wenruo ret = PTR_ERR(root); 222114033b08SQu Wenruo goto out; 222214033b08SQu Wenruo } 222314033b08SQu Wenruo } else { 222414033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 222514033b08SQu Wenruo fs_info->block_group_root = root; 222614033b08SQu Wenruo } 222714033b08SQu Wenruo } 222814033b08SQu Wenruo 222914033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2230a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2231f50f4353SLiu Bo if (IS_ERR(root)) { 223242437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2233f50f4353SLiu Bo ret = PTR_ERR(root); 2234f50f4353SLiu Bo goto out; 2235f50f4353SLiu Bo } 223642437a63SJosef Bacik } else { 2237a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2238a4f3d2c4SDavid Sterba fs_info->dev_root = root; 223942437a63SJosef Bacik } 2240820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2241a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2242a8d1b164SJohannes Thumshirn if (ret) 2243a8d1b164SJohannes Thumshirn goto out; 22444bbcaa64SEric Sandeen 2245aeb935a4SQu Wenruo /* 2246aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2247aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2248aeb935a4SQu Wenruo */ 224956e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 225056e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2251aeb935a4SQu Wenruo if (IS_ERR(root)) { 225242437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2253aeb935a4SQu Wenruo ret = PTR_ERR(root); 2254aeb935a4SQu Wenruo goto out; 2255aeb935a4SQu Wenruo } 225642437a63SJosef Bacik } else { 2257aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2258aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 225942437a63SJosef Bacik } 2260aeb935a4SQu Wenruo 22614bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2262a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2263a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2264a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2265afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2266a4f3d2c4SDavid Sterba fs_info->quota_root = root; 22674bbcaa64SEric Sandeen } 22684bbcaa64SEric Sandeen 22694bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2270a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2271a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 227242437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2273a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 22744bbcaa64SEric Sandeen if (ret != -ENOENT) 2275f50f4353SLiu Bo goto out; 227642437a63SJosef Bacik } 22774bbcaa64SEric Sandeen } else { 2278a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2279a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 22804bbcaa64SEric Sandeen } 22814bbcaa64SEric Sandeen 22824bbcaa64SEric Sandeen return 0; 2283f50f4353SLiu Bo out: 2284f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2285f50f4353SLiu Bo location.objectid, ret); 2286f50f4353SLiu Bo return ret; 22874bbcaa64SEric Sandeen } 22884bbcaa64SEric Sandeen 2289069ec957SQu Wenruo /* 2290069ec957SQu Wenruo * Real super block validation 2291069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2292069ec957SQu Wenruo * 2293069ec957SQu Wenruo * @sb: super block to check 2294069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2295069ec957SQu Wenruo * 0 the primary (1st) sb 2296069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2297069ec957SQu Wenruo * -1 skip bytenr check 2298069ec957SQu Wenruo */ 2299a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2300069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 230121a852b0SQu Wenruo { 230221a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 230321a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 230421a852b0SQu Wenruo int ret = 0; 230521a852b0SQu Wenruo 230621a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 230721a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 230821a852b0SQu Wenruo ret = -EINVAL; 230921a852b0SQu Wenruo } 231021a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 231121a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 231221a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 231321a852b0SQu Wenruo ret = -EINVAL; 231421a852b0SQu Wenruo } 231521a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 231621a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 231721a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 231821a852b0SQu Wenruo ret = -EINVAL; 231921a852b0SQu Wenruo } 232021a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 232121a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 232221a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 232321a852b0SQu Wenruo ret = -EINVAL; 232421a852b0SQu Wenruo } 232521a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 232621a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 232721a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 232821a852b0SQu Wenruo ret = -EINVAL; 232921a852b0SQu Wenruo } 233021a852b0SQu Wenruo 233121a852b0SQu Wenruo /* 233221a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 233321a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 233421a852b0SQu Wenruo */ 233521a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 233621a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 233721a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 233821a852b0SQu Wenruo ret = -EINVAL; 233921a852b0SQu Wenruo } 23400bb3eb3eSQu Wenruo 23410bb3eb3eSQu Wenruo /* 23421a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 23431a42daabSQu Wenruo * 23441a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 23451a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 23461a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 23471a42daabSQu Wenruo * beginning. 23480bb3eb3eSQu Wenruo */ 23491a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 235021a852b0SQu Wenruo btrfs_err(fs_info, 23510bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 235221a852b0SQu Wenruo sectorsize, PAGE_SIZE); 235321a852b0SQu Wenruo ret = -EINVAL; 235421a852b0SQu Wenruo } 23550bb3eb3eSQu Wenruo 235621a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 235721a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 235821a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 235921a852b0SQu Wenruo ret = -EINVAL; 236021a852b0SQu Wenruo } 236121a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 236221a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 236321a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 236421a852b0SQu Wenruo ret = -EINVAL; 236521a852b0SQu Wenruo } 236621a852b0SQu Wenruo 236721a852b0SQu Wenruo /* Root alignment check */ 236821a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 236921a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 237021a852b0SQu Wenruo btrfs_super_root(sb)); 237121a852b0SQu Wenruo ret = -EINVAL; 237221a852b0SQu Wenruo } 237321a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 237421a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 237521a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 237621a852b0SQu Wenruo ret = -EINVAL; 237721a852b0SQu Wenruo } 237821a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 237921a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 238021a852b0SQu Wenruo btrfs_super_log_root(sb)); 238121a852b0SQu Wenruo ret = -EINVAL; 238221a852b0SQu Wenruo } 238321a852b0SQu Wenruo 2384aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2385aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2386aefd7f70SNikolay Borisov btrfs_err(fs_info, 2387aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2388aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2389aefd7f70SNikolay Borisov ret = -EINVAL; 2390aefd7f70SNikolay Borisov } 2391aefd7f70SNikolay Borisov 2392aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2393aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2394aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2395aefd7f70SNikolay Borisov btrfs_err(fs_info, 2396aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2397aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2398aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2399aefd7f70SNikolay Borisov ret = -EINVAL; 2400aefd7f70SNikolay Borisov } 2401aefd7f70SNikolay Borisov 24021c56ab99SQu Wenruo /* 24031c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 24041c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 24051c56ab99SQu Wenruo */ 24061c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 24071c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 24081c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 24091c56ab99SQu Wenruo btrfs_err(fs_info, 24101c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 24111c56ab99SQu Wenruo ret = -EINVAL; 24121c56ab99SQu Wenruo } 24131c56ab99SQu Wenruo 2414de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 24157239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 241621a852b0SQu Wenruo btrfs_err(fs_info, 24177239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2418de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 241921a852b0SQu Wenruo ret = -EINVAL; 242021a852b0SQu Wenruo } 242121a852b0SQu Wenruo 242221a852b0SQu Wenruo /* 242321a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 242421a852b0SQu Wenruo * done later 242521a852b0SQu Wenruo */ 242621a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 242721a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 242821a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 242921a852b0SQu Wenruo ret = -EINVAL; 243021a852b0SQu Wenruo } 243121a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 243221a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 243321a852b0SQu Wenruo btrfs_super_stripesize(sb)); 243421a852b0SQu Wenruo ret = -EINVAL; 243521a852b0SQu Wenruo } 243621a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 243721a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 243821a852b0SQu Wenruo btrfs_super_num_devices(sb)); 243921a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 244021a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 244121a852b0SQu Wenruo ret = -EINVAL; 244221a852b0SQu Wenruo } 244321a852b0SQu Wenruo 2444069ec957SQu Wenruo if (mirror_num >= 0 && 2445069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 244621a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 244721a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 244821a852b0SQu Wenruo ret = -EINVAL; 244921a852b0SQu Wenruo } 245021a852b0SQu Wenruo 245121a852b0SQu Wenruo /* 245221a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 245321a852b0SQu Wenruo * and one chunk 245421a852b0SQu Wenruo */ 245521a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 245621a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 245721a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 245821a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 245921a852b0SQu Wenruo ret = -EINVAL; 246021a852b0SQu Wenruo } 246121a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 246221a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 246321a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 246421a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 246521a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 246621a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 246721a852b0SQu Wenruo ret = -EINVAL; 246821a852b0SQu Wenruo } 246921a852b0SQu Wenruo 247021a852b0SQu Wenruo /* 247121a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 247221a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 247321a852b0SQu Wenruo */ 247421a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 247521a852b0SQu Wenruo btrfs_warn(fs_info, 247621a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 247721a852b0SQu Wenruo btrfs_super_generation(sb), 247821a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 247921a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 248021a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 248121a852b0SQu Wenruo btrfs_warn(fs_info, 248221a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 248321a852b0SQu Wenruo btrfs_super_generation(sb), 248421a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 248521a852b0SQu Wenruo 248621a852b0SQu Wenruo return ret; 248721a852b0SQu Wenruo } 248821a852b0SQu Wenruo 2489069ec957SQu Wenruo /* 2490069ec957SQu Wenruo * Validation of super block at mount time. 2491069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2492069ec957SQu Wenruo * flags will be skipped. 2493069ec957SQu Wenruo */ 2494069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2495069ec957SQu Wenruo { 2496a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2497069ec957SQu Wenruo } 2498069ec957SQu Wenruo 249975cb857dSQu Wenruo /* 250075cb857dSQu Wenruo * Validation of super block at write time. 250175cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 250275cb857dSQu Wenruo * overwritten soon. 250375cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 250475cb857dSQu Wenruo */ 250575cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 250675cb857dSQu Wenruo struct btrfs_super_block *sb) 250775cb857dSQu Wenruo { 250875cb857dSQu Wenruo int ret; 250975cb857dSQu Wenruo 2510a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 251175cb857dSQu Wenruo if (ret < 0) 251275cb857dSQu Wenruo goto out; 2513e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 251475cb857dSQu Wenruo ret = -EUCLEAN; 251575cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 251675cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 251775cb857dSQu Wenruo goto out; 251875cb857dSQu Wenruo } 251975cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 252075cb857dSQu Wenruo ret = -EUCLEAN; 252175cb857dSQu Wenruo btrfs_err(fs_info, 252275cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 252375cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 252475cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 252575cb857dSQu Wenruo goto out; 252675cb857dSQu Wenruo } 252775cb857dSQu Wenruo out: 252875cb857dSQu Wenruo if (ret < 0) 252975cb857dSQu Wenruo btrfs_err(fs_info, 253075cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 253175cb857dSQu Wenruo return ret; 253275cb857dSQu Wenruo } 253375cb857dSQu Wenruo 2534bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2535bd676446SJosef Bacik { 2536789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 2537789d6a3aSQu Wenruo .level = level, 2538789d6a3aSQu Wenruo .transid = gen, 2539789d6a3aSQu Wenruo .owner_root = root->root_key.objectid 2540789d6a3aSQu Wenruo }; 2541bd676446SJosef Bacik int ret = 0; 2542bd676446SJosef Bacik 2543789d6a3aSQu Wenruo root->node = read_tree_block(root->fs_info, bytenr, &check); 2544bd676446SJosef Bacik if (IS_ERR(root->node)) { 2545bd676446SJosef Bacik ret = PTR_ERR(root->node); 2546bd676446SJosef Bacik root->node = NULL; 25474eb150d6SQu Wenruo return ret; 25484eb150d6SQu Wenruo } 25494eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2550bd676446SJosef Bacik free_extent_buffer(root->node); 2551bd676446SJosef Bacik root->node = NULL; 25524eb150d6SQu Wenruo return -EIO; 2553bd676446SJosef Bacik } 2554bd676446SJosef Bacik 2555bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2556bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2557bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2558bd676446SJosef Bacik return ret; 2559bd676446SJosef Bacik } 2560bd676446SJosef Bacik 2561bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2562bd676446SJosef Bacik { 2563bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2564bd676446SJosef Bacik u64 gen, bytenr; 2565bd676446SJosef Bacik int level, ret; 2566bd676446SJosef Bacik 2567bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2568bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2569bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2570bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 25719c54e80dSJosef Bacik if (ret) { 2572bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2573bd676446SJosef Bacik return ret; 2574bd676446SJosef Bacik } 25759c54e80dSJosef Bacik return 0; 25769c54e80dSJosef Bacik } 25779c54e80dSJosef Bacik 25786ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2579b8522a1eSNikolay Borisov { 25806ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2581b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2582b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2583b8522a1eSNikolay Borisov bool handle_error = false; 2584b8522a1eSNikolay Borisov int ret = 0; 2585b8522a1eSNikolay Borisov int i; 2586b8522a1eSNikolay Borisov 2587b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2588b8522a1eSNikolay Borisov if (handle_error) { 2589b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2590b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2591b8522a1eSNikolay Borisov tree_root->node = NULL; 2592b8522a1eSNikolay Borisov 2593b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2594b8522a1eSNikolay Borisov break; 2595b8522a1eSNikolay Borisov 2596b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2597b8522a1eSNikolay Borisov 2598b8522a1eSNikolay Borisov /* 2599b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2600b8522a1eSNikolay Borisov * valid 2601b8522a1eSNikolay Borisov */ 2602b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2603b8522a1eSNikolay Borisov 2604b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2605b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2606b8522a1eSNikolay Borisov 2607745806fbSQu Wenruo btrfs_warn(fs_info, "try to load backup roots slot %d", i); 2608b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26096ef108ddSNikolay Borisov backup_index = ret; 2610b8522a1eSNikolay Borisov if (ret < 0) 2611b8522a1eSNikolay Borisov return ret; 2612b8522a1eSNikolay Borisov } 2613b8522a1eSNikolay Borisov 2614bd676446SJosef Bacik ret = load_important_roots(fs_info); 2615bd676446SJosef Bacik if (ret) { 2616217f5004SNikolay Borisov handle_error = true; 2617b8522a1eSNikolay Borisov continue; 2618b8522a1eSNikolay Borisov } 2619b8522a1eSNikolay Borisov 2620336a0d8dSNikolay Borisov /* 2621336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2622336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2623336a0d8dSNikolay Borisov */ 2624453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2625b8522a1eSNikolay Borisov if (ret < 0) { 2626b8522a1eSNikolay Borisov handle_error = true; 2627b8522a1eSNikolay Borisov continue; 2628b8522a1eSNikolay Borisov } 2629b8522a1eSNikolay Borisov 26306b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2631b8522a1eSNikolay Borisov 2632b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2633b8522a1eSNikolay Borisov if (ret < 0) { 2634b8522a1eSNikolay Borisov handle_error = true; 2635b8522a1eSNikolay Borisov continue; 2636b8522a1eSNikolay Borisov } 2637b8522a1eSNikolay Borisov 2638b8522a1eSNikolay Borisov /* All successful */ 2639bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2640bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2641d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 26426ef108ddSNikolay Borisov 26436ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26446ef108ddSNikolay Borisov if (backup_index < 0) { 26456ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26466ef108ddSNikolay Borisov } else { 26476ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26486ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26496ef108ddSNikolay Borisov } 2650b8522a1eSNikolay Borisov break; 2651b8522a1eSNikolay Borisov } 2652b8522a1eSNikolay Borisov 2653b8522a1eSNikolay Borisov return ret; 2654b8522a1eSNikolay Borisov } 2655b8522a1eSNikolay Borisov 26568260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2657eb60ceacSChris Mason { 2658fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 265901cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26608fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2661facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 266224bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2663eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 266411833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2665eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2666a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2667fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 266824bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26694cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2670ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2671f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 267247ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 267340ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2674afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2675c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2676f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2677abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2678d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2679f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 26807585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2681573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 26820bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 26835f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 2684de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 268519c00ddcSChris Mason 2686e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 26875a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 26888b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 26895f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 26903e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 26913e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 26923e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 26933e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 26943e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 26953e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 26963e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 26973e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 2698e1489b4fSIoannis Angelakopoulos 26990b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 27006324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2701f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 270247ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 270318bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2704afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2705bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2706bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 27073fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27083fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2709bd647ce3SJosef Bacik #endif 2710c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 271166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 271266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 271366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 271466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 271566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 271666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 271766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2718ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2719ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2720ba2c4d4eSJosef Bacik 2721771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27224cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2723034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2724fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2725abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 272695ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27279ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27284cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2729a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2730f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27318b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2732fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2733c8b97818SChris Mason 2734b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2735b34b086cSChris Mason num_online_cpus() + 2, 8); 27360afbaf8cSChris Mason 2737199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2738199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 273969fe2d75SJosef Bacik 2740638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 274121adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 274221adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 274321adbd5cSStefan Behrens #endif 2744779a65a4SEric Sandeen btrfs_init_balance(fs_info); 274557056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2746a2de733cSArne Jansen 274716b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 274808dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 27490f9dd46cSJosef Bacik 2750fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 275135da5a7eSDavid Sterba IO_TREE_FS_EXCLUDED_EXTENTS); 275239279cc3SChris Mason 27535a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2754e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2755925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2756a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2757a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27581bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27599e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2760c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 276176dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2762803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2763fa9c0d79SChris Mason 2764ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2765f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2766b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2767416ac51dSArne Jansen 2768fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2769fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2770fa9c0d79SChris Mason 2771e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2772f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2773bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27744854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2775034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 27763768f368SChris Mason 2777da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2778da17066cSJeff Mahoney fs_info->nodesize = 4096; 2779da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2780ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2781da17066cSJeff Mahoney fs_info->stripesize = 4096; 2782da17066cSJeff Mahoney 2783f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 2784f7b12a62SNaohiro Aota 2785eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2786eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2787eede2bf3SOmar Sandoval 278818bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 278918bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 27908260edbaSJosef Bacik } 27918260edbaSJosef Bacik 27928260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 27938260edbaSJosef Bacik { 27948260edbaSJosef Bacik int ret; 27958260edbaSJosef Bacik 27968260edbaSJosef Bacik fs_info->sb = sb; 27978260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 27988260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 27999e967495SFilipe Manana 28005deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 2801ae18c37aSJosef Bacik if (ret) 2802ae18c37aSJosef Bacik return ret; 2803ae18c37aSJosef Bacik 2804ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2805ae18c37aSJosef Bacik if (ret) 2806c75e8394SJosef Bacik return ret; 2807ae18c37aSJosef Bacik 2808ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2809ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2810ae18c37aSJosef Bacik 2811ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2812ae18c37aSJosef Bacik if (ret) 2813c75e8394SJosef Bacik return ret; 2814ae18c37aSJosef Bacik 2815ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2816ae18c37aSJosef Bacik GFP_KERNEL); 2817ae18c37aSJosef Bacik if (ret) 2818c75e8394SJosef Bacik return ret; 2819ae18c37aSJosef Bacik 2820ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2821ae18c37aSJosef Bacik GFP_KERNEL); 2822c75e8394SJosef Bacik if (!fs_info->delayed_root) 2823c75e8394SJosef Bacik return -ENOMEM; 2824ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2825ae18c37aSJosef Bacik 2826a0a1db70SFilipe Manana if (sb_rdonly(sb)) 2827a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 2828a0a1db70SFilipe Manana 2829c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2830ae18c37aSJosef Bacik } 2831ae18c37aSJosef Bacik 283297f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 283397f4dd09SNikolay Borisov { 28340d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 283597f4dd09SNikolay Borisov int ret; 283697f4dd09SNikolay Borisov 283797f4dd09SNikolay Borisov /* 283897f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 283997f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 284097f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 284197f4dd09SNikolay Borisov */ 284297f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 284397f4dd09SNikolay Borisov if (ret < 0) { 2844c94bec2cSJosef Bacik if (ret != -EINTR) 2845c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2846c94bec2cSJosef Bacik ret); 284797f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 284897f4dd09SNikolay Borisov return ret; 284997f4dd09SNikolay Borisov } 285097f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 285197f4dd09SNikolay Borisov } 285297f4dd09SNikolay Borisov 285397f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 285497f4dd09SNikolay Borisov { 285597f4dd09SNikolay Borisov struct task_struct *task; 285697f4dd09SNikolay Borisov 285797f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 285897f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 285997f4dd09SNikolay Borisov if (IS_ERR(task)) { 286097f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 286197f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 286297f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 286397f4dd09SNikolay Borisov return PTR_ERR(task); 286497f4dd09SNikolay Borisov } 286597f4dd09SNikolay Borisov 286697f4dd09SNikolay Borisov return 0; 286797f4dd09SNikolay Borisov } 286897f4dd09SNikolay Borisov 286944c0ca21SBoris Burkov /* 28708cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 28718cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 28728cd29088SBoris Burkov * code paths. 28738cd29088SBoris Burkov */ 28748cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 28758cd29088SBoris Burkov { 28768cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 28778b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 28788cd29088SBoris Burkov } 28798cd29088SBoris Burkov 28808cd29088SBoris Burkov /* 288144c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 288244c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 288344c0ca21SBoris Burkov */ 288444c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 288544c0ca21SBoris Burkov { 288644c0ca21SBoris Burkov int ret; 288794846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 28881d6a4fc8SQu Wenruo bool rebuild_free_space_tree = false; 28898b228324SBoris Burkov 28908b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 28918b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 28921d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 28938b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 28948b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 28958b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 28961d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 28978b228324SBoris Burkov } 28988b228324SBoris Burkov 28991d6a4fc8SQu Wenruo if (rebuild_free_space_tree) { 29001d6a4fc8SQu Wenruo btrfs_info(fs_info, "rebuilding free space tree"); 29011d6a4fc8SQu Wenruo ret = btrfs_rebuild_free_space_tree(fs_info); 29028b228324SBoris Burkov if (ret) { 29038b228324SBoris Burkov btrfs_warn(fs_info, 29041d6a4fc8SQu Wenruo "failed to rebuild free space tree: %d", ret); 29051d6a4fc8SQu Wenruo goto out; 29061d6a4fc8SQu Wenruo } 29071d6a4fc8SQu Wenruo } 29081d6a4fc8SQu Wenruo 29091d6a4fc8SQu Wenruo if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 29101d6a4fc8SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { 29111d6a4fc8SQu Wenruo btrfs_info(fs_info, "disabling free space tree"); 29121d6a4fc8SQu Wenruo ret = btrfs_delete_free_space_tree(fs_info); 29131d6a4fc8SQu Wenruo if (ret) { 29141d6a4fc8SQu Wenruo btrfs_warn(fs_info, 29151d6a4fc8SQu Wenruo "failed to disable free space tree: %d", ret); 29168b228324SBoris Burkov goto out; 29178b228324SBoris Burkov } 29188b228324SBoris Burkov } 291944c0ca21SBoris Burkov 29208d488a8cSFilipe Manana /* 29218d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 29228d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 2923fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 29248d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 29258d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 29268d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 29278d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 29288d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 29298d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 29308d488a8cSFilipe Manana */ 29318d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 29328d488a8cSFilipe Manana if (ret) 29338d488a8cSFilipe Manana goto out; 29348d488a8cSFilipe Manana 293544c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 293644c0ca21SBoris Burkov if (ret) 293744c0ca21SBoris Burkov goto out; 293844c0ca21SBoris Burkov 29398f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 29408f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29418f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29428f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29438f1c21d7SBoris Burkov goto out; 29448f1c21d7SBoris Burkov } 29458f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29468f1c21d7SBoris Burkov 294744c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 29487eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 294944c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 295044c0ca21SBoris Burkov if (ret < 0) { 295144c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 295244c0ca21SBoris Burkov goto out; 295344c0ca21SBoris Burkov } 295444c0ca21SBoris Burkov 29555011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 29565011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 29575011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 29585011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 29595011139aSBoris Burkov if (ret) { 29605011139aSBoris Burkov btrfs_warn(fs_info, 29615011139aSBoris Burkov "failed to create free space tree: %d", ret); 29625011139aSBoris Burkov goto out; 29635011139aSBoris Burkov } 29645011139aSBoris Burkov } 29655011139aSBoris Burkov 296694846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 296794846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 296894846229SBoris Burkov if (ret) 296994846229SBoris Burkov goto out; 297094846229SBoris Burkov } 297194846229SBoris Burkov 297244c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 297344c0ca21SBoris Burkov if (ret) 297444c0ca21SBoris Burkov goto out; 297544c0ca21SBoris Burkov 297644c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 297744c0ca21SBoris Burkov if (ret) { 297844c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 297944c0ca21SBoris Burkov goto out; 298044c0ca21SBoris Burkov } 298144c0ca21SBoris Burkov 298244c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 298344c0ca21SBoris Burkov 298444c0ca21SBoris Burkov if (!fs_info->uuid_root) { 298544c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 298644c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 298744c0ca21SBoris Burkov if (ret) { 298844c0ca21SBoris Burkov btrfs_warn(fs_info, 298944c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 299044c0ca21SBoris Burkov goto out; 299144c0ca21SBoris Burkov } 299244c0ca21SBoris Burkov } 299344c0ca21SBoris Burkov 299444c0ca21SBoris Burkov out: 299544c0ca21SBoris Burkov return ret; 299644c0ca21SBoris Burkov } 299744c0ca21SBoris Burkov 2998d7f67ac9SQu Wenruo /* 2999d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3000d7f67ac9SQu Wenruo * 30012ba48b20SQu Wenruo * @is_rw_mount: If the mount is read-write. 30022ba48b20SQu Wenruo * 3003d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3004d7f67ac9SQu Wenruo * feature dependencies). 3005d7f67ac9SQu Wenruo * 3006d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3007d7f67ac9SQu Wenruo * btrfs_validate_super(). 3008d7f67ac9SQu Wenruo * 3009d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3010d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3011d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3012d7f67ac9SQu Wenruo */ 30132ba48b20SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) 3014d7f67ac9SQu Wenruo { 3015d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3016d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3017d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3018d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3019d7f67ac9SQu Wenruo 3020d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3021d7f67ac9SQu Wenruo btrfs_err(fs_info, 3022d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3023d7f67ac9SQu Wenruo incompat); 3024d7f67ac9SQu Wenruo return -EINVAL; 3025d7f67ac9SQu Wenruo } 3026d7f67ac9SQu Wenruo 3027d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3028d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3029d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3030d7f67ac9SQu Wenruo btrfs_err(fs_info, 3031d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3032d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3033d7f67ac9SQu Wenruo return -EINVAL; 3034d7f67ac9SQu Wenruo } 3035d7f67ac9SQu Wenruo 3036d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3037d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3038d7f67ac9SQu Wenruo 3039d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3040d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3041d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3042d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3043d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3044d7f67ac9SQu Wenruo 3045d7f67ac9SQu Wenruo /* 3046d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3047d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3048d7f67ac9SQu Wenruo */ 3049d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3050d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3051d7f67ac9SQu Wenruo 30522ba48b20SQu Wenruo if (compat_ro_unsupp && is_rw_mount) { 3053d7f67ac9SQu Wenruo btrfs_err(fs_info, 3054d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3055d7f67ac9SQu Wenruo compat_ro); 3056d7f67ac9SQu Wenruo return -EINVAL; 3057d7f67ac9SQu Wenruo } 3058d7f67ac9SQu Wenruo 3059d7f67ac9SQu Wenruo /* 3060d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3061d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3062d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3063d7f67ac9SQu Wenruo */ 3064d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3065d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3066d7f67ac9SQu Wenruo btrfs_err(fs_info, 3067d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3068d7f67ac9SQu Wenruo compat_ro); 3069d7f67ac9SQu Wenruo return -EINVAL; 3070d7f67ac9SQu Wenruo } 3071d7f67ac9SQu Wenruo 3072d7f67ac9SQu Wenruo /* 3073d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3074d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3075d7f67ac9SQu Wenruo */ 3076d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3077d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3078d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3079d7f67ac9SQu Wenruo btrfs_err(fs_info, 3080d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3081d7f67ac9SQu Wenruo return -EINVAL; 3082d7f67ac9SQu Wenruo } 3083d7f67ac9SQu Wenruo 3084d7f67ac9SQu Wenruo /* 3085d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3086d7f67ac9SQu Wenruo * 3087d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3088d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3089d7f67ac9SQu Wenruo * going to be deprecated anyway. 3090d7f67ac9SQu Wenruo */ 3091d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3092d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3093d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3094d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3095d7f67ac9SQu Wenruo return -EINVAL; 3096d7f67ac9SQu Wenruo } 3097d7f67ac9SQu Wenruo 3098d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3099d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3100d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3101d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3102d7f67ac9SQu Wenruo 3103d7f67ac9SQu Wenruo return 0; 3104d7f67ac9SQu Wenruo } 3105d7f67ac9SQu Wenruo 3106ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3107ae18c37aSJosef Bacik char *options) 3108ae18c37aSJosef Bacik { 3109ae18c37aSJosef Bacik u32 sectorsize; 3110ae18c37aSJosef Bacik u32 nodesize; 3111ae18c37aSJosef Bacik u32 stripesize; 3112ae18c37aSJosef Bacik u64 generation; 3113ae18c37aSJosef Bacik u64 features; 3114ae18c37aSJosef Bacik u16 csum_type; 3115ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3116ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3117ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3118ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3119ae18c37aSJosef Bacik int ret; 3120ae18c37aSJosef Bacik int level; 3121ae18c37aSJosef Bacik 31228260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 31234871c33bSQu Wenruo if (ret) 3124ae18c37aSJosef Bacik goto fail; 312553b381b3SDavid Woodhouse 3126ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3127ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3128ae18c37aSJosef Bacik GFP_KERNEL); 3129ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3130ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3131ae18c37aSJosef Bacik GFP_KERNEL); 3132ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3133ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 31344871c33bSQu Wenruo ret = -ENOMEM; 3135c75e8394SJosef Bacik goto fail; 3136ae18c37aSJosef Bacik } 3137ae18c37aSJosef Bacik 3138dcb2137cSChristoph Hellwig ret = btrfs_init_btree_inode(sb); 31394871c33bSQu Wenruo if (ret) 3140c75e8394SJosef Bacik goto fail; 3141ae18c37aSJosef Bacik 3142d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 31431104a885SDavid Sterba 31441104a885SDavid Sterba /* 31451104a885SDavid Sterba * Read super block and check the signature bytes only 31461104a885SDavid Sterba */ 3147d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 31488f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 31494871c33bSQu Wenruo ret = PTR_ERR(disk_super); 315016cdcec7SMiao Xie goto fail_alloc; 315120b45077SDave Young } 315239279cc3SChris Mason 31531104a885SDavid Sterba /* 3154260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 31558dc3f22cSJohannes Thumshirn * corrupted, we'll find out 31568dc3f22cSJohannes Thumshirn */ 31578f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 315851bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 31598dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 316051bce6c9SJohannes Thumshirn csum_type); 31614871c33bSQu Wenruo ret = -EINVAL; 31628f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31638dc3f22cSJohannes Thumshirn goto fail_alloc; 31648dc3f22cSJohannes Thumshirn } 31658dc3f22cSJohannes Thumshirn 316683c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 316783c68bbcSSu Yue 31686d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 31696d97c6e3SJohannes Thumshirn if (ret) { 31708f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31716d97c6e3SJohannes Thumshirn goto fail_alloc; 31726d97c6e3SJohannes Thumshirn } 31736d97c6e3SJohannes Thumshirn 31748dc3f22cSJohannes Thumshirn /* 31751104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 31761104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 31771104a885SDavid Sterba */ 31783d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 317905135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 31804871c33bSQu Wenruo ret = -EINVAL; 31818f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3182141386e1SJosef Bacik goto fail_alloc; 31831104a885SDavid Sterba } 31841104a885SDavid Sterba 31851104a885SDavid Sterba /* 31861104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 31871104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 31881104a885SDavid Sterba * the whole block of INFO_SIZE 31891104a885SDavid Sterba */ 31908f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 31918f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 31925f39d397SChris Mason 3193fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3194fbc6feaeSNikolay Borisov 31950b86a832SChris Mason 3196fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3197fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3198fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3199fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3200fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3201fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3202fbc6feaeSNikolay Borisov } 3203fbc6feaeSNikolay Borisov 3204fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3205fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3206de37aa51SNikolay Borisov 3207069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 32081104a885SDavid Sterba if (ret) { 320905135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 32104871c33bSQu Wenruo ret = -EINVAL; 3211141386e1SJosef Bacik goto fail_alloc; 32121104a885SDavid Sterba } 32131104a885SDavid Sterba 32144871c33bSQu Wenruo if (!btrfs_super_root(disk_super)) { 32154871c33bSQu Wenruo btrfs_err(fs_info, "invalid superblock tree root bytenr"); 32164871c33bSQu Wenruo ret = -EINVAL; 3217141386e1SJosef Bacik goto fail_alloc; 32184871c33bSQu Wenruo } 32190f7d52f4SChris Mason 3220acce952bSliubo /* check FS state, whether FS is broken. */ 322187533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 322287533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3223acce952bSliubo 322475e7cb7fSLiu Bo /* 322575e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 322675e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 322775e7cb7fSLiu Bo */ 322875e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 322975e7cb7fSLiu Bo 32306f93e834SAnand Jain 32316f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 32326f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 32336f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 32346f93e834SAnand Jain stripesize = sectorsize; 32356f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 32366f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 32376f93e834SAnand Jain 32386f93e834SAnand Jain fs_info->nodesize = nodesize; 32396f93e834SAnand Jain fs_info->sectorsize = sectorsize; 32406f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 32416f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 32426f93e834SAnand Jain fs_info->stripesize = stripesize; 32436f93e834SAnand Jain 32442ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 32454871c33bSQu Wenruo if (ret) 3246141386e1SJosef Bacik goto fail_alloc; 3247dfe25020SChris Mason 32482ba48b20SQu Wenruo ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); 32494871c33bSQu Wenruo if (ret < 0) 3250141386e1SJosef Bacik goto fail_alloc; 3251f2b636e8SJosef Bacik 32528481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 32538481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 32548481dd80SQu Wenruo 32559f73f1aeSQu Wenruo /* 32569f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 32579f73f1aeSQu Wenruo * going to be deprecated. 32589f73f1aeSQu Wenruo * 32599f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 32609f73f1aeSQu Wenruo */ 32619f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 32629f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 32639f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 32649f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 32659f73f1aeSQu Wenruo 326695ea0486SQu Wenruo btrfs_warn(fs_info, 326795ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 32680bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 32698481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 32704871c33bSQu Wenruo if (!subpage_info) { 32714871c33bSQu Wenruo ret = -ENOMEM; 32728481dd80SQu Wenruo goto fail_alloc; 32734871c33bSQu Wenruo } 32748481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 32758481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3276c8050b3bSQu Wenruo } 32770bb3eb3eSQu Wenruo 3278d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 32794871c33bSQu Wenruo if (ret) 32800dc3b84aSJosef Bacik goto fail_sb_buffer; 32814543df7eSChris Mason 32829e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 32839e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 32844575c9ccSChris Mason 3285a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3286a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3287de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3288db94535dSChris Mason 3289925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 32906bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3291925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 329284eed90fSChris Mason if (ret) { 329305135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 32945d4f98a2SYan Zheng goto fail_sb_buffer; 329584eed90fSChris Mason } 32960b86a832SChris Mason 329784234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3298581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3299bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3300bd676446SJosef Bacik generation, level); 3301bd676446SJosef Bacik if (ret) { 330205135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3303af31f5e5SChris Mason goto fail_tree_roots; 330483121942SDavid Woodhouse } 33050b86a832SChris Mason 3306e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3307c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3308c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3309e17cade2SChris Mason 33105b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 33112b82032cSYan Zheng if (ret) { 331205135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3313af31f5e5SChris Mason goto fail_tree_roots; 33142b82032cSYan Zheng } 33150b86a832SChris Mason 33168dabb742SStefan Behrens /* 3317bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3318bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3319bacce86aSAnand Jain * target is required. So free devices that are not part of this 33201a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3321bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 33228dabb742SStefan Behrens */ 3323bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3324d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 332505135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 33264871c33bSQu Wenruo ret = -EIO; 3327a6b0d5c8SChris Mason goto fail_tree_roots; 3328a6b0d5c8SChris Mason } 3329a6b0d5c8SChris Mason 3330b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 33314bbcaa64SEric Sandeen if (ret) 3332b8522a1eSNikolay Borisov goto fail_tree_roots; 33338929ecfaSYan, Zheng 333475ec1db8SJosef Bacik /* 333573651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 333673651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 333773651042SNaohiro Aota * zoned incompat feature flag set. 333873651042SNaohiro Aota */ 333973651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 334073651042SNaohiro Aota if (ret) { 334173651042SNaohiro Aota btrfs_err(fs_info, 33424871c33bSQu Wenruo "zoned: failed to read device zone info: %d", ret); 334373651042SNaohiro Aota goto fail_block_groups; 334473651042SNaohiro Aota } 334573651042SNaohiro Aota 334673651042SNaohiro Aota /* 334775ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 334875ec1db8SJosef Bacik * check the generation here so we can set the 334975ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 335075ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 335175ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 335275ec1db8SJosef Bacik * even though it was perfectly fine. 335375ec1db8SJosef Bacik */ 335475ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 335575ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 335675ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 335775ec1db8SJosef Bacik 3358cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3359cf90d884SQu Wenruo if (ret) { 3360cf90d884SQu Wenruo btrfs_err(fs_info, 3361cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3362cf90d884SQu Wenruo ret); 3363cf90d884SQu Wenruo goto fail_block_groups; 3364cf90d884SQu Wenruo } 336568310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 336668310a5eSIlya Dryomov if (ret) { 336705135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 336868310a5eSIlya Dryomov goto fail_block_groups; 336968310a5eSIlya Dryomov } 337068310a5eSIlya Dryomov 3371733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3372733f4fbbSStefan Behrens if (ret) { 337305135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3374733f4fbbSStefan Behrens goto fail_block_groups; 3375733f4fbbSStefan Behrens } 3376733f4fbbSStefan Behrens 33778dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 33788dabb742SStefan Behrens if (ret) { 337905135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 33808dabb742SStefan Behrens goto fail_block_groups; 33818dabb742SStefan Behrens } 33828dabb742SStefan Behrens 3383b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3384b70f5097SNaohiro Aota if (ret) { 3385b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3386b70f5097SNaohiro Aota ret); 3387b70f5097SNaohiro Aota goto fail_block_groups; 3388b70f5097SNaohiro Aota } 3389b70f5097SNaohiro Aota 3390c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3391b7c35e81SAnand Jain if (ret) { 339205135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 339305135f59SDavid Sterba ret); 3394b7c35e81SAnand Jain goto fail_block_groups; 3395b7c35e81SAnand Jain } 3396b7c35e81SAnand Jain 339796f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 33985ac1d209SJeff Mahoney if (ret) { 339905135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3400b7c35e81SAnand Jain goto fail_fsdev_sysfs; 34015ac1d209SJeff Mahoney } 34025ac1d209SJeff Mahoney 3403c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3404c59021f8Sliubo if (ret) { 340505135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 34062365dd3cSAnand Jain goto fail_sysfs; 3407c59021f8Sliubo } 3408c59021f8Sliubo 34095b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 34101b1d1f66SJosef Bacik if (ret) { 341105135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 34122365dd3cSAnand Jain goto fail_sysfs; 34131b1d1f66SJosef Bacik } 34144330e183SQu Wenruo 341516beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 341616beac87SNaohiro Aota 34175c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 34185c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 341905135f59SDavid Sterba btrfs_warn(fs_info, 342052042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 34214871c33bSQu Wenruo ret = -EINVAL; 34222365dd3cSAnand Jain goto fail_sysfs; 3423292fd7fcSStefan Behrens } 34249078a3e1SChris Mason 342533c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3426a74a4b97SChris Mason "btrfs-cleaner"); 34274871c33bSQu Wenruo if (IS_ERR(fs_info->cleaner_kthread)) { 34284871c33bSQu Wenruo ret = PTR_ERR(fs_info->cleaner_kthread); 34292365dd3cSAnand Jain goto fail_sysfs; 34304871c33bSQu Wenruo } 3431a74a4b97SChris Mason 3432a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3433a74a4b97SChris Mason tree_root, 3434a74a4b97SChris Mason "btrfs-transaction"); 34354871c33bSQu Wenruo if (IS_ERR(fs_info->transaction_kthread)) { 34364871c33bSQu Wenruo ret = PTR_ERR(fs_info->transaction_kthread); 34373f157a2fSChris Mason goto fail_cleaner; 34384871c33bSQu Wenruo } 3439a74a4b97SChris Mason 3440583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3441c289811cSChris Mason !fs_info->fs_devices->rotating) { 3442583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3443c289811cSChris Mason } 3444c289811cSChris Mason 3445572d9ab7SDavid Sterba /* 344663a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 344763a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 344863a7cb13SDavid Sterba * nodiscard for the same mount. 344963a7cb13SDavid Sterba */ 345063a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 345163a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 345263a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 345363a7cb13SDavid Sterba fs_info->fs_devices->discardable) { 345463a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 345563a7cb13SDavid Sterba "auto enabling async discard"); 345663a7cb13SDavid Sterba } 345763a7cb13SDavid Sterba 345821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 34590b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 34602ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 34610b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3462cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 346321adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 346421adbd5cSStefan Behrens if (ret) 346505135f59SDavid Sterba btrfs_warn(fs_info, 346605135f59SDavid Sterba "failed to initialize integrity check module: %d", 346705135f59SDavid Sterba ret); 346821adbd5cSStefan Behrens } 346921adbd5cSStefan Behrens #endif 3470bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3471bcef60f2SArne Jansen if (ret) 3472bcef60f2SArne Jansen goto fail_trans_kthread; 347321adbd5cSStefan Behrens 3474fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3475fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3476fd708b81SJosef Bacik 347796da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 347896da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 34790b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3480e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 348163443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 34824871c33bSQu Wenruo if (ret) 348328c16cbbSWang Shilong goto fail_qgroup; 3484e556ce2cSYan Zheng } 34851a40e23bSZheng Yan 348656e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 34873140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 34884871c33bSQu Wenruo ret = PTR_ERR(fs_info->fs_root); 34894871c33bSQu Wenruo btrfs_warn(fs_info, "failed to read fs tree: %d", ret); 3490315bf8efSJosef Bacik fs_info->fs_root = NULL; 3491bcef60f2SArne Jansen goto fail_qgroup; 34923140c9a3SDan Carpenter } 3493c289811cSChris Mason 3494bc98a42cSDavid Howells if (sb_rdonly(sb)) 34958cd29088SBoris Burkov goto clear_oneshot; 34962b6ba629SIlya Dryomov 349744c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 34982b6ba629SIlya Dryomov if (ret) { 34996bccf3abSJeff Mahoney close_ctree(fs_info); 35002b6ba629SIlya Dryomov return ret; 3501e3acc2a6SJosef Bacik } 3502b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3503b382a324SJan Schmidt 350444c0ca21SBoris Burkov if (fs_info->uuid_root && 350544c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 350644c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 350705135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 350870f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 350970f80175SStefan Behrens if (ret) { 351005135f59SDavid Sterba btrfs_warn(fs_info, 351105135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 35126bccf3abSJeff Mahoney close_ctree(fs_info); 351370f80175SStefan Behrens return ret; 351470f80175SStefan Behrens } 3515f7a81ea4SStefan Behrens } 351694846229SBoris Burkov 3517afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 351847ab2a6cSJosef Bacik 3519b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3520b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3521b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3522b4be6aefSJosef Bacik 35238cd29088SBoris Burkov clear_oneshot: 35248cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3525ad2b2c80SAl Viro return 0; 352639279cc3SChris Mason 3527bcef60f2SArne Jansen fail_qgroup: 3528bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 35297c2ca468SChris Mason fail_trans_kthread: 35307c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 35312ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3532faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 35333f157a2fSChris Mason fail_cleaner: 3534a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 35357c2ca468SChris Mason 35367c2ca468SChris Mason /* 35377c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 35387c2ca468SChris Mason * kthreads 35397c2ca468SChris Mason */ 35407c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 35417c2ca468SChris Mason 35422365dd3cSAnand Jain fail_sysfs: 35436618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 35442365dd3cSAnand Jain 3545b7c35e81SAnand Jain fail_fsdev_sysfs: 3546b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3547b7c35e81SAnand Jain 35481b1d1f66SJosef Bacik fail_block_groups: 354954067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3550af31f5e5SChris Mason 3551af31f5e5SChris Mason fail_tree_roots: 35529e3aa805SJosef Bacik if (fs_info->data_reloc_root) 35539e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 35544273eaffSAnand Jain free_root_pointers(fs_info, true); 35552b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3556af31f5e5SChris Mason 355739279cc3SChris Mason fail_sb_buffer: 35587abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 35595cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 356016cdcec7SMiao Xie fail_alloc: 3561586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3562586e46e2SIlya Dryomov 35634543df7eSChris Mason iput(fs_info->btree_inode); 35647e662854SQinghuang Feng fail: 3565586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 35664871c33bSQu Wenruo ASSERT(ret < 0); 35674871c33bSQu Wenruo return ret; 3568eb60ceacSChris Mason } 3569663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3570eb60ceacSChris Mason 3571314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3572f2984462SChris Mason { 3573314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3574314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3575314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3576314b6dd0SJohannes Thumshirn struct page *page; 3577442a4f63SStefan Behrens 3578314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3579314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3580314b6dd0SJohannes Thumshirn 3581314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3582fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3583314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3584cb3e217bSQu Wenruo btrfs_dev_name(device), 3585314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3586314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3587314b6dd0SJohannes Thumshirn SetPageError(page); 3588314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3589314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3590314b6dd0SJohannes Thumshirn } else { 3591314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3592f2984462SChris Mason } 3593314b6dd0SJohannes Thumshirn 3594314b6dd0SJohannes Thumshirn put_page(page); 3595314b6dd0SJohannes Thumshirn unlock_page(page); 3596314b6dd0SJohannes Thumshirn } 3597314b6dd0SJohannes Thumshirn 3598314b6dd0SJohannes Thumshirn bio_put(bio); 3599f2984462SChris Mason } 3600f2984462SChris Mason 36018f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3602a05d3c91SQu Wenruo int copy_num, bool drop_cache) 360329c36d72SAnand Jain { 360429c36d72SAnand Jain struct btrfs_super_block *super; 36058f32380dSJohannes Thumshirn struct page *page; 360612659251SNaohiro Aota u64 bytenr, bytenr_orig; 36078f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 360812659251SNaohiro Aota int ret; 360929c36d72SAnand Jain 361012659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 361112659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 361212659251SNaohiro Aota if (ret == -ENOENT) 361312659251SNaohiro Aota return ERR_PTR(-EINVAL); 361412659251SNaohiro Aota else if (ret) 361512659251SNaohiro Aota return ERR_PTR(ret); 361612659251SNaohiro Aota 3617cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 36188f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 361929c36d72SAnand Jain 3620a05d3c91SQu Wenruo if (drop_cache) { 3621a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3622a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3623a05d3c91SQu Wenruo 3624a05d3c91SQu Wenruo /* 3625a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3626a05d3c91SQu Wenruo * always read from the device. 3627a05d3c91SQu Wenruo */ 3628a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3629a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3630a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3631a05d3c91SQu Wenruo } 3632a05d3c91SQu Wenruo 36338f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 36348f32380dSJohannes Thumshirn if (IS_ERR(page)) 36358f32380dSJohannes Thumshirn return ERR_CAST(page); 363629c36d72SAnand Jain 36378f32380dSJohannes Thumshirn super = page_address(page); 363896c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 363996c2e067SAnand Jain btrfs_release_disk_super(super); 364096c2e067SAnand Jain return ERR_PTR(-ENODATA); 364196c2e067SAnand Jain } 364296c2e067SAnand Jain 364312659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 36448f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36458f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 364629c36d72SAnand Jain } 364729c36d72SAnand Jain 36488f32380dSJohannes Thumshirn return super; 364929c36d72SAnand Jain } 365029c36d72SAnand Jain 365129c36d72SAnand Jain 36528f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3653a512bbf8SYan Zheng { 36548f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3655a512bbf8SYan Zheng int i; 3656a512bbf8SYan Zheng u64 transid = 0; 3657a512bbf8SYan Zheng 3658a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3659a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3660a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3661a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3662a512bbf8SYan Zheng */ 3663a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3664a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 36658f32380dSJohannes Thumshirn if (IS_ERR(super)) 3666a512bbf8SYan Zheng continue; 3667a512bbf8SYan Zheng 3668a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 36698f32380dSJohannes Thumshirn if (latest) 36708f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 36718f32380dSJohannes Thumshirn 36728f32380dSJohannes Thumshirn latest = super; 3673a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3674a512bbf8SYan Zheng } 3675a512bbf8SYan Zheng } 367692fc03fbSAnand Jain 36778f32380dSJohannes Thumshirn return super; 3678a512bbf8SYan Zheng } 3679a512bbf8SYan Zheng 36804eedeb75SHisashi Hifumi /* 3681abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3682314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 36834eedeb75SHisashi Hifumi * 3684abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3685abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3686abbb3b8eSDavid Sterba * same for write and wait phases. 36874eedeb75SHisashi Hifumi * 3688314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 36894eedeb75SHisashi Hifumi */ 3690a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3691abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3692a512bbf8SYan Zheng { 3693d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3694314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3695d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3696a512bbf8SYan Zheng int i; 3697a512bbf8SYan Zheng int errors = 0; 369812659251SNaohiro Aota int ret; 369912659251SNaohiro Aota u64 bytenr, bytenr_orig; 3700a512bbf8SYan Zheng 3701a512bbf8SYan Zheng if (max_mirrors == 0) 3702a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3703a512bbf8SYan Zheng 3704d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3705d5178578SJohannes Thumshirn 3706a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3707314b6dd0SJohannes Thumshirn struct page *page; 3708314b6dd0SJohannes Thumshirn struct bio *bio; 3709314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3710314b6dd0SJohannes Thumshirn 371112659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 371212659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 371312659251SNaohiro Aota if (ret == -ENOENT) { 371412659251SNaohiro Aota continue; 371512659251SNaohiro Aota } else if (ret < 0) { 371612659251SNaohiro Aota btrfs_err(device->fs_info, 371712659251SNaohiro Aota "couldn't get super block location for mirror %d", 371812659251SNaohiro Aota i); 371912659251SNaohiro Aota errors++; 372012659251SNaohiro Aota continue; 372112659251SNaohiro Aota } 3722935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3723935e5cc9SMiao Xie device->commit_total_bytes) 3724a512bbf8SYan Zheng break; 3725a512bbf8SYan Zheng 372612659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3727a512bbf8SYan Zheng 3728fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3729fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3730fd08001fSEric Biggers sb->csum); 3731a512bbf8SYan Zheng 3732314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3733314b6dd0SJohannes Thumshirn GFP_NOFS); 3734314b6dd0SJohannes Thumshirn if (!page) { 3735fb456252SJeff Mahoney btrfs_err(device->fs_info, 3736314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3737f14d104dSDavid Sterba bytenr); 3738634554dcSJosef Bacik errors++; 3739634554dcSJosef Bacik continue; 3740634554dcSJosef Bacik } 3741634554dcSJosef Bacik 3742314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3743314b6dd0SJohannes Thumshirn get_page(page); 3744a512bbf8SYan Zheng 3745314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3746314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3747a512bbf8SYan Zheng 3748387125fcSChris Mason /* 3749314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3750314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3751314b6dd0SJohannes Thumshirn * checking. 3752387125fcSChris Mason */ 375307888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 375407888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 375507888c66SChristoph Hellwig GFP_NOFS); 3756314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3757314b6dd0SJohannes Thumshirn bio->bi_private = device; 3758314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3759314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3760314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3761314b6dd0SJohannes Thumshirn 3762314b6dd0SJohannes Thumshirn /* 3763314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3764314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3765314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3766314b6dd0SJohannes Thumshirn */ 37671b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3768314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3769314b6dd0SJohannes Thumshirn 377058ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 377158ff51f1SChristoph Hellwig submit_bio(bio); 37728376d9e1SNaohiro Aota 37738376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 37748376d9e1SNaohiro Aota errors++; 3775a512bbf8SYan Zheng } 3776a512bbf8SYan Zheng return errors < i ? 0 : -1; 3777a512bbf8SYan Zheng } 3778a512bbf8SYan Zheng 3779387125fcSChris Mason /* 3780abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3781abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3782abbb3b8eSDavid Sterba * 3783314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3784abbb3b8eSDavid Sterba * date. 3785abbb3b8eSDavid Sterba */ 3786abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3787abbb3b8eSDavid Sterba { 3788abbb3b8eSDavid Sterba int i; 3789abbb3b8eSDavid Sterba int errors = 0; 3790b6a535faSHoward McLauchlan bool primary_failed = false; 379112659251SNaohiro Aota int ret; 3792abbb3b8eSDavid Sterba u64 bytenr; 3793abbb3b8eSDavid Sterba 3794abbb3b8eSDavid Sterba if (max_mirrors == 0) 3795abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3796abbb3b8eSDavid Sterba 3797abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3798314b6dd0SJohannes Thumshirn struct page *page; 3799314b6dd0SJohannes Thumshirn 380012659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 380112659251SNaohiro Aota if (ret == -ENOENT) { 380212659251SNaohiro Aota break; 380312659251SNaohiro Aota } else if (ret < 0) { 380412659251SNaohiro Aota errors++; 380512659251SNaohiro Aota if (i == 0) 380612659251SNaohiro Aota primary_failed = true; 380712659251SNaohiro Aota continue; 380812659251SNaohiro Aota } 3809abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3810abbb3b8eSDavid Sterba device->commit_total_bytes) 3811abbb3b8eSDavid Sterba break; 3812abbb3b8eSDavid Sterba 3813314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3814314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3815314b6dd0SJohannes Thumshirn if (!page) { 3816abbb3b8eSDavid Sterba errors++; 3817b6a535faSHoward McLauchlan if (i == 0) 3818b6a535faSHoward McLauchlan primary_failed = true; 3819abbb3b8eSDavid Sterba continue; 3820abbb3b8eSDavid Sterba } 3821314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3822314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3823314b6dd0SJohannes Thumshirn if (PageError(page)) { 3824abbb3b8eSDavid Sterba errors++; 3825b6a535faSHoward McLauchlan if (i == 0) 3826b6a535faSHoward McLauchlan primary_failed = true; 3827b6a535faSHoward McLauchlan } 3828abbb3b8eSDavid Sterba 3829314b6dd0SJohannes Thumshirn /* Drop our reference */ 3830314b6dd0SJohannes Thumshirn put_page(page); 3831abbb3b8eSDavid Sterba 3832314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3833314b6dd0SJohannes Thumshirn put_page(page); 3834abbb3b8eSDavid Sterba } 3835abbb3b8eSDavid Sterba 3836b6a535faSHoward McLauchlan /* log error, force error return */ 3837b6a535faSHoward McLauchlan if (primary_failed) { 3838b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3839b6a535faSHoward McLauchlan device->devid); 3840b6a535faSHoward McLauchlan return -1; 3841b6a535faSHoward McLauchlan } 3842b6a535faSHoward McLauchlan 3843abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3844abbb3b8eSDavid Sterba } 3845abbb3b8eSDavid Sterba 3846abbb3b8eSDavid Sterba /* 3847387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3848387125fcSChris Mason * for the barrier when it is done 3849387125fcSChris Mason */ 38504246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3851387125fcSChris Mason { 3852f9e69aa9SChristoph Hellwig bio_uninit(bio); 3853387125fcSChris Mason complete(bio->bi_private); 3854387125fcSChris Mason } 3855387125fcSChris Mason 3856387125fcSChris Mason /* 38574fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 38584fc6441aSAnand Jain * done in the waiting counterpart. 3859387125fcSChris Mason */ 38604fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3861387125fcSChris Mason { 3862f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 3863387125fcSChris Mason 3864bfd3ea94SAnand Jain device->last_flush_error = BLK_STS_OK; 3865bfd3ea94SAnand Jain 3866a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3867a91cf0ffSWang Yugui /* 3868a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 3869a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 3870a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 3871a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 3872a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 3873a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 3874a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 3875a91cf0ffSWang Yugui * non-debug builds. 3876a91cf0ffSWang Yugui */ 387708e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 38784fc6441aSAnand Jain return; 3879a91cf0ffSWang Yugui #endif 3880387125fcSChris Mason 3881f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 3882f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 3883387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3884387125fcSChris Mason init_completion(&device->flush_wait); 3885387125fcSChris Mason bio->bi_private = &device->flush_wait; 3886387125fcSChris Mason 388758ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 388858ff51f1SChristoph Hellwig submit_bio(bio); 38891c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 38904fc6441aSAnand Jain } 3891387125fcSChris Mason 38924fc6441aSAnand Jain /* 38934fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 38941b465784SAnand Jain * Return true for any error, and false otherwise. 38954fc6441aSAnand Jain */ 38961b465784SAnand Jain static bool wait_dev_flush(struct btrfs_device *device) 38974fc6441aSAnand Jain { 3898f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 38994fc6441aSAnand Jain 39007e812f20SAnand Jain if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 39011b465784SAnand Jain return false; 39024fc6441aSAnand Jain 39032980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 39044fc6441aSAnand Jain 3905bfd3ea94SAnand Jain if (bio->bi_status) { 3906bfd3ea94SAnand Jain device->last_flush_error = bio->bi_status; 3907bfd3ea94SAnand Jain btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); 39081b465784SAnand Jain return true; 3909bfd3ea94SAnand Jain } 3910bfd3ea94SAnand Jain 39111b465784SAnand Jain return false; 3912387125fcSChris Mason } 3913387125fcSChris Mason 3914387125fcSChris Mason /* 3915387125fcSChris Mason * send an empty flush down to each device in parallel, 3916387125fcSChris Mason * then wait for them 3917387125fcSChris Mason */ 3918387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3919387125fcSChris Mason { 3920387125fcSChris Mason struct list_head *head; 3921387125fcSChris Mason struct btrfs_device *dev; 39225af3e8ccSStefan Behrens int errors_wait = 0; 3923387125fcSChris Mason 39241538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3925387125fcSChris Mason /* send down all the barriers */ 3926387125fcSChris Mason head = &info->fs_devices->devices; 39271538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3928e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3929f88ba6a2SHidetoshi Seto continue; 3930cea7c8bfSAnand Jain if (!dev->bdev) 3931387125fcSChris Mason continue; 3932e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3933ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3934387125fcSChris Mason continue; 3935387125fcSChris Mason 39364fc6441aSAnand Jain write_dev_flush(dev); 3937387125fcSChris Mason } 3938387125fcSChris Mason 3939387125fcSChris Mason /* wait for all the barriers */ 39401538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3941e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3942f88ba6a2SHidetoshi Seto continue; 3943387125fcSChris Mason if (!dev->bdev) { 39445af3e8ccSStefan Behrens errors_wait++; 3945387125fcSChris Mason continue; 3946387125fcSChris Mason } 3947e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3948ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3949387125fcSChris Mason continue; 3950387125fcSChris Mason 39511b465784SAnand Jain if (wait_dev_flush(dev)) 39525af3e8ccSStefan Behrens errors_wait++; 3953387125fcSChris Mason } 3954401b41e5SAnand Jain 3955401b41e5SAnand Jain /* 3956de38a206SAnand Jain * Checks last_flush_error of disks in order to determine the device 3957de38a206SAnand Jain * state. 3958401b41e5SAnand Jain */ 3959de38a206SAnand Jain if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) 3960de38a206SAnand Jain return -EIO; 3961de38a206SAnand Jain 3962387125fcSChris Mason return 0; 3963387125fcSChris Mason } 3964387125fcSChris Mason 3965943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3966943c6e99SZhao Lei { 39678789f4feSZhao Lei int raid_type; 39688789f4feSZhao Lei int min_tolerated = INT_MAX; 3969943c6e99SZhao Lei 39708789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 39718789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 39728c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 39738789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 39748789f4feSZhao Lei tolerated_failures); 3975943c6e99SZhao Lei 39768789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 39778789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 39788789f4feSZhao Lei continue; 397941a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 39808789f4feSZhao Lei continue; 39818c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 39828789f4feSZhao Lei btrfs_raid_array[raid_type]. 39838789f4feSZhao Lei tolerated_failures); 39848789f4feSZhao Lei } 3985943c6e99SZhao Lei 39868789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3987ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 39888789f4feSZhao Lei min_tolerated = 0; 39898789f4feSZhao Lei } 39908789f4feSZhao Lei 39918789f4feSZhao Lei return min_tolerated; 3992943c6e99SZhao Lei } 3993943c6e99SZhao Lei 3994eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3995f2984462SChris Mason { 3996e5e9a520SChris Mason struct list_head *head; 3997f2984462SChris Mason struct btrfs_device *dev; 3998a061fc8dSChris Mason struct btrfs_super_block *sb; 3999f2984462SChris Mason struct btrfs_dev_item *dev_item; 4000f2984462SChris Mason int ret; 4001f2984462SChris Mason int do_barriers; 4002a236aed1SChris Mason int max_errors; 4003a236aed1SChris Mason int total_errors = 0; 4004a061fc8dSChris Mason u64 flags; 4005f2984462SChris Mason 40060b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4007fed3b381SLiu Bo 4008fed3b381SLiu Bo /* 4009fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4010fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4011fed3b381SLiu Bo * been consistent on disk. 4012fed3b381SLiu Bo */ 4013fed3b381SLiu Bo if (max_mirrors == 0) 40140b246afaSJeff Mahoney backup_super_roots(fs_info); 4015f2984462SChris Mason 40160b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4017a061fc8dSChris Mason dev_item = &sb->dev_item; 4018e5e9a520SChris Mason 40190b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 40200b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 40210b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4022387125fcSChris Mason 40235af3e8ccSStefan Behrens if (do_barriers) { 40240b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 40255af3e8ccSStefan Behrens if (ret) { 40265af3e8ccSStefan Behrens mutex_unlock( 40270b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 40280b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 40295af3e8ccSStefan Behrens "errors while submitting device barriers."); 40305af3e8ccSStefan Behrens return ret; 40315af3e8ccSStefan Behrens } 40325af3e8ccSStefan Behrens } 4033387125fcSChris Mason 40341538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4035dfe25020SChris Mason if (!dev->bdev) { 4036dfe25020SChris Mason total_errors++; 4037dfe25020SChris Mason continue; 4038dfe25020SChris Mason } 4039e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4040ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4041dfe25020SChris Mason continue; 4042dfe25020SChris Mason 40432b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4044a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4045a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 40467df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4047935e5cc9SMiao Xie dev->commit_total_bytes); 4048ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4049ce7213c7SMiao Xie dev->commit_bytes_used); 4050a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4051a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4052a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4053a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 40547239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 40557239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4056a512bbf8SYan Zheng 4057a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4058a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4059f2984462SChris Mason 406075cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 406175cb857dSQu Wenruo if (ret < 0) { 406275cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 406375cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 406475cb857dSQu Wenruo "unexpected superblock corruption detected"); 406575cb857dSQu Wenruo return -EUCLEAN; 406675cb857dSQu Wenruo } 406775cb857dSQu Wenruo 4068abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4069a236aed1SChris Mason if (ret) 4070a236aed1SChris Mason total_errors++; 4071f2984462SChris Mason } 4072a236aed1SChris Mason if (total_errors > max_errors) { 40730b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4074d397712bSChris Mason total_errors); 40750b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 407679787eaaSJeff Mahoney 40779d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 40780b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 40790b246afaSJeff Mahoney "%d errors while writing supers", 40800b246afaSJeff Mahoney total_errors); 40819d565ba4SStefan Behrens return -EIO; 4082a236aed1SChris Mason } 4083f2984462SChris Mason 4084a512bbf8SYan Zheng total_errors = 0; 40851538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4086dfe25020SChris Mason if (!dev->bdev) 4087dfe25020SChris Mason continue; 4088e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4089ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4090dfe25020SChris Mason continue; 4091dfe25020SChris Mason 4092abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4093a512bbf8SYan Zheng if (ret) 40941259ab75SChris Mason total_errors++; 4095f2984462SChris Mason } 40960b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4097a236aed1SChris Mason if (total_errors > max_errors) { 40980b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 40990b246afaSJeff Mahoney "%d errors while writing supers", 41000b246afaSJeff Mahoney total_errors); 410179787eaaSJeff Mahoney return -EIO; 4102a236aed1SChris Mason } 4103f2984462SChris Mason return 0; 4104f2984462SChris Mason } 4105f2984462SChris Mason 4106cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4107cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4108cb517eabSMiao Xie struct btrfs_root *root) 41092619ba1fSChris Mason { 41104785e24fSJosef Bacik bool drop_ref = false; 41114785e24fSJosef Bacik 4112fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4113fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4114fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4115fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 41164785e24fSJosef Bacik drop_ref = true; 4117fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 411876dda93cSYan, Zheng 411984961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4120ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 41211c1ea4f7SLiu Bo if (root->reloc_root) { 412200246528SJosef Bacik btrfs_put_root(root->reloc_root); 41231c1ea4f7SLiu Bo root->reloc_root = NULL; 41241c1ea4f7SLiu Bo } 41251c1ea4f7SLiu Bo } 41263321719eSLiu Bo 41274785e24fSJosef Bacik if (drop_ref) 412800246528SJosef Bacik btrfs_put_root(root); 41292619ba1fSChris Mason } 41302619ba1fSChris Mason 4131c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4132c146afadSYan Zheng { 4133fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4134fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4135fc7cbcd4SDavid Sterba int i = 0; 413665d33fd7SQu Wenruo int err = 0; 4137fc7cbcd4SDavid Sterba unsigned int ret = 0; 4138c146afadSYan Zheng 4139c146afadSYan Zheng while (1) { 4140fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4141fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4142fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4143fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4144fc7cbcd4SDavid Sterba if (!ret) { 4145fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4146c146afadSYan Zheng break; 414765d33fd7SQu Wenruo } 4148fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 414966b4ffd1SJosef Bacik 4150fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4151fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4152fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4153fc7cbcd4SDavid Sterba gang[i] = NULL; 415465d33fd7SQu Wenruo continue; 4155c146afadSYan Zheng } 4156fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4157fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4158fc7cbcd4SDavid Sterba } 4159fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4160fc7cbcd4SDavid Sterba 4161fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4162fc7cbcd4SDavid Sterba if (!gang[i]) 4163fc7cbcd4SDavid Sterba continue; 4164fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4165fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4166fc7cbcd4SDavid Sterba if (err) 41676989627dSJosef Bacik goto out; 4168fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4169fc7cbcd4SDavid Sterba } 4170fc7cbcd4SDavid Sterba root_objectid++; 4171c146afadSYan Zheng } 41726989627dSJosef Bacik out: 4173fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4174fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4175fc7cbcd4SDavid Sterba if (gang[i]) 4176fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 417765d33fd7SQu Wenruo } 417865d33fd7SQu Wenruo return err; 4179c146afadSYan Zheng } 4180c146afadSYan Zheng 41816bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4182c146afadSYan Zheng { 41836bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4184c146afadSYan Zheng struct btrfs_trans_handle *trans; 4185c146afadSYan Zheng 41860b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 41872ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 41880b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 41890b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4190c71bf099SYan, Zheng 4191c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 41920b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 41930b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4194c71bf099SYan, Zheng 41957a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 41963612b495STsutomu Itoh if (IS_ERR(trans)) 41973612b495STsutomu Itoh return PTR_ERR(trans); 41983a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4199c146afadSYan Zheng } 4200c146afadSYan Zheng 420136c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 420236c86a9eSQu Wenruo { 420336c86a9eSQu Wenruo struct btrfs_transaction *trans; 420436c86a9eSQu Wenruo struct btrfs_transaction *tmp; 420536c86a9eSQu Wenruo bool found = false; 420636c86a9eSQu Wenruo 420736c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 420836c86a9eSQu Wenruo return; 420936c86a9eSQu Wenruo 421036c86a9eSQu Wenruo /* 421136c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 421236c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 421336c86a9eSQu Wenruo */ 421436c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 421536c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 421636c86a9eSQu Wenruo struct extent_state *cached = NULL; 421736c86a9eSQu Wenruo u64 dirty_bytes = 0; 421836c86a9eSQu Wenruo u64 cur = 0; 421936c86a9eSQu Wenruo u64 found_start; 422036c86a9eSQu Wenruo u64 found_end; 422136c86a9eSQu Wenruo 422236c86a9eSQu Wenruo found = true; 422336c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 422436c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 422536c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 422636c86a9eSQu Wenruo cur = found_end + 1; 422736c86a9eSQu Wenruo } 422836c86a9eSQu Wenruo btrfs_warn(fs_info, 422936c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 423036c86a9eSQu Wenruo trans->transid, dirty_bytes); 423136c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 423236c86a9eSQu Wenruo 423336c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 423436c86a9eSQu Wenruo fs_info->running_transaction = NULL; 423536c86a9eSQu Wenruo list_del_init(&trans->list); 423636c86a9eSQu Wenruo 423736c86a9eSQu Wenruo btrfs_put_transaction(trans); 423836c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 423936c86a9eSQu Wenruo } 424036c86a9eSQu Wenruo ASSERT(!found); 424136c86a9eSQu Wenruo } 424236c86a9eSQu Wenruo 4243b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4244eb60ceacSChris Mason { 4245c146afadSYan Zheng int ret; 4246e089f05cSChris Mason 4247afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 424831e70e52SFilipe Manana 424931e70e52SFilipe Manana /* 42508a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 42518a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 42528a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 42538a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 42548a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 42558a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 42568a1f1e3dSFilipe Manana * return 1. 42578a1f1e3dSFilipe Manana */ 42588a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 42598a1f1e3dSFilipe Manana 42608a1f1e3dSFilipe Manana /* 426131e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 426231e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 426331e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 426431e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 426531e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 426631e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 426731e70e52SFilipe Manana * trying to stop the async reclaim task. 426831e70e52SFilipe Manana */ 426931e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4270d6fd0ae2SOmar Sandoval /* 4271d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4272d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4273d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4274d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4275d6fd0ae2SOmar Sandoval */ 4276d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4277a2135011SChris Mason 42787343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4279d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 42807343dd61SJustin Maggard 4281803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4282803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4283803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4284803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4285803b2f54SStefan Behrens 4286837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4287aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4288837d5b6eSIlya Dryomov 42898dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 42908dabb742SStefan Behrens 4291aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 42924cb5300bSChris Mason 42934cb5300bSChris Mason /* wait for any defraggers to finish */ 42944cb5300bSChris Mason wait_event(fs_info->transaction_wait, 42954cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 42964cb5300bSChris Mason 42974cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 429826176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 42994cb5300bSChris Mason 4300a362bb86SFilipe Manana /* 4301a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4302a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4303a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4304a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4305a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4306a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4307a362bb86SFilipe Manana * no one else to run iputs. 4308a362bb86SFilipe Manana * 4309a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4310a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4311a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4312a362bb86SFilipe Manana * through some other means. 4313a362bb86SFilipe Manana * 4314a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4315a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4316a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4317a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4318a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4319a362bb86SFilipe Manana */ 4320a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4321a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4322a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4323a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4324a362bb86SFilipe Manana 432521c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 432657056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4327576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 432821c7e756SMiao Xie 4329b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4330b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4331b0643e59SDennis Zhou 4332bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4333e44163e1SJeff Mahoney /* 4334d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4335d6fd0ae2SOmar Sandoval * unused block groups. 4336e44163e1SJeff Mahoney */ 43370b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4338e44163e1SJeff Mahoney 4339f0cc2cd7SFilipe Manana /* 4340f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4341f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4342f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4343f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4344f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4345f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4346f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4347f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4348f0cc2cd7SFilipe Manana * This is a very rare case. 4349f0cc2cd7SFilipe Manana */ 4350f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4351f0cc2cd7SFilipe Manana 43526bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4353d397712bSChris Mason if (ret) 435404892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4355c146afadSYan Zheng } 4356ed2ff2cbSChris Mason 435784961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 43582ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4359acce952bSliubo 4360e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4361e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 43628929ecfaSYan, Zheng 4363e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4364afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4365f25784b3SYan Zheng 43665958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 43675958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 43685958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 43695958253cSQu Wenruo } 43705958253cSQu Wenruo 437104892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4372fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4373bcef60f2SArne Jansen 4374963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 437504892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4376963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4377b0c68f8bSChris Mason } 437831153d81SYan Zheng 43795deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 43804297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 43815deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 43824297ff84SJosef Bacik 43836618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4384b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 43855ac1d209SJeff Mahoney 43861a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 43871a4319ccSLiu Bo 4388de348ee0SWang Shilong /* 4389de348ee0SWang Shilong * we must make sure there is not any read request to 4390de348ee0SWang Shilong * submit after we stopping all workers. 4391de348ee0SWang Shilong */ 4392de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 439396192499SJosef Bacik btrfs_stop_all_workers(fs_info); 439496192499SJosef Bacik 43950a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 439636c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 43970a31daa4SFilipe Manana 4398afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 43994273eaffSAnand Jain free_root_pointers(fs_info, true); 44008c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 44019ad6b7bcSChris Mason 44024e19443dSJosef Bacik /* 44034e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 44044e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 44054e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 44064e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 44074e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 44084e19443dSJosef Bacik */ 44094e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 44104e19443dSJosef Bacik 441113e6c37bSJosef Bacik iput(fs_info->btree_inode); 4412d6bfde87SChris Mason 441321adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 44140b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 44152ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 441621adbd5cSStefan Behrens #endif 441721adbd5cSStefan Behrens 44180b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 441968c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4420eb60ceacSChris Mason } 4421eb60ceacSChris Mason 44225f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 44235f39d397SChris Mason { 44242f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 44255f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4426b4ce94deSChris Mason 442706ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 442806ea65a3SJosef Bacik /* 442906ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 443052042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 443106ea65a3SJosef Bacik * outside of the sanity tests. 443206ea65a3SJosef Bacik */ 4433b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 443406ea65a3SJosef Bacik return; 443506ea65a3SJosef Bacik #endif 443649d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 44370b246afaSJeff Mahoney if (transid != fs_info->generation) 44385d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 44390b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 4440f18cc978SChristoph Hellwig set_extent_buffer_dirty(buf); 44411f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 444269fc6cbbSQu Wenruo /* 444385d8a826SJosef Bacik * btrfs_check_leaf() won't check item data if we don't have WRITTEN 444485d8a826SJosef Bacik * set, so this will only validate the basic structure of the items. 444569fc6cbbSQu Wenruo */ 444685d8a826SJosef Bacik if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(buf)) { 4447a4f78750SDavid Sterba btrfs_print_leaf(buf); 44481f21ef0aSFilipe Manana ASSERT(0); 44491f21ef0aSFilipe Manana } 44501f21ef0aSFilipe Manana #endif 4451eb60ceacSChris Mason } 4452eb60ceacSChris Mason 44532ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4454b53d3f5dSLiu Bo int flush_delayed) 445535b7e476SChris Mason { 4456188de649SChris Mason /* 4457188de649SChris Mason * looks as though older kernels can get into trouble with 4458188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4459188de649SChris Mason */ 4460e2d84521SMiao Xie int ret; 4461d6bfde87SChris Mason 44626933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4463d6bfde87SChris Mason return; 4464d6bfde87SChris Mason 4465b53d3f5dSLiu Bo if (flush_delayed) 44662ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 446716cdcec7SMiao Xie 4468d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4469d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4470d814a491SEthan Lien fs_info->dirty_metadata_batch); 4471e2d84521SMiao Xie if (ret > 0) { 44720b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 447316cdcec7SMiao Xie } 447416cdcec7SMiao Xie } 447516cdcec7SMiao Xie 44762ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 447716cdcec7SMiao Xie { 44782ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 447935b7e476SChris Mason } 4480b53d3f5dSLiu Bo 44812ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4482b53d3f5dSLiu Bo { 44832ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4484d6bfde87SChris Mason } 44856b80053dSChris Mason 44862ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4487acce952bSliubo { 4488fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4489fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4490fe816d0fSNikolay Borisov 44910b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 44922ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 44930b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4494acce952bSliubo 44950b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 44960b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4497acce952bSliubo } 4498acce952bSliubo 4499ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4500ef67963dSJosef Bacik { 4501fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4502fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4503fc7cbcd4SDavid Sterba int ret; 4504ef67963dSJosef Bacik 4505fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4506fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4507fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4508fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4509fc7cbcd4SDavid Sterba int i; 4510ef67963dSJosef Bacik 4511fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4512fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4513fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4514fc7cbcd4SDavid Sterba 4515fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4516fc7cbcd4SDavid Sterba if (!gang[i]) 4517ef67963dSJosef Bacik continue; 4518fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4519fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4520fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4521ef67963dSJosef Bacik } 4522fc7cbcd4SDavid Sterba root_objectid++; 4523fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4524ef67963dSJosef Bacik } 4525fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4526ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4527ef67963dSJosef Bacik } 4528ef67963dSJosef Bacik 4529143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4530acce952bSliubo { 4531acce952bSliubo struct btrfs_ordered_extent *ordered; 4532acce952bSliubo 4533199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4534779880efSJosef Bacik /* 4535779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4536779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4537779880efSJosef Bacik */ 4538199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4539779880efSJosef Bacik root_extent_list) 4540779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4541199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4542199c2a9cSMiao Xie } 4543199c2a9cSMiao Xie 4544199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4545199c2a9cSMiao Xie { 4546199c2a9cSMiao Xie struct btrfs_root *root; 4547199c2a9cSMiao Xie struct list_head splice; 4548199c2a9cSMiao Xie 4549199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4550199c2a9cSMiao Xie 4551199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4552199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4553199c2a9cSMiao Xie while (!list_empty(&splice)) { 4554199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4555199c2a9cSMiao Xie ordered_root); 45561de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 45571de2cfdeSJosef Bacik &fs_info->ordered_roots); 4558199c2a9cSMiao Xie 45592a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4560199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4561199c2a9cSMiao Xie 45622a85d9caSLiu Bo cond_resched(); 45632a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4564199c2a9cSMiao Xie } 4565199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 456674d5d229SJosef Bacik 456774d5d229SJosef Bacik /* 456874d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 456974d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 457074d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 457174d5d229SJosef Bacik * actually get run and error out properly. 457274d5d229SJosef Bacik */ 457374d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4574acce952bSliubo } 4575acce952bSliubo 457635a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 45772ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4578acce952bSliubo { 4579acce952bSliubo struct rb_node *node; 4580acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4581acce952bSliubo struct btrfs_delayed_ref_node *ref; 4582acce952bSliubo int ret = 0; 4583acce952bSliubo 4584acce952bSliubo delayed_refs = &trans->delayed_refs; 4585acce952bSliubo 4586acce952bSliubo spin_lock(&delayed_refs->lock); 4587d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4588cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4589b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4590acce952bSliubo return ret; 4591acce952bSliubo } 4592acce952bSliubo 45935c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4594d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 45950e0adbcfSJosef Bacik struct rb_node *n; 4596e78417d1SJosef Bacik bool pin_bytes = false; 4597acce952bSliubo 4598d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4599d7df2c79SJosef Bacik href_node); 46003069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4601b939d1abSJosef Bacik continue; 46023069bd26SJosef Bacik 4603d7df2c79SJosef Bacik spin_lock(&head->lock); 4604e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 46050e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 46060e0adbcfSJosef Bacik ref_node); 4607d7df2c79SJosef Bacik ref->in_tree = 0; 4608e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 46090e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 46101d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 46111d57ee94SWang Xiaoguang list_del(&ref->add_list); 4612d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4613d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4614d7df2c79SJosef Bacik } 461554067ae9SJosef Bacik if (head->must_insert_reserved) 4616e78417d1SJosef Bacik pin_bytes = true; 461778a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4618fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4619d7df2c79SJosef Bacik spin_unlock(&head->lock); 4620acce952bSliubo spin_unlock(&delayed_refs->lock); 4621e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4622acce952bSliubo 4623f603bb94SNikolay Borisov if (pin_bytes) { 4624f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4625f603bb94SNikolay Borisov 4626f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4627f603bb94SNikolay Borisov BUG_ON(!cache); 4628f603bb94SNikolay Borisov 4629f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4630f603bb94SNikolay Borisov spin_lock(&cache->lock); 4631f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4632f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4633f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4634f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4635f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4636f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4637f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4638f603bb94SNikolay Borisov 4639f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4640f603bb94SNikolay Borisov 4641f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4642f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4643f603bb94SNikolay Borisov } 464431890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4645d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4646acce952bSliubo cond_resched(); 4647acce952bSliubo spin_lock(&delayed_refs->lock); 4648acce952bSliubo } 464981f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4650acce952bSliubo 4651acce952bSliubo spin_unlock(&delayed_refs->lock); 4652acce952bSliubo 4653acce952bSliubo return ret; 4654acce952bSliubo } 4655acce952bSliubo 4656143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4657acce952bSliubo { 4658acce952bSliubo struct btrfs_inode *btrfs_inode; 4659acce952bSliubo struct list_head splice; 4660acce952bSliubo 4661acce952bSliubo INIT_LIST_HEAD(&splice); 4662acce952bSliubo 4663eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4664eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4665acce952bSliubo 4666acce952bSliubo while (!list_empty(&splice)) { 4667fe816d0fSNikolay Borisov struct inode *inode = NULL; 4668eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4669acce952bSliubo delalloc_inodes); 4670fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4671eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4672acce952bSliubo 4673fe816d0fSNikolay Borisov /* 4674fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4675fe816d0fSNikolay Borisov * meanwhile. 4676fe816d0fSNikolay Borisov */ 4677fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4678fe816d0fSNikolay Borisov if (inode) { 4679597441b3SJosef Bacik unsigned int nofs_flag; 4680597441b3SJosef Bacik 4681597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4682fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4683597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4684fe816d0fSNikolay Borisov iput(inode); 4685fe816d0fSNikolay Borisov } 4686eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4687acce952bSliubo } 4688eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4689eb73c1b7SMiao Xie } 4690eb73c1b7SMiao Xie 4691eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4692eb73c1b7SMiao Xie { 4693eb73c1b7SMiao Xie struct btrfs_root *root; 4694eb73c1b7SMiao Xie struct list_head splice; 4695eb73c1b7SMiao Xie 4696eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4697eb73c1b7SMiao Xie 4698eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4699eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4700eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4701eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4702eb73c1b7SMiao Xie delalloc_root); 470300246528SJosef Bacik root = btrfs_grab_root(root); 4704eb73c1b7SMiao Xie BUG_ON(!root); 4705eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4706eb73c1b7SMiao Xie 4707eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 470800246528SJosef Bacik btrfs_put_root(root); 4709eb73c1b7SMiao Xie 4710eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4711eb73c1b7SMiao Xie } 4712eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4713acce952bSliubo } 4714acce952bSliubo 47152ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4716acce952bSliubo struct extent_io_tree *dirty_pages, 4717acce952bSliubo int mark) 4718acce952bSliubo { 4719acce952bSliubo int ret; 4720acce952bSliubo struct extent_buffer *eb; 4721acce952bSliubo u64 start = 0; 4722acce952bSliubo u64 end; 4723acce952bSliubo 4724acce952bSliubo while (1) { 4725acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4726e6138876SJosef Bacik mark, NULL); 4727acce952bSliubo if (ret) 4728acce952bSliubo break; 4729acce952bSliubo 473091166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4731acce952bSliubo while (start <= end) { 47320b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 47330b246afaSJeff Mahoney start += fs_info->nodesize; 4734fd8b2b61SJosef Bacik if (!eb) 4735acce952bSliubo continue; 4736acce952bSliubo 4737c4e54a65SJosef Bacik btrfs_tree_lock(eb); 4738c4e54a65SJosef Bacik wait_on_extent_buffer_writeback(eb); 4739190a8339SJosef Bacik btrfs_clear_buffer_dirty(NULL, eb); 4740c4e54a65SJosef Bacik btrfs_tree_unlock(eb); 4741c4e54a65SJosef Bacik 4742fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4743acce952bSliubo } 4744acce952bSliubo } 4745acce952bSliubo 4746acce952bSliubo return ret; 4747acce952bSliubo } 4748acce952bSliubo 47492ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4750fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4751acce952bSliubo { 4752acce952bSliubo u64 start; 4753acce952bSliubo u64 end; 4754acce952bSliubo int ret; 4755acce952bSliubo 4756acce952bSliubo while (1) { 47570e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 47580e6ec385SFilipe Manana 4759fcd5e742SLu Fengqi /* 4760fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4761fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4762fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4763fcd5e742SLu Fengqi * the same extent range. 4764fcd5e742SLu Fengqi */ 4765fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4766acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 47670e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4768fcd5e742SLu Fengqi if (ret) { 4769fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4770acce952bSliubo break; 4771fcd5e742SLu Fengqi } 4772acce952bSliubo 47730e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 47740e6ec385SFilipe Manana free_extent_state(cached_state); 47752ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4776fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4777acce952bSliubo cond_resched(); 4778acce952bSliubo } 4779acce952bSliubo 4780acce952bSliubo return 0; 4781acce952bSliubo } 4782acce952bSliubo 478332da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4784c79a1751SLiu Bo { 4785c79a1751SLiu Bo struct inode *inode; 4786c79a1751SLiu Bo 4787c79a1751SLiu Bo inode = cache->io_ctl.inode; 4788c79a1751SLiu Bo if (inode) { 4789597441b3SJosef Bacik unsigned int nofs_flag; 4790597441b3SJosef Bacik 4791597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4792c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4793597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4794597441b3SJosef Bacik 4795c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4796c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4797c79a1751SLiu Bo iput(inode); 4798c79a1751SLiu Bo } 4799bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4800c79a1751SLiu Bo btrfs_put_block_group(cache); 4801c79a1751SLiu Bo } 4802c79a1751SLiu Bo 4803c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 48042ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4805c79a1751SLiu Bo { 480632da5386SDavid Sterba struct btrfs_block_group *cache; 4807c79a1751SLiu Bo 4808c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4809c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4810c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 481132da5386SDavid Sterba struct btrfs_block_group, 4812c79a1751SLiu Bo dirty_list); 4813c79a1751SLiu Bo 4814c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4815c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4816c79a1751SLiu Bo list_del_init(&cache->io_list); 4817c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4818c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4819c79a1751SLiu Bo } 4820c79a1751SLiu Bo 4821c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4822c79a1751SLiu Bo spin_lock(&cache->lock); 4823c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4824c79a1751SLiu Bo spin_unlock(&cache->lock); 4825c79a1751SLiu Bo 4826c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4827c79a1751SLiu Bo btrfs_put_block_group(cache); 4828ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4829c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4830c79a1751SLiu Bo } 4831c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4832c79a1751SLiu Bo 483345ae2c18SNikolay Borisov /* 483445ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 483545ae2c18SNikolay Borisov * to use it without any locking 483645ae2c18SNikolay Borisov */ 4837c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4838c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 483932da5386SDavid Sterba struct btrfs_block_group, 4840c79a1751SLiu Bo io_list); 4841c79a1751SLiu Bo 4842c79a1751SLiu Bo list_del_init(&cache->io_list); 4843c79a1751SLiu Bo spin_lock(&cache->lock); 4844c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4845c79a1751SLiu Bo spin_unlock(&cache->lock); 4846c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4847c79a1751SLiu Bo } 4848c79a1751SLiu Bo } 4849c79a1751SLiu Bo 485049b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 48512ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 485249b25e05SJeff Mahoney { 4853bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4854bbbf7243SNikolay Borisov 48552ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4856c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4857c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4858c79a1751SLiu Bo 4859bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4860bbbf7243SNikolay Borisov post_commit_list) { 4861bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4862bbbf7243SNikolay Borisov } 4863bbbf7243SNikolay Borisov 48642ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 486549b25e05SJeff Mahoney 48664a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 48670b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 486849b25e05SJeff Mahoney 48694a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 48700b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 487149b25e05SJeff Mahoney 4872ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 487367cde344SMiao Xie 48742ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 487549b25e05SJeff Mahoney EXTENT_DIRTY); 4876fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 487749b25e05SJeff Mahoney 48784a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 48794a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 488049b25e05SJeff Mahoney } 488149b25e05SJeff Mahoney 48822ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4883acce952bSliubo { 4884acce952bSliubo struct btrfs_transaction *t; 4885acce952bSliubo 48860b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4887acce952bSliubo 48880b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 48890b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 48900b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4891724e2315SJosef Bacik struct btrfs_transaction, list); 4892724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 48939b64f57dSElena Reshetova refcount_inc(&t->use_count); 48940b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 48952ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4896724e2315SJosef Bacik btrfs_put_transaction(t); 48970b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4898724e2315SJosef Bacik continue; 4899724e2315SJosef Bacik } 49000b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4901724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 49020b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4903724e2315SJosef Bacik /* 4904724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4905724e2315SJosef Bacik * handle open currently for this transaction. 4906724e2315SJosef Bacik */ 4907724e2315SJosef Bacik wait_event(t->writer_wait, 4908724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4909724e2315SJosef Bacik } else { 49100b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4911724e2315SJosef Bacik } 49122ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4913724e2315SJosef Bacik 49140b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 49150b246afaSJeff Mahoney if (t == fs_info->running_transaction) 49160b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4917724e2315SJosef Bacik list_del_init(&t->list); 49180b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4919a4abeea4SJosef Bacik 4920724e2315SJosef Bacik btrfs_put_transaction(t); 49212e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 49220b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4923724e2315SJosef Bacik } 49240b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 49250b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4926ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4927ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 49280b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4929ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 49300b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4931acce952bSliubo 4932acce952bSliubo return 0; 4933acce952bSliubo } 4934ec7d6dfdSNikolay Borisov 4935453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 4936ec7d6dfdSNikolay Borisov { 4937ec7d6dfdSNikolay Borisov struct btrfs_path *path; 4938ec7d6dfdSNikolay Borisov int ret; 4939ec7d6dfdSNikolay Borisov struct extent_buffer *l; 4940ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 4941ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 4942ec7d6dfdSNikolay Borisov int slot; 4943ec7d6dfdSNikolay Borisov 4944ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 4945ec7d6dfdSNikolay Borisov if (!path) 4946ec7d6dfdSNikolay Borisov return -ENOMEM; 4947ec7d6dfdSNikolay Borisov 4948ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 4949ec7d6dfdSNikolay Borisov search_key.type = -1; 4950ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 4951ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 4952ec7d6dfdSNikolay Borisov if (ret < 0) 4953ec7d6dfdSNikolay Borisov goto error; 4954ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 4955ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 4956ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 4957ec7d6dfdSNikolay Borisov l = path->nodes[0]; 4958ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 495923125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 496023125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 4961ec7d6dfdSNikolay Borisov } else { 496223125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 4963ec7d6dfdSNikolay Borisov } 4964ec7d6dfdSNikolay Borisov ret = 0; 4965ec7d6dfdSNikolay Borisov error: 4966ec7d6dfdSNikolay Borisov btrfs_free_path(path); 4967ec7d6dfdSNikolay Borisov return ret; 4968ec7d6dfdSNikolay Borisov } 4969ec7d6dfdSNikolay Borisov 4970543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 4971ec7d6dfdSNikolay Borisov { 4972ec7d6dfdSNikolay Borisov int ret; 4973ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 4974ec7d6dfdSNikolay Borisov 49756b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 4976ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 4977ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 4978ec7d6dfdSNikolay Borisov root->root_key.objectid); 4979ec7d6dfdSNikolay Borisov ret = -ENOSPC; 4980ec7d6dfdSNikolay Borisov goto out; 4981ec7d6dfdSNikolay Borisov } 4982ec7d6dfdSNikolay Borisov 498323125104SNikolay Borisov *objectid = root->free_objectid++; 4984ec7d6dfdSNikolay Borisov ret = 0; 4985ec7d6dfdSNikolay Borisov out: 4986ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 4987ec7d6dfdSNikolay Borisov return ret; 4988ec7d6dfdSNikolay Borisov } 4989