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 */ 113*d87e6575SChristoph Hellwig int btrfs_buffer_uptodate(struct extent_buffer *eb, u64 parent_transid, int atomic) 1141259ab75SChris Mason { 115*d87e6575SChristoph Hellwig struct inode *btree_inode = eb->pages[0]->mapping->host; 116*d87e6575SChristoph Hellwig struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1172ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 118*d87e6575SChristoph Hellwig int ret = 1; 119*d87e6575SChristoph Hellwig 120*d87e6575SChristoph Hellwig if (!extent_buffer_uptodate(eb)) 121*d87e6575SChristoph Hellwig return 0; 1221259ab75SChris Mason 1231259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 124*d87e6575SChristoph Hellwig return 1; 1251259ab75SChris Mason 126b9fab919SChris Mason if (atomic) 127b9fab919SChris Mason return -EAGAIN; 128b9fab919SChris Mason 129570eb97bSJosef Bacik lock_extent(io_tree, eb->start, eb->start + eb->len - 1, &cached_state); 130*d87e6575SChristoph Hellwig if (!extent_buffer_uptodate(eb) || 131*d87e6575SChristoph Hellwig btrfs_header_generation(eb) != parent_transid) { 13294647322SDavid Sterba btrfs_err_rl(eb->fs_info, 1338f0ed7d4SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 1348f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 13529549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 1360b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 137*d87e6575SChristoph Hellwig ret = 0; 138*d87e6575SChristoph Hellwig } 139570eb97bSJosef Bacik unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, 140e43bbe5eSDavid Sterba &cached_state); 1411259ab75SChris Mason return ret; 1421259ab75SChris Mason } 1431259ab75SChris Mason 144e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 145e7e16f48SJohannes Thumshirn { 146e7e16f48SJohannes Thumshirn switch (csum_type) { 147e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 1483951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 1493831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 150352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 151e7e16f48SJohannes Thumshirn return true; 152e7e16f48SJohannes Thumshirn default: 153e7e16f48SJohannes Thumshirn return false; 154e7e16f48SJohannes Thumshirn } 155e7e16f48SJohannes Thumshirn } 156e7e16f48SJohannes Thumshirn 157d352ac68SChris Mason /* 1581104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 1591104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 1601104a885SDavid Sterba */ 1613d17adeaSQu Wenruo int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 1623d17adeaSQu Wenruo const struct btrfs_super_block *disk_sb) 1631104a885SDavid Sterba { 16451bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 165d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 166d5178578SJohannes Thumshirn 167d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 1681104a885SDavid Sterba 1691104a885SDavid Sterba /* 1701104a885SDavid Sterba * The super_block structure does not span the whole 17151bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 17251bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 1731104a885SDavid Sterba */ 1743d17adeaSQu Wenruo crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, 175fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 1761104a885SDavid Sterba 17755fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 178e7e16f48SJohannes Thumshirn return 1; 1791104a885SDavid Sterba 180e7e16f48SJohannes Thumshirn return 0; 1811104a885SDavid Sterba } 1821104a885SDavid Sterba 183bacf60e5SChristoph Hellwig static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, 184bacf60e5SChristoph Hellwig int mirror_num) 185bacf60e5SChristoph Hellwig { 186bacf60e5SChristoph Hellwig struct btrfs_fs_info *fs_info = eb->fs_info; 187bacf60e5SChristoph Hellwig int i, num_pages = num_extent_pages(eb); 188bacf60e5SChristoph Hellwig int ret = 0; 189bacf60e5SChristoph Hellwig 190bacf60e5SChristoph Hellwig if (sb_rdonly(fs_info->sb)) 191bacf60e5SChristoph Hellwig return -EROFS; 192bacf60e5SChristoph Hellwig 193bacf60e5SChristoph Hellwig for (i = 0; i < num_pages; i++) { 194bacf60e5SChristoph Hellwig struct page *p = eb->pages[i]; 195917ac778SQu Wenruo u64 start = max_t(u64, eb->start, page_offset(p)); 196917ac778SQu Wenruo u64 end = min_t(u64, eb->start + eb->len, page_offset(p) + PAGE_SIZE); 197917ac778SQu Wenruo u32 len = end - start; 198bacf60e5SChristoph Hellwig 199917ac778SQu Wenruo ret = btrfs_repair_io_failure(fs_info, 0, start, len, 200917ac778SQu Wenruo start, p, offset_in_page(start), mirror_num); 201bacf60e5SChristoph Hellwig if (ret) 202bacf60e5SChristoph Hellwig break; 203bacf60e5SChristoph Hellwig } 204bacf60e5SChristoph Hellwig 205bacf60e5SChristoph Hellwig return ret; 206bacf60e5SChristoph Hellwig } 207bacf60e5SChristoph Hellwig 2081104a885SDavid Sterba /* 209d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 210d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 211581c1760SQu Wenruo * 212789d6a3aSQu Wenruo * @check: expected tree parentness check, see the comments of the 213789d6a3aSQu Wenruo * structure for details. 214d352ac68SChris Mason */ 2156a2e9dc4SFilipe Manana int btrfs_read_extent_buffer(struct extent_buffer *eb, 216789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 217f188591eSChris Mason { 2185ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 219ea466794SJosef Bacik int failed = 0; 220f188591eSChris Mason int ret; 221f188591eSChris Mason int num_copies = 0; 222f188591eSChris Mason int mirror_num = 0; 223ea466794SJosef Bacik int failed_mirror = 0; 224f188591eSChris Mason 225789d6a3aSQu Wenruo ASSERT(check); 226789d6a3aSQu Wenruo 227f188591eSChris Mason while (1) { 228f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 229947a6299SQu Wenruo ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num, check); 230947a6299SQu Wenruo if (!ret) 231581c1760SQu Wenruo break; 232d397712bSChris Mason 2330b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 234f188591eSChris Mason eb->start, eb->len); 2354235298eSChris Mason if (num_copies == 1) 236ea466794SJosef Bacik break; 2374235298eSChris Mason 2385cf1ab56SJosef Bacik if (!failed_mirror) { 2395cf1ab56SJosef Bacik failed = 1; 2405cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 2415cf1ab56SJosef Bacik } 2425cf1ab56SJosef Bacik 243f188591eSChris Mason mirror_num++; 244ea466794SJosef Bacik if (mirror_num == failed_mirror) 245ea466794SJosef Bacik mirror_num++; 246ea466794SJosef Bacik 2474235298eSChris Mason if (mirror_num > num_copies) 248ea466794SJosef Bacik break; 249f188591eSChris Mason } 250ea466794SJosef Bacik 251c0901581SStefan Behrens if (failed && !ret && failed_mirror) 25220a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 253ea466794SJosef Bacik 254ea466794SJosef Bacik return ret; 255f188591eSChris Mason } 25619c00ddcSChris Mason 257eca0f6f6SQu Wenruo static int csum_one_extent_buffer(struct extent_buffer *eb) 258eca0f6f6SQu Wenruo { 259eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 260eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 261eca0f6f6SQu Wenruo int ret; 262eca0f6f6SQu Wenruo 263eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 264eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 265eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 266eca0f6f6SQu Wenruo csum_tree_block(eb, result); 267eca0f6f6SQu Wenruo 268eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 269eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 270eca0f6f6SQu Wenruo else 27185d8a826SJosef Bacik ret = btrfs_check_leaf(eb); 272eca0f6f6SQu Wenruo 2733777369fSQu Wenruo if (ret < 0) 2743777369fSQu Wenruo goto error; 2753777369fSQu Wenruo 2763777369fSQu Wenruo /* 2773777369fSQu Wenruo * Also check the generation, the eb reached here must be newer than 2783777369fSQu Wenruo * last committed. Or something seriously wrong happened. 2793777369fSQu Wenruo */ 2803777369fSQu Wenruo if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { 2813777369fSQu Wenruo ret = -EUCLEAN; 282eca0f6f6SQu Wenruo btrfs_err(fs_info, 2833777369fSQu Wenruo "block=%llu bad generation, have %llu expect > %llu", 2843777369fSQu Wenruo eb->start, btrfs_header_generation(eb), 2853777369fSQu Wenruo fs_info->last_trans_committed); 2863777369fSQu Wenruo goto error; 287eca0f6f6SQu Wenruo } 288eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 289eca0f6f6SQu Wenruo 290eca0f6f6SQu Wenruo return 0; 2913777369fSQu Wenruo 2923777369fSQu Wenruo error: 2933777369fSQu Wenruo btrfs_print_tree(eb, 0); 2943777369fSQu Wenruo btrfs_err(fs_info, "block=%llu write time tree block corruption detected", 2953777369fSQu Wenruo eb->start); 29616199ad9SFilipe Manana /* 29716199ad9SFilipe Manana * Be noisy if this is an extent buffer from a log tree. We don't abort 29816199ad9SFilipe Manana * a transaction in case there's a bad log tree extent buffer, we just 29916199ad9SFilipe Manana * fallback to a transaction commit. Still we want to know when there is 30016199ad9SFilipe Manana * a bad log tree extent buffer, as that may signal a bug somewhere. 30116199ad9SFilipe Manana */ 30216199ad9SFilipe Manana WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG) || 30316199ad9SFilipe Manana btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID); 3043777369fSQu Wenruo return ret; 305eca0f6f6SQu Wenruo } 306eca0f6f6SQu Wenruo 307eca0f6f6SQu Wenruo /* Checksum all dirty extent buffers in one bio_vec */ 308eca0f6f6SQu Wenruo static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, 309eca0f6f6SQu Wenruo struct bio_vec *bvec) 310eca0f6f6SQu Wenruo { 311eca0f6f6SQu Wenruo struct page *page = bvec->bv_page; 312eca0f6f6SQu Wenruo u64 bvec_start = page_offset(page) + bvec->bv_offset; 313eca0f6f6SQu Wenruo u64 cur; 314eca0f6f6SQu Wenruo int ret = 0; 315eca0f6f6SQu Wenruo 316eca0f6f6SQu Wenruo for (cur = bvec_start; cur < bvec_start + bvec->bv_len; 317eca0f6f6SQu Wenruo cur += fs_info->nodesize) { 318eca0f6f6SQu Wenruo struct extent_buffer *eb; 319eca0f6f6SQu Wenruo bool uptodate; 320eca0f6f6SQu Wenruo 321eca0f6f6SQu Wenruo eb = find_extent_buffer(fs_info, cur); 322eca0f6f6SQu Wenruo uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, 323eca0f6f6SQu Wenruo fs_info->nodesize); 324eca0f6f6SQu Wenruo 32501cd3909SDavid Sterba /* A dirty eb shouldn't disappear from buffer_radix */ 326eca0f6f6SQu Wenruo if (WARN_ON(!eb)) 327eca0f6f6SQu Wenruo return -EUCLEAN; 328eca0f6f6SQu Wenruo 329eca0f6f6SQu Wenruo if (WARN_ON(cur != btrfs_header_bytenr(eb))) { 330eca0f6f6SQu Wenruo free_extent_buffer(eb); 331eca0f6f6SQu Wenruo return -EUCLEAN; 332eca0f6f6SQu Wenruo } 333eca0f6f6SQu Wenruo if (WARN_ON(!uptodate)) { 334eca0f6f6SQu Wenruo free_extent_buffer(eb); 335eca0f6f6SQu Wenruo return -EUCLEAN; 336eca0f6f6SQu Wenruo } 337eca0f6f6SQu Wenruo 338eca0f6f6SQu Wenruo ret = csum_one_extent_buffer(eb); 339eca0f6f6SQu Wenruo free_extent_buffer(eb); 340eca0f6f6SQu Wenruo if (ret < 0) 341eca0f6f6SQu Wenruo return ret; 342eca0f6f6SQu Wenruo } 343eca0f6f6SQu Wenruo return ret; 344eca0f6f6SQu Wenruo } 345eca0f6f6SQu Wenruo 346d352ac68SChris Mason /* 347ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 348ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 349ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 350d352ac68SChris Mason */ 351ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 35219c00ddcSChris Mason { 353ac303b69SQu Wenruo struct page *page = bvec->bv_page; 3544eee4fa4SMiao Xie u64 start = page_offset(page); 35519c00ddcSChris Mason u64 found_start; 35619c00ddcSChris Mason struct extent_buffer *eb; 357eca0f6f6SQu Wenruo 358fbca46ebSQu Wenruo if (fs_info->nodesize < PAGE_SIZE) 359eca0f6f6SQu Wenruo return csum_dirty_subpage_buffers(fs_info, bvec); 360f188591eSChris Mason 3614f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 3624f2de97aSJosef Bacik if (page != eb->pages[0]) 3634f2de97aSJosef Bacik return 0; 3640f805531SAlex Lyakas 36519c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 366d3575156SNaohiro Aota 367d3575156SNaohiro Aota if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 368d3575156SNaohiro Aota WARN_ON(found_start != 0); 369d3575156SNaohiro Aota return 0; 370d3575156SNaohiro Aota } 371d3575156SNaohiro Aota 3720f805531SAlex Lyakas /* 3730f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 3740f805531SAlex Lyakas * It is useful to know what went wrong. 3750f805531SAlex Lyakas */ 3760f805531SAlex Lyakas if (WARN_ON(found_start != start)) 3770f805531SAlex Lyakas return -EUCLEAN; 3780f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 3790f805531SAlex Lyakas return -EUCLEAN; 3800f805531SAlex Lyakas 381eca0f6f6SQu Wenruo return csum_one_extent_buffer(eb); 38219c00ddcSChris Mason } 38319c00ddcSChris Mason 384542e300eSChristoph Hellwig blk_status_t btree_csum_one_bio(struct btrfs_bio *bbio) 385deb6216fSChristoph Hellwig { 386542e300eSChristoph Hellwig struct btrfs_fs_info *fs_info = bbio->inode->root->fs_info; 387542e300eSChristoph Hellwig struct bvec_iter iter; 388542e300eSChristoph Hellwig struct bio_vec bv; 389deb6216fSChristoph Hellwig int ret = 0; 390deb6216fSChristoph Hellwig 391542e300eSChristoph Hellwig bio_for_each_segment(bv, &bbio->bio, iter) { 392542e300eSChristoph Hellwig ret = csum_dirty_buffer(fs_info, &bv); 393deb6216fSChristoph Hellwig if (ret) 394deb6216fSChristoph Hellwig break; 395deb6216fSChristoph Hellwig } 396deb6216fSChristoph Hellwig 397deb6216fSChristoph Hellwig return errno_to_blk_status(ret); 398deb6216fSChristoph Hellwig } 399deb6216fSChristoph Hellwig 400b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 4012b82032cSYan Zheng { 402b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 403944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 40444880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 405944d3f9fSNikolay Borisov u8 *metadata_uuid; 4062b82032cSYan Zheng 4079a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 4089a8658e3SDavid Sterba BTRFS_FSID_SIZE); 4097239ff4bSNikolay Borisov /* 410944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 411944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 412944d3f9fSNikolay Borisov * ->fsid in this case is fine 4137239ff4bSNikolay Borisov */ 414944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 4157239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 4167239ff4bSNikolay Borisov else 4177239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 4187239ff4bSNikolay Borisov 419944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 420944d3f9fSNikolay Borisov return 0; 421944d3f9fSNikolay Borisov 422944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 423944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 424944d3f9fSNikolay Borisov return 0; 425944d3f9fSNikolay Borisov 426944d3f9fSNikolay Borisov return 1; 4272b82032cSYan Zheng } 4282b82032cSYan Zheng 42977bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 430947a6299SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb, 431947a6299SQu Wenruo struct btrfs_tree_parent_check *check) 432ce9adaa5SChris Mason { 43377bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 434ce9adaa5SChris Mason u64 found_start; 43577bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 43677bf40a2SQu Wenruo u8 found_level; 4372996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 438dfd29eedSDavid Sterba const u8 *header_csum; 43977bf40a2SQu Wenruo int ret = 0; 440ea466794SJosef Bacik 441947a6299SQu Wenruo ASSERT(check); 442947a6299SQu Wenruo 443ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 444727011e0SChris Mason if (found_start != eb->start) { 4458f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, 4468f0ed7d4SQu Wenruo "bad tree block start, mirror %u want %llu have %llu", 4478f0ed7d4SQu Wenruo eb->read_mirror, eb->start, found_start); 448f188591eSChris Mason ret = -EIO; 44977bf40a2SQu Wenruo goto out; 450ce9adaa5SChris Mason } 451b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 4528f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", 4538f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 4541259ab75SChris Mason ret = -EIO; 45577bf40a2SQu Wenruo goto out; 4561259ab75SChris Mason } 457ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 4581c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 4598f0ed7d4SQu Wenruo btrfs_err(fs_info, 4608f0ed7d4SQu Wenruo "bad tree block level, mirror %u level %d on logical %llu", 4618f0ed7d4SQu Wenruo eb->read_mirror, btrfs_header_level(eb), eb->start); 4621c24c3ceSJosef Bacik ret = -EIO; 46377bf40a2SQu Wenruo goto out; 4641c24c3ceSJosef Bacik } 465ce9adaa5SChris Mason 466c67b3892SDavid Sterba csum_tree_block(eb, result); 467dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 468dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 469a826d6dcSJosef Bacik 470dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 4712996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 4728f0ed7d4SQu Wenruo "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", 4738f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 474dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 47535be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 47635be8851SJohannes Thumshirn btrfs_header_level(eb)); 4772996e1f8SJohannes Thumshirn ret = -EUCLEAN; 47877bf40a2SQu Wenruo goto out; 4792996e1f8SJohannes Thumshirn } 4802996e1f8SJohannes Thumshirn 481947a6299SQu Wenruo if (found_level != check->level) { 48277177ed1SQu Wenruo btrfs_err(fs_info, 48377177ed1SQu Wenruo "level verify failed on logical %llu mirror %u wanted %u found %u", 48477177ed1SQu Wenruo eb->start, eb->read_mirror, check->level, found_level); 485947a6299SQu Wenruo ret = -EIO; 486947a6299SQu Wenruo goto out; 487947a6299SQu Wenruo } 488947a6299SQu Wenruo if (unlikely(check->transid && 489947a6299SQu Wenruo btrfs_header_generation(eb) != check->transid)) { 490947a6299SQu Wenruo btrfs_err_rl(eb->fs_info, 491947a6299SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 492947a6299SQu Wenruo eb->start, eb->read_mirror, check->transid, 493947a6299SQu Wenruo btrfs_header_generation(eb)); 494947a6299SQu Wenruo ret = -EIO; 495947a6299SQu Wenruo goto out; 496947a6299SQu Wenruo } 497947a6299SQu Wenruo if (check->has_first_key) { 498947a6299SQu Wenruo struct btrfs_key *expect_key = &check->first_key; 499947a6299SQu Wenruo struct btrfs_key found_key; 500947a6299SQu Wenruo 501947a6299SQu Wenruo if (found_level) 502947a6299SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 503947a6299SQu Wenruo else 504947a6299SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 505947a6299SQu Wenruo if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { 506947a6299SQu Wenruo btrfs_err(fs_info, 507947a6299SQu Wenruo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 508947a6299SQu Wenruo eb->start, check->transid, 509947a6299SQu Wenruo expect_key->objectid, 510947a6299SQu Wenruo expect_key->type, expect_key->offset, 511947a6299SQu Wenruo found_key.objectid, found_key.type, 512947a6299SQu Wenruo found_key.offset); 513947a6299SQu Wenruo ret = -EUCLEAN; 514947a6299SQu Wenruo goto out; 515947a6299SQu Wenruo } 516947a6299SQu Wenruo } 517947a6299SQu Wenruo if (check->owner_root) { 518947a6299SQu Wenruo ret = btrfs_check_eb_owner(eb, check->owner_root); 519947a6299SQu Wenruo if (ret < 0) 520947a6299SQu Wenruo goto out; 521947a6299SQu Wenruo } 522947a6299SQu Wenruo 523a826d6dcSJosef Bacik /* 524a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 525a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 526a826d6dcSJosef Bacik * return -EIO. 527a826d6dcSJosef Bacik */ 52885d8a826SJosef Bacik if (found_level == 0 && btrfs_check_leaf(eb)) { 529a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 530a826d6dcSJosef Bacik ret = -EIO; 531a826d6dcSJosef Bacik } 532ce9adaa5SChris Mason 533813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 534053ab70fSLiu Bo ret = -EIO; 535053ab70fSLiu Bo 536aebcc159SChristoph Hellwig if (ret) 53775391f0dSQu Wenruo btrfs_err(fs_info, 5388f0ed7d4SQu Wenruo "read time tree block corruption detected on logical %llu mirror %u", 5398f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 54077bf40a2SQu Wenruo out: 54177bf40a2SQu Wenruo return ret; 54277bf40a2SQu Wenruo } 54377bf40a2SQu Wenruo 544371cdc07SQu Wenruo static int validate_subpage_buffer(struct page *page, u64 start, u64 end, 545947a6299SQu Wenruo int mirror, struct btrfs_tree_parent_check *check) 546371cdc07SQu Wenruo { 547371cdc07SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 548371cdc07SQu Wenruo struct extent_buffer *eb; 549371cdc07SQu Wenruo bool reads_done; 550371cdc07SQu Wenruo int ret = 0; 551371cdc07SQu Wenruo 552947a6299SQu Wenruo ASSERT(check); 553947a6299SQu Wenruo 554371cdc07SQu Wenruo /* 555371cdc07SQu Wenruo * We don't allow bio merge for subpage metadata read, so we should 556371cdc07SQu Wenruo * only get one eb for each endio hook. 557371cdc07SQu Wenruo */ 558371cdc07SQu Wenruo ASSERT(end == start + fs_info->nodesize - 1); 559371cdc07SQu Wenruo ASSERT(PagePrivate(page)); 560371cdc07SQu Wenruo 561371cdc07SQu Wenruo eb = find_extent_buffer(fs_info, start); 562371cdc07SQu Wenruo /* 563371cdc07SQu Wenruo * When we are reading one tree block, eb must have been inserted into 564371cdc07SQu Wenruo * the radix tree. If not, something is wrong. 565371cdc07SQu Wenruo */ 566371cdc07SQu Wenruo ASSERT(eb); 567371cdc07SQu Wenruo 568371cdc07SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 569371cdc07SQu Wenruo /* Subpage read must finish in page read */ 570371cdc07SQu Wenruo ASSERT(reads_done); 571371cdc07SQu Wenruo 572371cdc07SQu Wenruo eb->read_mirror = mirror; 573371cdc07SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 574371cdc07SQu Wenruo ret = -EIO; 575371cdc07SQu Wenruo goto err; 576371cdc07SQu Wenruo } 577947a6299SQu Wenruo ret = validate_extent_buffer(eb, check); 578371cdc07SQu Wenruo if (ret < 0) 579371cdc07SQu Wenruo goto err; 580371cdc07SQu Wenruo 581371cdc07SQu Wenruo set_extent_buffer_uptodate(eb); 582371cdc07SQu Wenruo 583371cdc07SQu Wenruo free_extent_buffer(eb); 584371cdc07SQu Wenruo return ret; 585371cdc07SQu Wenruo err: 586371cdc07SQu Wenruo /* 587371cdc07SQu Wenruo * end_bio_extent_readpage decrements io_pages in case of error, 588371cdc07SQu Wenruo * make sure it has something to decrement. 589371cdc07SQu Wenruo */ 590371cdc07SQu Wenruo atomic_inc(&eb->io_pages); 591371cdc07SQu Wenruo clear_extent_buffer_uptodate(eb); 592371cdc07SQu Wenruo free_extent_buffer(eb); 593371cdc07SQu Wenruo return ret; 594371cdc07SQu Wenruo } 595371cdc07SQu Wenruo 596c3a3b19bSQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, 59777bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 59877bf40a2SQu Wenruo int mirror) 59977bf40a2SQu Wenruo { 60077bf40a2SQu Wenruo struct extent_buffer *eb; 60177bf40a2SQu Wenruo int ret = 0; 60277bf40a2SQu Wenruo int reads_done; 60377bf40a2SQu Wenruo 60477bf40a2SQu Wenruo ASSERT(page->private); 605371cdc07SQu Wenruo 606fbca46ebSQu Wenruo if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) 607947a6299SQu Wenruo return validate_subpage_buffer(page, start, end, mirror, 608947a6299SQu Wenruo &bbio->parent_check); 609371cdc07SQu Wenruo 61077bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 61177bf40a2SQu Wenruo 61277bf40a2SQu Wenruo /* 61377bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 61477bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 61577bf40a2SQu Wenruo */ 61677bf40a2SQu Wenruo atomic_inc(&eb->refs); 61777bf40a2SQu Wenruo 61877bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 61977bf40a2SQu Wenruo if (!reads_done) 62077bf40a2SQu Wenruo goto err; 62177bf40a2SQu Wenruo 62277bf40a2SQu Wenruo eb->read_mirror = mirror; 62377bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 62477bf40a2SQu Wenruo ret = -EIO; 62577bf40a2SQu Wenruo goto err; 62677bf40a2SQu Wenruo } 627947a6299SQu Wenruo ret = validate_extent_buffer(eb, &bbio->parent_check); 628aebcc159SChristoph Hellwig if (!ret) 629aebcc159SChristoph Hellwig set_extent_buffer_uptodate(eb); 630ce9adaa5SChris Mason err: 63153b381b3SDavid Woodhouse if (ret) { 63253b381b3SDavid Woodhouse /* 63353b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 63453b381b3SDavid Woodhouse * again, we have to make sure it has something 63553b381b3SDavid Woodhouse * to decrement 63653b381b3SDavid Woodhouse */ 63753b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6380b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 63953b381b3SDavid Woodhouse } 6400b32f4bbSJosef Bacik free_extent_buffer(eb); 64177bf40a2SQu Wenruo 642f188591eSChris Mason return ret; 643ce9adaa5SChris Mason } 644ce9adaa5SChris Mason 6453dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 6468958b551SMatthew Wilcox (Oracle) static int btree_migrate_folio(struct address_space *mapping, 6478958b551SMatthew Wilcox (Oracle) struct folio *dst, struct folio *src, enum migrate_mode mode) 648784b4e29SChris Mason { 649784b4e29SChris Mason /* 650784b4e29SChris Mason * we can't safely write a btree page from here, 651784b4e29SChris Mason * we haven't done the locking hook 652784b4e29SChris Mason */ 6538958b551SMatthew Wilcox (Oracle) if (folio_test_dirty(src)) 654784b4e29SChris Mason return -EAGAIN; 655784b4e29SChris Mason /* 656784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 657784b4e29SChris Mason * We must have no buffers or drop them. 658784b4e29SChris Mason */ 6598958b551SMatthew Wilcox (Oracle) if (folio_get_private(src) && 6608958b551SMatthew Wilcox (Oracle) !filemap_release_folio(src, GFP_KERNEL)) 661784b4e29SChris Mason return -EAGAIN; 66254184650SMatthew Wilcox (Oracle) return migrate_folio(mapping, dst, src, mode); 663784b4e29SChris Mason } 6648958b551SMatthew Wilcox (Oracle) #else 6658958b551SMatthew Wilcox (Oracle) #define btree_migrate_folio NULL 6663dd1462eSJan Beulich #endif 667784b4e29SChris Mason 6680da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 6690da5468fSChris Mason struct writeback_control *wbc) 6700da5468fSChris Mason { 671e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 672e2d84521SMiao Xie int ret; 673e2d84521SMiao Xie 674d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 675448d640bSChris Mason 676448d640bSChris Mason if (wbc->for_kupdate) 677448d640bSChris Mason return 0; 678448d640bSChris Mason 679e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 680b9473439SChris Mason /* this is a bit racy, but that's ok */ 681d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 682d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 683d814a491SEthan Lien fs_info->dirty_metadata_batch); 684e2d84521SMiao Xie if (ret < 0) 685793955bcSChris Mason return 0; 686793955bcSChris Mason } 6870b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 6880da5468fSChris Mason } 6890da5468fSChris Mason 690f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 6915f39d397SChris Mason { 692f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 693f913cff3SMatthew Wilcox (Oracle) return false; 6940c4e538bSDavid Sterba 695f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 696d98237b3SChris Mason } 697d98237b3SChris Mason 698895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 699895586ebSMatthew Wilcox (Oracle) size_t length) 700d98237b3SChris Mason { 701d1310b2eSChris Mason struct extent_io_tree *tree; 702895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 703895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 704f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 705895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 706895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 707895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 708895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 709895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 7109ad6b7bcSChris Mason } 711d98237b3SChris Mason } 712d98237b3SChris Mason 713bb146eb2SJosef Bacik #ifdef DEBUG 7140079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 7150079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 7160079c3b1SMatthew Wilcox (Oracle) { 7170079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 718139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 7190b32f4bbSJosef Bacik struct extent_buffer *eb; 720139e8cd3SQu Wenruo int cur_bit = 0; 7210079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 7220b32f4bbSJosef Bacik 723139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 7240079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 7250b32f4bbSJosef Bacik BUG_ON(!eb); 7260b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 7270b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 72849d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 7290079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 730139e8cd3SQu Wenruo } 7310079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 732139e8cd3SQu Wenruo 733139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 734139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 735139e8cd3SQu Wenruo unsigned long flags; 736139e8cd3SQu Wenruo u64 cur; 737139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 738139e8cd3SQu Wenruo 739139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 740139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 741139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 742139e8cd3SQu Wenruo cur_bit++; 743139e8cd3SQu Wenruo continue; 744139e8cd3SQu Wenruo } 745139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 746139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 747139e8cd3SQu Wenruo 748139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 749139e8cd3SQu Wenruo ASSERT(eb); 750139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 751139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 75249d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 753139e8cd3SQu Wenruo free_extent_buffer(eb); 754139e8cd3SQu Wenruo 755139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 756139e8cd3SQu Wenruo } 7570079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 7580b32f4bbSJosef Bacik } 7590079c3b1SMatthew Wilcox (Oracle) #else 7600079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 7610079c3b1SMatthew Wilcox (Oracle) #endif 7620b32f4bbSJosef Bacik 7637f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 7640da5468fSChris Mason .writepages = btree_writepages, 765f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 766895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 7678958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 7680079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 769d98237b3SChris Mason }; 770123abc88SChris Mason 7712ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 7722ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 7733fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 7743fbaf258SJosef Bacik int level) 7750999df54SChris Mason { 7760b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 7770b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 7783fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 7790999df54SChris Mason } 7800999df54SChris Mason 781581c1760SQu Wenruo /* 782581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 783581c1760SQu Wenruo * verification. 784581c1760SQu Wenruo * 785789d6a3aSQu Wenruo * @check: expected tree parentness check, see comments of the 786789d6a3aSQu Wenruo * structure for details. 787581c1760SQu Wenruo */ 7882ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 789789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 790e20d96d6SChris Mason { 7915f39d397SChris Mason struct extent_buffer *buf = NULL; 79219c00ddcSChris Mason int ret; 79319c00ddcSChris Mason 794789d6a3aSQu Wenruo ASSERT(check); 795789d6a3aSQu Wenruo 796789d6a3aSQu Wenruo buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, 797789d6a3aSQu Wenruo check->level); 798c871b0f2SLiu Bo if (IS_ERR(buf)) 799c871b0f2SLiu Bo return buf; 800e4204dedSChris Mason 801789d6a3aSQu Wenruo ret = btrfs_read_extent_buffer(buf, check); 8020f0fe8f7SFilipe David Borba Manana if (ret) { 803537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 80464c043deSLiu Bo return ERR_PTR(ret); 8050f0fe8f7SFilipe David Borba Manana } 806789d6a3aSQu Wenruo if (btrfs_check_eb_owner(buf, check->owner_root)) { 80788c602abSQu Wenruo free_extent_buffer_stale(buf); 80888c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 80988c602abSQu Wenruo } 8105f39d397SChris Mason return buf; 811ce9adaa5SChris Mason 812eb60ceacSChris Mason } 813eb60ceacSChris Mason 814da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 815e20d96d6SChris Mason u64 objectid) 816d97e63b6SChris Mason { 8177c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 8182e608bd1SJosef Bacik 8192e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 8202e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 8212e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 82296dfcb46SJosef Bacik root->fs_info = fs_info; 8232e608bd1SJosef Bacik root->root_key.objectid = objectid; 824cfaa7295SChris Mason root->node = NULL; 825a28ec197SChris Mason root->commit_root = NULL; 82627cdeb70SMiao Xie root->state = 0; 827abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 8280b86a832SChris Mason 8290f7d52f4SChris Mason root->last_trans = 0; 8306b8fad57SNikolay Borisov root->free_objectid = 0; 831eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 832199c2a9cSMiao Xie root->nr_ordered_extents = 0; 8336bef4d31SEric Paris root->inode_tree = RB_ROOT; 834088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 8352e608bd1SJosef Bacik 8362e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 8370b86a832SChris Mason 8380b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 8395d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 840eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 841eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 842199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 843199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 844d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 8452ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 8462ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 8475d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 848eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 849199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 850f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 8512ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 8522ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 8538287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 854a2135011SChris Mason mutex_init(&root->objectid_mutex); 855e02119d5SChris Mason mutex_init(&root->log_mutex); 85631f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 857573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 858c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 8597237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 8607237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 8617237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 8628b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 8638b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 8647237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 8657237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 8667237f183SYan Zheng atomic_set(&root->log_writers, 0); 8672ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 8680700cea7SElena Reshetova refcount_set(&root->refs, 1); 8698ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 870eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 8717237f183SYan Zheng root->log_transid = 0; 872d1433debSMiao Xie root->log_transid_committed = -1; 873257c62e1SChris Mason root->last_log_commit = 0; 8742e608bd1SJosef Bacik root->anon_dev = 0; 875e289f03eSFilipe Manana if (!dummy) { 87643eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 87735da5a7eSDavid Sterba IO_TREE_ROOT_DIRTY_LOG_PAGES); 878e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 87935da5a7eSDavid Sterba IO_TREE_LOG_CSUM_RANGE); 880e289f03eSFilipe Manana } 881017e5369SChris Mason 8825f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 883370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 884bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 885bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 886fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 887bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 888fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 889bd647ce3SJosef Bacik #endif 8903768f368SChris Mason } 8913768f368SChris Mason 89274e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 89396dfcb46SJosef Bacik u64 objectid, gfp_t flags) 8946f07e42eSAl Viro { 89574e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 8966f07e42eSAl Viro if (root) 89796dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 8986f07e42eSAl Viro return root; 8996f07e42eSAl Viro } 9006f07e42eSAl Viro 90106ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 90206ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 903da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 90406ea65a3SJosef Bacik { 90506ea65a3SJosef Bacik struct btrfs_root *root; 90606ea65a3SJosef Bacik 9077c0260eeSJeff Mahoney if (!fs_info) 9087c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 9097c0260eeSJeff Mahoney 91096dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 91106ea65a3SJosef Bacik if (!root) 91206ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 913da17066cSJeff Mahoney 914b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 915faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 91606ea65a3SJosef Bacik 91706ea65a3SJosef Bacik return root; 91806ea65a3SJosef Bacik } 91906ea65a3SJosef Bacik #endif 92006ea65a3SJosef Bacik 921abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 922abed4aaaSJosef Bacik { 923abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 924abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 925abed4aaaSJosef Bacik 926abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 927abed4aaaSJosef Bacik } 928abed4aaaSJosef Bacik 929abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 930abed4aaaSJosef Bacik { 931abed4aaaSJosef Bacik const struct btrfs_key *key = k; 932abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 933abed4aaaSJosef Bacik 934abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 935abed4aaaSJosef Bacik } 936abed4aaaSJosef Bacik 937abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 938abed4aaaSJosef Bacik { 939abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 940abed4aaaSJosef Bacik struct rb_node *tmp; 941745806fbSQu Wenruo int ret = 0; 942abed4aaaSJosef Bacik 943abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 944abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 945abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 946abed4aaaSJosef Bacik 947745806fbSQu Wenruo if (tmp) { 948745806fbSQu Wenruo ret = -EEXIST; 949745806fbSQu Wenruo btrfs_warn(fs_info, "global root %llu %llu already exists", 950745806fbSQu Wenruo root->root_key.objectid, root->root_key.offset); 951745806fbSQu Wenruo } 952745806fbSQu Wenruo return ret; 953abed4aaaSJosef Bacik } 954abed4aaaSJosef Bacik 955abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 956abed4aaaSJosef Bacik { 957abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 958abed4aaaSJosef Bacik 959abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 960abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 961abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 962abed4aaaSJosef Bacik } 963abed4aaaSJosef Bacik 964abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 965abed4aaaSJosef Bacik struct btrfs_key *key) 966abed4aaaSJosef Bacik { 967abed4aaaSJosef Bacik struct rb_node *node; 968abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 969abed4aaaSJosef Bacik 970abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 971abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 972abed4aaaSJosef Bacik if (node) 973abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 974abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 975abed4aaaSJosef Bacik 976abed4aaaSJosef Bacik return root; 977abed4aaaSJosef Bacik } 978abed4aaaSJosef Bacik 979f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 980f7238e50SJosef Bacik { 981f7238e50SJosef Bacik struct btrfs_block_group *block_group; 982f7238e50SJosef Bacik u64 ret; 983f7238e50SJosef Bacik 984f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 985f7238e50SJosef Bacik return 0; 986f7238e50SJosef Bacik 987f7238e50SJosef Bacik if (bytenr) 988f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 989f7238e50SJosef Bacik else 990f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 991f7238e50SJosef Bacik ASSERT(block_group); 992f7238e50SJosef Bacik if (!block_group) 993f7238e50SJosef Bacik return 0; 994f7238e50SJosef Bacik ret = block_group->global_root_id; 995f7238e50SJosef Bacik btrfs_put_block_group(block_group); 996f7238e50SJosef Bacik 997f7238e50SJosef Bacik return ret; 998f7238e50SJosef Bacik } 999f7238e50SJosef Bacik 1000abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1001abed4aaaSJosef Bacik { 1002abed4aaaSJosef Bacik struct btrfs_key key = { 1003abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 1004abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1005f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1006abed4aaaSJosef Bacik }; 1007abed4aaaSJosef Bacik 1008abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1009abed4aaaSJosef Bacik } 1010abed4aaaSJosef Bacik 1011abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1012abed4aaaSJosef Bacik { 1013abed4aaaSJosef Bacik struct btrfs_key key = { 1014abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 1015abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1016f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1017abed4aaaSJosef Bacik }; 1018abed4aaaSJosef Bacik 1019abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1020abed4aaaSJosef Bacik } 1021abed4aaaSJosef Bacik 102251129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 102351129b33SJosef Bacik { 102451129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 102551129b33SJosef Bacik return fs_info->block_group_root; 102651129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 102751129b33SJosef Bacik } 102851129b33SJosef Bacik 102920897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 103020897f5cSArne Jansen u64 objectid) 103120897f5cSArne Jansen { 10329b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 103320897f5cSArne Jansen struct extent_buffer *leaf; 103420897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 103520897f5cSArne Jansen struct btrfs_root *root; 103620897f5cSArne Jansen struct btrfs_key key; 1037b89f6d1fSFilipe Manana unsigned int nofs_flag; 103820897f5cSArne Jansen int ret = 0; 103920897f5cSArne Jansen 1040b89f6d1fSFilipe Manana /* 1041b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1042b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1043b89f6d1fSFilipe Manana */ 1044b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 104596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1046b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 104720897f5cSArne Jansen if (!root) 104820897f5cSArne Jansen return ERR_PTR(-ENOMEM); 104920897f5cSArne Jansen 105020897f5cSArne Jansen root->root_key.objectid = objectid; 105120897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 105220897f5cSArne Jansen root->root_key.offset = 0; 105320897f5cSArne Jansen 10549631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 10559631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 105620897f5cSArne Jansen if (IS_ERR(leaf)) { 105720897f5cSArne Jansen ret = PTR_ERR(leaf); 10581dd05682STsutomu Itoh leaf = NULL; 1059c1b07854SPeng Hao goto fail; 106020897f5cSArne Jansen } 106120897f5cSArne Jansen 106220897f5cSArne Jansen root->node = leaf; 106320897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 106420897f5cSArne Jansen 106520897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 106627cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 106720897f5cSArne Jansen 1068f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1069f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 107020897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 107120897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 107220897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 107320897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 107420897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 107520897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 107620897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 107733d85fdaSQu Wenruo if (is_fstree(objectid)) 1078807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1079807fc790SAndy Shevchenko else 1080807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1081c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 108220897f5cSArne Jansen 10838a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 10848a6a87cdSBoris Burkov 108520897f5cSArne Jansen key.objectid = objectid; 108620897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 108720897f5cSArne Jansen key.offset = 0; 108820897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 108920897f5cSArne Jansen if (ret) 109020897f5cSArne Jansen goto fail; 109120897f5cSArne Jansen 109220897f5cSArne Jansen return root; 10931dd05682STsutomu Itoh 10948a6a87cdSBoris Burkov fail: 109500246528SJosef Bacik btrfs_put_root(root); 10961dd05682STsutomu Itoh 10971dd05682STsutomu Itoh return ERR_PTR(ret); 109820897f5cSArne Jansen } 109920897f5cSArne Jansen 11007237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1101e02119d5SChris Mason struct btrfs_fs_info *fs_info) 11020f7d52f4SChris Mason { 11030f7d52f4SChris Mason struct btrfs_root *root; 1104e02119d5SChris Mason 110596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1106e02119d5SChris Mason if (!root) 11077237f183SYan Zheng return ERR_PTR(-ENOMEM); 1108e02119d5SChris Mason 1109e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1110e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1111e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 111227cdeb70SMiao Xie 11136ab6ebb7SNaohiro Aota return root; 11146ab6ebb7SNaohiro Aota } 11156ab6ebb7SNaohiro Aota 11166ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 11176ab6ebb7SNaohiro Aota struct btrfs_root *root) 11186ab6ebb7SNaohiro Aota { 11196ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 11206ab6ebb7SNaohiro Aota 11217237f183SYan Zheng /* 112292a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 112327cdeb70SMiao Xie * 112492a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 112592a7cc42SQu Wenruo * and they go away before a real commit is actually done. 112692a7cc42SQu Wenruo * 112792a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 112892a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 11297237f183SYan Zheng */ 1130e02119d5SChris Mason 11314d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 11329631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 11336ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 11346ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 1135e02119d5SChris Mason 11367237f183SYan Zheng root->node = leaf; 1137e02119d5SChris Mason 1138e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1139e02119d5SChris Mason btrfs_tree_unlock(root->node); 11406ab6ebb7SNaohiro Aota 11416ab6ebb7SNaohiro Aota return 0; 11427237f183SYan Zheng } 11437237f183SYan Zheng 11447237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 11457237f183SYan Zheng struct btrfs_fs_info *fs_info) 11467237f183SYan Zheng { 11477237f183SYan Zheng struct btrfs_root *log_root; 11487237f183SYan Zheng 11497237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 11507237f183SYan Zheng if (IS_ERR(log_root)) 11517237f183SYan Zheng return PTR_ERR(log_root); 11526ab6ebb7SNaohiro Aota 11533ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 11543ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 11553ddebf27SNaohiro Aota 11566ab6ebb7SNaohiro Aota if (ret) { 11576ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 11586ab6ebb7SNaohiro Aota return ret; 11596ab6ebb7SNaohiro Aota } 11603ddebf27SNaohiro Aota } 11616ab6ebb7SNaohiro Aota 11627237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 11637237f183SYan Zheng fs_info->log_root_tree = log_root; 11647237f183SYan Zheng return 0; 11657237f183SYan Zheng } 11667237f183SYan Zheng 11677237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 11687237f183SYan Zheng struct btrfs_root *root) 11697237f183SYan Zheng { 11700b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 11717237f183SYan Zheng struct btrfs_root *log_root; 11727237f183SYan Zheng struct btrfs_inode_item *inode_item; 11736ab6ebb7SNaohiro Aota int ret; 11747237f183SYan Zheng 11750b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 11767237f183SYan Zheng if (IS_ERR(log_root)) 11777237f183SYan Zheng return PTR_ERR(log_root); 11787237f183SYan Zheng 11796ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 11806ab6ebb7SNaohiro Aota if (ret) { 11816ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 11826ab6ebb7SNaohiro Aota return ret; 11836ab6ebb7SNaohiro Aota } 11846ab6ebb7SNaohiro Aota 11857237f183SYan Zheng log_root->last_trans = trans->transid; 11867237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 11877237f183SYan Zheng 11887237f183SYan Zheng inode_item = &log_root->root_item.inode; 11893cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 11903cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 11913cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1192da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 11930b246afaSJeff Mahoney fs_info->nodesize); 11943cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 11957237f183SYan Zheng 11965d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 11977237f183SYan Zheng 11987237f183SYan Zheng WARN_ON(root->log_root); 11997237f183SYan Zheng root->log_root = log_root; 12007237f183SYan Zheng root->log_transid = 0; 1201d1433debSMiao Xie root->log_transid_committed = -1; 1202257c62e1SChris Mason root->last_log_commit = 0; 1203e02119d5SChris Mason return 0; 1204e02119d5SChris Mason } 1205e02119d5SChris Mason 120649d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 120749d11beaSJosef Bacik struct btrfs_path *path, 1208cb517eabSMiao Xie struct btrfs_key *key) 1209e02119d5SChris Mason { 1210e02119d5SChris Mason struct btrfs_root *root; 1211789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 1212e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 121384234f3aSYan Zheng u64 generation; 1214cb517eabSMiao Xie int ret; 1215581c1760SQu Wenruo int level; 1216cb517eabSMiao Xie 121796dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 121849d11beaSJosef Bacik if (!root) 121949d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 12200f7d52f4SChris Mason 1221cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1222cb517eabSMiao Xie &root->root_item, &root->root_key); 12230f7d52f4SChris Mason if (ret) { 122413a8a7c8SYan, Zheng if (ret > 0) 122513a8a7c8SYan, Zheng ret = -ENOENT; 122649d11beaSJosef Bacik goto fail; 12270f7d52f4SChris Mason } 122813a8a7c8SYan, Zheng 122984234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1230581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 1231789d6a3aSQu Wenruo check.level = level; 1232789d6a3aSQu Wenruo check.transid = generation; 1233789d6a3aSQu Wenruo check.owner_root = key->objectid; 1234789d6a3aSQu Wenruo root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), 1235789d6a3aSQu Wenruo &check); 123664c043deSLiu Bo if (IS_ERR(root->node)) { 123764c043deSLiu Bo ret = PTR_ERR(root->node); 12388c38938cSJosef Bacik root->node = NULL; 123949d11beaSJosef Bacik goto fail; 12404eb150d6SQu Wenruo } 12414eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1242cb517eabSMiao Xie ret = -EIO; 124349d11beaSJosef Bacik goto fail; 1244416bc658SJosef Bacik } 124588c602abSQu Wenruo 124688c602abSQu Wenruo /* 124788c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 124888c602abSQu Wenruo * match its root node owner 124988c602abSQu Wenruo */ 125088c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 125188c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 125288c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 125388c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 125488c602abSQu Wenruo btrfs_crit(fs_info, 125588c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 125688c602abSQu Wenruo root->root_key.objectid, root->node->start, 125788c602abSQu Wenruo btrfs_header_owner(root->node), 125888c602abSQu Wenruo root->root_key.objectid); 125988c602abSQu Wenruo ret = -EUCLEAN; 126088c602abSQu Wenruo goto fail; 126188c602abSQu Wenruo } 12625d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1263cb517eabSMiao Xie return root; 126449d11beaSJosef Bacik fail: 126500246528SJosef Bacik btrfs_put_root(root); 126649d11beaSJosef Bacik return ERR_PTR(ret); 126749d11beaSJosef Bacik } 126849d11beaSJosef Bacik 126949d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 127049d11beaSJosef Bacik struct btrfs_key *key) 127149d11beaSJosef Bacik { 127249d11beaSJosef Bacik struct btrfs_root *root; 127349d11beaSJosef Bacik struct btrfs_path *path; 127449d11beaSJosef Bacik 127549d11beaSJosef Bacik path = btrfs_alloc_path(); 127649d11beaSJosef Bacik if (!path) 127749d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 127849d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 127949d11beaSJosef Bacik btrfs_free_path(path); 128049d11beaSJosef Bacik 128149d11beaSJosef Bacik return root; 1282cb517eabSMiao Xie } 1283cb517eabSMiao Xie 12842dfb1e43SQu Wenruo /* 12852dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 12862dfb1e43SQu Wenruo * 12872dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 12882dfb1e43SQu Wenruo */ 12892dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1290cb517eabSMiao Xie { 1291cb517eabSMiao Xie int ret; 1292cb517eabSMiao Xie 12930b548539SDavid Sterba btrfs_drew_lock_init(&root->snapshot_lock); 12948257b2dcSMiao Xie 1295aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 129637f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 129792a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1298f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1299f39e4571SJosef Bacik } 1300f39e4571SJosef Bacik 1301851fd730SQu Wenruo /* 1302851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1303851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1304851fd730SQu Wenruo */ 1305851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1306851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 13072dfb1e43SQu Wenruo if (!anon_dev) { 1308cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1309cb517eabSMiao Xie if (ret) 1310876d2cf1SLiu Bo goto fail; 13112dfb1e43SQu Wenruo } else { 13122dfb1e43SQu Wenruo root->anon_dev = anon_dev; 13132dfb1e43SQu Wenruo } 1314851fd730SQu Wenruo } 1315f32e48e9SChandan Rajendra 1316f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1317453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1318f32e48e9SChandan Rajendra if (ret) { 1319f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1320876d2cf1SLiu Bo goto fail; 1321f32e48e9SChandan Rajendra } 1322f32e48e9SChandan Rajendra 13236b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1324f32e48e9SChandan Rajendra 1325f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1326f32e48e9SChandan Rajendra 1327cb517eabSMiao Xie return 0; 1328cb517eabSMiao Xie fail: 132984db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1330cb517eabSMiao Xie return ret; 1331cb517eabSMiao Xie } 1332cb517eabSMiao Xie 1333a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1334cb517eabSMiao Xie u64 root_id) 1335cb517eabSMiao Xie { 1336cb517eabSMiao Xie struct btrfs_root *root; 1337cb517eabSMiao Xie 1338fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1339fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1340fc7cbcd4SDavid Sterba (unsigned long)root_id); 1341bc44d7c4SJosef Bacik if (root) 134200246528SJosef Bacik root = btrfs_grab_root(root); 1343fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1344cb517eabSMiao Xie return root; 1345cb517eabSMiao Xie } 1346cb517eabSMiao Xie 134749d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 134849d11beaSJosef Bacik u64 objectid) 134949d11beaSJosef Bacik { 1350abed4aaaSJosef Bacik struct btrfs_key key = { 1351abed4aaaSJosef Bacik .objectid = objectid, 1352abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1353abed4aaaSJosef Bacik .offset = 0, 1354abed4aaaSJosef Bacik }; 1355abed4aaaSJosef Bacik 135649d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 135749d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 135849d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 1359abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 136049d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 136149d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 136249d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 136349d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 136449d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 1365abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 136649d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 136749d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 136849d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 136949d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 137049d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 137149d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 137214033b08SQu Wenruo if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) 137314033b08SQu Wenruo return btrfs_grab_root(fs_info->block_group_root) ? 137414033b08SQu Wenruo fs_info->block_group_root : ERR_PTR(-ENOENT); 1375abed4aaaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { 1376abed4aaaSJosef Bacik struct btrfs_root *root = btrfs_global_root(fs_info, &key); 1377abed4aaaSJosef Bacik 1378abed4aaaSJosef Bacik return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); 1379abed4aaaSJosef Bacik } 138049d11beaSJosef Bacik return NULL; 138149d11beaSJosef Bacik } 138249d11beaSJosef Bacik 1383cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1384cb517eabSMiao Xie struct btrfs_root *root) 1385cb517eabSMiao Xie { 1386cb517eabSMiao Xie int ret; 1387cb517eabSMiao Xie 1388fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1389fc7cbcd4SDavid Sterba if (ret) 1390fc7cbcd4SDavid Sterba return ret; 1391fc7cbcd4SDavid Sterba 1392fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1393fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1394fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1395fc7cbcd4SDavid Sterba root); 1396af01d2e5SJosef Bacik if (ret == 0) { 139700246528SJosef Bacik btrfs_grab_root(root); 1398fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1399af01d2e5SJosef Bacik } 1400fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1401fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1402cb517eabSMiao Xie 1403cb517eabSMiao Xie return ret; 1404cb517eabSMiao Xie } 1405cb517eabSMiao Xie 1406bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1407bd647ce3SJosef Bacik { 1408bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1409bd647ce3SJosef Bacik struct btrfs_root *root; 1410bd647ce3SJosef Bacik 1411bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1412457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1413457f1864SJosef Bacik 1414bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1415bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1416457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 141771008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1418bd647ce3SJosef Bacik refcount_read(&root->refs)); 1419bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 142000246528SJosef Bacik btrfs_put_root(root); 142100246528SJosef Bacik btrfs_put_root(root); 1422bd647ce3SJosef Bacik } 1423bd647ce3SJosef Bacik #endif 1424bd647ce3SJosef Bacik } 1425bd647ce3SJosef Bacik 1426abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1427abed4aaaSJosef Bacik { 1428abed4aaaSJosef Bacik struct btrfs_root *root; 1429abed4aaaSJosef Bacik struct rb_node *node; 1430abed4aaaSJosef Bacik 1431abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1432abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1433abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1434abed4aaaSJosef Bacik btrfs_put_root(root); 1435abed4aaaSJosef Bacik } 1436abed4aaaSJosef Bacik } 1437abed4aaaSJosef Bacik 14380d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 14390d4b0463SJosef Bacik { 1440141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1441141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 14425deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1443141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1444141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1445141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1446141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 14470d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 14480d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1449abed4aaaSJosef Bacik free_global_roots(fs_info); 145000246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 145100246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 145200246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 145300246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 145400246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 145500246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1456aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 14579c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1458bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 14593fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 14600d4b0463SJosef Bacik kfree(fs_info->super_copy); 14610d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 14628481dd80SQu Wenruo kfree(fs_info->subpage_info); 14630d4b0463SJosef Bacik kvfree(fs_info); 14640d4b0463SJosef Bacik } 14650d4b0463SJosef Bacik 14660d4b0463SJosef Bacik 14672dfb1e43SQu Wenruo /* 14682dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 14692dfb1e43SQu Wenruo * 14702dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 14712dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 14722dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 14732dfb1e43SQu Wenruo * 14742dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 14752dfb1e43SQu Wenruo * 14762dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 14772dfb1e43SQu Wenruo * same root objectid. 14782dfb1e43SQu Wenruo * 14792dfb1e43SQu Wenruo * @objectid: root id 14802dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 14812dfb1e43SQu Wenruo * pass 0 for new allocation. 14822dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 14832dfb1e43SQu Wenruo * for orphan roots 14842dfb1e43SQu Wenruo */ 14852dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 14862dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 14872dfb1e43SQu Wenruo bool check_ref) 14885eda7b5eSChris Mason { 14895eda7b5eSChris Mason struct btrfs_root *root; 1490381cf658SDavid Sterba struct btrfs_path *path; 14911d4c08e0SDavid Sterba struct btrfs_key key; 14925eda7b5eSChris Mason int ret; 14935eda7b5eSChris Mason 149449d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 149549d11beaSJosef Bacik if (root) 149649d11beaSJosef Bacik return root; 14974df27c4dSYan, Zheng again: 149856e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 149948475471SStefan Behrens if (root) { 15002dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 15012dfb1e43SQu Wenruo ASSERT(!anon_dev); 1502bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 150300246528SJosef Bacik btrfs_put_root(root); 150448475471SStefan Behrens return ERR_PTR(-ENOENT); 1505bc44d7c4SJosef Bacik } 15065eda7b5eSChris Mason return root; 150748475471SStefan Behrens } 15085eda7b5eSChris Mason 150956e9357aSDavid Sterba key.objectid = objectid; 151056e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 151156e9357aSDavid Sterba key.offset = (u64)-1; 151256e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 15135eda7b5eSChris Mason if (IS_ERR(root)) 15145eda7b5eSChris Mason return root; 15153394e160SChris Mason 1516c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1517d68fc57bSYan, Zheng ret = -ENOENT; 1518d68fc57bSYan, Zheng goto fail; 1519d68fc57bSYan, Zheng } 1520d68fc57bSYan, Zheng 15212dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1522cb517eabSMiao Xie if (ret) 1523cb517eabSMiao Xie goto fail; 1524cb517eabSMiao Xie 1525381cf658SDavid Sterba path = btrfs_alloc_path(); 1526381cf658SDavid Sterba if (!path) { 1527381cf658SDavid Sterba ret = -ENOMEM; 1528381cf658SDavid Sterba goto fail; 1529381cf658SDavid Sterba } 15301d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 15311d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 153256e9357aSDavid Sterba key.offset = objectid; 15331d4c08e0SDavid Sterba 15341d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1535381cf658SDavid Sterba btrfs_free_path(path); 1536d68fc57bSYan, Zheng if (ret < 0) 1537d68fc57bSYan, Zheng goto fail; 1538d68fc57bSYan, Zheng if (ret == 0) 153927cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1540d68fc57bSYan, Zheng 1541cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 15420f7d52f4SChris Mason if (ret) { 1543168a2f77SJia-Ju Bai if (ret == -EEXIST) { 154400246528SJosef Bacik btrfs_put_root(root); 15454df27c4dSYan, Zheng goto again; 1546168a2f77SJia-Ju Bai } 15474df27c4dSYan, Zheng goto fail; 15484df27c4dSYan, Zheng } 1549edbd8d4eSChris Mason return root; 15504df27c4dSYan, Zheng fail: 155133fab972SFilipe Manana /* 155233fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1553143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 155433fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 155533fab972SFilipe Manana * and once again by our caller. 155633fab972SFilipe Manana */ 155733fab972SFilipe Manana if (anon_dev) 155833fab972SFilipe Manana root->anon_dev = 0; 15598c38938cSJosef Bacik btrfs_put_root(root); 15604df27c4dSYan, Zheng return ERR_PTR(ret); 1561edbd8d4eSChris Mason } 1562edbd8d4eSChris Mason 15632dfb1e43SQu Wenruo /* 15642dfb1e43SQu Wenruo * Get in-memory reference of a root structure 15652dfb1e43SQu Wenruo * 15662dfb1e43SQu Wenruo * @objectid: tree objectid 15672dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 15682dfb1e43SQu Wenruo * one reference 15692dfb1e43SQu Wenruo */ 15702dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 15712dfb1e43SQu Wenruo u64 objectid, bool check_ref) 15722dfb1e43SQu Wenruo { 15732dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 15742dfb1e43SQu Wenruo } 15752dfb1e43SQu Wenruo 15762dfb1e43SQu Wenruo /* 15772dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 15782dfb1e43SQu Wenruo * the anonymous block device id 15792dfb1e43SQu Wenruo * 15802dfb1e43SQu Wenruo * @objectid: tree objectid 15812dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 15822dfb1e43SQu Wenruo * parameter value 15832dfb1e43SQu Wenruo */ 15842dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 15852dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 15862dfb1e43SQu Wenruo { 15872dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 15882dfb1e43SQu Wenruo } 15892dfb1e43SQu Wenruo 15908b712842SChris Mason /* 159149d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 159249d11beaSJosef Bacik * @fs_info: the fs_info 159349d11beaSJosef Bacik * @objectid: the objectid we need to lookup 159449d11beaSJosef Bacik * 159549d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 159649d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 159749d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 159849d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 159949d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 160049d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 160149d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 160249d11beaSJosef Bacik * backref code is finished using the root. 160349d11beaSJosef Bacik */ 160449d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 160549d11beaSJosef Bacik struct btrfs_path *path, 160649d11beaSJosef Bacik u64 objectid) 160749d11beaSJosef Bacik { 160849d11beaSJosef Bacik struct btrfs_root *root; 160949d11beaSJosef Bacik struct btrfs_key key; 161049d11beaSJosef Bacik 161149d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 161249d11beaSJosef Bacik 161349d11beaSJosef Bacik /* 161449d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 161549d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 161649d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 161749d11beaSJosef Bacik * != NULL just return it. 161849d11beaSJosef Bacik */ 161949d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 162049d11beaSJosef Bacik if (root) 162149d11beaSJosef Bacik return root; 162249d11beaSJosef Bacik 162349d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 162449d11beaSJosef Bacik if (root) 162549d11beaSJosef Bacik return root; 162649d11beaSJosef Bacik 162749d11beaSJosef Bacik key.objectid = objectid; 162849d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 162949d11beaSJosef Bacik key.offset = (u64)-1; 163049d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 163149d11beaSJosef Bacik btrfs_release_path(path); 163249d11beaSJosef Bacik 163349d11beaSJosef Bacik return root; 163449d11beaSJosef Bacik } 163549d11beaSJosef Bacik 1636a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1637a74a4b97SChris Mason { 16380d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1639d0278245SMiao Xie int again; 1640a74a4b97SChris Mason 1641d6fd0ae2SOmar Sandoval while (1) { 1642d0278245SMiao Xie again = 0; 16439d1a2a3aSDavid Sterba 1644fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1645fd340d0fSJosef Bacik 1646d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 16472ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1648d0278245SMiao Xie goto sleep; 1649d0278245SMiao Xie 165090c711abSZygo Blaxell /* 165190c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 165290c711abSZygo Blaxell * before we have finished mounting the filesystem. 165390c711abSZygo Blaxell */ 16540b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 165590c711abSZygo Blaxell goto sleep; 165690c711abSZygo Blaxell 16570b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1658d0278245SMiao Xie goto sleep; 1659d0278245SMiao Xie 1660dc7f370cSMiao Xie /* 1661dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1662dc7f370cSMiao Xie * during the above check and trylock. 1663dc7f370cSMiao Xie */ 16642ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 16650b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1666dc7f370cSMiao Xie goto sleep; 1667dc7f370cSMiao Xie } 1668dc7f370cSMiao Xie 1669b7625f46SQu Wenruo if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) 1670b7625f46SQu Wenruo btrfs_sysfs_feature_update(fs_info); 1671b7625f46SQu Wenruo 16722ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1673c2d6cb16SFilipe Manana 167433c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 16750b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1676a74a4b97SChris Mason 1677d0278245SMiao Xie /* 167805323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 167905323cd1SMiao Xie * needn't do anything special here. 1680d0278245SMiao Xie */ 16810b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 168267c5e7d4SFilipe Manana 168367c5e7d4SFilipe Manana /* 1684f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 168567c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 168667c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1687f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 168867c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 168967c5e7d4SFilipe Manana * unused block groups. 169067c5e7d4SFilipe Manana */ 16910b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 169218bb8bbfSJohannes Thumshirn 169318bb8bbfSJohannes Thumshirn /* 169418bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 169518bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 169618bb8bbfSJohannes Thumshirn * space. 169718bb8bbfSJohannes Thumshirn */ 169818bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1699d0278245SMiao Xie sleep: 1700a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1701d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1702d6fd0ae2SOmar Sandoval kthread_parkme(); 1703d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1704d6fd0ae2SOmar Sandoval return 0; 1705838fe188SJiri Kosina if (!again) { 1706a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1707a74a4b97SChris Mason schedule(); 1708a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1709a74a4b97SChris Mason } 1710da288d28SFilipe Manana } 1711a74a4b97SChris Mason } 1712a74a4b97SChris Mason 1713a74a4b97SChris Mason static int transaction_kthread(void *arg) 1714a74a4b97SChris Mason { 1715a74a4b97SChris Mason struct btrfs_root *root = arg; 17160b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1717a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1718a74a4b97SChris Mason struct btrfs_transaction *cur; 17198929ecfaSYan, Zheng u64 transid; 1720643900beSNikolay Borisov time64_t delta; 1721a74a4b97SChris Mason unsigned long delay; 1722914b2007SJan Kara bool cannot_commit; 1723a74a4b97SChris Mason 1724a74a4b97SChris Mason do { 1725914b2007SJan Kara cannot_commit = false; 1726ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 17270b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1728a74a4b97SChris Mason 17290b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 17300b246afaSJeff Mahoney cur = fs_info->running_transaction; 1731a74a4b97SChris Mason if (!cur) { 17320b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1733a74a4b97SChris Mason goto sleep; 1734a74a4b97SChris Mason } 173531153d81SYan Zheng 1736643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1737fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1738fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1739643900beSNikolay Borisov delta < fs_info->commit_interval) { 17400b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1741fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1742fb8a7e94SNikolay Borisov delay = min(delay, 1743fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1744a74a4b97SChris Mason goto sleep; 1745a74a4b97SChris Mason } 17468929ecfaSYan, Zheng transid = cur->transid; 17470b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 174856bec294SChris Mason 174979787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1750354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1751914b2007SJan Kara if (IS_ERR(trans)) { 1752354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1753914b2007SJan Kara cannot_commit = true; 175479787eaaSJeff Mahoney goto sleep; 1755914b2007SJan Kara } 17568929ecfaSYan, Zheng if (transid == trans->transid) { 17573a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 17588929ecfaSYan, Zheng } else { 17593a45bb20SJeff Mahoney btrfs_end_transaction(trans); 17608929ecfaSYan, Zheng } 1761a74a4b97SChris Mason sleep: 17620b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 17630b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1764a74a4b97SChris Mason 176584961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 17662ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 17678929ecfaSYan, Zheng if (!kthread_should_stop() && 17680b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1769914b2007SJan Kara cannot_commit)) 1770bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1771a74a4b97SChris Mason } while (!kthread_should_stop()); 1772a74a4b97SChris Mason return 0; 1773a74a4b97SChris Mason } 1774a74a4b97SChris Mason 1775af31f5e5SChris Mason /* 177601f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 177701f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 177801f0f9daSNikolay Borisov * anything. 1779af31f5e5SChris Mason * 1780af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1781af31f5e5SChris Mason * generation of the latest root in the array with the generation 1782af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1783af31f5e5SChris Mason */ 178401f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1785af31f5e5SChris Mason { 178601f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1787af31f5e5SChris Mason u64 cur; 1788af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1789af31f5e5SChris Mason int i; 1790af31f5e5SChris Mason 1791af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1792af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1793af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1794af31f5e5SChris Mason if (cur == newest_gen) 179501f0f9daSNikolay Borisov return i; 1796af31f5e5SChris Mason } 1797af31f5e5SChris Mason 179801f0f9daSNikolay Borisov return -EINVAL; 1799af31f5e5SChris Mason } 1800af31f5e5SChris Mason 1801af31f5e5SChris Mason /* 1802af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1803af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1804af31f5e5SChris Mason * done 1805af31f5e5SChris Mason */ 1806af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1807af31f5e5SChris Mason { 18086ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1809af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1810af31f5e5SChris Mason 1811af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1812af31f5e5SChris Mason 1813af31f5e5SChris Mason /* 1814af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1815af31f5e5SChris Mason * regardless of which ones we use today 1816af31f5e5SChris Mason */ 1817af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1818af31f5e5SChris Mason 1819af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1820af31f5e5SChris Mason 1821af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1822af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1823af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1824af31f5e5SChris Mason 1825af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1826af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1827af31f5e5SChris Mason 1828af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1829af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1830af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1831af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1832af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1833af31f5e5SChris Mason 18341c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 18359c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1836f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 18379c54e80dSJosef Bacik 18389c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 18399c54e80dSJosef Bacik extent_root->node->start); 1840af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 184129cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 1842af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 184329cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 1844af31f5e5SChris Mason 1845f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 1846f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 1847f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 1848f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 1849f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 18509c54e80dSJosef Bacik } 1851af31f5e5SChris Mason 18527c7e82a7SChris Mason /* 18537c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 18547c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 18557c7e82a7SChris Mason */ 18567c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 18577c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 18587c7e82a7SChris Mason info->fs_root->node->start); 1859af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1860af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1861af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1862af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 18637c7e82a7SChris Mason } 1864af31f5e5SChris Mason 1865af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1866af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1867af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1868af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1869af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1870af31f5e5SChris Mason 1871af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1872af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1873af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1874af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1875af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1876af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1877af31f5e5SChris Mason 1878af31f5e5SChris Mason /* 1879af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1880af31f5e5SChris Mason * for the next commit 1881af31f5e5SChris Mason */ 1882af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1883af31f5e5SChris Mason &info->super_for_commit->super_roots, 1884af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1885af31f5e5SChris Mason } 1886af31f5e5SChris Mason 1887af31f5e5SChris Mason /* 1888bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1889bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1890bd2336b2SNikolay Borisov * 1891bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1892bd2336b2SNikolay Borisov * priority - priority of backup root required 1893bd2336b2SNikolay Borisov * 1894bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1895bd2336b2SNikolay Borisov */ 1896bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1897bd2336b2SNikolay Borisov { 1898bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1899bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1900bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1901bd2336b2SNikolay Borisov 1902bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1903bd2336b2SNikolay Borisov if (priority == 0) 1904bd2336b2SNikolay Borisov return backup_index; 1905bd2336b2SNikolay Borisov 1906bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1907bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1908bd2336b2SNikolay Borisov } else { 1909bd2336b2SNikolay Borisov return -EINVAL; 1910bd2336b2SNikolay Borisov } 1911bd2336b2SNikolay Borisov 1912bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1913bd2336b2SNikolay Borisov 1914bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1915bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1916bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1917bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1918bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1919bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1920bd2336b2SNikolay Borisov 1921bd2336b2SNikolay Borisov /* 1922bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1923bd2336b2SNikolay Borisov * need a fsck 1924bd2336b2SNikolay Borisov */ 1925bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1926bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1927bd2336b2SNikolay Borisov 1928bd2336b2SNikolay Borisov return backup_index; 1929bd2336b2SNikolay Borisov } 1930bd2336b2SNikolay Borisov 19317abadb64SLiu Bo /* helper to cleanup workers */ 19327abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 19337abadb64SLiu Bo { 1934dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1935afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 19365cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1937d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 1938d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 1939385de0efSChristoph Hellwig if (fs_info->rmw_workers) 1940385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 1941fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 1942fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 1943fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1944fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 19455b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1946e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1947a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1948fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1949b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1950b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1951a9b9477dSFilipe Manana /* 1952a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 1953a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 1954a9b9477dSFilipe Manana * queues can do metadata I/O operations. 1955a9b9477dSFilipe Manana */ 1956d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 1957d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 19587abadb64SLiu Bo } 19597abadb64SLiu Bo 19602e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 19612e9f5954SRashika { 19622e9f5954SRashika if (root) { 19632e9f5954SRashika free_extent_buffer(root->node); 19642e9f5954SRashika free_extent_buffer(root->commit_root); 19652e9f5954SRashika root->node = NULL; 19662e9f5954SRashika root->commit_root = NULL; 19672e9f5954SRashika } 19682e9f5954SRashika } 19692e9f5954SRashika 1970abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 1971abed4aaaSJosef Bacik { 1972abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 1973abed4aaaSJosef Bacik 1974abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 1975abed4aaaSJosef Bacik &fs_info->global_root_tree, 1976abed4aaaSJosef Bacik rb_node) 1977abed4aaaSJosef Bacik free_root_extent_buffers(root); 1978abed4aaaSJosef Bacik } 1979abed4aaaSJosef Bacik 1980af31f5e5SChris Mason /* helper to cleanup tree roots */ 19814273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 1982af31f5e5SChris Mason { 19832e9f5954SRashika free_root_extent_buffers(info->tree_root); 1984655b09feSJosef Bacik 1985abed4aaaSJosef Bacik free_global_root_pointers(info); 19862e9f5954SRashika free_root_extent_buffers(info->dev_root); 19872e9f5954SRashika free_root_extent_buffers(info->quota_root); 19882e9f5954SRashika free_root_extent_buffers(info->uuid_root); 19898c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 1990aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 19919c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 19924273eaffSAnand Jain if (free_chunk_root) 19932e9f5954SRashika free_root_extent_buffers(info->chunk_root); 1994af31f5e5SChris Mason } 1995af31f5e5SChris Mason 19968c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 19978c38938cSJosef Bacik { 19988c38938cSJosef Bacik if (!root) 19998c38938cSJosef Bacik return; 20008c38938cSJosef Bacik 20018c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 20028c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 20031dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 20048c38938cSJosef Bacik if (root->anon_dev) 20058c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 2006923eb523SJohannes Thumshirn free_root_extent_buffers(root); 20078c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2008fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 20098c38938cSJosef Bacik list_del_init(&root->leak_list); 2010fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 20118c38938cSJosef Bacik #endif 20128c38938cSJosef Bacik kfree(root); 20138c38938cSJosef Bacik } 20148c38938cSJosef Bacik } 20158c38938cSJosef Bacik 2016faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2017171f6537SJosef Bacik { 2018fc7cbcd4SDavid Sterba int ret; 2019fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 2020fc7cbcd4SDavid Sterba int i; 2021171f6537SJosef Bacik 2022171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2023fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 2024171f6537SJosef Bacik struct btrfs_root, root_list); 2025fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 2026171f6537SJosef Bacik 2027fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2028fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2029fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 2030171f6537SJosef Bacik } 2031171f6537SJosef Bacik 2032fc7cbcd4SDavid Sterba while (1) { 2033fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2034fc7cbcd4SDavid Sterba (void **)gang, 0, 2035fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 2036fc7cbcd4SDavid Sterba if (!ret) 2037fc7cbcd4SDavid Sterba break; 2038fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 2039fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2040171f6537SJosef Bacik } 2041171f6537SJosef Bacik } 2042af31f5e5SChris Mason 2043638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2044638aa7edSEric Sandeen { 2045638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2046638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2047638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2048638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2049638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2050638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2051ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2052638aa7edSEric Sandeen } 2053638aa7edSEric Sandeen 2054779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2055779a65a4SEric Sandeen { 2056779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2057779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2058779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2059779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2060779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2061779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2062907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 2063779a65a4SEric Sandeen } 2064779a65a4SEric Sandeen 2065dcb2137cSChristoph Hellwig static int btrfs_init_btree_inode(struct super_block *sb) 2066f37938e0SEric Sandeen { 2067dcb2137cSChristoph Hellwig struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2068e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 2069e256927bSJosef Bacik fs_info->tree_root); 2070dcb2137cSChristoph Hellwig struct inode *inode; 2071dcb2137cSChristoph Hellwig 2072dcb2137cSChristoph Hellwig inode = new_inode(sb); 2073dcb2137cSChristoph Hellwig if (!inode) 2074dcb2137cSChristoph Hellwig return -ENOMEM; 20752ff7e61eSJeff Mahoney 20762ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 20772ff7e61eSJeff Mahoney set_nlink(inode, 1); 2078f37938e0SEric Sandeen /* 2079f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2080f37938e0SEric Sandeen * the real end of the address space is determined by all of 2081f37938e0SEric Sandeen * the devices in the system 2082f37938e0SEric Sandeen */ 20832ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 20842ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2085dcb2137cSChristoph Hellwig mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); 2086f37938e0SEric Sandeen 20872ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 208843eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 208935da5a7eSDavid Sterba IO_TREE_BTREE_INODE_IO); 20902ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2091f37938e0SEric Sandeen 20925c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 2093adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 2094adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 2095adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 20962ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 2097e256927bSJosef Bacik __insert_inode_hash(inode, hash); 2098dcb2137cSChristoph Hellwig fs_info->btree_inode = inode; 2099dcb2137cSChristoph Hellwig 2100dcb2137cSChristoph Hellwig return 0; 2101f37938e0SEric Sandeen } 2102f37938e0SEric Sandeen 2103ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2104ad618368SEric Sandeen { 2105ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2106129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 21077f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2108ad618368SEric Sandeen } 2109ad618368SEric Sandeen 2110f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2111f9e92e40SEric Sandeen { 2112f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2113f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2114f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2115f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2116f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2117f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2118d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2119011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 2120f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2121f9e92e40SEric Sandeen } 2122f9e92e40SEric Sandeen 2123d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 21242a458198SEric Sandeen { 2125f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 21266f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 21272a458198SEric Sandeen 21282a458198SEric Sandeen fs_info->workers = 2129a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 21302a458198SEric Sandeen 21312a458198SEric Sandeen fs_info->delalloc_workers = 2132cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2133cb001095SJeff Mahoney flags, max_active, 2); 21342a458198SEric Sandeen 21352a458198SEric Sandeen fs_info->flush_workers = 2136cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2137cb001095SJeff Mahoney flags, max_active, 0); 21382a458198SEric Sandeen 21392a458198SEric Sandeen fs_info->caching_workers = 2140cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 21412a458198SEric Sandeen 21422a458198SEric Sandeen fs_info->fixup_workers = 2143cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 21442a458198SEric Sandeen 21452a458198SEric Sandeen fs_info->endio_workers = 2146d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 21472a458198SEric Sandeen fs_info->endio_meta_workers = 2148d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 2149385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 21502a458198SEric Sandeen fs_info->endio_write_workers = 2151cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2152cb001095SJeff Mahoney max_active, 2); 2153fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 2154fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 21552a458198SEric Sandeen fs_info->endio_freespace_worker = 2156cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2157cb001095SJeff Mahoney max_active, 0); 21582a458198SEric Sandeen fs_info->delayed_workers = 2159cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2160cb001095SJeff Mahoney max_active, 0); 21612a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2162cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2163b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2164b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 21652a458198SEric Sandeen 21668bfec2e4SChristoph Hellwig if (!(fs_info->workers && 2167a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 21682a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 2169fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 21701a1a2851SQu Wenruo fs_info->endio_write_workers && 21712a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 2172f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 2173f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 2174b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 21752a458198SEric Sandeen return -ENOMEM; 21762a458198SEric Sandeen } 21772a458198SEric Sandeen 21782a458198SEric Sandeen return 0; 21792a458198SEric Sandeen } 21802a458198SEric Sandeen 21816d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 21826d97c6e3SJohannes Thumshirn { 21836d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2184b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 21856d97c6e3SJohannes Thumshirn 2186b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 21876d97c6e3SJohannes Thumshirn 21886d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 21896d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2190b4e967beSDavid Sterba csum_driver); 21916d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 21926d97c6e3SJohannes Thumshirn } 21936d97c6e3SJohannes Thumshirn 21946d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 21956d97c6e3SJohannes Thumshirn 219668d99ab0SChristoph Hellwig /* 219768d99ab0SChristoph Hellwig * Check if the checksum implementation is a fast accelerated one. 219868d99ab0SChristoph Hellwig * As-is this is a bit of a hack and should be replaced once the csum 219968d99ab0SChristoph Hellwig * implementations provide that information themselves. 220068d99ab0SChristoph Hellwig */ 220168d99ab0SChristoph Hellwig switch (csum_type) { 220268d99ab0SChristoph Hellwig case BTRFS_CSUM_TYPE_CRC32: 220368d99ab0SChristoph Hellwig if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) 220468d99ab0SChristoph Hellwig set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); 220568d99ab0SChristoph Hellwig break; 220668d99ab0SChristoph Hellwig default: 220768d99ab0SChristoph Hellwig break; 220868d99ab0SChristoph Hellwig } 220968d99ab0SChristoph Hellwig 2210c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2211c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2212c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 22136d97c6e3SJohannes Thumshirn return 0; 22146d97c6e3SJohannes Thumshirn } 22156d97c6e3SJohannes Thumshirn 221663443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 221763443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 221863443bf5SEric Sandeen { 221963443bf5SEric Sandeen int ret; 2220789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 222163443bf5SEric Sandeen struct btrfs_root *log_tree_root; 222263443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 222363443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2224581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 222563443bf5SEric Sandeen 222663443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2227f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 222863443bf5SEric Sandeen return -EIO; 222963443bf5SEric Sandeen } 223063443bf5SEric Sandeen 223196dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 223296dfcb46SJosef Bacik GFP_KERNEL); 223363443bf5SEric Sandeen if (!log_tree_root) 223463443bf5SEric Sandeen return -ENOMEM; 223563443bf5SEric Sandeen 2236789d6a3aSQu Wenruo check.level = level; 2237789d6a3aSQu Wenruo check.transid = fs_info->generation + 1; 2238789d6a3aSQu Wenruo check.owner_root = BTRFS_TREE_LOG_OBJECTID; 2239789d6a3aSQu Wenruo log_tree_root->node = read_tree_block(fs_info, bytenr, &check); 224064c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2241f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 22420eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 22438c38938cSJosef Bacik log_tree_root->node = NULL; 224400246528SJosef Bacik btrfs_put_root(log_tree_root); 22450eeff236SLiu Bo return ret; 22464eb150d6SQu Wenruo } 22474eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2248f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 224900246528SJosef Bacik btrfs_put_root(log_tree_root); 225063443bf5SEric Sandeen return -EIO; 225163443bf5SEric Sandeen } 22524eb150d6SQu Wenruo 225363443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 225463443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 225563443bf5SEric Sandeen if (ret) { 22560b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 225763443bf5SEric Sandeen "Failed to recover log tree"); 225800246528SJosef Bacik btrfs_put_root(log_tree_root); 225963443bf5SEric Sandeen return ret; 226063443bf5SEric Sandeen } 226163443bf5SEric Sandeen 2262bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 22636bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 226463443bf5SEric Sandeen if (ret) 226563443bf5SEric Sandeen return ret; 226663443bf5SEric Sandeen } 226763443bf5SEric Sandeen 226863443bf5SEric Sandeen return 0; 226963443bf5SEric Sandeen } 227063443bf5SEric Sandeen 2271abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2272abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2273abed4aaaSJosef Bacik const char *name) 2274abed4aaaSJosef Bacik { 2275abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2276abed4aaaSJosef Bacik struct btrfs_root *root; 2277f7238e50SJosef Bacik u64 max_global_id = 0; 2278abed4aaaSJosef Bacik int ret; 2279abed4aaaSJosef Bacik struct btrfs_key key = { 2280abed4aaaSJosef Bacik .objectid = objectid, 2281abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2282abed4aaaSJosef Bacik .offset = 0, 2283abed4aaaSJosef Bacik }; 2284abed4aaaSJosef Bacik bool found = false; 2285abed4aaaSJosef Bacik 2286abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2287abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2288abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2289abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2290abed4aaaSJosef Bacik return 0; 2291abed4aaaSJosef Bacik } 2292abed4aaaSJosef Bacik 2293abed4aaaSJosef Bacik while (1) { 2294abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2295abed4aaaSJosef Bacik if (ret < 0) 2296abed4aaaSJosef Bacik break; 2297abed4aaaSJosef Bacik 2298abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2299abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2300abed4aaaSJosef Bacik if (ret) { 2301abed4aaaSJosef Bacik if (ret > 0) 2302abed4aaaSJosef Bacik ret = 0; 2303abed4aaaSJosef Bacik break; 2304abed4aaaSJosef Bacik } 2305abed4aaaSJosef Bacik } 2306abed4aaaSJosef Bacik ret = 0; 2307abed4aaaSJosef Bacik 2308abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2309abed4aaaSJosef Bacik if (key.objectid != objectid) 2310abed4aaaSJosef Bacik break; 2311abed4aaaSJosef Bacik btrfs_release_path(path); 2312abed4aaaSJosef Bacik 2313f7238e50SJosef Bacik /* 2314f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2315f7238e50SJosef Bacik * everybody. 2316f7238e50SJosef Bacik */ 2317f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2318f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2319f7238e50SJosef Bacik 2320abed4aaaSJosef Bacik found = true; 2321abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2322abed4aaaSJosef Bacik if (IS_ERR(root)) { 2323abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2324abed4aaaSJosef Bacik ret = PTR_ERR(root); 2325abed4aaaSJosef Bacik break; 2326abed4aaaSJosef Bacik } 2327abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2328abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2329abed4aaaSJosef Bacik if (ret) { 2330abed4aaaSJosef Bacik btrfs_put_root(root); 2331abed4aaaSJosef Bacik break; 2332abed4aaaSJosef Bacik } 2333abed4aaaSJosef Bacik key.offset++; 2334abed4aaaSJosef Bacik } 2335abed4aaaSJosef Bacik btrfs_release_path(path); 2336abed4aaaSJosef Bacik 2337f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2338f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2339f7238e50SJosef Bacik 2340abed4aaaSJosef Bacik if (!found || ret) { 2341abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2342abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2343abed4aaaSJosef Bacik 2344abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2345abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2346abed4aaaSJosef Bacik else 2347abed4aaaSJosef Bacik ret = 0; 2348abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2349abed4aaaSJosef Bacik } 2350abed4aaaSJosef Bacik return ret; 2351abed4aaaSJosef Bacik } 2352abed4aaaSJosef Bacik 2353abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2354abed4aaaSJosef Bacik { 2355abed4aaaSJosef Bacik struct btrfs_path *path; 2356abed4aaaSJosef Bacik int ret = 0; 2357abed4aaaSJosef Bacik 2358abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2359abed4aaaSJosef Bacik if (!path) 2360abed4aaaSJosef Bacik return -ENOMEM; 2361abed4aaaSJosef Bacik 2362abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2363abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2364abed4aaaSJosef Bacik if (ret) 2365abed4aaaSJosef Bacik goto out; 2366abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2367abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2368abed4aaaSJosef Bacik if (ret) 2369abed4aaaSJosef Bacik goto out; 2370abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2371abed4aaaSJosef Bacik goto out; 2372abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2373abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2374abed4aaaSJosef Bacik "free space"); 2375abed4aaaSJosef Bacik out: 2376abed4aaaSJosef Bacik btrfs_free_path(path); 2377abed4aaaSJosef Bacik return ret; 2378abed4aaaSJosef Bacik } 2379abed4aaaSJosef Bacik 23806bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 23814bbcaa64SEric Sandeen { 23826bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2383a4f3d2c4SDavid Sterba struct btrfs_root *root; 23844bbcaa64SEric Sandeen struct btrfs_key location; 23854bbcaa64SEric Sandeen int ret; 23864bbcaa64SEric Sandeen 23876bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 23886bccf3abSJeff Mahoney 2389abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2390abed4aaaSJosef Bacik if (ret) 2391abed4aaaSJosef Bacik return ret; 2392abed4aaaSJosef Bacik 23934bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 23944bbcaa64SEric Sandeen location.offset = 0; 23954bbcaa64SEric Sandeen 23961c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 239714033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 239814033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 239914033b08SQu Wenruo if (IS_ERR(root)) { 240014033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 240114033b08SQu Wenruo ret = PTR_ERR(root); 240214033b08SQu Wenruo goto out; 240314033b08SQu Wenruo } 240414033b08SQu Wenruo } else { 240514033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 240614033b08SQu Wenruo fs_info->block_group_root = root; 240714033b08SQu Wenruo } 240814033b08SQu Wenruo } 240914033b08SQu Wenruo 241014033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2411a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2412f50f4353SLiu Bo if (IS_ERR(root)) { 241342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2414f50f4353SLiu Bo ret = PTR_ERR(root); 2415f50f4353SLiu Bo goto out; 2416f50f4353SLiu Bo } 241742437a63SJosef Bacik } else { 2418a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2419a4f3d2c4SDavid Sterba fs_info->dev_root = root; 242042437a63SJosef Bacik } 2421820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2422a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2423a8d1b164SJohannes Thumshirn if (ret) 2424a8d1b164SJohannes Thumshirn goto out; 24254bbcaa64SEric Sandeen 2426aeb935a4SQu Wenruo /* 2427aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2428aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2429aeb935a4SQu Wenruo */ 243056e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 243156e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2432aeb935a4SQu Wenruo if (IS_ERR(root)) { 243342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2434aeb935a4SQu Wenruo ret = PTR_ERR(root); 2435aeb935a4SQu Wenruo goto out; 2436aeb935a4SQu Wenruo } 243742437a63SJosef Bacik } else { 2438aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2439aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 244042437a63SJosef Bacik } 2441aeb935a4SQu Wenruo 24424bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2443a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2444a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2445a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2446afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2447a4f3d2c4SDavid Sterba fs_info->quota_root = root; 24484bbcaa64SEric Sandeen } 24494bbcaa64SEric Sandeen 24504bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2451a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2452a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 245342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2454a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 24554bbcaa64SEric Sandeen if (ret != -ENOENT) 2456f50f4353SLiu Bo goto out; 245742437a63SJosef Bacik } 24584bbcaa64SEric Sandeen } else { 2459a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2460a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 24614bbcaa64SEric Sandeen } 24624bbcaa64SEric Sandeen 24634bbcaa64SEric Sandeen return 0; 2464f50f4353SLiu Bo out: 2465f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2466f50f4353SLiu Bo location.objectid, ret); 2467f50f4353SLiu Bo return ret; 24684bbcaa64SEric Sandeen } 24694bbcaa64SEric Sandeen 2470069ec957SQu Wenruo /* 2471069ec957SQu Wenruo * Real super block validation 2472069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2473069ec957SQu Wenruo * 2474069ec957SQu Wenruo * @sb: super block to check 2475069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2476069ec957SQu Wenruo * 0 the primary (1st) sb 2477069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2478069ec957SQu Wenruo * -1 skip bytenr check 2479069ec957SQu Wenruo */ 2480a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2481069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 248221a852b0SQu Wenruo { 248321a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 248421a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 248521a852b0SQu Wenruo int ret = 0; 248621a852b0SQu Wenruo 248721a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 248821a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 248921a852b0SQu Wenruo ret = -EINVAL; 249021a852b0SQu Wenruo } 249121a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 249221a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 249321a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 249421a852b0SQu Wenruo ret = -EINVAL; 249521a852b0SQu Wenruo } 249621a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 249721a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 249821a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 249921a852b0SQu Wenruo ret = -EINVAL; 250021a852b0SQu Wenruo } 250121a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 250221a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 250321a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 250421a852b0SQu Wenruo ret = -EINVAL; 250521a852b0SQu Wenruo } 250621a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 250721a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 250821a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 250921a852b0SQu Wenruo ret = -EINVAL; 251021a852b0SQu Wenruo } 251121a852b0SQu Wenruo 251221a852b0SQu Wenruo /* 251321a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 251421a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 251521a852b0SQu Wenruo */ 251621a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 251721a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 251821a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 251921a852b0SQu Wenruo ret = -EINVAL; 252021a852b0SQu Wenruo } 25210bb3eb3eSQu Wenruo 25220bb3eb3eSQu Wenruo /* 25231a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 25241a42daabSQu Wenruo * 25251a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 25261a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 25271a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 25281a42daabSQu Wenruo * beginning. 25290bb3eb3eSQu Wenruo */ 25301a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 253121a852b0SQu Wenruo btrfs_err(fs_info, 25320bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 253321a852b0SQu Wenruo sectorsize, PAGE_SIZE); 253421a852b0SQu Wenruo ret = -EINVAL; 253521a852b0SQu Wenruo } 25360bb3eb3eSQu Wenruo 253721a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 253821a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 253921a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 254021a852b0SQu Wenruo ret = -EINVAL; 254121a852b0SQu Wenruo } 254221a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 254321a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 254421a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 254521a852b0SQu Wenruo ret = -EINVAL; 254621a852b0SQu Wenruo } 254721a852b0SQu Wenruo 254821a852b0SQu Wenruo /* Root alignment check */ 254921a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 255021a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 255121a852b0SQu Wenruo btrfs_super_root(sb)); 255221a852b0SQu Wenruo ret = -EINVAL; 255321a852b0SQu Wenruo } 255421a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 255521a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 255621a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 255721a852b0SQu Wenruo ret = -EINVAL; 255821a852b0SQu Wenruo } 255921a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 256021a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 256121a852b0SQu Wenruo btrfs_super_log_root(sb)); 256221a852b0SQu Wenruo ret = -EINVAL; 256321a852b0SQu Wenruo } 256421a852b0SQu Wenruo 2565aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2566aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2567aefd7f70SNikolay Borisov btrfs_err(fs_info, 2568aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2569aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2570aefd7f70SNikolay Borisov ret = -EINVAL; 2571aefd7f70SNikolay Borisov } 2572aefd7f70SNikolay Borisov 2573aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2574aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2575aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2576aefd7f70SNikolay Borisov btrfs_err(fs_info, 2577aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2578aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2579aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2580aefd7f70SNikolay Borisov ret = -EINVAL; 2581aefd7f70SNikolay Borisov } 2582aefd7f70SNikolay Borisov 25831c56ab99SQu Wenruo /* 25841c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 25851c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 25861c56ab99SQu Wenruo */ 25871c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 25881c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 25891c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 25901c56ab99SQu Wenruo btrfs_err(fs_info, 25911c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 25921c56ab99SQu Wenruo ret = -EINVAL; 25931c56ab99SQu Wenruo } 25941c56ab99SQu Wenruo 2595de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 25967239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 259721a852b0SQu Wenruo btrfs_err(fs_info, 25987239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2599de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 260021a852b0SQu Wenruo ret = -EINVAL; 260121a852b0SQu Wenruo } 260221a852b0SQu Wenruo 260321a852b0SQu Wenruo /* 260421a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 260521a852b0SQu Wenruo * done later 260621a852b0SQu Wenruo */ 260721a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 260821a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 260921a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 261021a852b0SQu Wenruo ret = -EINVAL; 261121a852b0SQu Wenruo } 261221a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 261321a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 261421a852b0SQu Wenruo btrfs_super_stripesize(sb)); 261521a852b0SQu Wenruo ret = -EINVAL; 261621a852b0SQu Wenruo } 261721a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 261821a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 261921a852b0SQu Wenruo btrfs_super_num_devices(sb)); 262021a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 262121a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 262221a852b0SQu Wenruo ret = -EINVAL; 262321a852b0SQu Wenruo } 262421a852b0SQu Wenruo 2625069ec957SQu Wenruo if (mirror_num >= 0 && 2626069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 262721a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 262821a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 262921a852b0SQu Wenruo ret = -EINVAL; 263021a852b0SQu Wenruo } 263121a852b0SQu Wenruo 263221a852b0SQu Wenruo /* 263321a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 263421a852b0SQu Wenruo * and one chunk 263521a852b0SQu Wenruo */ 263621a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 263721a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 263821a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 263921a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 264021a852b0SQu Wenruo ret = -EINVAL; 264121a852b0SQu Wenruo } 264221a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 264321a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 264421a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 264521a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 264621a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 264721a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 264821a852b0SQu Wenruo ret = -EINVAL; 264921a852b0SQu Wenruo } 265021a852b0SQu Wenruo 265121a852b0SQu Wenruo /* 265221a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 265321a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 265421a852b0SQu Wenruo */ 265521a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 265621a852b0SQu Wenruo btrfs_warn(fs_info, 265721a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 265821a852b0SQu Wenruo btrfs_super_generation(sb), 265921a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 266021a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 266121a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 266221a852b0SQu Wenruo btrfs_warn(fs_info, 266321a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 266421a852b0SQu Wenruo btrfs_super_generation(sb), 266521a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 266621a852b0SQu Wenruo 266721a852b0SQu Wenruo return ret; 266821a852b0SQu Wenruo } 266921a852b0SQu Wenruo 2670069ec957SQu Wenruo /* 2671069ec957SQu Wenruo * Validation of super block at mount time. 2672069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2673069ec957SQu Wenruo * flags will be skipped. 2674069ec957SQu Wenruo */ 2675069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2676069ec957SQu Wenruo { 2677a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2678069ec957SQu Wenruo } 2679069ec957SQu Wenruo 268075cb857dSQu Wenruo /* 268175cb857dSQu Wenruo * Validation of super block at write time. 268275cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 268375cb857dSQu Wenruo * overwritten soon. 268475cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 268575cb857dSQu Wenruo */ 268675cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 268775cb857dSQu Wenruo struct btrfs_super_block *sb) 268875cb857dSQu Wenruo { 268975cb857dSQu Wenruo int ret; 269075cb857dSQu Wenruo 2691a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 269275cb857dSQu Wenruo if (ret < 0) 269375cb857dSQu Wenruo goto out; 2694e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 269575cb857dSQu Wenruo ret = -EUCLEAN; 269675cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 269775cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 269875cb857dSQu Wenruo goto out; 269975cb857dSQu Wenruo } 270075cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 270175cb857dSQu Wenruo ret = -EUCLEAN; 270275cb857dSQu Wenruo btrfs_err(fs_info, 270375cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 270475cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 270575cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 270675cb857dSQu Wenruo goto out; 270775cb857dSQu Wenruo } 270875cb857dSQu Wenruo out: 270975cb857dSQu Wenruo if (ret < 0) 271075cb857dSQu Wenruo btrfs_err(fs_info, 271175cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 271275cb857dSQu Wenruo return ret; 271375cb857dSQu Wenruo } 271475cb857dSQu Wenruo 2715bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2716bd676446SJosef Bacik { 2717789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 2718789d6a3aSQu Wenruo .level = level, 2719789d6a3aSQu Wenruo .transid = gen, 2720789d6a3aSQu Wenruo .owner_root = root->root_key.objectid 2721789d6a3aSQu Wenruo }; 2722bd676446SJosef Bacik int ret = 0; 2723bd676446SJosef Bacik 2724789d6a3aSQu Wenruo root->node = read_tree_block(root->fs_info, bytenr, &check); 2725bd676446SJosef Bacik if (IS_ERR(root->node)) { 2726bd676446SJosef Bacik ret = PTR_ERR(root->node); 2727bd676446SJosef Bacik root->node = NULL; 27284eb150d6SQu Wenruo return ret; 27294eb150d6SQu Wenruo } 27304eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2731bd676446SJosef Bacik free_extent_buffer(root->node); 2732bd676446SJosef Bacik root->node = NULL; 27334eb150d6SQu Wenruo return -EIO; 2734bd676446SJosef Bacik } 2735bd676446SJosef Bacik 2736bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2737bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2738bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2739bd676446SJosef Bacik return ret; 2740bd676446SJosef Bacik } 2741bd676446SJosef Bacik 2742bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2743bd676446SJosef Bacik { 2744bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2745bd676446SJosef Bacik u64 gen, bytenr; 2746bd676446SJosef Bacik int level, ret; 2747bd676446SJosef Bacik 2748bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2749bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2750bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2751bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 27529c54e80dSJosef Bacik if (ret) { 2753bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2754bd676446SJosef Bacik return ret; 2755bd676446SJosef Bacik } 27569c54e80dSJosef Bacik return 0; 27579c54e80dSJosef Bacik } 27589c54e80dSJosef Bacik 27596ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2760b8522a1eSNikolay Borisov { 27616ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2762b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2763b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2764b8522a1eSNikolay Borisov bool handle_error = false; 2765b8522a1eSNikolay Borisov int ret = 0; 2766b8522a1eSNikolay Borisov int i; 2767b8522a1eSNikolay Borisov 2768b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2769b8522a1eSNikolay Borisov if (handle_error) { 2770b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2771b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2772b8522a1eSNikolay Borisov tree_root->node = NULL; 2773b8522a1eSNikolay Borisov 2774b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2775b8522a1eSNikolay Borisov break; 2776b8522a1eSNikolay Borisov 2777b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2778b8522a1eSNikolay Borisov 2779b8522a1eSNikolay Borisov /* 2780b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2781b8522a1eSNikolay Borisov * valid 2782b8522a1eSNikolay Borisov */ 2783b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2784b8522a1eSNikolay Borisov 2785b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2786b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2787b8522a1eSNikolay Borisov 2788745806fbSQu Wenruo btrfs_warn(fs_info, "try to load backup roots slot %d", i); 2789b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 27906ef108ddSNikolay Borisov backup_index = ret; 2791b8522a1eSNikolay Borisov if (ret < 0) 2792b8522a1eSNikolay Borisov return ret; 2793b8522a1eSNikolay Borisov } 2794b8522a1eSNikolay Borisov 2795bd676446SJosef Bacik ret = load_important_roots(fs_info); 2796bd676446SJosef Bacik if (ret) { 2797217f5004SNikolay Borisov handle_error = true; 2798b8522a1eSNikolay Borisov continue; 2799b8522a1eSNikolay Borisov } 2800b8522a1eSNikolay Borisov 2801336a0d8dSNikolay Borisov /* 2802336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2803336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2804336a0d8dSNikolay Borisov */ 2805453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2806b8522a1eSNikolay Borisov if (ret < 0) { 2807b8522a1eSNikolay Borisov handle_error = true; 2808b8522a1eSNikolay Borisov continue; 2809b8522a1eSNikolay Borisov } 2810b8522a1eSNikolay Borisov 28116b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2812b8522a1eSNikolay Borisov 2813b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2814b8522a1eSNikolay Borisov if (ret < 0) { 2815b8522a1eSNikolay Borisov handle_error = true; 2816b8522a1eSNikolay Borisov continue; 2817b8522a1eSNikolay Borisov } 2818b8522a1eSNikolay Borisov 2819b8522a1eSNikolay Borisov /* All successful */ 2820bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2821bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2822d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 28236ef108ddSNikolay Borisov 28246ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 28256ef108ddSNikolay Borisov if (backup_index < 0) { 28266ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 28276ef108ddSNikolay Borisov } else { 28286ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 28296ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 28306ef108ddSNikolay Borisov } 2831b8522a1eSNikolay Borisov break; 2832b8522a1eSNikolay Borisov } 2833b8522a1eSNikolay Borisov 2834b8522a1eSNikolay Borisov return ret; 2835b8522a1eSNikolay Borisov } 2836b8522a1eSNikolay Borisov 28378260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2838eb60ceacSChris Mason { 2839fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 284001cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 28418fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2842facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 284324bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2844eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 284511833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2846eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2847a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2848fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 284924bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 28504cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2851ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2852f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 285347ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 285440ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2855afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2856c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2857f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2858abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2859d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2860f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 28617585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2862573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 28630bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 28645f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 2865de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 286619c00ddcSChris Mason 2867e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 28685a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 28698b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 28705f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 28713e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 28723e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 28733e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 28743e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 28753e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 28763e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 28773e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 28783e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 2879e1489b4fSIoannis Angelakopoulos 28800b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 28816324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2882f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 288347ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 288418bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 2885afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 2886bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2887bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 28883fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 28893fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2890bd647ce3SJosef Bacik #endif 2891c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 289266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 289366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 289466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 289566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 289666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 289766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 289866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2899ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2900ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2901ba2c4d4eSJosef Bacik 2902771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 29034cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2904034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2905fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2906abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 290795ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 29089ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 29094cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2910a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2911f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 29128b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2913fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2914c8b97818SChris Mason 2915b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2916b34b086cSChris Mason num_online_cpus() + 2, 8); 29170afbaf8cSChris Mason 2918199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2919199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 292069fe2d75SJosef Bacik 2921638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 292221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 292321adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 292421adbd5cSStefan Behrens #endif 2925779a65a4SEric Sandeen btrfs_init_balance(fs_info); 292657056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2927a2de733cSArne Jansen 292816b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 292908dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 29300f9dd46cSJosef Bacik 2931fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 293235da5a7eSDavid Sterba IO_TREE_FS_EXCLUDED_EXTENTS); 293339279cc3SChris Mason 29345a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2935e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2936925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2937a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2938a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 29391bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 29409e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2941c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 294276dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2943803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2944fa9c0d79SChris Mason 2945ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2946f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2947b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2948416ac51dSArne Jansen 2949fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2950fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2951fa9c0d79SChris Mason 2952e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2953f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2954bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 29554854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2956034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 29573768f368SChris Mason 2958da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2959da17066cSJeff Mahoney fs_info->nodesize = 4096; 2960da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2961ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2962da17066cSJeff Mahoney fs_info->stripesize = 4096; 2963da17066cSJeff Mahoney 2964f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 2965f7b12a62SNaohiro Aota 2966eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2967eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2968eede2bf3SOmar Sandoval 296918bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 297018bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 29718260edbaSJosef Bacik } 29728260edbaSJosef Bacik 29738260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 29748260edbaSJosef Bacik { 29758260edbaSJosef Bacik int ret; 29768260edbaSJosef Bacik 29778260edbaSJosef Bacik fs_info->sb = sb; 29788260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 29798260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 29809e967495SFilipe Manana 29815deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 2982ae18c37aSJosef Bacik if (ret) 2983ae18c37aSJosef Bacik return ret; 2984ae18c37aSJosef Bacik 2985ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2986ae18c37aSJosef Bacik if (ret) 2987c75e8394SJosef Bacik return ret; 2988ae18c37aSJosef Bacik 2989ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2990ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2991ae18c37aSJosef Bacik 2992ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2993ae18c37aSJosef Bacik if (ret) 2994c75e8394SJosef Bacik return ret; 2995ae18c37aSJosef Bacik 2996ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2997ae18c37aSJosef Bacik GFP_KERNEL); 2998ae18c37aSJosef Bacik if (ret) 2999c75e8394SJosef Bacik return ret; 3000ae18c37aSJosef Bacik 3001ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 3002ae18c37aSJosef Bacik GFP_KERNEL); 3003c75e8394SJosef Bacik if (!fs_info->delayed_root) 3004c75e8394SJosef Bacik return -ENOMEM; 3005ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 3006ae18c37aSJosef Bacik 3007a0a1db70SFilipe Manana if (sb_rdonly(sb)) 3008a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 3009a0a1db70SFilipe Manana 3010c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 3011ae18c37aSJosef Bacik } 3012ae18c37aSJosef Bacik 301397f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 301497f4dd09SNikolay Borisov { 30150d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 301697f4dd09SNikolay Borisov int ret; 301797f4dd09SNikolay Borisov 301897f4dd09SNikolay Borisov /* 301997f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 302097f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 302197f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 302297f4dd09SNikolay Borisov */ 302397f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 302497f4dd09SNikolay Borisov if (ret < 0) { 3025c94bec2cSJosef Bacik if (ret != -EINTR) 3026c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 3027c94bec2cSJosef Bacik ret); 302897f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 302997f4dd09SNikolay Borisov return ret; 303097f4dd09SNikolay Borisov } 303197f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 303297f4dd09SNikolay Borisov } 303397f4dd09SNikolay Borisov 303497f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 303597f4dd09SNikolay Borisov { 303697f4dd09SNikolay Borisov struct task_struct *task; 303797f4dd09SNikolay Borisov 303897f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 303997f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 304097f4dd09SNikolay Borisov if (IS_ERR(task)) { 304197f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 304297f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 304397f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 304497f4dd09SNikolay Borisov return PTR_ERR(task); 304597f4dd09SNikolay Borisov } 304697f4dd09SNikolay Borisov 304797f4dd09SNikolay Borisov return 0; 304897f4dd09SNikolay Borisov } 304997f4dd09SNikolay Borisov 305044c0ca21SBoris Burkov /* 30518cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 30528cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 30538cd29088SBoris Burkov * code paths. 30548cd29088SBoris Burkov */ 30558cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 30568cd29088SBoris Burkov { 30578cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 30588b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 30598cd29088SBoris Burkov } 30608cd29088SBoris Burkov 30618cd29088SBoris Burkov /* 306244c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 306344c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 306444c0ca21SBoris Burkov */ 306544c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 306644c0ca21SBoris Burkov { 306744c0ca21SBoris Burkov int ret; 306894846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 30691d6a4fc8SQu Wenruo bool rebuild_free_space_tree = false; 30708b228324SBoris Burkov 30718b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 30728b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 30731d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 30748b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 30758b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 30768b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 30771d6a4fc8SQu Wenruo rebuild_free_space_tree = true; 30788b228324SBoris Burkov } 30798b228324SBoris Burkov 30801d6a4fc8SQu Wenruo if (rebuild_free_space_tree) { 30811d6a4fc8SQu Wenruo btrfs_info(fs_info, "rebuilding free space tree"); 30821d6a4fc8SQu Wenruo ret = btrfs_rebuild_free_space_tree(fs_info); 30838b228324SBoris Burkov if (ret) { 30848b228324SBoris Burkov btrfs_warn(fs_info, 30851d6a4fc8SQu Wenruo "failed to rebuild free space tree: %d", ret); 30861d6a4fc8SQu Wenruo goto out; 30871d6a4fc8SQu Wenruo } 30881d6a4fc8SQu Wenruo } 30891d6a4fc8SQu Wenruo 30901d6a4fc8SQu Wenruo if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 30911d6a4fc8SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { 30921d6a4fc8SQu Wenruo btrfs_info(fs_info, "disabling free space tree"); 30931d6a4fc8SQu Wenruo ret = btrfs_delete_free_space_tree(fs_info); 30941d6a4fc8SQu Wenruo if (ret) { 30951d6a4fc8SQu Wenruo btrfs_warn(fs_info, 30961d6a4fc8SQu Wenruo "failed to disable free space tree: %d", ret); 30978b228324SBoris Burkov goto out; 30988b228324SBoris Burkov } 30998b228324SBoris Burkov } 310044c0ca21SBoris Burkov 31018d488a8cSFilipe Manana /* 31028d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 31038d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 3104fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 31058d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 31068d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 31078d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 31088d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 31098d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 31108d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 31118d488a8cSFilipe Manana */ 31128d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 31138d488a8cSFilipe Manana if (ret) 31148d488a8cSFilipe Manana goto out; 31158d488a8cSFilipe Manana 311644c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 311744c0ca21SBoris Burkov if (ret) 311844c0ca21SBoris Burkov goto out; 311944c0ca21SBoris Burkov 31208f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 31218f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 31228f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 31238f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 31248f1c21d7SBoris Burkov goto out; 31258f1c21d7SBoris Burkov } 31268f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 31278f1c21d7SBoris Burkov 312844c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 31297eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 313044c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 313144c0ca21SBoris Burkov if (ret < 0) { 313244c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 313344c0ca21SBoris Burkov goto out; 313444c0ca21SBoris Burkov } 313544c0ca21SBoris Burkov 31365011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 31375011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 31385011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 31395011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 31405011139aSBoris Burkov if (ret) { 31415011139aSBoris Burkov btrfs_warn(fs_info, 31425011139aSBoris Burkov "failed to create free space tree: %d", ret); 31435011139aSBoris Burkov goto out; 31445011139aSBoris Burkov } 31455011139aSBoris Burkov } 31465011139aSBoris Burkov 314794846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 314894846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 314994846229SBoris Burkov if (ret) 315094846229SBoris Burkov goto out; 315194846229SBoris Burkov } 315294846229SBoris Burkov 315344c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 315444c0ca21SBoris Burkov if (ret) 315544c0ca21SBoris Burkov goto out; 315644c0ca21SBoris Burkov 315744c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 315844c0ca21SBoris Burkov if (ret) { 315944c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 316044c0ca21SBoris Burkov goto out; 316144c0ca21SBoris Burkov } 316244c0ca21SBoris Burkov 316344c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 316444c0ca21SBoris Burkov 316544c0ca21SBoris Burkov if (!fs_info->uuid_root) { 316644c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 316744c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 316844c0ca21SBoris Burkov if (ret) { 316944c0ca21SBoris Burkov btrfs_warn(fs_info, 317044c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 317144c0ca21SBoris Burkov goto out; 317244c0ca21SBoris Burkov } 317344c0ca21SBoris Burkov } 317444c0ca21SBoris Burkov 317544c0ca21SBoris Burkov out: 317644c0ca21SBoris Burkov return ret; 317744c0ca21SBoris Burkov } 317844c0ca21SBoris Burkov 3179d7f67ac9SQu Wenruo /* 3180d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3181d7f67ac9SQu Wenruo * 31822ba48b20SQu Wenruo * @is_rw_mount: If the mount is read-write. 31832ba48b20SQu Wenruo * 3184d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3185d7f67ac9SQu Wenruo * feature dependencies). 3186d7f67ac9SQu Wenruo * 3187d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3188d7f67ac9SQu Wenruo * btrfs_validate_super(). 3189d7f67ac9SQu Wenruo * 3190d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3191d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3192d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3193d7f67ac9SQu Wenruo */ 31942ba48b20SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) 3195d7f67ac9SQu Wenruo { 3196d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3197d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3198d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3199d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3200d7f67ac9SQu Wenruo 3201d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3202d7f67ac9SQu Wenruo btrfs_err(fs_info, 3203d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3204d7f67ac9SQu Wenruo incompat); 3205d7f67ac9SQu Wenruo return -EINVAL; 3206d7f67ac9SQu Wenruo } 3207d7f67ac9SQu Wenruo 3208d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3209d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3210d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3211d7f67ac9SQu Wenruo btrfs_err(fs_info, 3212d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3213d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3214d7f67ac9SQu Wenruo return -EINVAL; 3215d7f67ac9SQu Wenruo } 3216d7f67ac9SQu Wenruo 3217d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3218d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3219d7f67ac9SQu Wenruo 3220d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3221d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3222d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3223d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3224d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3225d7f67ac9SQu Wenruo 3226d7f67ac9SQu Wenruo /* 3227d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3228d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3229d7f67ac9SQu Wenruo */ 3230d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3231d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3232d7f67ac9SQu Wenruo 32332ba48b20SQu Wenruo if (compat_ro_unsupp && is_rw_mount) { 3234d7f67ac9SQu Wenruo btrfs_err(fs_info, 3235d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3236d7f67ac9SQu Wenruo compat_ro); 3237d7f67ac9SQu Wenruo return -EINVAL; 3238d7f67ac9SQu Wenruo } 3239d7f67ac9SQu Wenruo 3240d7f67ac9SQu Wenruo /* 3241d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3242d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3243d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3244d7f67ac9SQu Wenruo */ 3245d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3246d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3247d7f67ac9SQu Wenruo btrfs_err(fs_info, 3248d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3249d7f67ac9SQu Wenruo compat_ro); 3250d7f67ac9SQu Wenruo return -EINVAL; 3251d7f67ac9SQu Wenruo } 3252d7f67ac9SQu Wenruo 3253d7f67ac9SQu Wenruo /* 3254d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3255d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3256d7f67ac9SQu Wenruo */ 3257d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3258d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3259d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3260d7f67ac9SQu Wenruo btrfs_err(fs_info, 3261d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3262d7f67ac9SQu Wenruo return -EINVAL; 3263d7f67ac9SQu Wenruo } 3264d7f67ac9SQu Wenruo 3265d7f67ac9SQu Wenruo /* 3266d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3267d7f67ac9SQu Wenruo * 3268d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3269d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3270d7f67ac9SQu Wenruo * going to be deprecated anyway. 3271d7f67ac9SQu Wenruo */ 3272d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3273d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3274d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3275d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3276d7f67ac9SQu Wenruo return -EINVAL; 3277d7f67ac9SQu Wenruo } 3278d7f67ac9SQu Wenruo 3279d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3280d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3281d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3282d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3283d7f67ac9SQu Wenruo 3284d7f67ac9SQu Wenruo return 0; 3285d7f67ac9SQu Wenruo } 3286d7f67ac9SQu Wenruo 3287ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3288ae18c37aSJosef Bacik char *options) 3289ae18c37aSJosef Bacik { 3290ae18c37aSJosef Bacik u32 sectorsize; 3291ae18c37aSJosef Bacik u32 nodesize; 3292ae18c37aSJosef Bacik u32 stripesize; 3293ae18c37aSJosef Bacik u64 generation; 3294ae18c37aSJosef Bacik u64 features; 3295ae18c37aSJosef Bacik u16 csum_type; 3296ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3297ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3298ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3299ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3300ae18c37aSJosef Bacik int ret; 3301ae18c37aSJosef Bacik int level; 3302ae18c37aSJosef Bacik 33038260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 33044871c33bSQu Wenruo if (ret) 3305ae18c37aSJosef Bacik goto fail; 330653b381b3SDavid Woodhouse 3307ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3308ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3309ae18c37aSJosef Bacik GFP_KERNEL); 3310ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3311ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3312ae18c37aSJosef Bacik GFP_KERNEL); 3313ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3314ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 33154871c33bSQu Wenruo ret = -ENOMEM; 3316c75e8394SJosef Bacik goto fail; 3317ae18c37aSJosef Bacik } 3318ae18c37aSJosef Bacik 3319dcb2137cSChristoph Hellwig ret = btrfs_init_btree_inode(sb); 33204871c33bSQu Wenruo if (ret) 3321c75e8394SJosef Bacik goto fail; 3322ae18c37aSJosef Bacik 3323d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 33241104a885SDavid Sterba 33251104a885SDavid Sterba /* 33261104a885SDavid Sterba * Read super block and check the signature bytes only 33271104a885SDavid Sterba */ 3328d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 33298f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 33304871c33bSQu Wenruo ret = PTR_ERR(disk_super); 333116cdcec7SMiao Xie goto fail_alloc; 333220b45077SDave Young } 333339279cc3SChris Mason 33341104a885SDavid Sterba /* 3335260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 33368dc3f22cSJohannes Thumshirn * corrupted, we'll find out 33378dc3f22cSJohannes Thumshirn */ 33388f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 333951bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 33408dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 334151bce6c9SJohannes Thumshirn csum_type); 33424871c33bSQu Wenruo ret = -EINVAL; 33438f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 33448dc3f22cSJohannes Thumshirn goto fail_alloc; 33458dc3f22cSJohannes Thumshirn } 33468dc3f22cSJohannes Thumshirn 334783c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 334883c68bbcSSu Yue 33496d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 33506d97c6e3SJohannes Thumshirn if (ret) { 33518f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 33526d97c6e3SJohannes Thumshirn goto fail_alloc; 33536d97c6e3SJohannes Thumshirn } 33546d97c6e3SJohannes Thumshirn 33558dc3f22cSJohannes Thumshirn /* 33561104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 33571104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 33581104a885SDavid Sterba */ 33593d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 336005135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 33614871c33bSQu Wenruo ret = -EINVAL; 33628f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3363141386e1SJosef Bacik goto fail_alloc; 33641104a885SDavid Sterba } 33651104a885SDavid Sterba 33661104a885SDavid Sterba /* 33671104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 33681104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 33691104a885SDavid Sterba * the whole block of INFO_SIZE 33701104a885SDavid Sterba */ 33718f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 33728f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 33735f39d397SChris Mason 3374fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3375fbc6feaeSNikolay Borisov 33760b86a832SChris Mason 3377fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3378fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3379fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3380fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3381fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3382fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3383fbc6feaeSNikolay Borisov } 3384fbc6feaeSNikolay Borisov 3385fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3386fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3387de37aa51SNikolay Borisov 3388069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 33891104a885SDavid Sterba if (ret) { 339005135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 33914871c33bSQu Wenruo ret = -EINVAL; 3392141386e1SJosef Bacik goto fail_alloc; 33931104a885SDavid Sterba } 33941104a885SDavid Sterba 33954871c33bSQu Wenruo if (!btrfs_super_root(disk_super)) { 33964871c33bSQu Wenruo btrfs_err(fs_info, "invalid superblock tree root bytenr"); 33974871c33bSQu Wenruo ret = -EINVAL; 3398141386e1SJosef Bacik goto fail_alloc; 33994871c33bSQu Wenruo } 34000f7d52f4SChris Mason 3401acce952bSliubo /* check FS state, whether FS is broken. */ 340287533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 340387533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3404acce952bSliubo 340575e7cb7fSLiu Bo /* 340675e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 340775e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 340875e7cb7fSLiu Bo */ 340975e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 341075e7cb7fSLiu Bo 34116f93e834SAnand Jain 34126f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 34136f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 34146f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 34156f93e834SAnand Jain stripesize = sectorsize; 34166f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 34176f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 34186f93e834SAnand Jain 34196f93e834SAnand Jain fs_info->nodesize = nodesize; 34206f93e834SAnand Jain fs_info->sectorsize = sectorsize; 34216f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 34226f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 34236f93e834SAnand Jain fs_info->stripesize = stripesize; 34246f93e834SAnand Jain 34252ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 34264871c33bSQu Wenruo if (ret) 3427141386e1SJosef Bacik goto fail_alloc; 3428dfe25020SChris Mason 34292ba48b20SQu Wenruo ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); 34304871c33bSQu Wenruo if (ret < 0) 3431141386e1SJosef Bacik goto fail_alloc; 3432f2b636e8SJosef Bacik 34338481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 34348481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 34358481dd80SQu Wenruo 34369f73f1aeSQu Wenruo /* 34379f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 34389f73f1aeSQu Wenruo * going to be deprecated. 34399f73f1aeSQu Wenruo * 34409f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 34419f73f1aeSQu Wenruo */ 34429f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 34439f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 34449f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 34459f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 34469f73f1aeSQu Wenruo 344795ea0486SQu Wenruo btrfs_warn(fs_info, 344895ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 34490bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 34508481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 34514871c33bSQu Wenruo if (!subpage_info) { 34524871c33bSQu Wenruo ret = -ENOMEM; 34538481dd80SQu Wenruo goto fail_alloc; 34544871c33bSQu Wenruo } 34558481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 34568481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3457c8050b3bSQu Wenruo } 34580bb3eb3eSQu Wenruo 3459d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 34604871c33bSQu Wenruo if (ret) 34610dc3b84aSJosef Bacik goto fail_sb_buffer; 34624543df7eSChris Mason 34639e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 34649e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 34654575c9ccSChris Mason 3466a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3467a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3468de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3469db94535dSChris Mason 3470925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 34716bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3472925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 347384eed90fSChris Mason if (ret) { 347405135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 34755d4f98a2SYan Zheng goto fail_sb_buffer; 347684eed90fSChris Mason } 34770b86a832SChris Mason 347884234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3479581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3480bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3481bd676446SJosef Bacik generation, level); 3482bd676446SJosef Bacik if (ret) { 348305135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3484af31f5e5SChris Mason goto fail_tree_roots; 348583121942SDavid Woodhouse } 34860b86a832SChris Mason 3487e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3488c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3489c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3490e17cade2SChris Mason 34915b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 34922b82032cSYan Zheng if (ret) { 349305135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3494af31f5e5SChris Mason goto fail_tree_roots; 34952b82032cSYan Zheng } 34960b86a832SChris Mason 34978dabb742SStefan Behrens /* 3498bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3499bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3500bacce86aSAnand Jain * target is required. So free devices that are not part of this 35011a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3502bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 35038dabb742SStefan Behrens */ 3504bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3505d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 350605135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 35074871c33bSQu Wenruo ret = -EIO; 3508a6b0d5c8SChris Mason goto fail_tree_roots; 3509a6b0d5c8SChris Mason } 3510a6b0d5c8SChris Mason 3511b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 35124bbcaa64SEric Sandeen if (ret) 3513b8522a1eSNikolay Borisov goto fail_tree_roots; 35148929ecfaSYan, Zheng 351575ec1db8SJosef Bacik /* 351673651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 351773651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 351873651042SNaohiro Aota * zoned incompat feature flag set. 351973651042SNaohiro Aota */ 352073651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 352173651042SNaohiro Aota if (ret) { 352273651042SNaohiro Aota btrfs_err(fs_info, 35234871c33bSQu Wenruo "zoned: failed to read device zone info: %d", ret); 352473651042SNaohiro Aota goto fail_block_groups; 352573651042SNaohiro Aota } 352673651042SNaohiro Aota 352773651042SNaohiro Aota /* 352875ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 352975ec1db8SJosef Bacik * check the generation here so we can set the 353075ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 353175ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 353275ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 353375ec1db8SJosef Bacik * even though it was perfectly fine. 353475ec1db8SJosef Bacik */ 353575ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 353675ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 353775ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 353875ec1db8SJosef Bacik 3539cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3540cf90d884SQu Wenruo if (ret) { 3541cf90d884SQu Wenruo btrfs_err(fs_info, 3542cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3543cf90d884SQu Wenruo ret); 3544cf90d884SQu Wenruo goto fail_block_groups; 3545cf90d884SQu Wenruo } 354668310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 354768310a5eSIlya Dryomov if (ret) { 354805135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 354968310a5eSIlya Dryomov goto fail_block_groups; 355068310a5eSIlya Dryomov } 355168310a5eSIlya Dryomov 3552733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3553733f4fbbSStefan Behrens if (ret) { 355405135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3555733f4fbbSStefan Behrens goto fail_block_groups; 3556733f4fbbSStefan Behrens } 3557733f4fbbSStefan Behrens 35588dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 35598dabb742SStefan Behrens if (ret) { 356005135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 35618dabb742SStefan Behrens goto fail_block_groups; 35628dabb742SStefan Behrens } 35638dabb742SStefan Behrens 3564b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3565b70f5097SNaohiro Aota if (ret) { 3566b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3567b70f5097SNaohiro Aota ret); 3568b70f5097SNaohiro Aota goto fail_block_groups; 3569b70f5097SNaohiro Aota } 3570b70f5097SNaohiro Aota 3571c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3572b7c35e81SAnand Jain if (ret) { 357305135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 357405135f59SDavid Sterba ret); 3575b7c35e81SAnand Jain goto fail_block_groups; 3576b7c35e81SAnand Jain } 3577b7c35e81SAnand Jain 357896f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 35795ac1d209SJeff Mahoney if (ret) { 358005135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3581b7c35e81SAnand Jain goto fail_fsdev_sysfs; 35825ac1d209SJeff Mahoney } 35835ac1d209SJeff Mahoney 3584c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3585c59021f8Sliubo if (ret) { 358605135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 35872365dd3cSAnand Jain goto fail_sysfs; 3588c59021f8Sliubo } 3589c59021f8Sliubo 35905b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 35911b1d1f66SJosef Bacik if (ret) { 359205135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 35932365dd3cSAnand Jain goto fail_sysfs; 35941b1d1f66SJosef Bacik } 35954330e183SQu Wenruo 359616beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 359716beac87SNaohiro Aota 35985c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 35995c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 360005135f59SDavid Sterba btrfs_warn(fs_info, 360152042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 36024871c33bSQu Wenruo ret = -EINVAL; 36032365dd3cSAnand Jain goto fail_sysfs; 3604292fd7fcSStefan Behrens } 36059078a3e1SChris Mason 360633c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3607a74a4b97SChris Mason "btrfs-cleaner"); 36084871c33bSQu Wenruo if (IS_ERR(fs_info->cleaner_kthread)) { 36094871c33bSQu Wenruo ret = PTR_ERR(fs_info->cleaner_kthread); 36102365dd3cSAnand Jain goto fail_sysfs; 36114871c33bSQu Wenruo } 3612a74a4b97SChris Mason 3613a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3614a74a4b97SChris Mason tree_root, 3615a74a4b97SChris Mason "btrfs-transaction"); 36164871c33bSQu Wenruo if (IS_ERR(fs_info->transaction_kthread)) { 36174871c33bSQu Wenruo ret = PTR_ERR(fs_info->transaction_kthread); 36183f157a2fSChris Mason goto fail_cleaner; 36194871c33bSQu Wenruo } 3620a74a4b97SChris Mason 3621583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3622c289811cSChris Mason !fs_info->fs_devices->rotating) { 3623583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3624c289811cSChris Mason } 3625c289811cSChris Mason 3626572d9ab7SDavid Sterba /* 362763a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 362863a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 362963a7cb13SDavid Sterba * nodiscard for the same mount. 363063a7cb13SDavid Sterba */ 363163a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 363263a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 363363a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 363463a7cb13SDavid Sterba fs_info->fs_devices->discardable) { 363563a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 363663a7cb13SDavid Sterba "auto enabling async discard"); 363763a7cb13SDavid Sterba } 363863a7cb13SDavid Sterba 363921adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 36400b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 36412ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 36420b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3643cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 364421adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 364521adbd5cSStefan Behrens if (ret) 364605135f59SDavid Sterba btrfs_warn(fs_info, 364705135f59SDavid Sterba "failed to initialize integrity check module: %d", 364805135f59SDavid Sterba ret); 364921adbd5cSStefan Behrens } 365021adbd5cSStefan Behrens #endif 3651bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3652bcef60f2SArne Jansen if (ret) 3653bcef60f2SArne Jansen goto fail_trans_kthread; 365421adbd5cSStefan Behrens 3655fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3656fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3657fd708b81SJosef Bacik 365896da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 365996da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 36600b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3661e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 366263443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 36634871c33bSQu Wenruo if (ret) 366428c16cbbSWang Shilong goto fail_qgroup; 3665e556ce2cSYan Zheng } 36661a40e23bSZheng Yan 366756e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 36683140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 36694871c33bSQu Wenruo ret = PTR_ERR(fs_info->fs_root); 36704871c33bSQu Wenruo btrfs_warn(fs_info, "failed to read fs tree: %d", ret); 3671315bf8efSJosef Bacik fs_info->fs_root = NULL; 3672bcef60f2SArne Jansen goto fail_qgroup; 36733140c9a3SDan Carpenter } 3674c289811cSChris Mason 3675bc98a42cSDavid Howells if (sb_rdonly(sb)) 36768cd29088SBoris Burkov goto clear_oneshot; 36772b6ba629SIlya Dryomov 367844c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 36792b6ba629SIlya Dryomov if (ret) { 36806bccf3abSJeff Mahoney close_ctree(fs_info); 36812b6ba629SIlya Dryomov return ret; 3682e3acc2a6SJosef Bacik } 3683b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3684b382a324SJan Schmidt 368544c0ca21SBoris Burkov if (fs_info->uuid_root && 368644c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 368744c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 368805135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 368970f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 369070f80175SStefan Behrens if (ret) { 369105135f59SDavid Sterba btrfs_warn(fs_info, 369205135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 36936bccf3abSJeff Mahoney close_ctree(fs_info); 369470f80175SStefan Behrens return ret; 369570f80175SStefan Behrens } 3696f7a81ea4SStefan Behrens } 369794846229SBoris Burkov 3698afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 369947ab2a6cSJosef Bacik 3700b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3701b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3702b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3703b4be6aefSJosef Bacik 37048cd29088SBoris Burkov clear_oneshot: 37058cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3706ad2b2c80SAl Viro return 0; 370739279cc3SChris Mason 3708bcef60f2SArne Jansen fail_qgroup: 3709bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 37107c2ca468SChris Mason fail_trans_kthread: 37117c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 37122ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3713faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 37143f157a2fSChris Mason fail_cleaner: 3715a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 37167c2ca468SChris Mason 37177c2ca468SChris Mason /* 37187c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 37197c2ca468SChris Mason * kthreads 37207c2ca468SChris Mason */ 37217c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 37227c2ca468SChris Mason 37232365dd3cSAnand Jain fail_sysfs: 37246618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 37252365dd3cSAnand Jain 3726b7c35e81SAnand Jain fail_fsdev_sysfs: 3727b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3728b7c35e81SAnand Jain 37291b1d1f66SJosef Bacik fail_block_groups: 373054067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3731af31f5e5SChris Mason 3732af31f5e5SChris Mason fail_tree_roots: 37339e3aa805SJosef Bacik if (fs_info->data_reloc_root) 37349e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 37354273eaffSAnand Jain free_root_pointers(fs_info, true); 37362b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3737af31f5e5SChris Mason 373839279cc3SChris Mason fail_sb_buffer: 37397abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 37405cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 374116cdcec7SMiao Xie fail_alloc: 3742586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3743586e46e2SIlya Dryomov 37444543df7eSChris Mason iput(fs_info->btree_inode); 37457e662854SQinghuang Feng fail: 3746586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 37474871c33bSQu Wenruo ASSERT(ret < 0); 37484871c33bSQu Wenruo return ret; 3749eb60ceacSChris Mason } 3750663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3751eb60ceacSChris Mason 3752314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3753f2984462SChris Mason { 3754314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3755314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3756314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3757314b6dd0SJohannes Thumshirn struct page *page; 3758442a4f63SStefan Behrens 3759314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3760314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3761314b6dd0SJohannes Thumshirn 3762314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3763fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3764314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3765cb3e217bSQu Wenruo btrfs_dev_name(device), 3766314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3767314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3768314b6dd0SJohannes Thumshirn SetPageError(page); 3769314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3770314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3771314b6dd0SJohannes Thumshirn } else { 3772314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3773f2984462SChris Mason } 3774314b6dd0SJohannes Thumshirn 3775314b6dd0SJohannes Thumshirn put_page(page); 3776314b6dd0SJohannes Thumshirn unlock_page(page); 3777314b6dd0SJohannes Thumshirn } 3778314b6dd0SJohannes Thumshirn 3779314b6dd0SJohannes Thumshirn bio_put(bio); 3780f2984462SChris Mason } 3781f2984462SChris Mason 37828f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3783a05d3c91SQu Wenruo int copy_num, bool drop_cache) 378429c36d72SAnand Jain { 378529c36d72SAnand Jain struct btrfs_super_block *super; 37868f32380dSJohannes Thumshirn struct page *page; 378712659251SNaohiro Aota u64 bytenr, bytenr_orig; 37888f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 378912659251SNaohiro Aota int ret; 379029c36d72SAnand Jain 379112659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 379212659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 379312659251SNaohiro Aota if (ret == -ENOENT) 379412659251SNaohiro Aota return ERR_PTR(-EINVAL); 379512659251SNaohiro Aota else if (ret) 379612659251SNaohiro Aota return ERR_PTR(ret); 379712659251SNaohiro Aota 3798cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 37998f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 380029c36d72SAnand Jain 3801a05d3c91SQu Wenruo if (drop_cache) { 3802a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3803a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3804a05d3c91SQu Wenruo 3805a05d3c91SQu Wenruo /* 3806a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3807a05d3c91SQu Wenruo * always read from the device. 3808a05d3c91SQu Wenruo */ 3809a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3810a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3811a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3812a05d3c91SQu Wenruo } 3813a05d3c91SQu Wenruo 38148f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 38158f32380dSJohannes Thumshirn if (IS_ERR(page)) 38168f32380dSJohannes Thumshirn return ERR_CAST(page); 381729c36d72SAnand Jain 38188f32380dSJohannes Thumshirn super = page_address(page); 381996c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 382096c2e067SAnand Jain btrfs_release_disk_super(super); 382196c2e067SAnand Jain return ERR_PTR(-ENODATA); 382296c2e067SAnand Jain } 382396c2e067SAnand Jain 382412659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 38258f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 38268f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 382729c36d72SAnand Jain } 382829c36d72SAnand Jain 38298f32380dSJohannes Thumshirn return super; 383029c36d72SAnand Jain } 383129c36d72SAnand Jain 383229c36d72SAnand Jain 38338f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3834a512bbf8SYan Zheng { 38358f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3836a512bbf8SYan Zheng int i; 3837a512bbf8SYan Zheng u64 transid = 0; 3838a512bbf8SYan Zheng 3839a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3840a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3841a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3842a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3843a512bbf8SYan Zheng */ 3844a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3845a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 38468f32380dSJohannes Thumshirn if (IS_ERR(super)) 3847a512bbf8SYan Zheng continue; 3848a512bbf8SYan Zheng 3849a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 38508f32380dSJohannes Thumshirn if (latest) 38518f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 38528f32380dSJohannes Thumshirn 38538f32380dSJohannes Thumshirn latest = super; 3854a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3855a512bbf8SYan Zheng } 3856a512bbf8SYan Zheng } 385792fc03fbSAnand Jain 38588f32380dSJohannes Thumshirn return super; 3859a512bbf8SYan Zheng } 3860a512bbf8SYan Zheng 38614eedeb75SHisashi Hifumi /* 3862abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3863314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 38644eedeb75SHisashi Hifumi * 3865abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3866abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3867abbb3b8eSDavid Sterba * same for write and wait phases. 38684eedeb75SHisashi Hifumi * 3869314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 38704eedeb75SHisashi Hifumi */ 3871a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3872abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3873a512bbf8SYan Zheng { 3874d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3875314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3876d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3877a512bbf8SYan Zheng int i; 3878a512bbf8SYan Zheng int errors = 0; 387912659251SNaohiro Aota int ret; 388012659251SNaohiro Aota u64 bytenr, bytenr_orig; 3881a512bbf8SYan Zheng 3882a512bbf8SYan Zheng if (max_mirrors == 0) 3883a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3884a512bbf8SYan Zheng 3885d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3886d5178578SJohannes Thumshirn 3887a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3888314b6dd0SJohannes Thumshirn struct page *page; 3889314b6dd0SJohannes Thumshirn struct bio *bio; 3890314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3891314b6dd0SJohannes Thumshirn 389212659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 389312659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 389412659251SNaohiro Aota if (ret == -ENOENT) { 389512659251SNaohiro Aota continue; 389612659251SNaohiro Aota } else if (ret < 0) { 389712659251SNaohiro Aota btrfs_err(device->fs_info, 389812659251SNaohiro Aota "couldn't get super block location for mirror %d", 389912659251SNaohiro Aota i); 390012659251SNaohiro Aota errors++; 390112659251SNaohiro Aota continue; 390212659251SNaohiro Aota } 3903935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3904935e5cc9SMiao Xie device->commit_total_bytes) 3905a512bbf8SYan Zheng break; 3906a512bbf8SYan Zheng 390712659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3908a512bbf8SYan Zheng 3909fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3910fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3911fd08001fSEric Biggers sb->csum); 3912a512bbf8SYan Zheng 3913314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3914314b6dd0SJohannes Thumshirn GFP_NOFS); 3915314b6dd0SJohannes Thumshirn if (!page) { 3916fb456252SJeff Mahoney btrfs_err(device->fs_info, 3917314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3918f14d104dSDavid Sterba bytenr); 3919634554dcSJosef Bacik errors++; 3920634554dcSJosef Bacik continue; 3921634554dcSJosef Bacik } 3922634554dcSJosef Bacik 3923314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3924314b6dd0SJohannes Thumshirn get_page(page); 3925a512bbf8SYan Zheng 3926314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3927314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3928a512bbf8SYan Zheng 3929387125fcSChris Mason /* 3930314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3931314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3932314b6dd0SJohannes Thumshirn * checking. 3933387125fcSChris Mason */ 393407888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 393507888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 393607888c66SChristoph Hellwig GFP_NOFS); 3937314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3938314b6dd0SJohannes Thumshirn bio->bi_private = device; 3939314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3940314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3941314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3942314b6dd0SJohannes Thumshirn 3943314b6dd0SJohannes Thumshirn /* 3944314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3945314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3946314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3947314b6dd0SJohannes Thumshirn */ 39481b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3949314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3950314b6dd0SJohannes Thumshirn 395158ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 395258ff51f1SChristoph Hellwig submit_bio(bio); 39538376d9e1SNaohiro Aota 39548376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 39558376d9e1SNaohiro Aota errors++; 3956a512bbf8SYan Zheng } 3957a512bbf8SYan Zheng return errors < i ? 0 : -1; 3958a512bbf8SYan Zheng } 3959a512bbf8SYan Zheng 3960387125fcSChris Mason /* 3961abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3962abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3963abbb3b8eSDavid Sterba * 3964314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3965abbb3b8eSDavid Sterba * date. 3966abbb3b8eSDavid Sterba */ 3967abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3968abbb3b8eSDavid Sterba { 3969abbb3b8eSDavid Sterba int i; 3970abbb3b8eSDavid Sterba int errors = 0; 3971b6a535faSHoward McLauchlan bool primary_failed = false; 397212659251SNaohiro Aota int ret; 3973abbb3b8eSDavid Sterba u64 bytenr; 3974abbb3b8eSDavid Sterba 3975abbb3b8eSDavid Sterba if (max_mirrors == 0) 3976abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3977abbb3b8eSDavid Sterba 3978abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3979314b6dd0SJohannes Thumshirn struct page *page; 3980314b6dd0SJohannes Thumshirn 398112659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 398212659251SNaohiro Aota if (ret == -ENOENT) { 398312659251SNaohiro Aota break; 398412659251SNaohiro Aota } else if (ret < 0) { 398512659251SNaohiro Aota errors++; 398612659251SNaohiro Aota if (i == 0) 398712659251SNaohiro Aota primary_failed = true; 398812659251SNaohiro Aota continue; 398912659251SNaohiro Aota } 3990abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3991abbb3b8eSDavid Sterba device->commit_total_bytes) 3992abbb3b8eSDavid Sterba break; 3993abbb3b8eSDavid Sterba 3994314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3995314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3996314b6dd0SJohannes Thumshirn if (!page) { 3997abbb3b8eSDavid Sterba errors++; 3998b6a535faSHoward McLauchlan if (i == 0) 3999b6a535faSHoward McLauchlan primary_failed = true; 4000abbb3b8eSDavid Sterba continue; 4001abbb3b8eSDavid Sterba } 4002314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 4003314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 4004314b6dd0SJohannes Thumshirn if (PageError(page)) { 4005abbb3b8eSDavid Sterba errors++; 4006b6a535faSHoward McLauchlan if (i == 0) 4007b6a535faSHoward McLauchlan primary_failed = true; 4008b6a535faSHoward McLauchlan } 4009abbb3b8eSDavid Sterba 4010314b6dd0SJohannes Thumshirn /* Drop our reference */ 4011314b6dd0SJohannes Thumshirn put_page(page); 4012abbb3b8eSDavid Sterba 4013314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 4014314b6dd0SJohannes Thumshirn put_page(page); 4015abbb3b8eSDavid Sterba } 4016abbb3b8eSDavid Sterba 4017b6a535faSHoward McLauchlan /* log error, force error return */ 4018b6a535faSHoward McLauchlan if (primary_failed) { 4019b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 4020b6a535faSHoward McLauchlan device->devid); 4021b6a535faSHoward McLauchlan return -1; 4022b6a535faSHoward McLauchlan } 4023b6a535faSHoward McLauchlan 4024abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 4025abbb3b8eSDavid Sterba } 4026abbb3b8eSDavid Sterba 4027abbb3b8eSDavid Sterba /* 4028387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 4029387125fcSChris Mason * for the barrier when it is done 4030387125fcSChris Mason */ 40314246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 4032387125fcSChris Mason { 4033f9e69aa9SChristoph Hellwig bio_uninit(bio); 4034387125fcSChris Mason complete(bio->bi_private); 4035387125fcSChris Mason } 4036387125fcSChris Mason 4037387125fcSChris Mason /* 40384fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 40394fc6441aSAnand Jain * done in the waiting counterpart. 4040387125fcSChris Mason */ 40414fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 4042387125fcSChris Mason { 4043f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 4044387125fcSChris Mason 4045bfd3ea94SAnand Jain device->last_flush_error = BLK_STS_OK; 4046bfd3ea94SAnand Jain 4047a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 4048a91cf0ffSWang Yugui /* 4049a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 4050a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 4051a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 4052a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 4053a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 4054a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 4055a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 4056a91cf0ffSWang Yugui * non-debug builds. 4057a91cf0ffSWang Yugui */ 405808e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 40594fc6441aSAnand Jain return; 4060a91cf0ffSWang Yugui #endif 4061387125fcSChris Mason 4062f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 4063f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 4064387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 4065387125fcSChris Mason init_completion(&device->flush_wait); 4066387125fcSChris Mason bio->bi_private = &device->flush_wait; 4067387125fcSChris Mason 406858ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 406958ff51f1SChristoph Hellwig submit_bio(bio); 40701c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 40714fc6441aSAnand Jain } 4072387125fcSChris Mason 40734fc6441aSAnand Jain /* 40744fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 40751b465784SAnand Jain * Return true for any error, and false otherwise. 40764fc6441aSAnand Jain */ 40771b465784SAnand Jain static bool wait_dev_flush(struct btrfs_device *device) 40784fc6441aSAnand Jain { 4079f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 40804fc6441aSAnand Jain 40817e812f20SAnand Jain if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 40821b465784SAnand Jain return false; 40834fc6441aSAnand Jain 40842980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 40854fc6441aSAnand Jain 4086bfd3ea94SAnand Jain if (bio->bi_status) { 4087bfd3ea94SAnand Jain device->last_flush_error = bio->bi_status; 4088bfd3ea94SAnand Jain btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); 40891b465784SAnand Jain return true; 4090bfd3ea94SAnand Jain } 4091bfd3ea94SAnand Jain 40921b465784SAnand Jain return false; 4093387125fcSChris Mason } 4094387125fcSChris Mason 4095387125fcSChris Mason /* 4096387125fcSChris Mason * send an empty flush down to each device in parallel, 4097387125fcSChris Mason * then wait for them 4098387125fcSChris Mason */ 4099387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 4100387125fcSChris Mason { 4101387125fcSChris Mason struct list_head *head; 4102387125fcSChris Mason struct btrfs_device *dev; 41035af3e8ccSStefan Behrens int errors_wait = 0; 4104387125fcSChris Mason 41051538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 4106387125fcSChris Mason /* send down all the barriers */ 4107387125fcSChris Mason head = &info->fs_devices->devices; 41081538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4109e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4110f88ba6a2SHidetoshi Seto continue; 4111cea7c8bfSAnand Jain if (!dev->bdev) 4112387125fcSChris Mason continue; 4113e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4114ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4115387125fcSChris Mason continue; 4116387125fcSChris Mason 41174fc6441aSAnand Jain write_dev_flush(dev); 4118387125fcSChris Mason } 4119387125fcSChris Mason 4120387125fcSChris Mason /* wait for all the barriers */ 41211538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4122e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4123f88ba6a2SHidetoshi Seto continue; 4124387125fcSChris Mason if (!dev->bdev) { 41255af3e8ccSStefan Behrens errors_wait++; 4126387125fcSChris Mason continue; 4127387125fcSChris Mason } 4128e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4129ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4130387125fcSChris Mason continue; 4131387125fcSChris Mason 41321b465784SAnand Jain if (wait_dev_flush(dev)) 41335af3e8ccSStefan Behrens errors_wait++; 4134387125fcSChris Mason } 4135401b41e5SAnand Jain 4136401b41e5SAnand Jain /* 4137de38a206SAnand Jain * Checks last_flush_error of disks in order to determine the device 4138de38a206SAnand Jain * state. 4139401b41e5SAnand Jain */ 4140de38a206SAnand Jain if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) 4141de38a206SAnand Jain return -EIO; 4142de38a206SAnand Jain 4143387125fcSChris Mason return 0; 4144387125fcSChris Mason } 4145387125fcSChris Mason 4146943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4147943c6e99SZhao Lei { 41488789f4feSZhao Lei int raid_type; 41498789f4feSZhao Lei int min_tolerated = INT_MAX; 4150943c6e99SZhao Lei 41518789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 41528789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 41538c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 41548789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 41558789f4feSZhao Lei tolerated_failures); 4156943c6e99SZhao Lei 41578789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 41588789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 41598789f4feSZhao Lei continue; 416041a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 41618789f4feSZhao Lei continue; 41628c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 41638789f4feSZhao Lei btrfs_raid_array[raid_type]. 41648789f4feSZhao Lei tolerated_failures); 41658789f4feSZhao Lei } 4166943c6e99SZhao Lei 41678789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4168ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 41698789f4feSZhao Lei min_tolerated = 0; 41708789f4feSZhao Lei } 41718789f4feSZhao Lei 41728789f4feSZhao Lei return min_tolerated; 4173943c6e99SZhao Lei } 4174943c6e99SZhao Lei 4175eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4176f2984462SChris Mason { 4177e5e9a520SChris Mason struct list_head *head; 4178f2984462SChris Mason struct btrfs_device *dev; 4179a061fc8dSChris Mason struct btrfs_super_block *sb; 4180f2984462SChris Mason struct btrfs_dev_item *dev_item; 4181f2984462SChris Mason int ret; 4182f2984462SChris Mason int do_barriers; 4183a236aed1SChris Mason int max_errors; 4184a236aed1SChris Mason int total_errors = 0; 4185a061fc8dSChris Mason u64 flags; 4186f2984462SChris Mason 41870b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4188fed3b381SLiu Bo 4189fed3b381SLiu Bo /* 4190fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4191fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4192fed3b381SLiu Bo * been consistent on disk. 4193fed3b381SLiu Bo */ 4194fed3b381SLiu Bo if (max_mirrors == 0) 41950b246afaSJeff Mahoney backup_super_roots(fs_info); 4196f2984462SChris Mason 41970b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4198a061fc8dSChris Mason dev_item = &sb->dev_item; 4199e5e9a520SChris Mason 42000b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 42010b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 42020b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4203387125fcSChris Mason 42045af3e8ccSStefan Behrens if (do_barriers) { 42050b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 42065af3e8ccSStefan Behrens if (ret) { 42075af3e8ccSStefan Behrens mutex_unlock( 42080b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 42090b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 42105af3e8ccSStefan Behrens "errors while submitting device barriers."); 42115af3e8ccSStefan Behrens return ret; 42125af3e8ccSStefan Behrens } 42135af3e8ccSStefan Behrens } 4214387125fcSChris Mason 42151538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4216dfe25020SChris Mason if (!dev->bdev) { 4217dfe25020SChris Mason total_errors++; 4218dfe25020SChris Mason continue; 4219dfe25020SChris Mason } 4220e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4221ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4222dfe25020SChris Mason continue; 4223dfe25020SChris Mason 42242b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4225a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4226a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 42277df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4228935e5cc9SMiao Xie dev->commit_total_bytes); 4229ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4230ce7213c7SMiao Xie dev->commit_bytes_used); 4231a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4232a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4233a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4234a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 42357239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 42367239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4237a512bbf8SYan Zheng 4238a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4239a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4240f2984462SChris Mason 424175cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 424275cb857dSQu Wenruo if (ret < 0) { 424375cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 424475cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 424575cb857dSQu Wenruo "unexpected superblock corruption detected"); 424675cb857dSQu Wenruo return -EUCLEAN; 424775cb857dSQu Wenruo } 424875cb857dSQu Wenruo 4249abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4250a236aed1SChris Mason if (ret) 4251a236aed1SChris Mason total_errors++; 4252f2984462SChris Mason } 4253a236aed1SChris Mason if (total_errors > max_errors) { 42540b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4255d397712bSChris Mason total_errors); 42560b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 425779787eaaSJeff Mahoney 42589d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 42590b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 42600b246afaSJeff Mahoney "%d errors while writing supers", 42610b246afaSJeff Mahoney total_errors); 42629d565ba4SStefan Behrens return -EIO; 4263a236aed1SChris Mason } 4264f2984462SChris Mason 4265a512bbf8SYan Zheng total_errors = 0; 42661538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4267dfe25020SChris Mason if (!dev->bdev) 4268dfe25020SChris Mason continue; 4269e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4270ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4271dfe25020SChris Mason continue; 4272dfe25020SChris Mason 4273abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4274a512bbf8SYan Zheng if (ret) 42751259ab75SChris Mason total_errors++; 4276f2984462SChris Mason } 42770b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4278a236aed1SChris Mason if (total_errors > max_errors) { 42790b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 42800b246afaSJeff Mahoney "%d errors while writing supers", 42810b246afaSJeff Mahoney total_errors); 428279787eaaSJeff Mahoney return -EIO; 4283a236aed1SChris Mason } 4284f2984462SChris Mason return 0; 4285f2984462SChris Mason } 4286f2984462SChris Mason 4287cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4288cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4289cb517eabSMiao Xie struct btrfs_root *root) 42902619ba1fSChris Mason { 42914785e24fSJosef Bacik bool drop_ref = false; 42924785e24fSJosef Bacik 4293fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4294fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4295fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4296fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 42974785e24fSJosef Bacik drop_ref = true; 4298fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 429976dda93cSYan, Zheng 430084961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4301ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 43021c1ea4f7SLiu Bo if (root->reloc_root) { 430300246528SJosef Bacik btrfs_put_root(root->reloc_root); 43041c1ea4f7SLiu Bo root->reloc_root = NULL; 43051c1ea4f7SLiu Bo } 43061c1ea4f7SLiu Bo } 43073321719eSLiu Bo 43084785e24fSJosef Bacik if (drop_ref) 430900246528SJosef Bacik btrfs_put_root(root); 43102619ba1fSChris Mason } 43112619ba1fSChris Mason 4312c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4313c146afadSYan Zheng { 4314fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4315fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4316fc7cbcd4SDavid Sterba int i = 0; 431765d33fd7SQu Wenruo int err = 0; 4318fc7cbcd4SDavid Sterba unsigned int ret = 0; 4319c146afadSYan Zheng 4320c146afadSYan Zheng while (1) { 4321fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4322fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4323fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4324fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4325fc7cbcd4SDavid Sterba if (!ret) { 4326fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4327c146afadSYan Zheng break; 432865d33fd7SQu Wenruo } 4329fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 433066b4ffd1SJosef Bacik 4331fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4332fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4333fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4334fc7cbcd4SDavid Sterba gang[i] = NULL; 433565d33fd7SQu Wenruo continue; 4336c146afadSYan Zheng } 4337fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4338fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4339fc7cbcd4SDavid Sterba } 4340fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4341fc7cbcd4SDavid Sterba 4342fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4343fc7cbcd4SDavid Sterba if (!gang[i]) 4344fc7cbcd4SDavid Sterba continue; 4345fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4346fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4347fc7cbcd4SDavid Sterba if (err) 43486989627dSJosef Bacik goto out; 4349fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4350fc7cbcd4SDavid Sterba } 4351fc7cbcd4SDavid Sterba root_objectid++; 4352c146afadSYan Zheng } 43536989627dSJosef Bacik out: 4354fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4355fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4356fc7cbcd4SDavid Sterba if (gang[i]) 4357fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 435865d33fd7SQu Wenruo } 435965d33fd7SQu Wenruo return err; 4360c146afadSYan Zheng } 4361c146afadSYan Zheng 43626bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4363c146afadSYan Zheng { 43646bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4365c146afadSYan Zheng struct btrfs_trans_handle *trans; 4366c146afadSYan Zheng 43670b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 43682ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 43690b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 43700b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4371c71bf099SYan, Zheng 4372c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 43730b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 43740b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4375c71bf099SYan, Zheng 43767a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 43773612b495STsutomu Itoh if (IS_ERR(trans)) 43783612b495STsutomu Itoh return PTR_ERR(trans); 43793a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4380c146afadSYan Zheng } 4381c146afadSYan Zheng 438236c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 438336c86a9eSQu Wenruo { 438436c86a9eSQu Wenruo struct btrfs_transaction *trans; 438536c86a9eSQu Wenruo struct btrfs_transaction *tmp; 438636c86a9eSQu Wenruo bool found = false; 438736c86a9eSQu Wenruo 438836c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 438936c86a9eSQu Wenruo return; 439036c86a9eSQu Wenruo 439136c86a9eSQu Wenruo /* 439236c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 439336c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 439436c86a9eSQu Wenruo */ 439536c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 439636c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 439736c86a9eSQu Wenruo struct extent_state *cached = NULL; 439836c86a9eSQu Wenruo u64 dirty_bytes = 0; 439936c86a9eSQu Wenruo u64 cur = 0; 440036c86a9eSQu Wenruo u64 found_start; 440136c86a9eSQu Wenruo u64 found_end; 440236c86a9eSQu Wenruo 440336c86a9eSQu Wenruo found = true; 440436c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 440536c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 440636c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 440736c86a9eSQu Wenruo cur = found_end + 1; 440836c86a9eSQu Wenruo } 440936c86a9eSQu Wenruo btrfs_warn(fs_info, 441036c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 441136c86a9eSQu Wenruo trans->transid, dirty_bytes); 441236c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 441336c86a9eSQu Wenruo 441436c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 441536c86a9eSQu Wenruo fs_info->running_transaction = NULL; 441636c86a9eSQu Wenruo list_del_init(&trans->list); 441736c86a9eSQu Wenruo 441836c86a9eSQu Wenruo btrfs_put_transaction(trans); 441936c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 442036c86a9eSQu Wenruo } 442136c86a9eSQu Wenruo ASSERT(!found); 442236c86a9eSQu Wenruo } 442336c86a9eSQu Wenruo 4424b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4425eb60ceacSChris Mason { 4426c146afadSYan Zheng int ret; 4427e089f05cSChris Mason 4428afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 442931e70e52SFilipe Manana 443031e70e52SFilipe Manana /* 44318a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 44328a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 44338a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 44348a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 44358a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 44368a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 44378a1f1e3dSFilipe Manana * return 1. 44388a1f1e3dSFilipe Manana */ 44398a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 44408a1f1e3dSFilipe Manana 44418a1f1e3dSFilipe Manana /* 444231e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 444331e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 444431e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 444531e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 444631e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 444731e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 444831e70e52SFilipe Manana * trying to stop the async reclaim task. 444931e70e52SFilipe Manana */ 445031e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4451d6fd0ae2SOmar Sandoval /* 4452d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4453d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4454d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4455d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4456d6fd0ae2SOmar Sandoval */ 4457d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4458a2135011SChris Mason 44597343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4460d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 44617343dd61SJustin Maggard 4462803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4463803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4464803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4465803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4466803b2f54SStefan Behrens 4467837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4468aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4469837d5b6eSIlya Dryomov 44708dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 44718dabb742SStefan Behrens 4472aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 44734cb5300bSChris Mason 44744cb5300bSChris Mason /* wait for any defraggers to finish */ 44754cb5300bSChris Mason wait_event(fs_info->transaction_wait, 44764cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 44774cb5300bSChris Mason 44784cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 447926176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 44804cb5300bSChris Mason 4481a362bb86SFilipe Manana /* 4482a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4483a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4484a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4485a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4486a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4487a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4488a362bb86SFilipe Manana * no one else to run iputs. 4489a362bb86SFilipe Manana * 4490a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4491a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4492a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4493a362bb86SFilipe Manana * through some other means. 4494a362bb86SFilipe Manana * 4495a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4496a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4497a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4498a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4499a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4500a362bb86SFilipe Manana */ 4501a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4502a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4503a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4504a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4505a362bb86SFilipe Manana 450621c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 450757056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4508576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 450921c7e756SMiao Xie 4510b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4511b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4512b0643e59SDennis Zhou 4513bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4514e44163e1SJeff Mahoney /* 4515d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4516d6fd0ae2SOmar Sandoval * unused block groups. 4517e44163e1SJeff Mahoney */ 45180b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4519e44163e1SJeff Mahoney 4520f0cc2cd7SFilipe Manana /* 4521f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4522f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4523f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4524f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4525f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4526f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4527f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4528f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4529f0cc2cd7SFilipe Manana * This is a very rare case. 4530f0cc2cd7SFilipe Manana */ 4531f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4532f0cc2cd7SFilipe Manana 45336bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4534d397712bSChris Mason if (ret) 453504892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4536c146afadSYan Zheng } 4537ed2ff2cbSChris Mason 453884961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 45392ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4540acce952bSliubo 4541e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4542e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 45438929ecfaSYan, Zheng 4544e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4545afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4546f25784b3SYan Zheng 45475958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 45485958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 45495958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 45505958253cSQu Wenruo } 45515958253cSQu Wenruo 455204892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4553fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4554bcef60f2SArne Jansen 4555963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 455604892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4557963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4558b0c68f8bSChris Mason } 455931153d81SYan Zheng 45605deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 45614297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 45625deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 45634297ff84SJosef Bacik 45646618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4565b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 45665ac1d209SJeff Mahoney 45671a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 45681a4319ccSLiu Bo 4569de348ee0SWang Shilong /* 4570de348ee0SWang Shilong * we must make sure there is not any read request to 4571de348ee0SWang Shilong * submit after we stopping all workers. 4572de348ee0SWang Shilong */ 4573de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 457496192499SJosef Bacik btrfs_stop_all_workers(fs_info); 457596192499SJosef Bacik 45760a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 457736c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 45780a31daa4SFilipe Manana 4579afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 45804273eaffSAnand Jain free_root_pointers(fs_info, true); 45818c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 45829ad6b7bcSChris Mason 45834e19443dSJosef Bacik /* 45844e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 45854e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 45864e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 45874e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 45884e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 45894e19443dSJosef Bacik */ 45904e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 45914e19443dSJosef Bacik 459213e6c37bSJosef Bacik iput(fs_info->btree_inode); 4593d6bfde87SChris Mason 459421adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 45950b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 45962ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 459721adbd5cSStefan Behrens #endif 459821adbd5cSStefan Behrens 45990b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 460068c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4601eb60ceacSChris Mason } 4602eb60ceacSChris Mason 46035f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 46045f39d397SChris Mason { 46052f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 46065f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4607b4ce94deSChris Mason 460806ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 460906ea65a3SJosef Bacik /* 461006ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 461152042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 461206ea65a3SJosef Bacik * outside of the sanity tests. 461306ea65a3SJosef Bacik */ 4614b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 461506ea65a3SJosef Bacik return; 461606ea65a3SJosef Bacik #endif 461749d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 46180b246afaSJeff Mahoney if (transid != fs_info->generation) 46195d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 46200b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 4621f18cc978SChristoph Hellwig set_extent_buffer_dirty(buf); 46221f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 462369fc6cbbSQu Wenruo /* 462485d8a826SJosef Bacik * btrfs_check_leaf() won't check item data if we don't have WRITTEN 462585d8a826SJosef Bacik * set, so this will only validate the basic structure of the items. 462669fc6cbbSQu Wenruo */ 462785d8a826SJosef Bacik if (btrfs_header_level(buf) == 0 && btrfs_check_leaf(buf)) { 4628a4f78750SDavid Sterba btrfs_print_leaf(buf); 46291f21ef0aSFilipe Manana ASSERT(0); 46301f21ef0aSFilipe Manana } 46311f21ef0aSFilipe Manana #endif 4632eb60ceacSChris Mason } 4633eb60ceacSChris Mason 46342ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4635b53d3f5dSLiu Bo int flush_delayed) 463635b7e476SChris Mason { 4637188de649SChris Mason /* 4638188de649SChris Mason * looks as though older kernels can get into trouble with 4639188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4640188de649SChris Mason */ 4641e2d84521SMiao Xie int ret; 4642d6bfde87SChris Mason 46436933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4644d6bfde87SChris Mason return; 4645d6bfde87SChris Mason 4646b53d3f5dSLiu Bo if (flush_delayed) 46472ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 464816cdcec7SMiao Xie 4649d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4650d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4651d814a491SEthan Lien fs_info->dirty_metadata_batch); 4652e2d84521SMiao Xie if (ret > 0) { 46530b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 465416cdcec7SMiao Xie } 465516cdcec7SMiao Xie } 465616cdcec7SMiao Xie 46572ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 465816cdcec7SMiao Xie { 46592ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 466035b7e476SChris Mason } 4661b53d3f5dSLiu Bo 46622ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4663b53d3f5dSLiu Bo { 46642ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4665d6bfde87SChris Mason } 46666b80053dSChris Mason 46672ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4668acce952bSliubo { 4669fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4670fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4671fe816d0fSNikolay Borisov 46720b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 46732ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 46740b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4675acce952bSliubo 46760b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 46770b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4678acce952bSliubo } 4679acce952bSliubo 4680ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4681ef67963dSJosef Bacik { 4682fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4683fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4684fc7cbcd4SDavid Sterba int ret; 4685ef67963dSJosef Bacik 4686fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4687fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4688fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4689fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4690fc7cbcd4SDavid Sterba int i; 4691ef67963dSJosef Bacik 4692fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4693fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4694fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4695fc7cbcd4SDavid Sterba 4696fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4697fc7cbcd4SDavid Sterba if (!gang[i]) 4698ef67963dSJosef Bacik continue; 4699fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4700fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4701fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4702ef67963dSJosef Bacik } 4703fc7cbcd4SDavid Sterba root_objectid++; 4704fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4705ef67963dSJosef Bacik } 4706fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4707ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4708ef67963dSJosef Bacik } 4709ef67963dSJosef Bacik 4710143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4711acce952bSliubo { 4712acce952bSliubo struct btrfs_ordered_extent *ordered; 4713acce952bSliubo 4714199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4715779880efSJosef Bacik /* 4716779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4717779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4718779880efSJosef Bacik */ 4719199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4720779880efSJosef Bacik root_extent_list) 4721779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4722199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4723199c2a9cSMiao Xie } 4724199c2a9cSMiao Xie 4725199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4726199c2a9cSMiao Xie { 4727199c2a9cSMiao Xie struct btrfs_root *root; 4728199c2a9cSMiao Xie struct list_head splice; 4729199c2a9cSMiao Xie 4730199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4731199c2a9cSMiao Xie 4732199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4733199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4734199c2a9cSMiao Xie while (!list_empty(&splice)) { 4735199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4736199c2a9cSMiao Xie ordered_root); 47371de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 47381de2cfdeSJosef Bacik &fs_info->ordered_roots); 4739199c2a9cSMiao Xie 47402a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4741199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4742199c2a9cSMiao Xie 47432a85d9caSLiu Bo cond_resched(); 47442a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4745199c2a9cSMiao Xie } 4746199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 474774d5d229SJosef Bacik 474874d5d229SJosef Bacik /* 474974d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 475074d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 475174d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 475274d5d229SJosef Bacik * actually get run and error out properly. 475374d5d229SJosef Bacik */ 475474d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4755acce952bSliubo } 4756acce952bSliubo 475735a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 47582ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4759acce952bSliubo { 4760acce952bSliubo struct rb_node *node; 4761acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4762acce952bSliubo struct btrfs_delayed_ref_node *ref; 4763acce952bSliubo int ret = 0; 4764acce952bSliubo 4765acce952bSliubo delayed_refs = &trans->delayed_refs; 4766acce952bSliubo 4767acce952bSliubo spin_lock(&delayed_refs->lock); 4768d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4769cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4770b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4771acce952bSliubo return ret; 4772acce952bSliubo } 4773acce952bSliubo 47745c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4775d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 47760e0adbcfSJosef Bacik struct rb_node *n; 4777e78417d1SJosef Bacik bool pin_bytes = false; 4778acce952bSliubo 4779d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4780d7df2c79SJosef Bacik href_node); 47813069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4782b939d1abSJosef Bacik continue; 47833069bd26SJosef Bacik 4784d7df2c79SJosef Bacik spin_lock(&head->lock); 4785e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 47860e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 47870e0adbcfSJosef Bacik ref_node); 4788d7df2c79SJosef Bacik ref->in_tree = 0; 4789e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 47900e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 47911d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 47921d57ee94SWang Xiaoguang list_del(&ref->add_list); 4793d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4794d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4795d7df2c79SJosef Bacik } 479654067ae9SJosef Bacik if (head->must_insert_reserved) 4797e78417d1SJosef Bacik pin_bytes = true; 479878a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4799fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4800d7df2c79SJosef Bacik spin_unlock(&head->lock); 4801acce952bSliubo spin_unlock(&delayed_refs->lock); 4802e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4803acce952bSliubo 4804f603bb94SNikolay Borisov if (pin_bytes) { 4805f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4806f603bb94SNikolay Borisov 4807f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4808f603bb94SNikolay Borisov BUG_ON(!cache); 4809f603bb94SNikolay Borisov 4810f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4811f603bb94SNikolay Borisov spin_lock(&cache->lock); 4812f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4813f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4814f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4815f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4816f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4817f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4818f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4819f603bb94SNikolay Borisov 4820f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4821f603bb94SNikolay Borisov 4822f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4823f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4824f603bb94SNikolay Borisov } 482531890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4826d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4827acce952bSliubo cond_resched(); 4828acce952bSliubo spin_lock(&delayed_refs->lock); 4829acce952bSliubo } 483081f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4831acce952bSliubo 4832acce952bSliubo spin_unlock(&delayed_refs->lock); 4833acce952bSliubo 4834acce952bSliubo return ret; 4835acce952bSliubo } 4836acce952bSliubo 4837143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4838acce952bSliubo { 4839acce952bSliubo struct btrfs_inode *btrfs_inode; 4840acce952bSliubo struct list_head splice; 4841acce952bSliubo 4842acce952bSliubo INIT_LIST_HEAD(&splice); 4843acce952bSliubo 4844eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4845eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4846acce952bSliubo 4847acce952bSliubo while (!list_empty(&splice)) { 4848fe816d0fSNikolay Borisov struct inode *inode = NULL; 4849eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4850acce952bSliubo delalloc_inodes); 4851fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4852eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4853acce952bSliubo 4854fe816d0fSNikolay Borisov /* 4855fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4856fe816d0fSNikolay Borisov * meanwhile. 4857fe816d0fSNikolay Borisov */ 4858fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4859fe816d0fSNikolay Borisov if (inode) { 4860597441b3SJosef Bacik unsigned int nofs_flag; 4861597441b3SJosef Bacik 4862597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4863fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4864597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4865fe816d0fSNikolay Borisov iput(inode); 4866fe816d0fSNikolay Borisov } 4867eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4868acce952bSliubo } 4869eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4870eb73c1b7SMiao Xie } 4871eb73c1b7SMiao Xie 4872eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4873eb73c1b7SMiao Xie { 4874eb73c1b7SMiao Xie struct btrfs_root *root; 4875eb73c1b7SMiao Xie struct list_head splice; 4876eb73c1b7SMiao Xie 4877eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4878eb73c1b7SMiao Xie 4879eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4880eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4881eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4882eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4883eb73c1b7SMiao Xie delalloc_root); 488400246528SJosef Bacik root = btrfs_grab_root(root); 4885eb73c1b7SMiao Xie BUG_ON(!root); 4886eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4887eb73c1b7SMiao Xie 4888eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 488900246528SJosef Bacik btrfs_put_root(root); 4890eb73c1b7SMiao Xie 4891eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4892eb73c1b7SMiao Xie } 4893eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4894acce952bSliubo } 4895acce952bSliubo 48962ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4897acce952bSliubo struct extent_io_tree *dirty_pages, 4898acce952bSliubo int mark) 4899acce952bSliubo { 4900acce952bSliubo int ret; 4901acce952bSliubo struct extent_buffer *eb; 4902acce952bSliubo u64 start = 0; 4903acce952bSliubo u64 end; 4904acce952bSliubo 4905acce952bSliubo while (1) { 4906acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4907e6138876SJosef Bacik mark, NULL); 4908acce952bSliubo if (ret) 4909acce952bSliubo break; 4910acce952bSliubo 491191166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4912acce952bSliubo while (start <= end) { 49130b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 49140b246afaSJeff Mahoney start += fs_info->nodesize; 4915fd8b2b61SJosef Bacik if (!eb) 4916acce952bSliubo continue; 4917acce952bSliubo 4918c4e54a65SJosef Bacik btrfs_tree_lock(eb); 4919c4e54a65SJosef Bacik wait_on_extent_buffer_writeback(eb); 4920190a8339SJosef Bacik btrfs_clear_buffer_dirty(NULL, eb); 4921c4e54a65SJosef Bacik btrfs_tree_unlock(eb); 4922c4e54a65SJosef Bacik 4923fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4924acce952bSliubo } 4925acce952bSliubo } 4926acce952bSliubo 4927acce952bSliubo return ret; 4928acce952bSliubo } 4929acce952bSliubo 49302ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4931fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4932acce952bSliubo { 4933acce952bSliubo u64 start; 4934acce952bSliubo u64 end; 4935acce952bSliubo int ret; 4936acce952bSliubo 4937acce952bSliubo while (1) { 49380e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 49390e6ec385SFilipe Manana 4940fcd5e742SLu Fengqi /* 4941fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4942fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4943fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4944fcd5e742SLu Fengqi * the same extent range. 4945fcd5e742SLu Fengqi */ 4946fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4947acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 49480e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4949fcd5e742SLu Fengqi if (ret) { 4950fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4951acce952bSliubo break; 4952fcd5e742SLu Fengqi } 4953acce952bSliubo 49540e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 49550e6ec385SFilipe Manana free_extent_state(cached_state); 49562ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4957fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4958acce952bSliubo cond_resched(); 4959acce952bSliubo } 4960acce952bSliubo 4961acce952bSliubo return 0; 4962acce952bSliubo } 4963acce952bSliubo 496432da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4965c79a1751SLiu Bo { 4966c79a1751SLiu Bo struct inode *inode; 4967c79a1751SLiu Bo 4968c79a1751SLiu Bo inode = cache->io_ctl.inode; 4969c79a1751SLiu Bo if (inode) { 4970597441b3SJosef Bacik unsigned int nofs_flag; 4971597441b3SJosef Bacik 4972597441b3SJosef Bacik nofs_flag = memalloc_nofs_save(); 4973c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4974597441b3SJosef Bacik memalloc_nofs_restore(nofs_flag); 4975597441b3SJosef Bacik 4976c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4977c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4978c79a1751SLiu Bo iput(inode); 4979c79a1751SLiu Bo } 4980bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4981c79a1751SLiu Bo btrfs_put_block_group(cache); 4982c79a1751SLiu Bo } 4983c79a1751SLiu Bo 4984c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 49852ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4986c79a1751SLiu Bo { 498732da5386SDavid Sterba struct btrfs_block_group *cache; 4988c79a1751SLiu Bo 4989c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4990c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4991c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 499232da5386SDavid Sterba struct btrfs_block_group, 4993c79a1751SLiu Bo dirty_list); 4994c79a1751SLiu Bo 4995c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4996c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4997c79a1751SLiu Bo list_del_init(&cache->io_list); 4998c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4999c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5000c79a1751SLiu Bo } 5001c79a1751SLiu Bo 5002c79a1751SLiu Bo list_del_init(&cache->dirty_list); 5003c79a1751SLiu Bo spin_lock(&cache->lock); 5004c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5005c79a1751SLiu Bo spin_unlock(&cache->lock); 5006c79a1751SLiu Bo 5007c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5008c79a1751SLiu Bo btrfs_put_block_group(cache); 5009ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 5010c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5011c79a1751SLiu Bo } 5012c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5013c79a1751SLiu Bo 501445ae2c18SNikolay Borisov /* 501545ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 501645ae2c18SNikolay Borisov * to use it without any locking 501745ae2c18SNikolay Borisov */ 5018c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 5019c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 502032da5386SDavid Sterba struct btrfs_block_group, 5021c79a1751SLiu Bo io_list); 5022c79a1751SLiu Bo 5023c79a1751SLiu Bo list_del_init(&cache->io_list); 5024c79a1751SLiu Bo spin_lock(&cache->lock); 5025c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5026c79a1751SLiu Bo spin_unlock(&cache->lock); 5027c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5028c79a1751SLiu Bo } 5029c79a1751SLiu Bo } 5030c79a1751SLiu Bo 503149b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 50322ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 503349b25e05SJeff Mahoney { 5034bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 5035bbbf7243SNikolay Borisov 50362ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 5037c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 5038c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 5039c79a1751SLiu Bo 5040bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 5041bbbf7243SNikolay Borisov post_commit_list) { 5042bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 5043bbbf7243SNikolay Borisov } 5044bbbf7243SNikolay Borisov 50452ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 504649b25e05SJeff Mahoney 50474a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 50480b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 504949b25e05SJeff Mahoney 50504a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 50510b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 505249b25e05SJeff Mahoney 5053ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 505467cde344SMiao Xie 50552ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 505649b25e05SJeff Mahoney EXTENT_DIRTY); 5057fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 505849b25e05SJeff Mahoney 50594a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 50604a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 506149b25e05SJeff Mahoney } 506249b25e05SJeff Mahoney 50632ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 5064acce952bSliubo { 5065acce952bSliubo struct btrfs_transaction *t; 5066acce952bSliubo 50670b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 5068acce952bSliubo 50690b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 50700b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 50710b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 5072724e2315SJosef Bacik struct btrfs_transaction, list); 5073724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 50749b64f57dSElena Reshetova refcount_inc(&t->use_count); 50750b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 50762ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 5077724e2315SJosef Bacik btrfs_put_transaction(t); 50780b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5079724e2315SJosef Bacik continue; 5080724e2315SJosef Bacik } 50810b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 5082724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 50830b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5084724e2315SJosef Bacik /* 5085724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 5086724e2315SJosef Bacik * handle open currently for this transaction. 5087724e2315SJosef Bacik */ 5088724e2315SJosef Bacik wait_event(t->writer_wait, 5089724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 5090724e2315SJosef Bacik } else { 50910b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5092724e2315SJosef Bacik } 50932ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 5094724e2315SJosef Bacik 50950b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 50960b246afaSJeff Mahoney if (t == fs_info->running_transaction) 50970b246afaSJeff Mahoney fs_info->running_transaction = NULL; 5098724e2315SJosef Bacik list_del_init(&t->list); 50990b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5100a4abeea4SJosef Bacik 5101724e2315SJosef Bacik btrfs_put_transaction(t); 51022e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 51030b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5104724e2315SJosef Bacik } 51050b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 51060b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 5107ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 5108ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 51090b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 5110ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 51110b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 5112acce952bSliubo 5113acce952bSliubo return 0; 5114acce952bSliubo } 5115ec7d6dfdSNikolay Borisov 5116453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 5117ec7d6dfdSNikolay Borisov { 5118ec7d6dfdSNikolay Borisov struct btrfs_path *path; 5119ec7d6dfdSNikolay Borisov int ret; 5120ec7d6dfdSNikolay Borisov struct extent_buffer *l; 5121ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 5122ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 5123ec7d6dfdSNikolay Borisov int slot; 5124ec7d6dfdSNikolay Borisov 5125ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 5126ec7d6dfdSNikolay Borisov if (!path) 5127ec7d6dfdSNikolay Borisov return -ENOMEM; 5128ec7d6dfdSNikolay Borisov 5129ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 5130ec7d6dfdSNikolay Borisov search_key.type = -1; 5131ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 5132ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 5133ec7d6dfdSNikolay Borisov if (ret < 0) 5134ec7d6dfdSNikolay Borisov goto error; 5135ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 5136ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 5137ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 5138ec7d6dfdSNikolay Borisov l = path->nodes[0]; 5139ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 514023125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 514123125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 5142ec7d6dfdSNikolay Borisov } else { 514323125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 5144ec7d6dfdSNikolay Borisov } 5145ec7d6dfdSNikolay Borisov ret = 0; 5146ec7d6dfdSNikolay Borisov error: 5147ec7d6dfdSNikolay Borisov btrfs_free_path(path); 5148ec7d6dfdSNikolay Borisov return ret; 5149ec7d6dfdSNikolay Borisov } 5150ec7d6dfdSNikolay Borisov 5151543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 5152ec7d6dfdSNikolay Borisov { 5153ec7d6dfdSNikolay Borisov int ret; 5154ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5155ec7d6dfdSNikolay Borisov 51566b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5157ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5158ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5159ec7d6dfdSNikolay Borisov root->root_key.objectid); 5160ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5161ec7d6dfdSNikolay Borisov goto out; 5162ec7d6dfdSNikolay Borisov } 5163ec7d6dfdSNikolay Borisov 516423125104SNikolay Borisov *objectid = root->free_objectid++; 5165ec7d6dfdSNikolay Borisov ret = 0; 5166ec7d6dfdSNikolay Borisov out: 5167ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5168ec7d6dfdSNikolay Borisov return ret; 5169ec7d6dfdSNikolay Borisov } 5170