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" 260b86a832SChris Mason #include "volumes.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" 49*45c40c8fSJosef Bacik #include "root-tree.h" 50eb60ceacSChris Mason 51319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 52319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 53319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 54319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 55e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 56e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 57319e4d06SQu Wenruo 58143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 59acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 602ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 61143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 622ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 63acce952bSliubo struct extent_io_tree *dirty_pages, 64acce952bSliubo int mark); 652ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 66acce952bSliubo struct extent_io_tree *pinned_extents); 672ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 682ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 69ce9adaa5SChris Mason 70141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 71141386e1SJosef Bacik { 72141386e1SJosef Bacik if (fs_info->csum_shash) 73141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 74141386e1SJosef Bacik } 75141386e1SJosef Bacik 76d352ac68SChris Mason /* 77d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 78d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 79d352ac68SChris Mason * just before they are sent down the IO stack. 80d352ac68SChris Mason */ 8144b8bd7eSChris Mason struct async_submit_bio { 828896a08dSQu Wenruo struct inode *inode; 8344b8bd7eSChris Mason struct bio *bio; 84a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 8544b8bd7eSChris Mason int mirror_num; 861941b64bSQu Wenruo 871941b64bSQu Wenruo /* Optional parameter for submit_bio_start used by direct io */ 881941b64bSQu Wenruo u64 dio_file_offset; 898b712842SChris Mason struct btrfs_work work; 904e4cbee9SChristoph Hellwig blk_status_t status; 9144b8bd7eSChris Mason }; 9244b8bd7eSChris Mason 9385d4e461SChris Mason /* 942996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 95d352ac68SChris Mason */ 96c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 9719c00ddcSChris Mason { 98d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 997280305eSDavid Sterba const int num_pages = num_extent_pages(buf); 100a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 101d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 10219c00ddcSChris Mason char *kaddr; 103e9be5a30SDavid Sterba int i; 104d5178578SJohannes Thumshirn 105d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 106d5178578SJohannes Thumshirn crypto_shash_init(shash); 107a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 108e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 109a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 11019c00ddcSChris Mason 111e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 112e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 113e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 11419c00ddcSChris Mason } 11571a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 116d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 11719c00ddcSChris Mason } 11819c00ddcSChris Mason 119d352ac68SChris Mason /* 120d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 121d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 122d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 123d352ac68SChris Mason * in the wrong place. 124d352ac68SChris Mason */ 1251259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 126b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 127b9fab919SChris Mason int atomic) 1281259ab75SChris Mason { 1292ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 1301259ab75SChris Mason int ret; 1311259ab75SChris Mason 1321259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 1331259ab75SChris Mason return 0; 1341259ab75SChris Mason 135b9fab919SChris Mason if (atomic) 136b9fab919SChris Mason return -EAGAIN; 137b9fab919SChris Mason 138570eb97bSJosef Bacik lock_extent(io_tree, eb->start, eb->start + eb->len - 1, &cached_state); 1390b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 1401259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 1411259ab75SChris Mason ret = 0; 1421259ab75SChris Mason goto out; 1431259ab75SChris Mason } 14494647322SDavid Sterba btrfs_err_rl(eb->fs_info, 1458f0ed7d4SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 1468f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 14729549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 1481259ab75SChris Mason ret = 1; 1490b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 15033958dc6SChris Mason out: 151570eb97bSJosef Bacik unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, 152e43bbe5eSDavid Sterba &cached_state); 1531259ab75SChris Mason return ret; 1541259ab75SChris Mason } 1551259ab75SChris Mason 156e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 157e7e16f48SJohannes Thumshirn { 158e7e16f48SJohannes Thumshirn switch (csum_type) { 159e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 1603951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 1613831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 162352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 163e7e16f48SJohannes Thumshirn return true; 164e7e16f48SJohannes Thumshirn default: 165e7e16f48SJohannes Thumshirn return false; 166e7e16f48SJohannes Thumshirn } 167e7e16f48SJohannes Thumshirn } 168e7e16f48SJohannes Thumshirn 169d352ac68SChris Mason /* 1701104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 1711104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 1721104a885SDavid Sterba */ 1733d17adeaSQu Wenruo int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 1743d17adeaSQu Wenruo const struct btrfs_super_block *disk_sb) 1751104a885SDavid Sterba { 17651bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 177d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 178d5178578SJohannes Thumshirn 179d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 1801104a885SDavid Sterba 1811104a885SDavid Sterba /* 1821104a885SDavid Sterba * The super_block structure does not span the whole 18351bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 18451bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 1851104a885SDavid Sterba */ 1863d17adeaSQu Wenruo crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, 187fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 1881104a885SDavid Sterba 18955fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 190e7e16f48SJohannes Thumshirn return 1; 1911104a885SDavid Sterba 192e7e16f48SJohannes Thumshirn return 0; 1931104a885SDavid Sterba } 1941104a885SDavid Sterba 195e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 196ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 197581c1760SQu Wenruo { 198e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 199581c1760SQu Wenruo int found_level; 200581c1760SQu Wenruo struct btrfs_key found_key; 201581c1760SQu Wenruo int ret; 202581c1760SQu Wenruo 203581c1760SQu Wenruo found_level = btrfs_header_level(eb); 204581c1760SQu Wenruo if (found_level != level) { 20563489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 20663489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 207581c1760SQu Wenruo btrfs_err(fs_info, 208581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 209581c1760SQu Wenruo eb->start, level, found_level); 210581c1760SQu Wenruo return -EIO; 211581c1760SQu Wenruo } 212581c1760SQu Wenruo 213581c1760SQu Wenruo if (!first_key) 214581c1760SQu Wenruo return 0; 215581c1760SQu Wenruo 2165d41be6fSQu Wenruo /* 2175d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 2185d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 2195d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 2205d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 2215d41be6fSQu Wenruo */ 2225d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 2235d41be6fSQu Wenruo return 0; 22462fdaa52SQu Wenruo 22562fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 22662fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 22762fdaa52SQu Wenruo btrfs_err(fs_info, 22862fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 22962fdaa52SQu Wenruo eb->start); 23062fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 23162fdaa52SQu Wenruo return -EUCLEAN; 23262fdaa52SQu Wenruo } 23362fdaa52SQu Wenruo 234581c1760SQu Wenruo if (found_level) 235581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 236581c1760SQu Wenruo else 237581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 238581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 239581c1760SQu Wenruo 240581c1760SQu Wenruo if (ret) { 24163489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 24263489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 243581c1760SQu Wenruo btrfs_err(fs_info, 244ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 245ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 246ff76a864SLiu Bo first_key->type, first_key->offset, 247ff76a864SLiu Bo found_key.objectid, found_key.type, 248ff76a864SLiu Bo found_key.offset); 249581c1760SQu Wenruo } 250581c1760SQu Wenruo return ret; 251581c1760SQu Wenruo } 252581c1760SQu Wenruo 2531104a885SDavid Sterba /* 254d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 255d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 256581c1760SQu Wenruo * 257581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 258581c1760SQu Wenruo * @level: expected level, mandatory check 259581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 260d352ac68SChris Mason */ 2616a2e9dc4SFilipe Manana int btrfs_read_extent_buffer(struct extent_buffer *eb, 262581c1760SQu Wenruo u64 parent_transid, int level, 263581c1760SQu Wenruo struct btrfs_key *first_key) 264f188591eSChris Mason { 2655ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 266f188591eSChris Mason struct extent_io_tree *io_tree; 267ea466794SJosef Bacik int failed = 0; 268f188591eSChris Mason int ret; 269f188591eSChris Mason int num_copies = 0; 270f188591eSChris Mason int mirror_num = 0; 271ea466794SJosef Bacik int failed_mirror = 0; 272f188591eSChris Mason 2730b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 274f188591eSChris Mason while (1) { 275f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 276c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); 277256dd1bbSStefan Behrens if (!ret) { 278581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 279b9fab919SChris Mason parent_transid, 0)) 280256dd1bbSStefan Behrens ret = -EIO; 281e064d5e9SDavid Sterba else if (btrfs_verify_level_key(eb, level, 282ff76a864SLiu Bo first_key, parent_transid)) 283581c1760SQu Wenruo ret = -EUCLEAN; 284581c1760SQu Wenruo else 285581c1760SQu Wenruo break; 286256dd1bbSStefan Behrens } 287d397712bSChris Mason 2880b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 289f188591eSChris Mason eb->start, eb->len); 2904235298eSChris Mason if (num_copies == 1) 291ea466794SJosef Bacik break; 2924235298eSChris Mason 2935cf1ab56SJosef Bacik if (!failed_mirror) { 2945cf1ab56SJosef Bacik failed = 1; 2955cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 2965cf1ab56SJosef Bacik } 2975cf1ab56SJosef Bacik 298f188591eSChris Mason mirror_num++; 299ea466794SJosef Bacik if (mirror_num == failed_mirror) 300ea466794SJosef Bacik mirror_num++; 301ea466794SJosef Bacik 3024235298eSChris Mason if (mirror_num > num_copies) 303ea466794SJosef Bacik break; 304f188591eSChris Mason } 305ea466794SJosef Bacik 306c0901581SStefan Behrens if (failed && !ret && failed_mirror) 30720a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 308ea466794SJosef Bacik 309ea466794SJosef Bacik return ret; 310f188591eSChris Mason } 31119c00ddcSChris Mason 312eca0f6f6SQu Wenruo static int csum_one_extent_buffer(struct extent_buffer *eb) 313eca0f6f6SQu Wenruo { 314eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 315eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 316eca0f6f6SQu Wenruo int ret; 317eca0f6f6SQu Wenruo 318eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 319eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 320eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 321eca0f6f6SQu Wenruo csum_tree_block(eb, result); 322eca0f6f6SQu Wenruo 323eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 324eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 325eca0f6f6SQu Wenruo else 326eca0f6f6SQu Wenruo ret = btrfs_check_leaf_full(eb); 327eca0f6f6SQu Wenruo 3283777369fSQu Wenruo if (ret < 0) 3293777369fSQu Wenruo goto error; 3303777369fSQu Wenruo 3313777369fSQu Wenruo /* 3323777369fSQu Wenruo * Also check the generation, the eb reached here must be newer than 3333777369fSQu Wenruo * last committed. Or something seriously wrong happened. 3343777369fSQu Wenruo */ 3353777369fSQu Wenruo if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { 3363777369fSQu Wenruo ret = -EUCLEAN; 337eca0f6f6SQu Wenruo btrfs_err(fs_info, 3383777369fSQu Wenruo "block=%llu bad generation, have %llu expect > %llu", 3393777369fSQu Wenruo eb->start, btrfs_header_generation(eb), 3403777369fSQu Wenruo fs_info->last_trans_committed); 3413777369fSQu Wenruo goto error; 342eca0f6f6SQu Wenruo } 343eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 344eca0f6f6SQu Wenruo 345eca0f6f6SQu Wenruo return 0; 3463777369fSQu Wenruo 3473777369fSQu Wenruo error: 3483777369fSQu Wenruo btrfs_print_tree(eb, 0); 3493777369fSQu Wenruo btrfs_err(fs_info, "block=%llu write time tree block corruption detected", 3503777369fSQu Wenruo eb->start); 3513777369fSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 3523777369fSQu Wenruo return ret; 353eca0f6f6SQu Wenruo } 354eca0f6f6SQu Wenruo 355eca0f6f6SQu Wenruo /* Checksum all dirty extent buffers in one bio_vec */ 356eca0f6f6SQu Wenruo static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, 357eca0f6f6SQu Wenruo struct bio_vec *bvec) 358eca0f6f6SQu Wenruo { 359eca0f6f6SQu Wenruo struct page *page = bvec->bv_page; 360eca0f6f6SQu Wenruo u64 bvec_start = page_offset(page) + bvec->bv_offset; 361eca0f6f6SQu Wenruo u64 cur; 362eca0f6f6SQu Wenruo int ret = 0; 363eca0f6f6SQu Wenruo 364eca0f6f6SQu Wenruo for (cur = bvec_start; cur < bvec_start + bvec->bv_len; 365eca0f6f6SQu Wenruo cur += fs_info->nodesize) { 366eca0f6f6SQu Wenruo struct extent_buffer *eb; 367eca0f6f6SQu Wenruo bool uptodate; 368eca0f6f6SQu Wenruo 369eca0f6f6SQu Wenruo eb = find_extent_buffer(fs_info, cur); 370eca0f6f6SQu Wenruo uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, 371eca0f6f6SQu Wenruo fs_info->nodesize); 372eca0f6f6SQu Wenruo 37301cd3909SDavid Sterba /* A dirty eb shouldn't disappear from buffer_radix */ 374eca0f6f6SQu Wenruo if (WARN_ON(!eb)) 375eca0f6f6SQu Wenruo return -EUCLEAN; 376eca0f6f6SQu Wenruo 377eca0f6f6SQu Wenruo if (WARN_ON(cur != btrfs_header_bytenr(eb))) { 378eca0f6f6SQu Wenruo free_extent_buffer(eb); 379eca0f6f6SQu Wenruo return -EUCLEAN; 380eca0f6f6SQu Wenruo } 381eca0f6f6SQu Wenruo if (WARN_ON(!uptodate)) { 382eca0f6f6SQu Wenruo free_extent_buffer(eb); 383eca0f6f6SQu Wenruo return -EUCLEAN; 384eca0f6f6SQu Wenruo } 385eca0f6f6SQu Wenruo 386eca0f6f6SQu Wenruo ret = csum_one_extent_buffer(eb); 387eca0f6f6SQu Wenruo free_extent_buffer(eb); 388eca0f6f6SQu Wenruo if (ret < 0) 389eca0f6f6SQu Wenruo return ret; 390eca0f6f6SQu Wenruo } 391eca0f6f6SQu Wenruo return ret; 392eca0f6f6SQu Wenruo } 393eca0f6f6SQu Wenruo 394d352ac68SChris Mason /* 395ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 396ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 397ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 398d352ac68SChris Mason */ 399ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 40019c00ddcSChris Mason { 401ac303b69SQu Wenruo struct page *page = bvec->bv_page; 4024eee4fa4SMiao Xie u64 start = page_offset(page); 40319c00ddcSChris Mason u64 found_start; 40419c00ddcSChris Mason struct extent_buffer *eb; 405eca0f6f6SQu Wenruo 406fbca46ebSQu Wenruo if (fs_info->nodesize < PAGE_SIZE) 407eca0f6f6SQu Wenruo return csum_dirty_subpage_buffers(fs_info, bvec); 408f188591eSChris Mason 4094f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4104f2de97aSJosef Bacik if (page != eb->pages[0]) 4114f2de97aSJosef Bacik return 0; 4120f805531SAlex Lyakas 41319c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 414d3575156SNaohiro Aota 415d3575156SNaohiro Aota if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 416d3575156SNaohiro Aota WARN_ON(found_start != 0); 417d3575156SNaohiro Aota return 0; 418d3575156SNaohiro Aota } 419d3575156SNaohiro Aota 4200f805531SAlex Lyakas /* 4210f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 4220f805531SAlex Lyakas * It is useful to know what went wrong. 4230f805531SAlex Lyakas */ 4240f805531SAlex Lyakas if (WARN_ON(found_start != start)) 4250f805531SAlex Lyakas return -EUCLEAN; 4260f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 4270f805531SAlex Lyakas return -EUCLEAN; 4280f805531SAlex Lyakas 429eca0f6f6SQu Wenruo return csum_one_extent_buffer(eb); 43019c00ddcSChris Mason } 43119c00ddcSChris Mason 432b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 4332b82032cSYan Zheng { 434b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 435944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 43644880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 437944d3f9fSNikolay Borisov u8 *metadata_uuid; 4382b82032cSYan Zheng 4399a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 4409a8658e3SDavid Sterba BTRFS_FSID_SIZE); 4417239ff4bSNikolay Borisov /* 442944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 443944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 444944d3f9fSNikolay Borisov * ->fsid in this case is fine 4457239ff4bSNikolay Borisov */ 446944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 4477239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 4487239ff4bSNikolay Borisov else 4497239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 4507239ff4bSNikolay Borisov 451944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 452944d3f9fSNikolay Borisov return 0; 453944d3f9fSNikolay Borisov 454944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 455944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 456944d3f9fSNikolay Borisov return 0; 457944d3f9fSNikolay Borisov 458944d3f9fSNikolay Borisov return 1; 4592b82032cSYan Zheng } 4602b82032cSYan Zheng 46177bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 46277bf40a2SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb) 463ce9adaa5SChris Mason { 46477bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 465ce9adaa5SChris Mason u64 found_start; 46677bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 46777bf40a2SQu Wenruo u8 found_level; 4682996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 469dfd29eedSDavid Sterba const u8 *header_csum; 47077bf40a2SQu Wenruo int ret = 0; 471ea466794SJosef Bacik 472ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 473727011e0SChris Mason if (found_start != eb->start) { 4748f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, 4758f0ed7d4SQu Wenruo "bad tree block start, mirror %u want %llu have %llu", 4768f0ed7d4SQu Wenruo eb->read_mirror, eb->start, found_start); 477f188591eSChris Mason ret = -EIO; 47877bf40a2SQu Wenruo goto out; 479ce9adaa5SChris Mason } 480b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 4818f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", 4828f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 4831259ab75SChris Mason ret = -EIO; 48477bf40a2SQu Wenruo goto out; 4851259ab75SChris Mason } 486ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 4871c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 4888f0ed7d4SQu Wenruo btrfs_err(fs_info, 4898f0ed7d4SQu Wenruo "bad tree block level, mirror %u level %d on logical %llu", 4908f0ed7d4SQu Wenruo eb->read_mirror, btrfs_header_level(eb), eb->start); 4911c24c3ceSJosef Bacik ret = -EIO; 49277bf40a2SQu Wenruo goto out; 4931c24c3ceSJosef Bacik } 494ce9adaa5SChris Mason 495c67b3892SDavid Sterba csum_tree_block(eb, result); 496dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 497dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 498a826d6dcSJosef Bacik 499dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 5002996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 5018f0ed7d4SQu Wenruo "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", 5028f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 503dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 50435be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 50535be8851SJohannes Thumshirn btrfs_header_level(eb)); 5062996e1f8SJohannes Thumshirn ret = -EUCLEAN; 50777bf40a2SQu Wenruo goto out; 5082996e1f8SJohannes Thumshirn } 5092996e1f8SJohannes Thumshirn 510a826d6dcSJosef Bacik /* 511a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 512a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 513a826d6dcSJosef Bacik * return -EIO. 514a826d6dcSJosef Bacik */ 5151c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 516a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 517a826d6dcSJosef Bacik ret = -EIO; 518a826d6dcSJosef Bacik } 519ce9adaa5SChris Mason 520813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 521053ab70fSLiu Bo ret = -EIO; 522053ab70fSLiu Bo 5230b32f4bbSJosef Bacik if (!ret) 5240b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 52575391f0dSQu Wenruo else 52675391f0dSQu Wenruo btrfs_err(fs_info, 5278f0ed7d4SQu Wenruo "read time tree block corruption detected on logical %llu mirror %u", 5288f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 52977bf40a2SQu Wenruo out: 53077bf40a2SQu Wenruo return ret; 53177bf40a2SQu Wenruo } 53277bf40a2SQu Wenruo 533371cdc07SQu Wenruo static int validate_subpage_buffer(struct page *page, u64 start, u64 end, 534371cdc07SQu Wenruo int mirror) 535371cdc07SQu Wenruo { 536371cdc07SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 537371cdc07SQu Wenruo struct extent_buffer *eb; 538371cdc07SQu Wenruo bool reads_done; 539371cdc07SQu Wenruo int ret = 0; 540371cdc07SQu Wenruo 541371cdc07SQu Wenruo /* 542371cdc07SQu Wenruo * We don't allow bio merge for subpage metadata read, so we should 543371cdc07SQu Wenruo * only get one eb for each endio hook. 544371cdc07SQu Wenruo */ 545371cdc07SQu Wenruo ASSERT(end == start + fs_info->nodesize - 1); 546371cdc07SQu Wenruo ASSERT(PagePrivate(page)); 547371cdc07SQu Wenruo 548371cdc07SQu Wenruo eb = find_extent_buffer(fs_info, start); 549371cdc07SQu Wenruo /* 550371cdc07SQu Wenruo * When we are reading one tree block, eb must have been inserted into 551371cdc07SQu Wenruo * the radix tree. If not, something is wrong. 552371cdc07SQu Wenruo */ 553371cdc07SQu Wenruo ASSERT(eb); 554371cdc07SQu Wenruo 555371cdc07SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 556371cdc07SQu Wenruo /* Subpage read must finish in page read */ 557371cdc07SQu Wenruo ASSERT(reads_done); 558371cdc07SQu Wenruo 559371cdc07SQu Wenruo eb->read_mirror = mirror; 560371cdc07SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 561371cdc07SQu Wenruo ret = -EIO; 562371cdc07SQu Wenruo goto err; 563371cdc07SQu Wenruo } 564371cdc07SQu Wenruo ret = validate_extent_buffer(eb); 565371cdc07SQu Wenruo if (ret < 0) 566371cdc07SQu Wenruo goto err; 567371cdc07SQu Wenruo 568371cdc07SQu Wenruo set_extent_buffer_uptodate(eb); 569371cdc07SQu Wenruo 570371cdc07SQu Wenruo free_extent_buffer(eb); 571371cdc07SQu Wenruo return ret; 572371cdc07SQu Wenruo err: 573371cdc07SQu Wenruo /* 574371cdc07SQu Wenruo * end_bio_extent_readpage decrements io_pages in case of error, 575371cdc07SQu Wenruo * make sure it has something to decrement. 576371cdc07SQu Wenruo */ 577371cdc07SQu Wenruo atomic_inc(&eb->io_pages); 578371cdc07SQu Wenruo clear_extent_buffer_uptodate(eb); 579371cdc07SQu Wenruo free_extent_buffer(eb); 580371cdc07SQu Wenruo return ret; 581371cdc07SQu Wenruo } 582371cdc07SQu Wenruo 583c3a3b19bSQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, 58477bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 58577bf40a2SQu Wenruo int mirror) 58677bf40a2SQu Wenruo { 58777bf40a2SQu Wenruo struct extent_buffer *eb; 58877bf40a2SQu Wenruo int ret = 0; 58977bf40a2SQu Wenruo int reads_done; 59077bf40a2SQu Wenruo 59177bf40a2SQu Wenruo ASSERT(page->private); 592371cdc07SQu Wenruo 593fbca46ebSQu Wenruo if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) 594371cdc07SQu Wenruo return validate_subpage_buffer(page, start, end, mirror); 595371cdc07SQu Wenruo 59677bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 59777bf40a2SQu Wenruo 59877bf40a2SQu Wenruo /* 59977bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 60077bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 60177bf40a2SQu Wenruo */ 60277bf40a2SQu Wenruo atomic_inc(&eb->refs); 60377bf40a2SQu Wenruo 60477bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 60577bf40a2SQu Wenruo if (!reads_done) 60677bf40a2SQu Wenruo goto err; 60777bf40a2SQu Wenruo 60877bf40a2SQu Wenruo eb->read_mirror = mirror; 60977bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 61077bf40a2SQu Wenruo ret = -EIO; 61177bf40a2SQu Wenruo goto err; 61277bf40a2SQu Wenruo } 61377bf40a2SQu Wenruo ret = validate_extent_buffer(eb); 614ce9adaa5SChris Mason err: 61553b381b3SDavid Woodhouse if (ret) { 61653b381b3SDavid Woodhouse /* 61753b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 61853b381b3SDavid Woodhouse * again, we have to make sure it has something 61953b381b3SDavid Woodhouse * to decrement 62053b381b3SDavid Woodhouse */ 62153b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6220b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 62353b381b3SDavid Woodhouse } 6240b32f4bbSJosef Bacik free_extent_buffer(eb); 62577bf40a2SQu Wenruo 626f188591eSChris Mason return ret; 627ce9adaa5SChris Mason } 628ce9adaa5SChris Mason 6294a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 6304a69a410SChris Mason { 6314a69a410SChris Mason struct async_submit_bio *async; 6324e4cbee9SChristoph Hellwig blk_status_t ret; 6334a69a410SChris Mason 6344a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 6351941b64bSQu Wenruo ret = async->submit_bio_start(async->inode, async->bio, 6361941b64bSQu Wenruo async->dio_file_offset); 63779787eaaSJeff Mahoney if (ret) 6384e4cbee9SChristoph Hellwig async->status = ret; 6394a69a410SChris Mason } 6404a69a410SChris Mason 64106ea01b1SDavid Sterba /* 64206ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 64306ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 64406ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 64506ea01b1SDavid Sterba * 64606ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 64706ea01b1SDavid Sterba * inserted into the tree. 64806ea01b1SDavid Sterba */ 6494a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 6508b712842SChris Mason { 651917f32a2SChristoph Hellwig struct async_submit_bio *async = 652917f32a2SChristoph Hellwig container_of(work, struct async_submit_bio, work); 653917f32a2SChristoph Hellwig struct inode *inode = async->inode; 654917f32a2SChristoph Hellwig struct btrfs_bio *bbio = btrfs_bio(async->bio); 6554854ddd0SChris Mason 656bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 6574e4cbee9SChristoph Hellwig if (async->status) { 658917f32a2SChristoph Hellwig btrfs_bio_end_io(bbio, async->status); 65979787eaaSJeff Mahoney return; 66079787eaaSJeff Mahoney } 66179787eaaSJeff Mahoney 662ec39f769SChris Mason /* 663ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 664ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 665ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 666ec39f769SChris Mason */ 667ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 6681a722d8fSChristoph Hellwig btrfs_submit_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 6694a69a410SChris Mason } 6704a69a410SChris Mason 6714a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 6724a69a410SChris Mason { 6734a69a410SChris Mason struct async_submit_bio *async; 6744a69a410SChris Mason 6754a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 6768b712842SChris Mason kfree(async); 6778b712842SChris Mason } 6788b712842SChris Mason 679ea1f0cedSChristoph Hellwig /* 680ea1f0cedSChristoph Hellwig * Submit bio to an async queue. 681ea1f0cedSChristoph Hellwig * 682ea1f0cedSChristoph Hellwig * Retrun: 683ea1f0cedSChristoph Hellwig * - true if the work has been succesfuly submitted 684ea1f0cedSChristoph Hellwig * - false in case of error 685ea1f0cedSChristoph Hellwig */ 686ea1f0cedSChristoph Hellwig bool btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, int mirror_num, 687ea1f0cedSChristoph Hellwig u64 dio_file_offset, 688e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 68944b8bd7eSChris Mason { 6908896a08dSQu Wenruo struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 69144b8bd7eSChris Mason struct async_submit_bio *async; 69244b8bd7eSChris Mason 69344b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 69444b8bd7eSChris Mason if (!async) 695ea1f0cedSChristoph Hellwig return false; 69644b8bd7eSChris Mason 6978896a08dSQu Wenruo async->inode = inode; 69844b8bd7eSChris Mason async->bio = bio; 69944b8bd7eSChris Mason async->mirror_num = mirror_num; 7004a69a410SChris Mason async->submit_bio_start = submit_bio_start; 7014a69a410SChris Mason 702a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 703a0cac0ecSOmar Sandoval run_one_async_free); 7044a69a410SChris Mason 7051941b64bSQu Wenruo async->dio_file_offset = dio_file_offset; 7068c8bee1dSChris Mason 7074e4cbee9SChristoph Hellwig async->status = 0; 70879787eaaSJeff Mahoney 70967f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 710a31b4a43SChristoph Hellwig btrfs_queue_work(fs_info->hipri_workers, &async->work); 711a31b4a43SChristoph Hellwig else 7125cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 713ea1f0cedSChristoph Hellwig return true; 71444b8bd7eSChris Mason } 71544b8bd7eSChris Mason 7164e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 717ce3ed71aSChris Mason { 7182c30c71bSKent Overstreet struct bio_vec *bvec; 719ce3ed71aSChris Mason struct btrfs_root *root; 7202b070cfeSChristoph Hellwig int ret = 0; 7216dc4f100SMing Lei struct bvec_iter_all iter_all; 722ce3ed71aSChris Mason 723c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 7242b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 725ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 726ac303b69SQu Wenruo ret = csum_dirty_buffer(root->fs_info, bvec); 72779787eaaSJeff Mahoney if (ret) 72879787eaaSJeff Mahoney break; 729ce3ed71aSChris Mason } 7302c30c71bSKent Overstreet 7314e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 732ce3ed71aSChris Mason } 733ce3ed71aSChris Mason 7348896a08dSQu Wenruo static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, 7351941b64bSQu Wenruo u64 dio_file_offset) 73622c59948SChris Mason { 7378b712842SChris Mason /* 7388b712842SChris Mason * when we're called for a write, we're already in the async 7391a722d8fSChristoph Hellwig * submission context. Just jump into btrfs_submit_bio. 7408b712842SChris Mason */ 74179787eaaSJeff Mahoney return btree_csum_one_bio(bio); 74222c59948SChris Mason } 74322c59948SChris Mason 744f4dcfb30SJohannes Thumshirn static bool should_async_write(struct btrfs_fs_info *fs_info, 7459b4e675aSDavid Sterba struct btrfs_inode *bi) 746de0022b9SJosef Bacik { 7474eef29efSNaohiro Aota if (btrfs_is_zoned(fs_info)) 748f4dcfb30SJohannes Thumshirn return false; 7496300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 750f4dcfb30SJohannes Thumshirn return false; 7519b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 752f4dcfb30SJohannes Thumshirn return false; 753f4dcfb30SJohannes Thumshirn return true; 754de0022b9SJosef Bacik } 755de0022b9SJosef Bacik 75694d9e11bSChristoph Hellwig void btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, int mirror_num) 75744b8bd7eSChris Mason { 7580b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 759917f32a2SChristoph Hellwig struct btrfs_bio *bbio = btrfs_bio(bio); 7604e4cbee9SChristoph Hellwig blk_status_t ret; 761cad321adSChris Mason 76208a6f464SChristoph Hellwig bio->bi_opf |= REQ_META; 76308a6f464SChristoph Hellwig 764cfe94440SNaohiro Aota if (btrfs_op(bio) != BTRFS_MAP_WRITE) { 7651a722d8fSChristoph Hellwig btrfs_submit_bio(fs_info, bio, mirror_num); 7661a722d8fSChristoph Hellwig return; 76744b8bd7eSChris Mason } 76844b8bd7eSChris Mason 769ea1f0cedSChristoph Hellwig /* 770ea1f0cedSChristoph Hellwig * Kthread helpers are used to submit writes so that checksumming can 771ea1f0cedSChristoph Hellwig * happen in parallel across all CPUs. 772ea1f0cedSChristoph Hellwig */ 773ea1f0cedSChristoph Hellwig if (should_async_write(fs_info, BTRFS_I(inode)) && 774ea1f0cedSChristoph Hellwig btrfs_wq_submit_bio(inode, bio, mirror_num, 0, btree_submit_bio_start)) 775ea1f0cedSChristoph Hellwig return; 776ea1f0cedSChristoph Hellwig 777ea1f0cedSChristoph Hellwig ret = btree_csum_one_bio(bio); 77894d9e11bSChristoph Hellwig if (ret) { 779917f32a2SChristoph Hellwig btrfs_bio_end_io(bbio, ret); 780ea1f0cedSChristoph Hellwig return; 78194d9e11bSChristoph Hellwig } 782ea1f0cedSChristoph Hellwig 783ea1f0cedSChristoph Hellwig btrfs_submit_bio(fs_info, bio, mirror_num); 78461891923SStefan Behrens } 78561891923SStefan Behrens 7863dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 7878958b551SMatthew Wilcox (Oracle) static int btree_migrate_folio(struct address_space *mapping, 7888958b551SMatthew Wilcox (Oracle) struct folio *dst, struct folio *src, enum migrate_mode mode) 789784b4e29SChris Mason { 790784b4e29SChris Mason /* 791784b4e29SChris Mason * we can't safely write a btree page from here, 792784b4e29SChris Mason * we haven't done the locking hook 793784b4e29SChris Mason */ 7948958b551SMatthew Wilcox (Oracle) if (folio_test_dirty(src)) 795784b4e29SChris Mason return -EAGAIN; 796784b4e29SChris Mason /* 797784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 798784b4e29SChris Mason * We must have no buffers or drop them. 799784b4e29SChris Mason */ 8008958b551SMatthew Wilcox (Oracle) if (folio_get_private(src) && 8018958b551SMatthew Wilcox (Oracle) !filemap_release_folio(src, GFP_KERNEL)) 802784b4e29SChris Mason return -EAGAIN; 80354184650SMatthew Wilcox (Oracle) return migrate_folio(mapping, dst, src, mode); 804784b4e29SChris Mason } 8058958b551SMatthew Wilcox (Oracle) #else 8068958b551SMatthew Wilcox (Oracle) #define btree_migrate_folio NULL 8073dd1462eSJan Beulich #endif 808784b4e29SChris Mason 8090da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 8100da5468fSChris Mason struct writeback_control *wbc) 8110da5468fSChris Mason { 812e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 813e2d84521SMiao Xie int ret; 814e2d84521SMiao Xie 815d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 816448d640bSChris Mason 817448d640bSChris Mason if (wbc->for_kupdate) 818448d640bSChris Mason return 0; 819448d640bSChris Mason 820e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 821b9473439SChris Mason /* this is a bit racy, but that's ok */ 822d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 823d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 824d814a491SEthan Lien fs_info->dirty_metadata_batch); 825e2d84521SMiao Xie if (ret < 0) 826793955bcSChris Mason return 0; 827793955bcSChris Mason } 8280b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 8290da5468fSChris Mason } 8300da5468fSChris Mason 831f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 8325f39d397SChris Mason { 833f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 834f913cff3SMatthew Wilcox (Oracle) return false; 8350c4e538bSDavid Sterba 836f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 837d98237b3SChris Mason } 838d98237b3SChris Mason 839895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 840895586ebSMatthew Wilcox (Oracle) size_t length) 841d98237b3SChris Mason { 842d1310b2eSChris Mason struct extent_io_tree *tree; 843895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 844895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 845f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 846895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 847895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 848895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 849895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 850895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 8519ad6b7bcSChris Mason } 852d98237b3SChris Mason } 853d98237b3SChris Mason 854bb146eb2SJosef Bacik #ifdef DEBUG 8550079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 8560079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 8570079c3b1SMatthew Wilcox (Oracle) { 8580079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 859139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 8600b32f4bbSJosef Bacik struct extent_buffer *eb; 861139e8cd3SQu Wenruo int cur_bit = 0; 8620079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 8630b32f4bbSJosef Bacik 864139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 8650079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 8660b32f4bbSJosef Bacik BUG_ON(!eb); 8670b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 8680b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 86949d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 8700079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 871139e8cd3SQu Wenruo } 8720079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 873139e8cd3SQu Wenruo 874139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 875139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 876139e8cd3SQu Wenruo unsigned long flags; 877139e8cd3SQu Wenruo u64 cur; 878139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 879139e8cd3SQu Wenruo 880139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 881139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 882139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 883139e8cd3SQu Wenruo cur_bit++; 884139e8cd3SQu Wenruo continue; 885139e8cd3SQu Wenruo } 886139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 887139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 888139e8cd3SQu Wenruo 889139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 890139e8cd3SQu Wenruo ASSERT(eb); 891139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 892139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 89349d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 894139e8cd3SQu Wenruo free_extent_buffer(eb); 895139e8cd3SQu Wenruo 896139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 897139e8cd3SQu Wenruo } 8980079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 8990b32f4bbSJosef Bacik } 9000079c3b1SMatthew Wilcox (Oracle) #else 9010079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 9020079c3b1SMatthew Wilcox (Oracle) #endif 9030b32f4bbSJosef Bacik 9047f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 9050da5468fSChris Mason .writepages = btree_writepages, 906f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 907895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 9088958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 9090079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 910d98237b3SChris Mason }; 911123abc88SChris Mason 9122ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 9132ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 9143fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 9153fbaf258SJosef Bacik int level) 9160999df54SChris Mason { 9170b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 9180b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 9193fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 9200999df54SChris Mason } 9210999df54SChris Mason 922581c1760SQu Wenruo /* 923581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 924581c1760SQu Wenruo * verification. 925581c1760SQu Wenruo * 9261b7ec85eSJosef Bacik * @owner_root: the objectid of the root owner for this block. 927581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 928581c1760SQu Wenruo * @level: expected level, mandatory check 929581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 930581c1760SQu Wenruo */ 9312ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 9321b7ec85eSJosef Bacik u64 owner_root, u64 parent_transid, 9331b7ec85eSJosef Bacik int level, struct btrfs_key *first_key) 934e20d96d6SChris Mason { 9355f39d397SChris Mason struct extent_buffer *buf = NULL; 93619c00ddcSChris Mason int ret; 93719c00ddcSChris Mason 9383fbaf258SJosef Bacik buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); 939c871b0f2SLiu Bo if (IS_ERR(buf)) 940c871b0f2SLiu Bo return buf; 941e4204dedSChris Mason 9426a2e9dc4SFilipe Manana ret = btrfs_read_extent_buffer(buf, parent_transid, level, first_key); 9430f0fe8f7SFilipe David Borba Manana if (ret) { 944537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 94564c043deSLiu Bo return ERR_PTR(ret); 9460f0fe8f7SFilipe David Borba Manana } 94788c602abSQu Wenruo if (btrfs_check_eb_owner(buf, owner_root)) { 94888c602abSQu Wenruo free_extent_buffer_stale(buf); 94988c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 95088c602abSQu Wenruo } 9515f39d397SChris Mason return buf; 952ce9adaa5SChris Mason 953eb60ceacSChris Mason } 954eb60ceacSChris Mason 9556a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 956ed2ff2cbSChris Mason { 9576a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 95855c69072SChris Mason if (btrfs_header_generation(buf) == 959e2d84521SMiao Xie fs_info->running_transaction->transid) { 96049d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 961b4ce94deSChris Mason 962b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 963104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 964e2d84521SMiao Xie -buf->len, 965e2d84521SMiao Xie fs_info->dirty_metadata_batch); 9660b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 967925baeddSChris Mason } 9685f39d397SChris Mason } 969ed7b63ebSJosef Bacik } 9705f39d397SChris Mason 971da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 972e20d96d6SChris Mason u64 objectid) 973d97e63b6SChris Mason { 9747c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 9752e608bd1SJosef Bacik 9762e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 9772e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 9782e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 97996dfcb46SJosef Bacik root->fs_info = fs_info; 9802e608bd1SJosef Bacik root->root_key.objectid = objectid; 981cfaa7295SChris Mason root->node = NULL; 982a28ec197SChris Mason root->commit_root = NULL; 98327cdeb70SMiao Xie root->state = 0; 984abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 9850b86a832SChris Mason 9860f7d52f4SChris Mason root->last_trans = 0; 9876b8fad57SNikolay Borisov root->free_objectid = 0; 988eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 989199c2a9cSMiao Xie root->nr_ordered_extents = 0; 9906bef4d31SEric Paris root->inode_tree = RB_ROOT; 991088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 9922e608bd1SJosef Bacik 9932e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 9940b86a832SChris Mason 9950b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 9965d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 997eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 998eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 999199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1000199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1001d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 10022ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 10032ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 10045d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1005eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1006199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1007f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 10082ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 10092ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 10108287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1011a2135011SChris Mason mutex_init(&root->objectid_mutex); 1012e02119d5SChris Mason mutex_init(&root->log_mutex); 101331f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1014573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1015c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 10167237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 10177237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 10187237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 10198b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 10208b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 10217237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 10227237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 10237237f183SYan Zheng atomic_set(&root->log_writers, 0); 10242ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 10250700cea7SElena Reshetova refcount_set(&root->refs, 1); 10268ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1027eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 10287237f183SYan Zheng root->log_transid = 0; 1029d1433debSMiao Xie root->log_transid_committed = -1; 1030257c62e1SChris Mason root->last_log_commit = 0; 10312e608bd1SJosef Bacik root->anon_dev = 0; 1032e289f03eSFilipe Manana if (!dummy) { 103343eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 103443eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1035e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1036e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1037e289f03eSFilipe Manana } 1038017e5369SChris Mason 10395f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1040370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1041bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1042bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1043fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1044bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1045fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1046bd647ce3SJosef Bacik #endif 10473768f368SChris Mason } 10483768f368SChris Mason 104974e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 105096dfcb46SJosef Bacik u64 objectid, gfp_t flags) 10516f07e42eSAl Viro { 105274e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 10536f07e42eSAl Viro if (root) 105496dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 10556f07e42eSAl Viro return root; 10566f07e42eSAl Viro } 10576f07e42eSAl Viro 105806ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 105906ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1060da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 106106ea65a3SJosef Bacik { 106206ea65a3SJosef Bacik struct btrfs_root *root; 106306ea65a3SJosef Bacik 10647c0260eeSJeff Mahoney if (!fs_info) 10657c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 10667c0260eeSJeff Mahoney 106796dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 106806ea65a3SJosef Bacik if (!root) 106906ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1070da17066cSJeff Mahoney 1071b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1072faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 107306ea65a3SJosef Bacik 107406ea65a3SJosef Bacik return root; 107506ea65a3SJosef Bacik } 107606ea65a3SJosef Bacik #endif 107706ea65a3SJosef Bacik 1078abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 1079abed4aaaSJosef Bacik { 1080abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 1081abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 1082abed4aaaSJosef Bacik 1083abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 1084abed4aaaSJosef Bacik } 1085abed4aaaSJosef Bacik 1086abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 1087abed4aaaSJosef Bacik { 1088abed4aaaSJosef Bacik const struct btrfs_key *key = k; 1089abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 1090abed4aaaSJosef Bacik 1091abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 1092abed4aaaSJosef Bacik } 1093abed4aaaSJosef Bacik 1094abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 1095abed4aaaSJosef Bacik { 1096abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1097abed4aaaSJosef Bacik struct rb_node *tmp; 1098abed4aaaSJosef Bacik 1099abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1100abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 1101abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1102abed4aaaSJosef Bacik ASSERT(!tmp); 1103abed4aaaSJosef Bacik 1104abed4aaaSJosef Bacik return tmp ? -EEXIST : 0; 1105abed4aaaSJosef Bacik } 1106abed4aaaSJosef Bacik 1107abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 1108abed4aaaSJosef Bacik { 1109abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1110abed4aaaSJosef Bacik 1111abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1112abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1113abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1114abed4aaaSJosef Bacik } 1115abed4aaaSJosef Bacik 1116abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 1117abed4aaaSJosef Bacik struct btrfs_key *key) 1118abed4aaaSJosef Bacik { 1119abed4aaaSJosef Bacik struct rb_node *node; 1120abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 1121abed4aaaSJosef Bacik 1122abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 1123abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 1124abed4aaaSJosef Bacik if (node) 1125abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 1126abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 1127abed4aaaSJosef Bacik 1128abed4aaaSJosef Bacik return root; 1129abed4aaaSJosef Bacik } 1130abed4aaaSJosef Bacik 1131f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 1132f7238e50SJosef Bacik { 1133f7238e50SJosef Bacik struct btrfs_block_group *block_group; 1134f7238e50SJosef Bacik u64 ret; 1135f7238e50SJosef Bacik 1136f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 1137f7238e50SJosef Bacik return 0; 1138f7238e50SJosef Bacik 1139f7238e50SJosef Bacik if (bytenr) 1140f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 1141f7238e50SJosef Bacik else 1142f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 1143f7238e50SJosef Bacik ASSERT(block_group); 1144f7238e50SJosef Bacik if (!block_group) 1145f7238e50SJosef Bacik return 0; 1146f7238e50SJosef Bacik ret = block_group->global_root_id; 1147f7238e50SJosef Bacik btrfs_put_block_group(block_group); 1148f7238e50SJosef Bacik 1149f7238e50SJosef Bacik return ret; 1150f7238e50SJosef Bacik } 1151f7238e50SJosef Bacik 1152abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1153abed4aaaSJosef Bacik { 1154abed4aaaSJosef Bacik struct btrfs_key key = { 1155abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 1156abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1157f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1158abed4aaaSJosef Bacik }; 1159abed4aaaSJosef Bacik 1160abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1161abed4aaaSJosef Bacik } 1162abed4aaaSJosef Bacik 1163abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1164abed4aaaSJosef Bacik { 1165abed4aaaSJosef Bacik struct btrfs_key key = { 1166abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 1167abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1168f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1169abed4aaaSJosef Bacik }; 1170abed4aaaSJosef Bacik 1171abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1172abed4aaaSJosef Bacik } 1173abed4aaaSJosef Bacik 117451129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 117551129b33SJosef Bacik { 117651129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 117751129b33SJosef Bacik return fs_info->block_group_root; 117851129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 117951129b33SJosef Bacik } 118051129b33SJosef Bacik 118120897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 118220897f5cSArne Jansen u64 objectid) 118320897f5cSArne Jansen { 11849b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 118520897f5cSArne Jansen struct extent_buffer *leaf; 118620897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 118720897f5cSArne Jansen struct btrfs_root *root; 118820897f5cSArne Jansen struct btrfs_key key; 1189b89f6d1fSFilipe Manana unsigned int nofs_flag; 119020897f5cSArne Jansen int ret = 0; 119120897f5cSArne Jansen 1192b89f6d1fSFilipe Manana /* 1193b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1194b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1195b89f6d1fSFilipe Manana */ 1196b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 119796dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1198b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 119920897f5cSArne Jansen if (!root) 120020897f5cSArne Jansen return ERR_PTR(-ENOMEM); 120120897f5cSArne Jansen 120220897f5cSArne Jansen root->root_key.objectid = objectid; 120320897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 120420897f5cSArne Jansen root->root_key.offset = 0; 120520897f5cSArne Jansen 12069631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 12079631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 120820897f5cSArne Jansen if (IS_ERR(leaf)) { 120920897f5cSArne Jansen ret = PTR_ERR(leaf); 12101dd05682STsutomu Itoh leaf = NULL; 1211c1b07854SPeng Hao goto fail; 121220897f5cSArne Jansen } 121320897f5cSArne Jansen 121420897f5cSArne Jansen root->node = leaf; 121520897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 121620897f5cSArne Jansen 121720897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 121827cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 121920897f5cSArne Jansen 1220f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1221f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 122220897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 122320897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 122420897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 122520897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 122620897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 122720897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 122820897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 122933d85fdaSQu Wenruo if (is_fstree(objectid)) 1230807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1231807fc790SAndy Shevchenko else 1232807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1233c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 123420897f5cSArne Jansen 12358a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 12368a6a87cdSBoris Burkov 123720897f5cSArne Jansen key.objectid = objectid; 123820897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 123920897f5cSArne Jansen key.offset = 0; 124020897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 124120897f5cSArne Jansen if (ret) 124220897f5cSArne Jansen goto fail; 124320897f5cSArne Jansen 124420897f5cSArne Jansen return root; 12451dd05682STsutomu Itoh 12468a6a87cdSBoris Burkov fail: 124700246528SJosef Bacik btrfs_put_root(root); 12481dd05682STsutomu Itoh 12491dd05682STsutomu Itoh return ERR_PTR(ret); 125020897f5cSArne Jansen } 125120897f5cSArne Jansen 12527237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1253e02119d5SChris Mason struct btrfs_fs_info *fs_info) 12540f7d52f4SChris Mason { 12550f7d52f4SChris Mason struct btrfs_root *root; 1256e02119d5SChris Mason 125796dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1258e02119d5SChris Mason if (!root) 12597237f183SYan Zheng return ERR_PTR(-ENOMEM); 1260e02119d5SChris Mason 1261e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1262e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1263e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 126427cdeb70SMiao Xie 12656ab6ebb7SNaohiro Aota return root; 12666ab6ebb7SNaohiro Aota } 12676ab6ebb7SNaohiro Aota 12686ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 12696ab6ebb7SNaohiro Aota struct btrfs_root *root) 12706ab6ebb7SNaohiro Aota { 12716ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 12726ab6ebb7SNaohiro Aota 12737237f183SYan Zheng /* 127492a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 127527cdeb70SMiao Xie * 127692a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 127792a7cc42SQu Wenruo * and they go away before a real commit is actually done. 127892a7cc42SQu Wenruo * 127992a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 128092a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 12817237f183SYan Zheng */ 1282e02119d5SChris Mason 12834d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 12849631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 12856ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 12866ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 1287e02119d5SChris Mason 12887237f183SYan Zheng root->node = leaf; 1289e02119d5SChris Mason 1290e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1291e02119d5SChris Mason btrfs_tree_unlock(root->node); 12926ab6ebb7SNaohiro Aota 12936ab6ebb7SNaohiro Aota return 0; 12947237f183SYan Zheng } 12957237f183SYan Zheng 12967237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 12977237f183SYan Zheng struct btrfs_fs_info *fs_info) 12987237f183SYan Zheng { 12997237f183SYan Zheng struct btrfs_root *log_root; 13007237f183SYan Zheng 13017237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13027237f183SYan Zheng if (IS_ERR(log_root)) 13037237f183SYan Zheng return PTR_ERR(log_root); 13046ab6ebb7SNaohiro Aota 13053ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 13063ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 13073ddebf27SNaohiro Aota 13086ab6ebb7SNaohiro Aota if (ret) { 13096ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13106ab6ebb7SNaohiro Aota return ret; 13116ab6ebb7SNaohiro Aota } 13123ddebf27SNaohiro Aota } 13136ab6ebb7SNaohiro Aota 13147237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13157237f183SYan Zheng fs_info->log_root_tree = log_root; 13167237f183SYan Zheng return 0; 13177237f183SYan Zheng } 13187237f183SYan Zheng 13197237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13207237f183SYan Zheng struct btrfs_root *root) 13217237f183SYan Zheng { 13220b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13237237f183SYan Zheng struct btrfs_root *log_root; 13247237f183SYan Zheng struct btrfs_inode_item *inode_item; 13256ab6ebb7SNaohiro Aota int ret; 13267237f183SYan Zheng 13270b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13287237f183SYan Zheng if (IS_ERR(log_root)) 13297237f183SYan Zheng return PTR_ERR(log_root); 13307237f183SYan Zheng 13316ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 13326ab6ebb7SNaohiro Aota if (ret) { 13336ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13346ab6ebb7SNaohiro Aota return ret; 13356ab6ebb7SNaohiro Aota } 13366ab6ebb7SNaohiro Aota 13377237f183SYan Zheng log_root->last_trans = trans->transid; 13387237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 13397237f183SYan Zheng 13407237f183SYan Zheng inode_item = &log_root->root_item.inode; 13413cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 13423cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 13433cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1344da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 13450b246afaSJeff Mahoney fs_info->nodesize); 13463cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 13477237f183SYan Zheng 13485d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 13497237f183SYan Zheng 13507237f183SYan Zheng WARN_ON(root->log_root); 13517237f183SYan Zheng root->log_root = log_root; 13527237f183SYan Zheng root->log_transid = 0; 1353d1433debSMiao Xie root->log_transid_committed = -1; 1354257c62e1SChris Mason root->last_log_commit = 0; 1355e02119d5SChris Mason return 0; 1356e02119d5SChris Mason } 1357e02119d5SChris Mason 135849d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 135949d11beaSJosef Bacik struct btrfs_path *path, 1360cb517eabSMiao Xie struct btrfs_key *key) 1361e02119d5SChris Mason { 1362e02119d5SChris Mason struct btrfs_root *root; 1363e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 136484234f3aSYan Zheng u64 generation; 1365cb517eabSMiao Xie int ret; 1366581c1760SQu Wenruo int level; 1367cb517eabSMiao Xie 136896dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 136949d11beaSJosef Bacik if (!root) 137049d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 13710f7d52f4SChris Mason 1372cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1373cb517eabSMiao Xie &root->root_item, &root->root_key); 13740f7d52f4SChris Mason if (ret) { 137513a8a7c8SYan, Zheng if (ret > 0) 137613a8a7c8SYan, Zheng ret = -ENOENT; 137749d11beaSJosef Bacik goto fail; 13780f7d52f4SChris Mason } 137913a8a7c8SYan, Zheng 138084234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1381581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 13822ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 13832ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 13841b7ec85eSJosef Bacik key->objectid, generation, level, NULL); 138564c043deSLiu Bo if (IS_ERR(root->node)) { 138664c043deSLiu Bo ret = PTR_ERR(root->node); 13878c38938cSJosef Bacik root->node = NULL; 138849d11beaSJosef Bacik goto fail; 13894eb150d6SQu Wenruo } 13904eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1391cb517eabSMiao Xie ret = -EIO; 139249d11beaSJosef Bacik goto fail; 1393416bc658SJosef Bacik } 139488c602abSQu Wenruo 139588c602abSQu Wenruo /* 139688c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 139788c602abSQu Wenruo * match its root node owner 139888c602abSQu Wenruo */ 139988c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 140088c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 140188c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 140288c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 140388c602abSQu Wenruo btrfs_crit(fs_info, 140488c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 140588c602abSQu Wenruo root->root_key.objectid, root->node->start, 140688c602abSQu Wenruo btrfs_header_owner(root->node), 140788c602abSQu Wenruo root->root_key.objectid); 140888c602abSQu Wenruo ret = -EUCLEAN; 140988c602abSQu Wenruo goto fail; 141088c602abSQu Wenruo } 14115d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1412cb517eabSMiao Xie return root; 141349d11beaSJosef Bacik fail: 141400246528SJosef Bacik btrfs_put_root(root); 141549d11beaSJosef Bacik return ERR_PTR(ret); 141649d11beaSJosef Bacik } 141749d11beaSJosef Bacik 141849d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 141949d11beaSJosef Bacik struct btrfs_key *key) 142049d11beaSJosef Bacik { 142149d11beaSJosef Bacik struct btrfs_root *root; 142249d11beaSJosef Bacik struct btrfs_path *path; 142349d11beaSJosef Bacik 142449d11beaSJosef Bacik path = btrfs_alloc_path(); 142549d11beaSJosef Bacik if (!path) 142649d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 142749d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 142849d11beaSJosef Bacik btrfs_free_path(path); 142949d11beaSJosef Bacik 143049d11beaSJosef Bacik return root; 1431cb517eabSMiao Xie } 1432cb517eabSMiao Xie 14332dfb1e43SQu Wenruo /* 14342dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 14352dfb1e43SQu Wenruo * 14362dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 14372dfb1e43SQu Wenruo */ 14382dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1439cb517eabSMiao Xie { 1440cb517eabSMiao Xie int ret; 1441dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1442cb517eabSMiao Xie 1443dcc3eb96SNikolay Borisov /* 1444dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1445dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1446dcc3eb96SNikolay Borisov */ 1447dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1448dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1449dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1450dcc3eb96SNikolay Borisov if (ret) 14518257b2dcSMiao Xie goto fail; 14528257b2dcSMiao Xie 1453aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 145437f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 145592a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1456f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1457f39e4571SJosef Bacik } 1458f39e4571SJosef Bacik 1459851fd730SQu Wenruo /* 1460851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1461851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1462851fd730SQu Wenruo */ 1463851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1464851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 14652dfb1e43SQu Wenruo if (!anon_dev) { 1466cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1467cb517eabSMiao Xie if (ret) 1468876d2cf1SLiu Bo goto fail; 14692dfb1e43SQu Wenruo } else { 14702dfb1e43SQu Wenruo root->anon_dev = anon_dev; 14712dfb1e43SQu Wenruo } 1472851fd730SQu Wenruo } 1473f32e48e9SChandan Rajendra 1474f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1475453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1476f32e48e9SChandan Rajendra if (ret) { 1477f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1478876d2cf1SLiu Bo goto fail; 1479f32e48e9SChandan Rajendra } 1480f32e48e9SChandan Rajendra 14816b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1482f32e48e9SChandan Rajendra 1483f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1484f32e48e9SChandan Rajendra 1485cb517eabSMiao Xie return 0; 1486cb517eabSMiao Xie fail: 148784db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1488cb517eabSMiao Xie return ret; 1489cb517eabSMiao Xie } 1490cb517eabSMiao Xie 1491a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1492cb517eabSMiao Xie u64 root_id) 1493cb517eabSMiao Xie { 1494cb517eabSMiao Xie struct btrfs_root *root; 1495cb517eabSMiao Xie 1496fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1497fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1498fc7cbcd4SDavid Sterba (unsigned long)root_id); 1499bc44d7c4SJosef Bacik if (root) 150000246528SJosef Bacik root = btrfs_grab_root(root); 1501fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1502cb517eabSMiao Xie return root; 1503cb517eabSMiao Xie } 1504cb517eabSMiao Xie 150549d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 150649d11beaSJosef Bacik u64 objectid) 150749d11beaSJosef Bacik { 1508abed4aaaSJosef Bacik struct btrfs_key key = { 1509abed4aaaSJosef Bacik .objectid = objectid, 1510abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1511abed4aaaSJosef Bacik .offset = 0, 1512abed4aaaSJosef Bacik }; 1513abed4aaaSJosef Bacik 151449d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 151549d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 151649d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 1517abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 151849d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 151949d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 152049d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 152149d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 152249d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 1523abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 152449d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 152549d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 152649d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 152749d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 152849d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 152949d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 153014033b08SQu Wenruo if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) 153114033b08SQu Wenruo return btrfs_grab_root(fs_info->block_group_root) ? 153214033b08SQu Wenruo fs_info->block_group_root : ERR_PTR(-ENOENT); 1533abed4aaaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { 1534abed4aaaSJosef Bacik struct btrfs_root *root = btrfs_global_root(fs_info, &key); 1535abed4aaaSJosef Bacik 1536abed4aaaSJosef Bacik return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); 1537abed4aaaSJosef Bacik } 153849d11beaSJosef Bacik return NULL; 153949d11beaSJosef Bacik } 154049d11beaSJosef Bacik 1541cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1542cb517eabSMiao Xie struct btrfs_root *root) 1543cb517eabSMiao Xie { 1544cb517eabSMiao Xie int ret; 1545cb517eabSMiao Xie 1546fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1547fc7cbcd4SDavid Sterba if (ret) 1548fc7cbcd4SDavid Sterba return ret; 1549fc7cbcd4SDavid Sterba 1550fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1551fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1552fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1553fc7cbcd4SDavid Sterba root); 1554af01d2e5SJosef Bacik if (ret == 0) { 155500246528SJosef Bacik btrfs_grab_root(root); 1556fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1557af01d2e5SJosef Bacik } 1558fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1559fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1560cb517eabSMiao Xie 1561cb517eabSMiao Xie return ret; 1562cb517eabSMiao Xie } 1563cb517eabSMiao Xie 1564bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1565bd647ce3SJosef Bacik { 1566bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1567bd647ce3SJosef Bacik struct btrfs_root *root; 1568bd647ce3SJosef Bacik 1569bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1570457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1571457f1864SJosef Bacik 1572bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1573bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1574457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 157571008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1576bd647ce3SJosef Bacik refcount_read(&root->refs)); 1577bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 157800246528SJosef Bacik btrfs_put_root(root); 157900246528SJosef Bacik btrfs_put_root(root); 1580bd647ce3SJosef Bacik } 1581bd647ce3SJosef Bacik #endif 1582bd647ce3SJosef Bacik } 1583bd647ce3SJosef Bacik 1584abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1585abed4aaaSJosef Bacik { 1586abed4aaaSJosef Bacik struct btrfs_root *root; 1587abed4aaaSJosef Bacik struct rb_node *node; 1588abed4aaaSJosef Bacik 1589abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1590abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1591abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1592abed4aaaSJosef Bacik btrfs_put_root(root); 1593abed4aaaSJosef Bacik } 1594abed4aaaSJosef Bacik } 1595abed4aaaSJosef Bacik 15960d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 15970d4b0463SJosef Bacik { 1598141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1599141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 16005deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1601141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1602141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1603141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1604141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 16050d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 16060d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1607abed4aaaSJosef Bacik free_global_roots(fs_info); 160800246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 160900246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 161000246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 161100246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 161200246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 161300246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1614aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 16159c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1616bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 16173fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 16180d4b0463SJosef Bacik kfree(fs_info->super_copy); 16190d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 16208481dd80SQu Wenruo kfree(fs_info->subpage_info); 16210d4b0463SJosef Bacik kvfree(fs_info); 16220d4b0463SJosef Bacik } 16230d4b0463SJosef Bacik 16240d4b0463SJosef Bacik 16252dfb1e43SQu Wenruo /* 16262dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 16272dfb1e43SQu Wenruo * 16282dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 16292dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 16302dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 16312dfb1e43SQu Wenruo * 16322dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 16332dfb1e43SQu Wenruo * 16342dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 16352dfb1e43SQu Wenruo * same root objectid. 16362dfb1e43SQu Wenruo * 16372dfb1e43SQu Wenruo * @objectid: root id 16382dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 16392dfb1e43SQu Wenruo * pass 0 for new allocation. 16402dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 16412dfb1e43SQu Wenruo * for orphan roots 16422dfb1e43SQu Wenruo */ 16432dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 16442dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 16452dfb1e43SQu Wenruo bool check_ref) 16465eda7b5eSChris Mason { 16475eda7b5eSChris Mason struct btrfs_root *root; 1648381cf658SDavid Sterba struct btrfs_path *path; 16491d4c08e0SDavid Sterba struct btrfs_key key; 16505eda7b5eSChris Mason int ret; 16515eda7b5eSChris Mason 165249d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 165349d11beaSJosef Bacik if (root) 165449d11beaSJosef Bacik return root; 16554df27c4dSYan, Zheng again: 165656e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 165748475471SStefan Behrens if (root) { 16582dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 16592dfb1e43SQu Wenruo ASSERT(!anon_dev); 1660bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 166100246528SJosef Bacik btrfs_put_root(root); 166248475471SStefan Behrens return ERR_PTR(-ENOENT); 1663bc44d7c4SJosef Bacik } 16645eda7b5eSChris Mason return root; 166548475471SStefan Behrens } 16665eda7b5eSChris Mason 166756e9357aSDavid Sterba key.objectid = objectid; 166856e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 166956e9357aSDavid Sterba key.offset = (u64)-1; 167056e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 16715eda7b5eSChris Mason if (IS_ERR(root)) 16725eda7b5eSChris Mason return root; 16733394e160SChris Mason 1674c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1675d68fc57bSYan, Zheng ret = -ENOENT; 1676d68fc57bSYan, Zheng goto fail; 1677d68fc57bSYan, Zheng } 1678d68fc57bSYan, Zheng 16792dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1680cb517eabSMiao Xie if (ret) 1681cb517eabSMiao Xie goto fail; 1682cb517eabSMiao Xie 1683381cf658SDavid Sterba path = btrfs_alloc_path(); 1684381cf658SDavid Sterba if (!path) { 1685381cf658SDavid Sterba ret = -ENOMEM; 1686381cf658SDavid Sterba goto fail; 1687381cf658SDavid Sterba } 16881d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 16891d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 169056e9357aSDavid Sterba key.offset = objectid; 16911d4c08e0SDavid Sterba 16921d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1693381cf658SDavid Sterba btrfs_free_path(path); 1694d68fc57bSYan, Zheng if (ret < 0) 1695d68fc57bSYan, Zheng goto fail; 1696d68fc57bSYan, Zheng if (ret == 0) 169727cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1698d68fc57bSYan, Zheng 1699cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 17000f7d52f4SChris Mason if (ret) { 1701168a2f77SJia-Ju Bai if (ret == -EEXIST) { 170200246528SJosef Bacik btrfs_put_root(root); 17034df27c4dSYan, Zheng goto again; 1704168a2f77SJia-Ju Bai } 17054df27c4dSYan, Zheng goto fail; 17064df27c4dSYan, Zheng } 1707edbd8d4eSChris Mason return root; 17084df27c4dSYan, Zheng fail: 170933fab972SFilipe Manana /* 171033fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1711143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 171233fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 171333fab972SFilipe Manana * and once again by our caller. 171433fab972SFilipe Manana */ 171533fab972SFilipe Manana if (anon_dev) 171633fab972SFilipe Manana root->anon_dev = 0; 17178c38938cSJosef Bacik btrfs_put_root(root); 17184df27c4dSYan, Zheng return ERR_PTR(ret); 1719edbd8d4eSChris Mason } 1720edbd8d4eSChris Mason 17212dfb1e43SQu Wenruo /* 17222dfb1e43SQu Wenruo * Get in-memory reference of a root structure 17232dfb1e43SQu Wenruo * 17242dfb1e43SQu Wenruo * @objectid: tree objectid 17252dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 17262dfb1e43SQu Wenruo * one reference 17272dfb1e43SQu Wenruo */ 17282dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 17292dfb1e43SQu Wenruo u64 objectid, bool check_ref) 17302dfb1e43SQu Wenruo { 17312dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 17322dfb1e43SQu Wenruo } 17332dfb1e43SQu Wenruo 17342dfb1e43SQu Wenruo /* 17352dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 17362dfb1e43SQu Wenruo * the anonymous block device id 17372dfb1e43SQu Wenruo * 17382dfb1e43SQu Wenruo * @objectid: tree objectid 17392dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 17402dfb1e43SQu Wenruo * parameter value 17412dfb1e43SQu Wenruo */ 17422dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 17432dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 17442dfb1e43SQu Wenruo { 17452dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 17462dfb1e43SQu Wenruo } 17472dfb1e43SQu Wenruo 17488b712842SChris Mason /* 174949d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 175049d11beaSJosef Bacik * @fs_info: the fs_info 175149d11beaSJosef Bacik * @objectid: the objectid we need to lookup 175249d11beaSJosef Bacik * 175349d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 175449d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 175549d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 175649d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 175749d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 175849d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 175949d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 176049d11beaSJosef Bacik * backref code is finished using the root. 176149d11beaSJosef Bacik */ 176249d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 176349d11beaSJosef Bacik struct btrfs_path *path, 176449d11beaSJosef Bacik u64 objectid) 176549d11beaSJosef Bacik { 176649d11beaSJosef Bacik struct btrfs_root *root; 176749d11beaSJosef Bacik struct btrfs_key key; 176849d11beaSJosef Bacik 176949d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 177049d11beaSJosef Bacik 177149d11beaSJosef Bacik /* 177249d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 177349d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 177449d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 177549d11beaSJosef Bacik * != NULL just return it. 177649d11beaSJosef Bacik */ 177749d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 177849d11beaSJosef Bacik if (root) 177949d11beaSJosef Bacik return root; 178049d11beaSJosef Bacik 178149d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 178249d11beaSJosef Bacik if (root) 178349d11beaSJosef Bacik return root; 178449d11beaSJosef Bacik 178549d11beaSJosef Bacik key.objectid = objectid; 178649d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 178749d11beaSJosef Bacik key.offset = (u64)-1; 178849d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 178949d11beaSJosef Bacik btrfs_release_path(path); 179049d11beaSJosef Bacik 179149d11beaSJosef Bacik return root; 179249d11beaSJosef Bacik } 179349d11beaSJosef Bacik 1794a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1795a74a4b97SChris Mason { 17960d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1797d0278245SMiao Xie int again; 1798a74a4b97SChris Mason 1799d6fd0ae2SOmar Sandoval while (1) { 1800d0278245SMiao Xie again = 0; 18019d1a2a3aSDavid Sterba 1802fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1803fd340d0fSJosef Bacik 1804d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 18052ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1806d0278245SMiao Xie goto sleep; 1807d0278245SMiao Xie 180890c711abSZygo Blaxell /* 180990c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 181090c711abSZygo Blaxell * before we have finished mounting the filesystem. 181190c711abSZygo Blaxell */ 18120b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 181390c711abSZygo Blaxell goto sleep; 181490c711abSZygo Blaxell 18150b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1816d0278245SMiao Xie goto sleep; 1817d0278245SMiao Xie 1818dc7f370cSMiao Xie /* 1819dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1820dc7f370cSMiao Xie * during the above check and trylock. 1821dc7f370cSMiao Xie */ 18222ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 18230b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1824dc7f370cSMiao Xie goto sleep; 1825dc7f370cSMiao Xie } 1826dc7f370cSMiao Xie 18272ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1828c2d6cb16SFilipe Manana 182933c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 18300b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1831a74a4b97SChris Mason 1832d0278245SMiao Xie /* 183305323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 183405323cd1SMiao Xie * needn't do anything special here. 1835d0278245SMiao Xie */ 18360b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 183767c5e7d4SFilipe Manana 183867c5e7d4SFilipe Manana /* 1839f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 184067c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 184167c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1842f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 184367c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 184467c5e7d4SFilipe Manana * unused block groups. 184567c5e7d4SFilipe Manana */ 18460b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 184718bb8bbfSJohannes Thumshirn 184818bb8bbfSJohannes Thumshirn /* 184918bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 185018bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 185118bb8bbfSJohannes Thumshirn * space. 185218bb8bbfSJohannes Thumshirn */ 185318bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1854d0278245SMiao Xie sleep: 1855a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1856d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1857d6fd0ae2SOmar Sandoval kthread_parkme(); 1858d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1859d6fd0ae2SOmar Sandoval return 0; 1860838fe188SJiri Kosina if (!again) { 1861a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1862a74a4b97SChris Mason schedule(); 1863a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1864a74a4b97SChris Mason } 1865da288d28SFilipe Manana } 1866a74a4b97SChris Mason } 1867a74a4b97SChris Mason 1868a74a4b97SChris Mason static int transaction_kthread(void *arg) 1869a74a4b97SChris Mason { 1870a74a4b97SChris Mason struct btrfs_root *root = arg; 18710b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1872a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1873a74a4b97SChris Mason struct btrfs_transaction *cur; 18748929ecfaSYan, Zheng u64 transid; 1875643900beSNikolay Borisov time64_t delta; 1876a74a4b97SChris Mason unsigned long delay; 1877914b2007SJan Kara bool cannot_commit; 1878a74a4b97SChris Mason 1879a74a4b97SChris Mason do { 1880914b2007SJan Kara cannot_commit = false; 1881ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 18820b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1883a74a4b97SChris Mason 18840b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 18850b246afaSJeff Mahoney cur = fs_info->running_transaction; 1886a74a4b97SChris Mason if (!cur) { 18870b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1888a74a4b97SChris Mason goto sleep; 1889a74a4b97SChris Mason } 189031153d81SYan Zheng 1891643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1892fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1893fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1894643900beSNikolay Borisov delta < fs_info->commit_interval) { 18950b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1896fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1897fb8a7e94SNikolay Borisov delay = min(delay, 1898fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1899a74a4b97SChris Mason goto sleep; 1900a74a4b97SChris Mason } 19018929ecfaSYan, Zheng transid = cur->transid; 19020b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 190356bec294SChris Mason 190479787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1905354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1906914b2007SJan Kara if (IS_ERR(trans)) { 1907354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1908914b2007SJan Kara cannot_commit = true; 190979787eaaSJeff Mahoney goto sleep; 1910914b2007SJan Kara } 19118929ecfaSYan, Zheng if (transid == trans->transid) { 19123a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 19138929ecfaSYan, Zheng } else { 19143a45bb20SJeff Mahoney btrfs_end_transaction(trans); 19158929ecfaSYan, Zheng } 1916a74a4b97SChris Mason sleep: 19170b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 19180b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1919a74a4b97SChris Mason 192084961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 19212ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 19228929ecfaSYan, Zheng if (!kthread_should_stop() && 19230b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1924914b2007SJan Kara cannot_commit)) 1925bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1926a74a4b97SChris Mason } while (!kthread_should_stop()); 1927a74a4b97SChris Mason return 0; 1928a74a4b97SChris Mason } 1929a74a4b97SChris Mason 1930af31f5e5SChris Mason /* 193101f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 193201f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 193301f0f9daSNikolay Borisov * anything. 1934af31f5e5SChris Mason * 1935af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1936af31f5e5SChris Mason * generation of the latest root in the array with the generation 1937af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1938af31f5e5SChris Mason */ 193901f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1940af31f5e5SChris Mason { 194101f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1942af31f5e5SChris Mason u64 cur; 1943af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1944af31f5e5SChris Mason int i; 1945af31f5e5SChris Mason 1946af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1947af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1948af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1949af31f5e5SChris Mason if (cur == newest_gen) 195001f0f9daSNikolay Borisov return i; 1951af31f5e5SChris Mason } 1952af31f5e5SChris Mason 195301f0f9daSNikolay Borisov return -EINVAL; 1954af31f5e5SChris Mason } 1955af31f5e5SChris Mason 1956af31f5e5SChris Mason /* 1957af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1958af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1959af31f5e5SChris Mason * done 1960af31f5e5SChris Mason */ 1961af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1962af31f5e5SChris Mason { 19636ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1964af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1965af31f5e5SChris Mason 1966af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1967af31f5e5SChris Mason 1968af31f5e5SChris Mason /* 1969af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1970af31f5e5SChris Mason * regardless of which ones we use today 1971af31f5e5SChris Mason */ 1972af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1973af31f5e5SChris Mason 1974af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1975af31f5e5SChris Mason 1976af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1977af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1978af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1979af31f5e5SChris Mason 1980af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1981af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1982af31f5e5SChris Mason 1983af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1984af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1985af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1986af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1987af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1988af31f5e5SChris Mason 19891c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 19909c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1991f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 19929c54e80dSJosef Bacik 19939c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 19949c54e80dSJosef Bacik extent_root->node->start); 1995af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 199629cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 1997af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 199829cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 1999af31f5e5SChris Mason 2000f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 2001f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 2002f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 2003f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 2004f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 20059c54e80dSJosef Bacik } 2006af31f5e5SChris Mason 20077c7e82a7SChris Mason /* 20087c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 20097c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 20107c7e82a7SChris Mason */ 20117c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 20127c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 20137c7e82a7SChris Mason info->fs_root->node->start); 2014af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2015af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2016af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2017af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 20187c7e82a7SChris Mason } 2019af31f5e5SChris Mason 2020af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2021af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2022af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2023af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2024af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2025af31f5e5SChris Mason 2026af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2027af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2028af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2029af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2030af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2031af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2032af31f5e5SChris Mason 2033af31f5e5SChris Mason /* 2034af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2035af31f5e5SChris Mason * for the next commit 2036af31f5e5SChris Mason */ 2037af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2038af31f5e5SChris Mason &info->super_for_commit->super_roots, 2039af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2040af31f5e5SChris Mason } 2041af31f5e5SChris Mason 2042af31f5e5SChris Mason /* 2043bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 2044bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 2045bd2336b2SNikolay Borisov * 2046bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 2047bd2336b2SNikolay Borisov * priority - priority of backup root required 2048bd2336b2SNikolay Borisov * 2049bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 2050bd2336b2SNikolay Borisov */ 2051bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 2052bd2336b2SNikolay Borisov { 2053bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2054bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 2055bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 2056bd2336b2SNikolay Borisov 2057bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 2058bd2336b2SNikolay Borisov if (priority == 0) 2059bd2336b2SNikolay Borisov return backup_index; 2060bd2336b2SNikolay Borisov 2061bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 2062bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 2063bd2336b2SNikolay Borisov } else { 2064bd2336b2SNikolay Borisov return -EINVAL; 2065bd2336b2SNikolay Borisov } 2066bd2336b2SNikolay Borisov 2067bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 2068bd2336b2SNikolay Borisov 2069bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 2070bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 2071bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2072bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 2073bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 2074bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2075bd2336b2SNikolay Borisov 2076bd2336b2SNikolay Borisov /* 2077bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 2078bd2336b2SNikolay Borisov * need a fsck 2079bd2336b2SNikolay Borisov */ 2080bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2081bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2082bd2336b2SNikolay Borisov 2083bd2336b2SNikolay Borisov return backup_index; 2084bd2336b2SNikolay Borisov } 2085bd2336b2SNikolay Borisov 20867abadb64SLiu Bo /* helper to cleanup workers */ 20877abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20887abadb64SLiu Bo { 2089dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2090afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 2091a31b4a43SChristoph Hellwig btrfs_destroy_workqueue(fs_info->hipri_workers); 20925cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2093d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 2094d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 2095d34e123dSChristoph Hellwig if (fs_info->endio_raid56_workers) 2096d34e123dSChristoph Hellwig destroy_workqueue(fs_info->endio_raid56_workers); 2097385de0efSChristoph Hellwig if (fs_info->rmw_workers) 2098385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 2099fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 2100fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 2101fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2102fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 21035b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2104e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2105a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2106fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2107b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 2108b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 2109a9b9477dSFilipe Manana /* 2110a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2111a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2112a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2113a9b9477dSFilipe Manana */ 2114d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 2115d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 21167abadb64SLiu Bo } 21177abadb64SLiu Bo 21182e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 21192e9f5954SRashika { 21202e9f5954SRashika if (root) { 21212e9f5954SRashika free_extent_buffer(root->node); 21222e9f5954SRashika free_extent_buffer(root->commit_root); 21232e9f5954SRashika root->node = NULL; 21242e9f5954SRashika root->commit_root = NULL; 21252e9f5954SRashika } 21262e9f5954SRashika } 21272e9f5954SRashika 2128abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 2129abed4aaaSJosef Bacik { 2130abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 2131abed4aaaSJosef Bacik 2132abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 2133abed4aaaSJosef Bacik &fs_info->global_root_tree, 2134abed4aaaSJosef Bacik rb_node) 2135abed4aaaSJosef Bacik free_root_extent_buffers(root); 2136abed4aaaSJosef Bacik } 2137abed4aaaSJosef Bacik 2138af31f5e5SChris Mason /* helper to cleanup tree roots */ 21394273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2140af31f5e5SChris Mason { 21412e9f5954SRashika free_root_extent_buffers(info->tree_root); 2142655b09feSJosef Bacik 2143abed4aaaSJosef Bacik free_global_root_pointers(info); 21442e9f5954SRashika free_root_extent_buffers(info->dev_root); 21452e9f5954SRashika free_root_extent_buffers(info->quota_root); 21462e9f5954SRashika free_root_extent_buffers(info->uuid_root); 21478c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2148aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 21499c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 21504273eaffSAnand Jain if (free_chunk_root) 21512e9f5954SRashika free_root_extent_buffers(info->chunk_root); 2152af31f5e5SChris Mason } 2153af31f5e5SChris Mason 21548c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 21558c38938cSJosef Bacik { 21568c38938cSJosef Bacik if (!root) 21578c38938cSJosef Bacik return; 21588c38938cSJosef Bacik 21598c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 21608c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 21611dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 21628c38938cSJosef Bacik if (root->anon_dev) 21638c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 21648c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2165923eb523SJohannes Thumshirn free_root_extent_buffers(root); 21668c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2167fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 21688c38938cSJosef Bacik list_del_init(&root->leak_list); 2169fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 21708c38938cSJosef Bacik #endif 21718c38938cSJosef Bacik kfree(root); 21728c38938cSJosef Bacik } 21738c38938cSJosef Bacik } 21748c38938cSJosef Bacik 2175faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2176171f6537SJosef Bacik { 2177fc7cbcd4SDavid Sterba int ret; 2178fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 2179fc7cbcd4SDavid Sterba int i; 2180171f6537SJosef Bacik 2181171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2182fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 2183171f6537SJosef Bacik struct btrfs_root, root_list); 2184fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 2185171f6537SJosef Bacik 2186fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2187fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2188fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 2189171f6537SJosef Bacik } 2190171f6537SJosef Bacik 2191fc7cbcd4SDavid Sterba while (1) { 2192fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2193fc7cbcd4SDavid Sterba (void **)gang, 0, 2194fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 2195fc7cbcd4SDavid Sterba if (!ret) 2196fc7cbcd4SDavid Sterba break; 2197fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 2198fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2199171f6537SJosef Bacik } 2200171f6537SJosef Bacik } 2201af31f5e5SChris Mason 2202638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2203638aa7edSEric Sandeen { 2204638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2205638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2206638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2207638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2208638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2209638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2210ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2211638aa7edSEric Sandeen } 2212638aa7edSEric Sandeen 2213779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2214779a65a4SEric Sandeen { 2215779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2216779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2217779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2218779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2219779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2220779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2221907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 2222779a65a4SEric Sandeen } 2223779a65a4SEric Sandeen 22246bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2225f37938e0SEric Sandeen { 22262ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 2227e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 2228e256927bSJosef Bacik fs_info->tree_root); 22292ff7e61eSJeff Mahoney 22302ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 22312ff7e61eSJeff Mahoney set_nlink(inode, 1); 2232f37938e0SEric Sandeen /* 2233f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2234f37938e0SEric Sandeen * the real end of the address space is determined by all of 2235f37938e0SEric Sandeen * the devices in the system 2236f37938e0SEric Sandeen */ 22372ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 22382ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2239f37938e0SEric Sandeen 22402ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 224143eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 2242efb0645bSJosef Bacik IO_TREE_BTREE_INODE_IO, NULL); 22432ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2244f37938e0SEric Sandeen 22455c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 2246adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 2247adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 2248adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 22492ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 2250e256927bSJosef Bacik __insert_inode_hash(inode, hash); 2251f37938e0SEric Sandeen } 2252f37938e0SEric Sandeen 2253ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2254ad618368SEric Sandeen { 2255ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2256129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 22577f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2258ad618368SEric Sandeen } 2259ad618368SEric Sandeen 2260f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2261f9e92e40SEric Sandeen { 2262f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2263f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2264f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2265f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2266f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2267f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2268d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2269011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 2270f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2271f9e92e40SEric Sandeen } 2272f9e92e40SEric Sandeen 2273d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 22742a458198SEric Sandeen { 2275f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 22766f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 22772a458198SEric Sandeen 22782a458198SEric Sandeen fs_info->workers = 2279a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 2280a31b4a43SChristoph Hellwig fs_info->hipri_workers = 2281a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker-high", 2282cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 22832a458198SEric Sandeen 22842a458198SEric Sandeen fs_info->delalloc_workers = 2285cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2286cb001095SJeff Mahoney flags, max_active, 2); 22872a458198SEric Sandeen 22882a458198SEric Sandeen fs_info->flush_workers = 2289cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2290cb001095SJeff Mahoney flags, max_active, 0); 22912a458198SEric Sandeen 22922a458198SEric Sandeen fs_info->caching_workers = 2293cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 22942a458198SEric Sandeen 22952a458198SEric Sandeen fs_info->fixup_workers = 2296cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 22972a458198SEric Sandeen 22982a458198SEric Sandeen fs_info->endio_workers = 2299d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 23002a458198SEric Sandeen fs_info->endio_meta_workers = 2301d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 23022a458198SEric Sandeen fs_info->endio_raid56_workers = 2303d34e123dSChristoph Hellwig alloc_workqueue("btrfs-endio-raid56", flags, max_active); 2304385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 23052a458198SEric Sandeen fs_info->endio_write_workers = 2306cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2307cb001095SJeff Mahoney max_active, 2); 2308fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 2309fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 23102a458198SEric Sandeen fs_info->endio_freespace_worker = 2311cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2312cb001095SJeff Mahoney max_active, 0); 23132a458198SEric Sandeen fs_info->delayed_workers = 2314cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2315cb001095SJeff Mahoney max_active, 0); 23162a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2317cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2318b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2319b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 23202a458198SEric Sandeen 2321a31b4a43SChristoph Hellwig if (!(fs_info->workers && fs_info->hipri_workers && 2322a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 23232a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 2324fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 23252a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 23262a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 2327f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 2328f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 2329b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 23302a458198SEric Sandeen return -ENOMEM; 23312a458198SEric Sandeen } 23322a458198SEric Sandeen 23332a458198SEric Sandeen return 0; 23342a458198SEric Sandeen } 23352a458198SEric Sandeen 23366d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 23376d97c6e3SJohannes Thumshirn { 23386d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2339b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 23406d97c6e3SJohannes Thumshirn 2341b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 23426d97c6e3SJohannes Thumshirn 23436d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 23446d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2345b4e967beSDavid Sterba csum_driver); 23466d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 23476d97c6e3SJohannes Thumshirn } 23486d97c6e3SJohannes Thumshirn 23496d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 23506d97c6e3SJohannes Thumshirn 2351c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2352c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2353c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 23546d97c6e3SJohannes Thumshirn return 0; 23556d97c6e3SJohannes Thumshirn } 23566d97c6e3SJohannes Thumshirn 235763443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 235863443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 235963443bf5SEric Sandeen { 236063443bf5SEric Sandeen int ret; 236163443bf5SEric Sandeen struct btrfs_root *log_tree_root; 236263443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 236363443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2364581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 236563443bf5SEric Sandeen 236663443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2367f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 236863443bf5SEric Sandeen return -EIO; 236963443bf5SEric Sandeen } 237063443bf5SEric Sandeen 237196dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 237296dfcb46SJosef Bacik GFP_KERNEL); 237363443bf5SEric Sandeen if (!log_tree_root) 237463443bf5SEric Sandeen return -ENOMEM; 237563443bf5SEric Sandeen 23762ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 23771b7ec85eSJosef Bacik BTRFS_TREE_LOG_OBJECTID, 23781b7ec85eSJosef Bacik fs_info->generation + 1, level, 23791b7ec85eSJosef Bacik NULL); 238064c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2381f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 23820eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 23838c38938cSJosef Bacik log_tree_root->node = NULL; 238400246528SJosef Bacik btrfs_put_root(log_tree_root); 23850eeff236SLiu Bo return ret; 23864eb150d6SQu Wenruo } 23874eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2388f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 238900246528SJosef Bacik btrfs_put_root(log_tree_root); 239063443bf5SEric Sandeen return -EIO; 239163443bf5SEric Sandeen } 23924eb150d6SQu Wenruo 239363443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 239463443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 239563443bf5SEric Sandeen if (ret) { 23960b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 239763443bf5SEric Sandeen "Failed to recover log tree"); 239800246528SJosef Bacik btrfs_put_root(log_tree_root); 239963443bf5SEric Sandeen return ret; 240063443bf5SEric Sandeen } 240163443bf5SEric Sandeen 2402bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 24036bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 240463443bf5SEric Sandeen if (ret) 240563443bf5SEric Sandeen return ret; 240663443bf5SEric Sandeen } 240763443bf5SEric Sandeen 240863443bf5SEric Sandeen return 0; 240963443bf5SEric Sandeen } 241063443bf5SEric Sandeen 2411abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2412abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2413abed4aaaSJosef Bacik const char *name) 2414abed4aaaSJosef Bacik { 2415abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2416abed4aaaSJosef Bacik struct btrfs_root *root; 2417f7238e50SJosef Bacik u64 max_global_id = 0; 2418abed4aaaSJosef Bacik int ret; 2419abed4aaaSJosef Bacik struct btrfs_key key = { 2420abed4aaaSJosef Bacik .objectid = objectid, 2421abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2422abed4aaaSJosef Bacik .offset = 0, 2423abed4aaaSJosef Bacik }; 2424abed4aaaSJosef Bacik bool found = false; 2425abed4aaaSJosef Bacik 2426abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2427abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2428abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2429abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2430abed4aaaSJosef Bacik return 0; 2431abed4aaaSJosef Bacik } 2432abed4aaaSJosef Bacik 2433abed4aaaSJosef Bacik while (1) { 2434abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2435abed4aaaSJosef Bacik if (ret < 0) 2436abed4aaaSJosef Bacik break; 2437abed4aaaSJosef Bacik 2438abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2439abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2440abed4aaaSJosef Bacik if (ret) { 2441abed4aaaSJosef Bacik if (ret > 0) 2442abed4aaaSJosef Bacik ret = 0; 2443abed4aaaSJosef Bacik break; 2444abed4aaaSJosef Bacik } 2445abed4aaaSJosef Bacik } 2446abed4aaaSJosef Bacik ret = 0; 2447abed4aaaSJosef Bacik 2448abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2449abed4aaaSJosef Bacik if (key.objectid != objectid) 2450abed4aaaSJosef Bacik break; 2451abed4aaaSJosef Bacik btrfs_release_path(path); 2452abed4aaaSJosef Bacik 2453f7238e50SJosef Bacik /* 2454f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2455f7238e50SJosef Bacik * everybody. 2456f7238e50SJosef Bacik */ 2457f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2458f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2459f7238e50SJosef Bacik 2460abed4aaaSJosef Bacik found = true; 2461abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2462abed4aaaSJosef Bacik if (IS_ERR(root)) { 2463abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2464abed4aaaSJosef Bacik ret = PTR_ERR(root); 2465abed4aaaSJosef Bacik break; 2466abed4aaaSJosef Bacik } 2467abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2468abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2469abed4aaaSJosef Bacik if (ret) { 2470abed4aaaSJosef Bacik btrfs_put_root(root); 2471abed4aaaSJosef Bacik break; 2472abed4aaaSJosef Bacik } 2473abed4aaaSJosef Bacik key.offset++; 2474abed4aaaSJosef Bacik } 2475abed4aaaSJosef Bacik btrfs_release_path(path); 2476abed4aaaSJosef Bacik 2477f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2478f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2479f7238e50SJosef Bacik 2480abed4aaaSJosef Bacik if (!found || ret) { 2481abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2482abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2483abed4aaaSJosef Bacik 2484abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2485abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2486abed4aaaSJosef Bacik else 2487abed4aaaSJosef Bacik ret = 0; 2488abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2489abed4aaaSJosef Bacik } 2490abed4aaaSJosef Bacik return ret; 2491abed4aaaSJosef Bacik } 2492abed4aaaSJosef Bacik 2493abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2494abed4aaaSJosef Bacik { 2495abed4aaaSJosef Bacik struct btrfs_path *path; 2496abed4aaaSJosef Bacik int ret = 0; 2497abed4aaaSJosef Bacik 2498abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2499abed4aaaSJosef Bacik if (!path) 2500abed4aaaSJosef Bacik return -ENOMEM; 2501abed4aaaSJosef Bacik 2502abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2503abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2504abed4aaaSJosef Bacik if (ret) 2505abed4aaaSJosef Bacik goto out; 2506abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2507abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2508abed4aaaSJosef Bacik if (ret) 2509abed4aaaSJosef Bacik goto out; 2510abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2511abed4aaaSJosef Bacik goto out; 2512abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2513abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2514abed4aaaSJosef Bacik "free space"); 2515abed4aaaSJosef Bacik out: 2516abed4aaaSJosef Bacik btrfs_free_path(path); 2517abed4aaaSJosef Bacik return ret; 2518abed4aaaSJosef Bacik } 2519abed4aaaSJosef Bacik 25206bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 25214bbcaa64SEric Sandeen { 25226bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2523a4f3d2c4SDavid Sterba struct btrfs_root *root; 25244bbcaa64SEric Sandeen struct btrfs_key location; 25254bbcaa64SEric Sandeen int ret; 25264bbcaa64SEric Sandeen 25276bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 25286bccf3abSJeff Mahoney 2529abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2530abed4aaaSJosef Bacik if (ret) 2531abed4aaaSJosef Bacik return ret; 2532abed4aaaSJosef Bacik 25334bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 25344bbcaa64SEric Sandeen location.offset = 0; 25354bbcaa64SEric Sandeen 25361c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 253714033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 253814033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 253914033b08SQu Wenruo if (IS_ERR(root)) { 254014033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 254114033b08SQu Wenruo ret = PTR_ERR(root); 254214033b08SQu Wenruo goto out; 254314033b08SQu Wenruo } 254414033b08SQu Wenruo } else { 254514033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 254614033b08SQu Wenruo fs_info->block_group_root = root; 254714033b08SQu Wenruo } 254814033b08SQu Wenruo } 254914033b08SQu Wenruo 255014033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2551a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2552f50f4353SLiu Bo if (IS_ERR(root)) { 255342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2554f50f4353SLiu Bo ret = PTR_ERR(root); 2555f50f4353SLiu Bo goto out; 2556f50f4353SLiu Bo } 255742437a63SJosef Bacik } else { 2558a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2559a4f3d2c4SDavid Sterba fs_info->dev_root = root; 256042437a63SJosef Bacik } 2561820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2562a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2563a8d1b164SJohannes Thumshirn if (ret) 2564a8d1b164SJohannes Thumshirn goto out; 25654bbcaa64SEric Sandeen 2566aeb935a4SQu Wenruo /* 2567aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2568aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2569aeb935a4SQu Wenruo */ 257056e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 257156e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2572aeb935a4SQu Wenruo if (IS_ERR(root)) { 257342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2574aeb935a4SQu Wenruo ret = PTR_ERR(root); 2575aeb935a4SQu Wenruo goto out; 2576aeb935a4SQu Wenruo } 257742437a63SJosef Bacik } else { 2578aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2579aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 258042437a63SJosef Bacik } 2581aeb935a4SQu Wenruo 25824bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2583a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2584a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2585a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2586afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2587a4f3d2c4SDavid Sterba fs_info->quota_root = root; 25884bbcaa64SEric Sandeen } 25894bbcaa64SEric Sandeen 25904bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2591a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2592a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 259342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2594a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 25954bbcaa64SEric Sandeen if (ret != -ENOENT) 2596f50f4353SLiu Bo goto out; 259742437a63SJosef Bacik } 25984bbcaa64SEric Sandeen } else { 2599a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2600a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 26014bbcaa64SEric Sandeen } 26024bbcaa64SEric Sandeen 26034bbcaa64SEric Sandeen return 0; 2604f50f4353SLiu Bo out: 2605f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2606f50f4353SLiu Bo location.objectid, ret); 2607f50f4353SLiu Bo return ret; 26084bbcaa64SEric Sandeen } 26094bbcaa64SEric Sandeen 2610069ec957SQu Wenruo /* 2611069ec957SQu Wenruo * Real super block validation 2612069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2613069ec957SQu Wenruo * 2614069ec957SQu Wenruo * @sb: super block to check 2615069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2616069ec957SQu Wenruo * 0 the primary (1st) sb 2617069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2618069ec957SQu Wenruo * -1 skip bytenr check 2619069ec957SQu Wenruo */ 2620a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2621069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 262221a852b0SQu Wenruo { 262321a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 262421a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 262521a852b0SQu Wenruo int ret = 0; 262621a852b0SQu Wenruo 262721a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 262821a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 262921a852b0SQu Wenruo ret = -EINVAL; 263021a852b0SQu Wenruo } 263121a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 263221a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 263321a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 263421a852b0SQu Wenruo ret = -EINVAL; 263521a852b0SQu Wenruo } 263621a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 263721a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 263821a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 263921a852b0SQu Wenruo ret = -EINVAL; 264021a852b0SQu Wenruo } 264121a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 264221a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 264321a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 264421a852b0SQu Wenruo ret = -EINVAL; 264521a852b0SQu Wenruo } 264621a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 264721a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 264821a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 264921a852b0SQu Wenruo ret = -EINVAL; 265021a852b0SQu Wenruo } 265121a852b0SQu Wenruo 265221a852b0SQu Wenruo /* 265321a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 265421a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 265521a852b0SQu Wenruo */ 265621a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 265721a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 265821a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 265921a852b0SQu Wenruo ret = -EINVAL; 266021a852b0SQu Wenruo } 26610bb3eb3eSQu Wenruo 26620bb3eb3eSQu Wenruo /* 26631a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 26641a42daabSQu Wenruo * 26651a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 26661a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 26671a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 26681a42daabSQu Wenruo * beginning. 26690bb3eb3eSQu Wenruo */ 26701a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 267121a852b0SQu Wenruo btrfs_err(fs_info, 26720bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 267321a852b0SQu Wenruo sectorsize, PAGE_SIZE); 267421a852b0SQu Wenruo ret = -EINVAL; 267521a852b0SQu Wenruo } 26760bb3eb3eSQu Wenruo 267721a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 267821a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 267921a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 268021a852b0SQu Wenruo ret = -EINVAL; 268121a852b0SQu Wenruo } 268221a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 268321a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 268421a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 268521a852b0SQu Wenruo ret = -EINVAL; 268621a852b0SQu Wenruo } 268721a852b0SQu Wenruo 268821a852b0SQu Wenruo /* Root alignment check */ 268921a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 269021a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 269121a852b0SQu Wenruo btrfs_super_root(sb)); 269221a852b0SQu Wenruo ret = -EINVAL; 269321a852b0SQu Wenruo } 269421a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 269521a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 269621a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 269721a852b0SQu Wenruo ret = -EINVAL; 269821a852b0SQu Wenruo } 269921a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 270021a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 270121a852b0SQu Wenruo btrfs_super_log_root(sb)); 270221a852b0SQu Wenruo ret = -EINVAL; 270321a852b0SQu Wenruo } 270421a852b0SQu Wenruo 2705aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2706aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2707aefd7f70SNikolay Borisov btrfs_err(fs_info, 2708aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2709aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2710aefd7f70SNikolay Borisov ret = -EINVAL; 2711aefd7f70SNikolay Borisov } 2712aefd7f70SNikolay Borisov 2713aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2714aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2715aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2716aefd7f70SNikolay Borisov btrfs_err(fs_info, 2717aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2718aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2719aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2720aefd7f70SNikolay Borisov ret = -EINVAL; 2721aefd7f70SNikolay Borisov } 2722aefd7f70SNikolay Borisov 27231c56ab99SQu Wenruo /* 27241c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 27251c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 27261c56ab99SQu Wenruo */ 27271c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 27281c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 27291c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 27301c56ab99SQu Wenruo btrfs_err(fs_info, 27311c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 27321c56ab99SQu Wenruo ret = -EINVAL; 27331c56ab99SQu Wenruo } 27341c56ab99SQu Wenruo 2735de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 27367239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 273721a852b0SQu Wenruo btrfs_err(fs_info, 27387239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2739de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 274021a852b0SQu Wenruo ret = -EINVAL; 274121a852b0SQu Wenruo } 274221a852b0SQu Wenruo 274321a852b0SQu Wenruo /* 274421a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 274521a852b0SQu Wenruo * done later 274621a852b0SQu Wenruo */ 274721a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 274821a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 274921a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 275021a852b0SQu Wenruo ret = -EINVAL; 275121a852b0SQu Wenruo } 275221a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 275321a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 275421a852b0SQu Wenruo btrfs_super_stripesize(sb)); 275521a852b0SQu Wenruo ret = -EINVAL; 275621a852b0SQu Wenruo } 275721a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 275821a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 275921a852b0SQu Wenruo btrfs_super_num_devices(sb)); 276021a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 276121a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 276221a852b0SQu Wenruo ret = -EINVAL; 276321a852b0SQu Wenruo } 276421a852b0SQu Wenruo 2765069ec957SQu Wenruo if (mirror_num >= 0 && 2766069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 276721a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 276821a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 276921a852b0SQu Wenruo ret = -EINVAL; 277021a852b0SQu Wenruo } 277121a852b0SQu Wenruo 277221a852b0SQu Wenruo /* 277321a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 277421a852b0SQu Wenruo * and one chunk 277521a852b0SQu Wenruo */ 277621a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 277721a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 277821a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 277921a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 278021a852b0SQu Wenruo ret = -EINVAL; 278121a852b0SQu Wenruo } 278221a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 278321a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 278421a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 278521a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 278621a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 278721a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 278821a852b0SQu Wenruo ret = -EINVAL; 278921a852b0SQu Wenruo } 279021a852b0SQu Wenruo 279121a852b0SQu Wenruo /* 279221a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 279321a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 279421a852b0SQu Wenruo */ 279521a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 279621a852b0SQu Wenruo btrfs_warn(fs_info, 279721a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 279821a852b0SQu Wenruo btrfs_super_generation(sb), 279921a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 280021a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 280121a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 280221a852b0SQu Wenruo btrfs_warn(fs_info, 280321a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 280421a852b0SQu Wenruo btrfs_super_generation(sb), 280521a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 280621a852b0SQu Wenruo 280721a852b0SQu Wenruo return ret; 280821a852b0SQu Wenruo } 280921a852b0SQu Wenruo 2810069ec957SQu Wenruo /* 2811069ec957SQu Wenruo * Validation of super block at mount time. 2812069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2813069ec957SQu Wenruo * flags will be skipped. 2814069ec957SQu Wenruo */ 2815069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2816069ec957SQu Wenruo { 2817a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2818069ec957SQu Wenruo } 2819069ec957SQu Wenruo 282075cb857dSQu Wenruo /* 282175cb857dSQu Wenruo * Validation of super block at write time. 282275cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 282375cb857dSQu Wenruo * overwritten soon. 282475cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 282575cb857dSQu Wenruo */ 282675cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 282775cb857dSQu Wenruo struct btrfs_super_block *sb) 282875cb857dSQu Wenruo { 282975cb857dSQu Wenruo int ret; 283075cb857dSQu Wenruo 2831a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 283275cb857dSQu Wenruo if (ret < 0) 283375cb857dSQu Wenruo goto out; 2834e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 283575cb857dSQu Wenruo ret = -EUCLEAN; 283675cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 283775cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 283875cb857dSQu Wenruo goto out; 283975cb857dSQu Wenruo } 284075cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 284175cb857dSQu Wenruo ret = -EUCLEAN; 284275cb857dSQu Wenruo btrfs_err(fs_info, 284375cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 284475cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 284575cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 284675cb857dSQu Wenruo goto out; 284775cb857dSQu Wenruo } 284875cb857dSQu Wenruo out: 284975cb857dSQu Wenruo if (ret < 0) 285075cb857dSQu Wenruo btrfs_err(fs_info, 285175cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 285275cb857dSQu Wenruo return ret; 285375cb857dSQu Wenruo } 285475cb857dSQu Wenruo 2855bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2856bd676446SJosef Bacik { 2857bd676446SJosef Bacik int ret = 0; 2858bd676446SJosef Bacik 2859bd676446SJosef Bacik root->node = read_tree_block(root->fs_info, bytenr, 2860bd676446SJosef Bacik root->root_key.objectid, gen, level, NULL); 2861bd676446SJosef Bacik if (IS_ERR(root->node)) { 2862bd676446SJosef Bacik ret = PTR_ERR(root->node); 2863bd676446SJosef Bacik root->node = NULL; 28644eb150d6SQu Wenruo return ret; 28654eb150d6SQu Wenruo } 28664eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2867bd676446SJosef Bacik free_extent_buffer(root->node); 2868bd676446SJosef Bacik root->node = NULL; 28694eb150d6SQu Wenruo return -EIO; 2870bd676446SJosef Bacik } 2871bd676446SJosef Bacik 2872bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2873bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2874bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2875bd676446SJosef Bacik return ret; 2876bd676446SJosef Bacik } 2877bd676446SJosef Bacik 2878bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2879bd676446SJosef Bacik { 2880bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2881bd676446SJosef Bacik u64 gen, bytenr; 2882bd676446SJosef Bacik int level, ret; 2883bd676446SJosef Bacik 2884bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2885bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2886bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2887bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 28889c54e80dSJosef Bacik if (ret) { 2889bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2890bd676446SJosef Bacik return ret; 2891bd676446SJosef Bacik } 28929c54e80dSJosef Bacik return 0; 28939c54e80dSJosef Bacik } 28949c54e80dSJosef Bacik 28956ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2896b8522a1eSNikolay Borisov { 28976ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2898b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2899b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2900b8522a1eSNikolay Borisov bool handle_error = false; 2901b8522a1eSNikolay Borisov int ret = 0; 2902b8522a1eSNikolay Borisov int i; 2903b8522a1eSNikolay Borisov 2904b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2905b8522a1eSNikolay Borisov if (handle_error) { 2906b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2907b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2908b8522a1eSNikolay Borisov tree_root->node = NULL; 2909b8522a1eSNikolay Borisov 2910b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2911b8522a1eSNikolay Borisov break; 2912b8522a1eSNikolay Borisov 2913b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2914b8522a1eSNikolay Borisov 2915b8522a1eSNikolay Borisov /* 2916b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2917b8522a1eSNikolay Borisov * valid 2918b8522a1eSNikolay Borisov */ 2919b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2920b8522a1eSNikolay Borisov 2921b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2922b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2923b8522a1eSNikolay Borisov 2924b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 29256ef108ddSNikolay Borisov backup_index = ret; 2926b8522a1eSNikolay Borisov if (ret < 0) 2927b8522a1eSNikolay Borisov return ret; 2928b8522a1eSNikolay Borisov } 2929b8522a1eSNikolay Borisov 2930bd676446SJosef Bacik ret = load_important_roots(fs_info); 2931bd676446SJosef Bacik if (ret) { 2932217f5004SNikolay Borisov handle_error = true; 2933b8522a1eSNikolay Borisov continue; 2934b8522a1eSNikolay Borisov } 2935b8522a1eSNikolay Borisov 2936336a0d8dSNikolay Borisov /* 2937336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2938336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2939336a0d8dSNikolay Borisov */ 2940453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2941b8522a1eSNikolay Borisov if (ret < 0) { 2942b8522a1eSNikolay Borisov handle_error = true; 2943b8522a1eSNikolay Borisov continue; 2944b8522a1eSNikolay Borisov } 2945b8522a1eSNikolay Borisov 29466b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2947b8522a1eSNikolay Borisov 2948b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2949b8522a1eSNikolay Borisov if (ret < 0) { 2950b8522a1eSNikolay Borisov handle_error = true; 2951b8522a1eSNikolay Borisov continue; 2952b8522a1eSNikolay Borisov } 2953b8522a1eSNikolay Borisov 2954b8522a1eSNikolay Borisov /* All successful */ 2955bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2956bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2957d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 29586ef108ddSNikolay Borisov 29596ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 29606ef108ddSNikolay Borisov if (backup_index < 0) { 29616ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 29626ef108ddSNikolay Borisov } else { 29636ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 29646ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 29656ef108ddSNikolay Borisov } 2966b8522a1eSNikolay Borisov break; 2967b8522a1eSNikolay Borisov } 2968b8522a1eSNikolay Borisov 2969b8522a1eSNikolay Borisov return ret; 2970b8522a1eSNikolay Borisov } 2971b8522a1eSNikolay Borisov 29728260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2973eb60ceacSChris Mason { 2974fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 297501cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 29768fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2977facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 297824bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2979eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 298011833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2981eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2982a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2983fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 298424bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 29854cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2986ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2987f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 298847ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 298940ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2990afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2991c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2992f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2993abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2994d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2995f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 29967585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2997573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 29980bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 29995f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 3000de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 300119c00ddcSChris Mason 3002e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 30035a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 30048b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 30055f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 30063e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 30073e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 30083e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 30093e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 30103e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 30113e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 30123e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 30133e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 3014e1489b4fSIoannis Angelakopoulos 30150b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 30166324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 3017f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 301847ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 301918bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 3020afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 3021bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 3022bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 30233fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 30243fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 3025bd647ce3SJosef Bacik #endif 3026c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 302766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 302866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 302966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 303066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 303166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 303266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 303366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 3034ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 3035ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 3036ba2c4d4eSJosef Bacik 3037771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 30384cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 3039034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 3040fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 3041abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 304295ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 30439ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 30444cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 3045a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 3046f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 30478b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 3048f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 3049fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 3050c8b97818SChris Mason 3051b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 3052b34b086cSChris Mason num_online_cpus() + 2, 8); 30530afbaf8cSChris Mason 3054199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 3055199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 305669fe2d75SJosef Bacik 3057638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 305821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 305921adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 306021adbd5cSStefan Behrens #endif 3061779a65a4SEric Sandeen btrfs_init_balance(fs_info); 306257056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 3063a2de733cSArne Jansen 306416b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 306508dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 30660f9dd46cSJosef Bacik 3067fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 3068fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 306939279cc3SChris Mason 30705a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 3071e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 3072925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 3073a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 3074a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 30751bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 30769e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 3077c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 307876dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 3079803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 3080fa9c0d79SChris Mason 3081ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 3082f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 3083b0643e59SDennis Zhou btrfs_discard_init(fs_info); 3084416ac51dSArne Jansen 3085fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 3086fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 3087fa9c0d79SChris Mason 3088e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 3089f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 3090bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 30914854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 3092034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 30933768f368SChris Mason 3094da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 3095da17066cSJeff Mahoney fs_info->nodesize = 4096; 3096da17066cSJeff Mahoney fs_info->sectorsize = 4096; 3097ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 3098da17066cSJeff Mahoney fs_info->stripesize = 4096; 3099da17066cSJeff Mahoney 3100f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 3101f7b12a62SNaohiro Aota 3102eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 3103eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 3104eede2bf3SOmar Sandoval 310518bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 310618bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 31078260edbaSJosef Bacik } 31088260edbaSJosef Bacik 31098260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 31108260edbaSJosef Bacik { 31118260edbaSJosef Bacik int ret; 31128260edbaSJosef Bacik 31138260edbaSJosef Bacik fs_info->sb = sb; 31148260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 31158260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 31169e967495SFilipe Manana 31175deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 3118ae18c37aSJosef Bacik if (ret) 3119ae18c37aSJosef Bacik return ret; 3120ae18c37aSJosef Bacik 3121ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 3122ae18c37aSJosef Bacik if (ret) 3123c75e8394SJosef Bacik return ret; 3124ae18c37aSJosef Bacik 3125ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 3126ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 3127ae18c37aSJosef Bacik 3128ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 3129ae18c37aSJosef Bacik if (ret) 3130c75e8394SJosef Bacik return ret; 3131ae18c37aSJosef Bacik 3132ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 3133ae18c37aSJosef Bacik GFP_KERNEL); 3134ae18c37aSJosef Bacik if (ret) 3135c75e8394SJosef Bacik return ret; 3136ae18c37aSJosef Bacik 3137ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 3138ae18c37aSJosef Bacik GFP_KERNEL); 3139c75e8394SJosef Bacik if (!fs_info->delayed_root) 3140c75e8394SJosef Bacik return -ENOMEM; 3141ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 3142ae18c37aSJosef Bacik 3143a0a1db70SFilipe Manana if (sb_rdonly(sb)) 3144a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 3145a0a1db70SFilipe Manana 3146c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 3147ae18c37aSJosef Bacik } 3148ae18c37aSJosef Bacik 314997f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 315097f4dd09SNikolay Borisov { 31510d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 315297f4dd09SNikolay Borisov int ret; 315397f4dd09SNikolay Borisov 315497f4dd09SNikolay Borisov /* 315597f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 315697f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 315797f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 315897f4dd09SNikolay Borisov */ 315997f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 316097f4dd09SNikolay Borisov if (ret < 0) { 3161c94bec2cSJosef Bacik if (ret != -EINTR) 3162c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 3163c94bec2cSJosef Bacik ret); 316497f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 316597f4dd09SNikolay Borisov return ret; 316697f4dd09SNikolay Borisov } 316797f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 316897f4dd09SNikolay Borisov } 316997f4dd09SNikolay Borisov 317097f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 317197f4dd09SNikolay Borisov { 317297f4dd09SNikolay Borisov struct task_struct *task; 317397f4dd09SNikolay Borisov 317497f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 317597f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 317697f4dd09SNikolay Borisov if (IS_ERR(task)) { 317797f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 317897f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 317997f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 318097f4dd09SNikolay Borisov return PTR_ERR(task); 318197f4dd09SNikolay Borisov } 318297f4dd09SNikolay Borisov 318397f4dd09SNikolay Borisov return 0; 318497f4dd09SNikolay Borisov } 318597f4dd09SNikolay Borisov 318644c0ca21SBoris Burkov /* 31878cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 31888cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 31898cd29088SBoris Burkov * code paths. 31908cd29088SBoris Burkov */ 31918cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 31928cd29088SBoris Burkov { 31938cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 31948b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 31958cd29088SBoris Burkov } 31968cd29088SBoris Burkov 31978cd29088SBoris Burkov /* 319844c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 319944c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 320044c0ca21SBoris Burkov */ 320144c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 320244c0ca21SBoris Burkov { 320344c0ca21SBoris Burkov int ret; 320494846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 32058b228324SBoris Burkov bool clear_free_space_tree = false; 32068b228324SBoris Burkov 32078b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 32088b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 32098b228324SBoris Burkov clear_free_space_tree = true; 32108b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 32118b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 32128b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 32138b228324SBoris Burkov clear_free_space_tree = true; 32148b228324SBoris Burkov } 32158b228324SBoris Burkov 32168b228324SBoris Burkov if (clear_free_space_tree) { 32178b228324SBoris Burkov btrfs_info(fs_info, "clearing free space tree"); 32188b228324SBoris Burkov ret = btrfs_clear_free_space_tree(fs_info); 32198b228324SBoris Burkov if (ret) { 32208b228324SBoris Burkov btrfs_warn(fs_info, 32218b228324SBoris Burkov "failed to clear free space tree: %d", ret); 32228b228324SBoris Burkov goto out; 32238b228324SBoris Burkov } 32248b228324SBoris Burkov } 322544c0ca21SBoris Burkov 32268d488a8cSFilipe Manana /* 32278d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 32288d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 3229fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 32308d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 32318d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 32328d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 32338d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 32348d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 32358d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 32368d488a8cSFilipe Manana */ 32378d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 32388d488a8cSFilipe Manana if (ret) 32398d488a8cSFilipe Manana goto out; 32408d488a8cSFilipe Manana 324144c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 324244c0ca21SBoris Burkov if (ret) 324344c0ca21SBoris Burkov goto out; 324444c0ca21SBoris Burkov 32458f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 32468f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 32478f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 32488f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 32498f1c21d7SBoris Burkov goto out; 32508f1c21d7SBoris Burkov } 32518f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 32528f1c21d7SBoris Burkov 325344c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 32547eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 325544c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 325644c0ca21SBoris Burkov if (ret < 0) { 325744c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 325844c0ca21SBoris Burkov goto out; 325944c0ca21SBoris Burkov } 326044c0ca21SBoris Burkov 32615011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 32625011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 32635011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 32645011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 32655011139aSBoris Burkov if (ret) { 32665011139aSBoris Burkov btrfs_warn(fs_info, 32675011139aSBoris Burkov "failed to create free space tree: %d", ret); 32685011139aSBoris Burkov goto out; 32695011139aSBoris Burkov } 32705011139aSBoris Burkov } 32715011139aSBoris Burkov 327294846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 327394846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 327494846229SBoris Burkov if (ret) 327594846229SBoris Burkov goto out; 327694846229SBoris Burkov } 327794846229SBoris Burkov 327844c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 327944c0ca21SBoris Burkov if (ret) 328044c0ca21SBoris Burkov goto out; 328144c0ca21SBoris Burkov 328244c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 328344c0ca21SBoris Burkov if (ret) { 328444c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 328544c0ca21SBoris Burkov goto out; 328644c0ca21SBoris Burkov } 328744c0ca21SBoris Burkov 328844c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 328944c0ca21SBoris Burkov 329044c0ca21SBoris Burkov if (!fs_info->uuid_root) { 329144c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 329244c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 329344c0ca21SBoris Burkov if (ret) { 329444c0ca21SBoris Burkov btrfs_warn(fs_info, 329544c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 329644c0ca21SBoris Burkov goto out; 329744c0ca21SBoris Burkov } 329844c0ca21SBoris Burkov } 329944c0ca21SBoris Burkov 330044c0ca21SBoris Burkov out: 330144c0ca21SBoris Burkov return ret; 330244c0ca21SBoris Burkov } 330344c0ca21SBoris Burkov 3304d7f67ac9SQu Wenruo /* 3305d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3306d7f67ac9SQu Wenruo * 3307d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3308d7f67ac9SQu Wenruo * feature dependencies). 3309d7f67ac9SQu Wenruo * 3310d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3311d7f67ac9SQu Wenruo * btrfs_validate_super(). 3312d7f67ac9SQu Wenruo * 3313d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3314d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3315d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3316d7f67ac9SQu Wenruo */ 3317d7f67ac9SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, struct super_block *sb) 3318d7f67ac9SQu Wenruo { 3319d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3320d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3321d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3322d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3323d7f67ac9SQu Wenruo 3324d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3325d7f67ac9SQu Wenruo btrfs_err(fs_info, 3326d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3327d7f67ac9SQu Wenruo incompat); 3328d7f67ac9SQu Wenruo return -EINVAL; 3329d7f67ac9SQu Wenruo } 3330d7f67ac9SQu Wenruo 3331d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3332d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3333d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3334d7f67ac9SQu Wenruo btrfs_err(fs_info, 3335d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3336d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3337d7f67ac9SQu Wenruo return -EINVAL; 3338d7f67ac9SQu Wenruo } 3339d7f67ac9SQu Wenruo 3340d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3341d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3342d7f67ac9SQu Wenruo 3343d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3344d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3345d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3346d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3347d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3348d7f67ac9SQu Wenruo 3349d7f67ac9SQu Wenruo /* 3350d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3351d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3352d7f67ac9SQu Wenruo */ 3353d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3354d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3355d7f67ac9SQu Wenruo 3356d7f67ac9SQu Wenruo if (compat_ro_unsupp && !sb_rdonly(sb)) { 3357d7f67ac9SQu Wenruo btrfs_err(fs_info, 3358d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3359d7f67ac9SQu Wenruo compat_ro); 3360d7f67ac9SQu Wenruo return -EINVAL; 3361d7f67ac9SQu Wenruo } 3362d7f67ac9SQu Wenruo 3363d7f67ac9SQu Wenruo /* 3364d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3365d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3366d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3367d7f67ac9SQu Wenruo */ 3368d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3369d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3370d7f67ac9SQu Wenruo btrfs_err(fs_info, 3371d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3372d7f67ac9SQu Wenruo compat_ro); 3373d7f67ac9SQu Wenruo return -EINVAL; 3374d7f67ac9SQu Wenruo } 3375d7f67ac9SQu Wenruo 3376d7f67ac9SQu Wenruo /* 3377d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3378d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3379d7f67ac9SQu Wenruo */ 3380d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3381d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3382d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3383d7f67ac9SQu Wenruo btrfs_err(fs_info, 3384d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3385d7f67ac9SQu Wenruo return -EINVAL; 3386d7f67ac9SQu Wenruo } 3387d7f67ac9SQu Wenruo 3388d7f67ac9SQu Wenruo /* 3389d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3390d7f67ac9SQu Wenruo * 3391d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3392d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3393d7f67ac9SQu Wenruo * going to be deprecated anyway. 3394d7f67ac9SQu Wenruo */ 3395d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3396d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3397d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3398d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3399d7f67ac9SQu Wenruo return -EINVAL; 3400d7f67ac9SQu Wenruo } 3401d7f67ac9SQu Wenruo 3402d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3403d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3404d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3405d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3406d7f67ac9SQu Wenruo 3407d7f67ac9SQu Wenruo return 0; 3408d7f67ac9SQu Wenruo } 3409d7f67ac9SQu Wenruo 3410ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3411ae18c37aSJosef Bacik char *options) 3412ae18c37aSJosef Bacik { 3413ae18c37aSJosef Bacik u32 sectorsize; 3414ae18c37aSJosef Bacik u32 nodesize; 3415ae18c37aSJosef Bacik u32 stripesize; 3416ae18c37aSJosef Bacik u64 generation; 3417ae18c37aSJosef Bacik u64 features; 3418ae18c37aSJosef Bacik u16 csum_type; 3419ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3420ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3421ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3422ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3423ae18c37aSJosef Bacik int ret; 3424ae18c37aSJosef Bacik int err = -EINVAL; 3425ae18c37aSJosef Bacik int level; 3426ae18c37aSJosef Bacik 34278260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 342853b381b3SDavid Woodhouse if (ret) { 342983c8266aSDavid Sterba err = ret; 3430ae18c37aSJosef Bacik goto fail; 343153b381b3SDavid Woodhouse } 343253b381b3SDavid Woodhouse 3433ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3434ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3435ae18c37aSJosef Bacik GFP_KERNEL); 3436ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3437ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3438ae18c37aSJosef Bacik GFP_KERNEL); 3439ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3440ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 3441ae18c37aSJosef Bacik err = -ENOMEM; 3442c75e8394SJosef Bacik goto fail; 3443ae18c37aSJosef Bacik } 3444ae18c37aSJosef Bacik 3445ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 3446ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 3447ae18c37aSJosef Bacik err = -ENOMEM; 3448c75e8394SJosef Bacik goto fail; 3449ae18c37aSJosef Bacik } 3450ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 3451ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 3452ae18c37aSJosef Bacik 3453d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 34541104a885SDavid Sterba 34551104a885SDavid Sterba /* 34561104a885SDavid Sterba * Read super block and check the signature bytes only 34571104a885SDavid Sterba */ 3458d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 34598f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 34608f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 346116cdcec7SMiao Xie goto fail_alloc; 346220b45077SDave Young } 346339279cc3SChris Mason 34641104a885SDavid Sterba /* 3465260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 34668dc3f22cSJohannes Thumshirn * corrupted, we'll find out 34678dc3f22cSJohannes Thumshirn */ 34688f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 346951bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 34708dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 347151bce6c9SJohannes Thumshirn csum_type); 34728dc3f22cSJohannes Thumshirn err = -EINVAL; 34738f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 34748dc3f22cSJohannes Thumshirn goto fail_alloc; 34758dc3f22cSJohannes Thumshirn } 34768dc3f22cSJohannes Thumshirn 347783c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 347883c68bbcSSu Yue 34796d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 34806d97c6e3SJohannes Thumshirn if (ret) { 34816d97c6e3SJohannes Thumshirn err = ret; 34828f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 34836d97c6e3SJohannes Thumshirn goto fail_alloc; 34846d97c6e3SJohannes Thumshirn } 34856d97c6e3SJohannes Thumshirn 34868dc3f22cSJohannes Thumshirn /* 34871104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 34881104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 34891104a885SDavid Sterba */ 34903d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 349105135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 34921104a885SDavid Sterba err = -EINVAL; 34938f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3494141386e1SJosef Bacik goto fail_alloc; 34951104a885SDavid Sterba } 34961104a885SDavid Sterba 34971104a885SDavid Sterba /* 34981104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 34991104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 35001104a885SDavid Sterba * the whole block of INFO_SIZE 35011104a885SDavid Sterba */ 35028f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 35038f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35045f39d397SChris Mason 3505fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3506fbc6feaeSNikolay Borisov 35070b86a832SChris Mason 3508fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3509fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3510fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3511fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3512fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3513fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3514fbc6feaeSNikolay Borisov } 3515fbc6feaeSNikolay Borisov 3516fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3517fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3518de37aa51SNikolay Borisov 3519069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 35201104a885SDavid Sterba if (ret) { 352105135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 35221104a885SDavid Sterba err = -EINVAL; 3523141386e1SJosef Bacik goto fail_alloc; 35241104a885SDavid Sterba } 35251104a885SDavid Sterba 35260f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3527141386e1SJosef Bacik goto fail_alloc; 35280f7d52f4SChris Mason 3529acce952bSliubo /* check FS state, whether FS is broken. */ 353087533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 353187533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3532acce952bSliubo 353375e7cb7fSLiu Bo /* 353475e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 353575e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 353675e7cb7fSLiu Bo */ 353775e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 353875e7cb7fSLiu Bo 35396f93e834SAnand Jain 35406f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 35416f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 35426f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 35436f93e834SAnand Jain stripesize = sectorsize; 35446f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 35456f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 35466f93e834SAnand Jain 35476f93e834SAnand Jain fs_info->nodesize = nodesize; 35486f93e834SAnand Jain fs_info->sectorsize = sectorsize; 35496f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 35506f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 35516f93e834SAnand Jain fs_info->stripesize = stripesize; 35526f93e834SAnand Jain 35532ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 35542b82032cSYan Zheng if (ret) { 35552b82032cSYan Zheng err = ret; 3556141386e1SJosef Bacik goto fail_alloc; 35572b82032cSYan Zheng } 3558dfe25020SChris Mason 3559d7f67ac9SQu Wenruo ret = btrfs_check_features(fs_info, sb); 3560d7f67ac9SQu Wenruo if (ret < 0) { 3561d7f67ac9SQu Wenruo err = ret; 3562141386e1SJosef Bacik goto fail_alloc; 3563f2b636e8SJosef Bacik } 3564f2b636e8SJosef Bacik 35658481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 35668481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 35678481dd80SQu Wenruo 35689f73f1aeSQu Wenruo /* 35699f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 35709f73f1aeSQu Wenruo * going to be deprecated. 35719f73f1aeSQu Wenruo * 35729f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 35739f73f1aeSQu Wenruo */ 35749f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 35759f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 35769f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 35779f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 35789f73f1aeSQu Wenruo 357995ea0486SQu Wenruo btrfs_warn(fs_info, 358095ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 35810bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 35828481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 35838481dd80SQu Wenruo if (!subpage_info) 35848481dd80SQu Wenruo goto fail_alloc; 35858481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 35868481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3587c8050b3bSQu Wenruo } 35880bb3eb3eSQu Wenruo 3589d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 35902a458198SEric Sandeen if (ret) { 35912a458198SEric Sandeen err = ret; 35920dc3b84aSJosef Bacik goto fail_sb_buffer; 35930dc3b84aSJosef Bacik } 35944543df7eSChris Mason 35959e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 35969e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 35974575c9ccSChris Mason 3598a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3599a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3600de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3601db94535dSChris Mason 3602925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 36036bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3604925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 360584eed90fSChris Mason if (ret) { 360605135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 36075d4f98a2SYan Zheng goto fail_sb_buffer; 360884eed90fSChris Mason } 36090b86a832SChris Mason 361084234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3611581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3612bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3613bd676446SJosef Bacik generation, level); 3614bd676446SJosef Bacik if (ret) { 361505135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3616af31f5e5SChris Mason goto fail_tree_roots; 361783121942SDavid Woodhouse } 36180b86a832SChris Mason 3619e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3620c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3621c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3622e17cade2SChris Mason 36235b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 36242b82032cSYan Zheng if (ret) { 362505135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3626af31f5e5SChris Mason goto fail_tree_roots; 36272b82032cSYan Zheng } 36280b86a832SChris Mason 36298dabb742SStefan Behrens /* 3630bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3631bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3632bacce86aSAnand Jain * target is required. So free devices that are not part of this 36331a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3634bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 36358dabb742SStefan Behrens */ 3636bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3637d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 363805135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3639a6b0d5c8SChris Mason goto fail_tree_roots; 3640a6b0d5c8SChris Mason } 3641a6b0d5c8SChris Mason 3642b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 36434bbcaa64SEric Sandeen if (ret) 3644b8522a1eSNikolay Borisov goto fail_tree_roots; 36458929ecfaSYan, Zheng 364675ec1db8SJosef Bacik /* 364773651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 364873651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 364973651042SNaohiro Aota * zoned incompat feature flag set. 365073651042SNaohiro Aota */ 365173651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 365273651042SNaohiro Aota if (ret) { 365373651042SNaohiro Aota btrfs_err(fs_info, 365473651042SNaohiro Aota "zoned: failed to read device zone info: %d", 365573651042SNaohiro Aota ret); 365673651042SNaohiro Aota goto fail_block_groups; 365773651042SNaohiro Aota } 365873651042SNaohiro Aota 365973651042SNaohiro Aota /* 366075ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 366175ec1db8SJosef Bacik * check the generation here so we can set the 366275ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 366375ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 366475ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 366575ec1db8SJosef Bacik * even though it was perfectly fine. 366675ec1db8SJosef Bacik */ 366775ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 366875ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 366975ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 367075ec1db8SJosef Bacik 3671cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3672cf90d884SQu Wenruo if (ret) { 3673cf90d884SQu Wenruo btrfs_err(fs_info, 3674cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3675cf90d884SQu Wenruo ret); 3676cf90d884SQu Wenruo goto fail_block_groups; 3677cf90d884SQu Wenruo } 367868310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 367968310a5eSIlya Dryomov if (ret) { 368005135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 368168310a5eSIlya Dryomov goto fail_block_groups; 368268310a5eSIlya Dryomov } 368368310a5eSIlya Dryomov 3684733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3685733f4fbbSStefan Behrens if (ret) { 368605135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3687733f4fbbSStefan Behrens goto fail_block_groups; 3688733f4fbbSStefan Behrens } 3689733f4fbbSStefan Behrens 36908dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 36918dabb742SStefan Behrens if (ret) { 369205135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 36938dabb742SStefan Behrens goto fail_block_groups; 36948dabb742SStefan Behrens } 36958dabb742SStefan Behrens 3696b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3697b70f5097SNaohiro Aota if (ret) { 3698b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3699b70f5097SNaohiro Aota ret); 3700b70f5097SNaohiro Aota goto fail_block_groups; 3701b70f5097SNaohiro Aota } 3702b70f5097SNaohiro Aota 3703c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3704b7c35e81SAnand Jain if (ret) { 370505135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 370605135f59SDavid Sterba ret); 3707b7c35e81SAnand Jain goto fail_block_groups; 3708b7c35e81SAnand Jain } 3709b7c35e81SAnand Jain 371096f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 37115ac1d209SJeff Mahoney if (ret) { 371205135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3713b7c35e81SAnand Jain goto fail_fsdev_sysfs; 37145ac1d209SJeff Mahoney } 37155ac1d209SJeff Mahoney 3716c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3717c59021f8Sliubo if (ret) { 371805135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 37192365dd3cSAnand Jain goto fail_sysfs; 3720c59021f8Sliubo } 3721c59021f8Sliubo 37225b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 37231b1d1f66SJosef Bacik if (ret) { 372405135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 37252365dd3cSAnand Jain goto fail_sysfs; 37261b1d1f66SJosef Bacik } 37274330e183SQu Wenruo 372816beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 372916beac87SNaohiro Aota 37305c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 37315c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 373205135f59SDavid Sterba btrfs_warn(fs_info, 373352042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 37342365dd3cSAnand Jain goto fail_sysfs; 3735292fd7fcSStefan Behrens } 37369078a3e1SChris Mason 373733c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3738a74a4b97SChris Mason "btrfs-cleaner"); 373957506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 37402365dd3cSAnand Jain goto fail_sysfs; 3741a74a4b97SChris Mason 3742a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3743a74a4b97SChris Mason tree_root, 3744a74a4b97SChris Mason "btrfs-transaction"); 374557506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 37463f157a2fSChris Mason goto fail_cleaner; 3747a74a4b97SChris Mason 3748583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3749c289811cSChris Mason !fs_info->fs_devices->rotating) { 3750583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3751c289811cSChris Mason } 3752c289811cSChris Mason 3753572d9ab7SDavid Sterba /* 375463a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 375563a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 375663a7cb13SDavid Sterba * nodiscard for the same mount. 375763a7cb13SDavid Sterba */ 375863a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 375963a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 376063a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 376163a7cb13SDavid Sterba fs_info->fs_devices->discardable) { 376263a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 376363a7cb13SDavid Sterba "auto enabling async discard"); 376463a7cb13SDavid Sterba btrfs_clear_opt(fs_info->mount_opt, NODISCARD); 376563a7cb13SDavid Sterba } 376663a7cb13SDavid Sterba 376721adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 37680b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 37692ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 37700b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3771cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 377221adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 377321adbd5cSStefan Behrens if (ret) 377405135f59SDavid Sterba btrfs_warn(fs_info, 377505135f59SDavid Sterba "failed to initialize integrity check module: %d", 377605135f59SDavid Sterba ret); 377721adbd5cSStefan Behrens } 377821adbd5cSStefan Behrens #endif 3779bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3780bcef60f2SArne Jansen if (ret) 3781bcef60f2SArne Jansen goto fail_trans_kthread; 378221adbd5cSStefan Behrens 3783fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3784fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3785fd708b81SJosef Bacik 378696da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 378796da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 37880b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3789e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 379063443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 379179787eaaSJeff Mahoney if (ret) { 379263443bf5SEric Sandeen err = ret; 379328c16cbbSWang Shilong goto fail_qgroup; 3794e556ce2cSYan Zheng } 3795e02119d5SChris Mason } 37961a40e23bSZheng Yan 379756e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 37983140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 37993140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3800f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3801315bf8efSJosef Bacik fs_info->fs_root = NULL; 3802bcef60f2SArne Jansen goto fail_qgroup; 38033140c9a3SDan Carpenter } 3804c289811cSChris Mason 3805bc98a42cSDavid Howells if (sb_rdonly(sb)) 38068cd29088SBoris Burkov goto clear_oneshot; 38072b6ba629SIlya Dryomov 380844c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 38092b6ba629SIlya Dryomov if (ret) { 38106bccf3abSJeff Mahoney close_ctree(fs_info); 38112b6ba629SIlya Dryomov return ret; 3812e3acc2a6SJosef Bacik } 3813b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3814b382a324SJan Schmidt 381544c0ca21SBoris Burkov if (fs_info->uuid_root && 381644c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 381744c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 381805135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 381970f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 382070f80175SStefan Behrens if (ret) { 382105135f59SDavid Sterba btrfs_warn(fs_info, 382205135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 38236bccf3abSJeff Mahoney close_ctree(fs_info); 382470f80175SStefan Behrens return ret; 382570f80175SStefan Behrens } 3826f7a81ea4SStefan Behrens } 382794846229SBoris Burkov 3828afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 382947ab2a6cSJosef Bacik 3830b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3831b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3832b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3833b4be6aefSJosef Bacik 38348cd29088SBoris Burkov clear_oneshot: 38358cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3836ad2b2c80SAl Viro return 0; 383739279cc3SChris Mason 3838bcef60f2SArne Jansen fail_qgroup: 3839bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 38407c2ca468SChris Mason fail_trans_kthread: 38417c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 38422ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3843faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 38443f157a2fSChris Mason fail_cleaner: 3845a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 38467c2ca468SChris Mason 38477c2ca468SChris Mason /* 38487c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 38497c2ca468SChris Mason * kthreads 38507c2ca468SChris Mason */ 38517c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 38527c2ca468SChris Mason 38532365dd3cSAnand Jain fail_sysfs: 38546618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 38552365dd3cSAnand Jain 3856b7c35e81SAnand Jain fail_fsdev_sysfs: 3857b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3858b7c35e81SAnand Jain 38591b1d1f66SJosef Bacik fail_block_groups: 386054067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3861af31f5e5SChris Mason 3862af31f5e5SChris Mason fail_tree_roots: 38639e3aa805SJosef Bacik if (fs_info->data_reloc_root) 38649e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 38654273eaffSAnand Jain free_root_pointers(fs_info, true); 38662b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3867af31f5e5SChris Mason 386839279cc3SChris Mason fail_sb_buffer: 38697abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 38705cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 387116cdcec7SMiao Xie fail_alloc: 3872586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3873586e46e2SIlya Dryomov 38744543df7eSChris Mason iput(fs_info->btree_inode); 38757e662854SQinghuang Feng fail: 3876586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3877ad2b2c80SAl Viro return err; 3878eb60ceacSChris Mason } 3879663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3880eb60ceacSChris Mason 3881314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3882f2984462SChris Mason { 3883314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3884314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3885314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3886314b6dd0SJohannes Thumshirn struct page *page; 3887442a4f63SStefan Behrens 3888314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3889314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3890314b6dd0SJohannes Thumshirn 3891314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3892fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3893314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3894314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3895314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3896314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3897314b6dd0SJohannes Thumshirn SetPageError(page); 3898314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3899314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3900314b6dd0SJohannes Thumshirn } else { 3901314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3902f2984462SChris Mason } 3903314b6dd0SJohannes Thumshirn 3904314b6dd0SJohannes Thumshirn put_page(page); 3905314b6dd0SJohannes Thumshirn unlock_page(page); 3906314b6dd0SJohannes Thumshirn } 3907314b6dd0SJohannes Thumshirn 3908314b6dd0SJohannes Thumshirn bio_put(bio); 3909f2984462SChris Mason } 3910f2984462SChris Mason 39118f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3912a05d3c91SQu Wenruo int copy_num, bool drop_cache) 391329c36d72SAnand Jain { 391429c36d72SAnand Jain struct btrfs_super_block *super; 39158f32380dSJohannes Thumshirn struct page *page; 391612659251SNaohiro Aota u64 bytenr, bytenr_orig; 39178f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 391812659251SNaohiro Aota int ret; 391929c36d72SAnand Jain 392012659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 392112659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 392212659251SNaohiro Aota if (ret == -ENOENT) 392312659251SNaohiro Aota return ERR_PTR(-EINVAL); 392412659251SNaohiro Aota else if (ret) 392512659251SNaohiro Aota return ERR_PTR(ret); 392612659251SNaohiro Aota 3927cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 39288f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 392929c36d72SAnand Jain 3930a05d3c91SQu Wenruo if (drop_cache) { 3931a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3932a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3933a05d3c91SQu Wenruo 3934a05d3c91SQu Wenruo /* 3935a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3936a05d3c91SQu Wenruo * always read from the device. 3937a05d3c91SQu Wenruo */ 3938a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3939a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3940a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3941a05d3c91SQu Wenruo } 3942a05d3c91SQu Wenruo 39438f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 39448f32380dSJohannes Thumshirn if (IS_ERR(page)) 39458f32380dSJohannes Thumshirn return ERR_CAST(page); 394629c36d72SAnand Jain 39478f32380dSJohannes Thumshirn super = page_address(page); 394896c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 394996c2e067SAnand Jain btrfs_release_disk_super(super); 395096c2e067SAnand Jain return ERR_PTR(-ENODATA); 395196c2e067SAnand Jain } 395296c2e067SAnand Jain 395312659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 39548f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 39558f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 395629c36d72SAnand Jain } 395729c36d72SAnand Jain 39588f32380dSJohannes Thumshirn return super; 395929c36d72SAnand Jain } 396029c36d72SAnand Jain 396129c36d72SAnand Jain 39628f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3963a512bbf8SYan Zheng { 39648f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3965a512bbf8SYan Zheng int i; 3966a512bbf8SYan Zheng u64 transid = 0; 3967a512bbf8SYan Zheng 3968a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3969a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3970a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3971a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3972a512bbf8SYan Zheng */ 3973a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3974a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 39758f32380dSJohannes Thumshirn if (IS_ERR(super)) 3976a512bbf8SYan Zheng continue; 3977a512bbf8SYan Zheng 3978a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 39798f32380dSJohannes Thumshirn if (latest) 39808f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 39818f32380dSJohannes Thumshirn 39828f32380dSJohannes Thumshirn latest = super; 3983a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3984a512bbf8SYan Zheng } 3985a512bbf8SYan Zheng } 398692fc03fbSAnand Jain 39878f32380dSJohannes Thumshirn return super; 3988a512bbf8SYan Zheng } 3989a512bbf8SYan Zheng 39904eedeb75SHisashi Hifumi /* 3991abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3992314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 39934eedeb75SHisashi Hifumi * 3994abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3995abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3996abbb3b8eSDavid Sterba * same for write and wait phases. 39974eedeb75SHisashi Hifumi * 3998314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 39994eedeb75SHisashi Hifumi */ 4000a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 4001abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 4002a512bbf8SYan Zheng { 4003d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 4004314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 4005d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 4006a512bbf8SYan Zheng int i; 4007a512bbf8SYan Zheng int errors = 0; 400812659251SNaohiro Aota int ret; 400912659251SNaohiro Aota u64 bytenr, bytenr_orig; 4010a512bbf8SYan Zheng 4011a512bbf8SYan Zheng if (max_mirrors == 0) 4012a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4013a512bbf8SYan Zheng 4014d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 4015d5178578SJohannes Thumshirn 4016a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 4017314b6dd0SJohannes Thumshirn struct page *page; 4018314b6dd0SJohannes Thumshirn struct bio *bio; 4019314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 4020314b6dd0SJohannes Thumshirn 402112659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 402212659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 402312659251SNaohiro Aota if (ret == -ENOENT) { 402412659251SNaohiro Aota continue; 402512659251SNaohiro Aota } else if (ret < 0) { 402612659251SNaohiro Aota btrfs_err(device->fs_info, 402712659251SNaohiro Aota "couldn't get super block location for mirror %d", 402812659251SNaohiro Aota i); 402912659251SNaohiro Aota errors++; 403012659251SNaohiro Aota continue; 403112659251SNaohiro Aota } 4032935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4033935e5cc9SMiao Xie device->commit_total_bytes) 4034a512bbf8SYan Zheng break; 4035a512bbf8SYan Zheng 403612659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 4037a512bbf8SYan Zheng 4038fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 4039fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 4040fd08001fSEric Biggers sb->csum); 4041a512bbf8SYan Zheng 4042314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 4043314b6dd0SJohannes Thumshirn GFP_NOFS); 4044314b6dd0SJohannes Thumshirn if (!page) { 4045fb456252SJeff Mahoney btrfs_err(device->fs_info, 4046314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 4047f14d104dSDavid Sterba bytenr); 4048634554dcSJosef Bacik errors++; 4049634554dcSJosef Bacik continue; 4050634554dcSJosef Bacik } 4051634554dcSJosef Bacik 4052314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 4053314b6dd0SJohannes Thumshirn get_page(page); 4054a512bbf8SYan Zheng 4055314b6dd0SJohannes Thumshirn disk_super = page_address(page); 4056314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 4057a512bbf8SYan Zheng 4058387125fcSChris Mason /* 4059314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 4060314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 4061314b6dd0SJohannes Thumshirn * checking. 4062387125fcSChris Mason */ 406307888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 406407888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 406507888c66SChristoph Hellwig GFP_NOFS); 4066314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 4067314b6dd0SJohannes Thumshirn bio->bi_private = device; 4068314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 4069314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 4070314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 4071314b6dd0SJohannes Thumshirn 4072314b6dd0SJohannes Thumshirn /* 4073314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 4074314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 4075314b6dd0SJohannes Thumshirn * copies might still contain the older version. 4076314b6dd0SJohannes Thumshirn */ 40771b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 4078314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 4079314b6dd0SJohannes Thumshirn 408058ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 408158ff51f1SChristoph Hellwig submit_bio(bio); 40828376d9e1SNaohiro Aota 40838376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 40848376d9e1SNaohiro Aota errors++; 4085a512bbf8SYan Zheng } 4086a512bbf8SYan Zheng return errors < i ? 0 : -1; 4087a512bbf8SYan Zheng } 4088a512bbf8SYan Zheng 4089387125fcSChris Mason /* 4090abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 4091abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 4092abbb3b8eSDavid Sterba * 4093314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 4094abbb3b8eSDavid Sterba * date. 4095abbb3b8eSDavid Sterba */ 4096abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 4097abbb3b8eSDavid Sterba { 4098abbb3b8eSDavid Sterba int i; 4099abbb3b8eSDavid Sterba int errors = 0; 4100b6a535faSHoward McLauchlan bool primary_failed = false; 410112659251SNaohiro Aota int ret; 4102abbb3b8eSDavid Sterba u64 bytenr; 4103abbb3b8eSDavid Sterba 4104abbb3b8eSDavid Sterba if (max_mirrors == 0) 4105abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4106abbb3b8eSDavid Sterba 4107abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 4108314b6dd0SJohannes Thumshirn struct page *page; 4109314b6dd0SJohannes Thumshirn 411012659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 411112659251SNaohiro Aota if (ret == -ENOENT) { 411212659251SNaohiro Aota break; 411312659251SNaohiro Aota } else if (ret < 0) { 411412659251SNaohiro Aota errors++; 411512659251SNaohiro Aota if (i == 0) 411612659251SNaohiro Aota primary_failed = true; 411712659251SNaohiro Aota continue; 411812659251SNaohiro Aota } 4119abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4120abbb3b8eSDavid Sterba device->commit_total_bytes) 4121abbb3b8eSDavid Sterba break; 4122abbb3b8eSDavid Sterba 4123314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 4124314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 4125314b6dd0SJohannes Thumshirn if (!page) { 4126abbb3b8eSDavid Sterba errors++; 4127b6a535faSHoward McLauchlan if (i == 0) 4128b6a535faSHoward McLauchlan primary_failed = true; 4129abbb3b8eSDavid Sterba continue; 4130abbb3b8eSDavid Sterba } 4131314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 4132314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 4133314b6dd0SJohannes Thumshirn if (PageError(page)) { 4134abbb3b8eSDavid Sterba errors++; 4135b6a535faSHoward McLauchlan if (i == 0) 4136b6a535faSHoward McLauchlan primary_failed = true; 4137b6a535faSHoward McLauchlan } 4138abbb3b8eSDavid Sterba 4139314b6dd0SJohannes Thumshirn /* Drop our reference */ 4140314b6dd0SJohannes Thumshirn put_page(page); 4141abbb3b8eSDavid Sterba 4142314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 4143314b6dd0SJohannes Thumshirn put_page(page); 4144abbb3b8eSDavid Sterba } 4145abbb3b8eSDavid Sterba 4146b6a535faSHoward McLauchlan /* log error, force error return */ 4147b6a535faSHoward McLauchlan if (primary_failed) { 4148b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 4149b6a535faSHoward McLauchlan device->devid); 4150b6a535faSHoward McLauchlan return -1; 4151b6a535faSHoward McLauchlan } 4152b6a535faSHoward McLauchlan 4153abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 4154abbb3b8eSDavid Sterba } 4155abbb3b8eSDavid Sterba 4156abbb3b8eSDavid Sterba /* 4157387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 4158387125fcSChris Mason * for the barrier when it is done 4159387125fcSChris Mason */ 41604246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 4161387125fcSChris Mason { 4162f9e69aa9SChristoph Hellwig bio_uninit(bio); 4163387125fcSChris Mason complete(bio->bi_private); 4164387125fcSChris Mason } 4165387125fcSChris Mason 4166387125fcSChris Mason /* 41674fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 41684fc6441aSAnand Jain * done in the waiting counterpart. 4169387125fcSChris Mason */ 41704fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 4171387125fcSChris Mason { 4172f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 4173387125fcSChris Mason 4174a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 4175a91cf0ffSWang Yugui /* 4176a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 4177a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 4178a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 4179a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 4180a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 4181a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 4182a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 4183a91cf0ffSWang Yugui * non-debug builds. 4184a91cf0ffSWang Yugui */ 418508e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 41864fc6441aSAnand Jain return; 4187a91cf0ffSWang Yugui #endif 4188387125fcSChris Mason 4189f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 4190f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 4191387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 4192387125fcSChris Mason init_completion(&device->flush_wait); 4193387125fcSChris Mason bio->bi_private = &device->flush_wait; 4194387125fcSChris Mason 419558ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 419658ff51f1SChristoph Hellwig submit_bio(bio); 41971c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 41984fc6441aSAnand Jain } 4199387125fcSChris Mason 42004fc6441aSAnand Jain /* 42014fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 42024fc6441aSAnand Jain */ 42038c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 42044fc6441aSAnand Jain { 4205f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 42064fc6441aSAnand Jain 42071c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 420858efbc9fSOmar Sandoval return BLK_STS_OK; 42094fc6441aSAnand Jain 42101c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 42112980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 42124fc6441aSAnand Jain 42138c27cb35SLinus Torvalds return bio->bi_status; 4214387125fcSChris Mason } 4215387125fcSChris Mason 4216d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 4217401b41e5SAnand Jain { 42186528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 4219401b41e5SAnand Jain return -EIO; 4220387125fcSChris Mason return 0; 4221387125fcSChris Mason } 4222387125fcSChris Mason 4223387125fcSChris Mason /* 4224387125fcSChris Mason * send an empty flush down to each device in parallel, 4225387125fcSChris Mason * then wait for them 4226387125fcSChris Mason */ 4227387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 4228387125fcSChris Mason { 4229387125fcSChris Mason struct list_head *head; 4230387125fcSChris Mason struct btrfs_device *dev; 42315af3e8ccSStefan Behrens int errors_wait = 0; 42324e4cbee9SChristoph Hellwig blk_status_t ret; 4233387125fcSChris Mason 42341538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 4235387125fcSChris Mason /* send down all the barriers */ 4236387125fcSChris Mason head = &info->fs_devices->devices; 42371538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4238e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4239f88ba6a2SHidetoshi Seto continue; 4240cea7c8bfSAnand Jain if (!dev->bdev) 4241387125fcSChris Mason continue; 4242e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4243ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4244387125fcSChris Mason continue; 4245387125fcSChris Mason 42464fc6441aSAnand Jain write_dev_flush(dev); 424758efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 4248387125fcSChris Mason } 4249387125fcSChris Mason 4250387125fcSChris Mason /* wait for all the barriers */ 42511538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4252e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4253f88ba6a2SHidetoshi Seto continue; 4254387125fcSChris Mason if (!dev->bdev) { 42555af3e8ccSStefan Behrens errors_wait++; 4256387125fcSChris Mason continue; 4257387125fcSChris Mason } 4258e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4259ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4260387125fcSChris Mason continue; 4261387125fcSChris Mason 42624fc6441aSAnand Jain ret = wait_dev_flush(dev); 4263401b41e5SAnand Jain if (ret) { 4264401b41e5SAnand Jain dev->last_flush_error = ret; 426566b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 426666b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 42675af3e8ccSStefan Behrens errors_wait++; 4268387125fcSChris Mason } 4269401b41e5SAnand Jain } 4270401b41e5SAnand Jain 4271cea7c8bfSAnand Jain if (errors_wait) { 4272401b41e5SAnand Jain /* 4273401b41e5SAnand Jain * At some point we need the status of all disks 4274401b41e5SAnand Jain * to arrive at the volume status. So error checking 4275401b41e5SAnand Jain * is being pushed to a separate loop. 4276401b41e5SAnand Jain */ 4277d10b82feSQu Wenruo return check_barrier_error(info); 4278401b41e5SAnand Jain } 4279387125fcSChris Mason return 0; 4280387125fcSChris Mason } 4281387125fcSChris Mason 4282943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4283943c6e99SZhao Lei { 42848789f4feSZhao Lei int raid_type; 42858789f4feSZhao Lei int min_tolerated = INT_MAX; 4286943c6e99SZhao Lei 42878789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 42888789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 42898c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 42908789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 42918789f4feSZhao Lei tolerated_failures); 4292943c6e99SZhao Lei 42938789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 42948789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 42958789f4feSZhao Lei continue; 429641a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 42978789f4feSZhao Lei continue; 42988c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 42998789f4feSZhao Lei btrfs_raid_array[raid_type]. 43008789f4feSZhao Lei tolerated_failures); 43018789f4feSZhao Lei } 4302943c6e99SZhao Lei 43038789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4304ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 43058789f4feSZhao Lei min_tolerated = 0; 43068789f4feSZhao Lei } 43078789f4feSZhao Lei 43088789f4feSZhao Lei return min_tolerated; 4309943c6e99SZhao Lei } 4310943c6e99SZhao Lei 4311eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4312f2984462SChris Mason { 4313e5e9a520SChris Mason struct list_head *head; 4314f2984462SChris Mason struct btrfs_device *dev; 4315a061fc8dSChris Mason struct btrfs_super_block *sb; 4316f2984462SChris Mason struct btrfs_dev_item *dev_item; 4317f2984462SChris Mason int ret; 4318f2984462SChris Mason int do_barriers; 4319a236aed1SChris Mason int max_errors; 4320a236aed1SChris Mason int total_errors = 0; 4321a061fc8dSChris Mason u64 flags; 4322f2984462SChris Mason 43230b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4324fed3b381SLiu Bo 4325fed3b381SLiu Bo /* 4326fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4327fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4328fed3b381SLiu Bo * been consistent on disk. 4329fed3b381SLiu Bo */ 4330fed3b381SLiu Bo if (max_mirrors == 0) 43310b246afaSJeff Mahoney backup_super_roots(fs_info); 4332f2984462SChris Mason 43330b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4334a061fc8dSChris Mason dev_item = &sb->dev_item; 4335e5e9a520SChris Mason 43360b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 43370b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 43380b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4339387125fcSChris Mason 43405af3e8ccSStefan Behrens if (do_barriers) { 43410b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 43425af3e8ccSStefan Behrens if (ret) { 43435af3e8ccSStefan Behrens mutex_unlock( 43440b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 43450b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 43465af3e8ccSStefan Behrens "errors while submitting device barriers."); 43475af3e8ccSStefan Behrens return ret; 43485af3e8ccSStefan Behrens } 43495af3e8ccSStefan Behrens } 4350387125fcSChris Mason 43511538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4352dfe25020SChris Mason if (!dev->bdev) { 4353dfe25020SChris Mason total_errors++; 4354dfe25020SChris Mason continue; 4355dfe25020SChris Mason } 4356e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4357ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4358dfe25020SChris Mason continue; 4359dfe25020SChris Mason 43602b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4361a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4362a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 43637df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4364935e5cc9SMiao Xie dev->commit_total_bytes); 4365ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4366ce7213c7SMiao Xie dev->commit_bytes_used); 4367a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4368a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4369a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4370a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 43717239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 43727239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4373a512bbf8SYan Zheng 4374a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4375a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4376f2984462SChris Mason 437775cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 437875cb857dSQu Wenruo if (ret < 0) { 437975cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 438075cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 438175cb857dSQu Wenruo "unexpected superblock corruption detected"); 438275cb857dSQu Wenruo return -EUCLEAN; 438375cb857dSQu Wenruo } 438475cb857dSQu Wenruo 4385abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4386a236aed1SChris Mason if (ret) 4387a236aed1SChris Mason total_errors++; 4388f2984462SChris Mason } 4389a236aed1SChris Mason if (total_errors > max_errors) { 43900b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4391d397712bSChris Mason total_errors); 43920b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 439379787eaaSJeff Mahoney 43949d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 43950b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 43960b246afaSJeff Mahoney "%d errors while writing supers", 43970b246afaSJeff Mahoney total_errors); 43989d565ba4SStefan Behrens return -EIO; 4399a236aed1SChris Mason } 4400f2984462SChris Mason 4401a512bbf8SYan Zheng total_errors = 0; 44021538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4403dfe25020SChris Mason if (!dev->bdev) 4404dfe25020SChris Mason continue; 4405e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4406ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4407dfe25020SChris Mason continue; 4408dfe25020SChris Mason 4409abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4410a512bbf8SYan Zheng if (ret) 44111259ab75SChris Mason total_errors++; 4412f2984462SChris Mason } 44130b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4414a236aed1SChris Mason if (total_errors > max_errors) { 44150b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 44160b246afaSJeff Mahoney "%d errors while writing supers", 44170b246afaSJeff Mahoney total_errors); 441879787eaaSJeff Mahoney return -EIO; 4419a236aed1SChris Mason } 4420f2984462SChris Mason return 0; 4421f2984462SChris Mason } 4422f2984462SChris Mason 4423cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4424cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4425cb517eabSMiao Xie struct btrfs_root *root) 44262619ba1fSChris Mason { 44274785e24fSJosef Bacik bool drop_ref = false; 44284785e24fSJosef Bacik 4429fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4430fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4431fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4432fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 44334785e24fSJosef Bacik drop_ref = true; 4434fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 443576dda93cSYan, Zheng 443684961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4437ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 44381c1ea4f7SLiu Bo if (root->reloc_root) { 443900246528SJosef Bacik btrfs_put_root(root->reloc_root); 44401c1ea4f7SLiu Bo root->reloc_root = NULL; 44411c1ea4f7SLiu Bo } 44421c1ea4f7SLiu Bo } 44433321719eSLiu Bo 44444785e24fSJosef Bacik if (drop_ref) 444500246528SJosef Bacik btrfs_put_root(root); 44462619ba1fSChris Mason } 44472619ba1fSChris Mason 4448c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4449c146afadSYan Zheng { 4450fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4451fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4452fc7cbcd4SDavid Sterba int i = 0; 445365d33fd7SQu Wenruo int err = 0; 4454fc7cbcd4SDavid Sterba unsigned int ret = 0; 4455c146afadSYan Zheng 4456c146afadSYan Zheng while (1) { 4457fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4458fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4459fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4460fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4461fc7cbcd4SDavid Sterba if (!ret) { 4462fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4463c146afadSYan Zheng break; 446465d33fd7SQu Wenruo } 4465fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 446666b4ffd1SJosef Bacik 4467fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4468fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4469fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4470fc7cbcd4SDavid Sterba gang[i] = NULL; 447165d33fd7SQu Wenruo continue; 4472c146afadSYan Zheng } 4473fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4474fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4475fc7cbcd4SDavid Sterba } 4476fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4477fc7cbcd4SDavid Sterba 4478fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4479fc7cbcd4SDavid Sterba if (!gang[i]) 4480fc7cbcd4SDavid Sterba continue; 4481fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4482fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4483fc7cbcd4SDavid Sterba if (err) 4484fc7cbcd4SDavid Sterba break; 4485fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4486fc7cbcd4SDavid Sterba } 4487fc7cbcd4SDavid Sterba root_objectid++; 4488c146afadSYan Zheng } 448965d33fd7SQu Wenruo 4490fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4491fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4492fc7cbcd4SDavid Sterba if (gang[i]) 4493fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 449465d33fd7SQu Wenruo } 449565d33fd7SQu Wenruo return err; 4496c146afadSYan Zheng } 4497c146afadSYan Zheng 44986bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4499c146afadSYan Zheng { 45006bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4501c146afadSYan Zheng struct btrfs_trans_handle *trans; 4502c146afadSYan Zheng 45030b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45042ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45050b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 45060b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4507c71bf099SYan, Zheng 4508c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 45090b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45100b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4511c71bf099SYan, Zheng 45127a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 45133612b495STsutomu Itoh if (IS_ERR(trans)) 45143612b495STsutomu Itoh return PTR_ERR(trans); 45153a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4516c146afadSYan Zheng } 4517c146afadSYan Zheng 451836c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 451936c86a9eSQu Wenruo { 452036c86a9eSQu Wenruo struct btrfs_transaction *trans; 452136c86a9eSQu Wenruo struct btrfs_transaction *tmp; 452236c86a9eSQu Wenruo bool found = false; 452336c86a9eSQu Wenruo 452436c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 452536c86a9eSQu Wenruo return; 452636c86a9eSQu Wenruo 452736c86a9eSQu Wenruo /* 452836c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 452936c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 453036c86a9eSQu Wenruo */ 453136c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 453236c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 453336c86a9eSQu Wenruo struct extent_state *cached = NULL; 453436c86a9eSQu Wenruo u64 dirty_bytes = 0; 453536c86a9eSQu Wenruo u64 cur = 0; 453636c86a9eSQu Wenruo u64 found_start; 453736c86a9eSQu Wenruo u64 found_end; 453836c86a9eSQu Wenruo 453936c86a9eSQu Wenruo found = true; 454036c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 454136c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 454236c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 454336c86a9eSQu Wenruo cur = found_end + 1; 454436c86a9eSQu Wenruo } 454536c86a9eSQu Wenruo btrfs_warn(fs_info, 454636c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 454736c86a9eSQu Wenruo trans->transid, dirty_bytes); 454836c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 454936c86a9eSQu Wenruo 455036c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 455136c86a9eSQu Wenruo fs_info->running_transaction = NULL; 455236c86a9eSQu Wenruo list_del_init(&trans->list); 455336c86a9eSQu Wenruo 455436c86a9eSQu Wenruo btrfs_put_transaction(trans); 455536c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 455636c86a9eSQu Wenruo } 455736c86a9eSQu Wenruo ASSERT(!found); 455836c86a9eSQu Wenruo } 455936c86a9eSQu Wenruo 4560b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4561eb60ceacSChris Mason { 4562c146afadSYan Zheng int ret; 4563e089f05cSChris Mason 4564afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 456531e70e52SFilipe Manana 456631e70e52SFilipe Manana /* 45678a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 45688a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 45698a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 45708a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 45718a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 45728a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 45738a1f1e3dSFilipe Manana * return 1. 45748a1f1e3dSFilipe Manana */ 45758a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 45768a1f1e3dSFilipe Manana 45778a1f1e3dSFilipe Manana /* 457831e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 457931e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 458031e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 458131e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 458231e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 458331e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 458431e70e52SFilipe Manana * trying to stop the async reclaim task. 458531e70e52SFilipe Manana */ 458631e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4587d6fd0ae2SOmar Sandoval /* 4588d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4589d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4590d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4591d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4592d6fd0ae2SOmar Sandoval */ 4593d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4594a2135011SChris Mason 45957343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4596d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 45977343dd61SJustin Maggard 4598803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4599803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4600803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4601803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4602803b2f54SStefan Behrens 4603837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4604aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4605837d5b6eSIlya Dryomov 46068dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 46078dabb742SStefan Behrens 4608aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 46094cb5300bSChris Mason 46104cb5300bSChris Mason /* wait for any defraggers to finish */ 46114cb5300bSChris Mason wait_event(fs_info->transaction_wait, 46124cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 46134cb5300bSChris Mason 46144cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 461526176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 46164cb5300bSChris Mason 4617a362bb86SFilipe Manana /* 4618a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4619a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4620a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4621a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4622a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4623a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4624a362bb86SFilipe Manana * no one else to run iputs. 4625a362bb86SFilipe Manana * 4626a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4627a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4628a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4629a362bb86SFilipe Manana * through some other means. 4630a362bb86SFilipe Manana * 4631a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4632a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4633a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4634a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4635a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4636a362bb86SFilipe Manana */ 4637a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4638a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4639a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4640a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4641a362bb86SFilipe Manana 464221c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 464357056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4644576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 464521c7e756SMiao Xie 4646b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4647b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4648b0643e59SDennis Zhou 4649bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4650e44163e1SJeff Mahoney /* 4651d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4652d6fd0ae2SOmar Sandoval * unused block groups. 4653e44163e1SJeff Mahoney */ 46540b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4655e44163e1SJeff Mahoney 4656f0cc2cd7SFilipe Manana /* 4657f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4658f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4659f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4660f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4661f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4662f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4663f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4664f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4665f0cc2cd7SFilipe Manana * This is a very rare case. 4666f0cc2cd7SFilipe Manana */ 4667f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4668f0cc2cd7SFilipe Manana 46696bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4670d397712bSChris Mason if (ret) 467104892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4672c146afadSYan Zheng } 4673ed2ff2cbSChris Mason 467484961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 46752ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4676acce952bSliubo 4677e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4678e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 46798929ecfaSYan, Zheng 4680e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4681afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4682f25784b3SYan Zheng 46835958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 46845958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 46855958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 46865958253cSQu Wenruo } 46875958253cSQu Wenruo 468804892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4689fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4690bcef60f2SArne Jansen 4691963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 469204892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4693963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4694b0c68f8bSChris Mason } 469531153d81SYan Zheng 46965deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 46974297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 46985deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 46994297ff84SJosef Bacik 47006618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4701b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 47025ac1d209SJeff Mahoney 47031a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 47041a4319ccSLiu Bo 4705de348ee0SWang Shilong /* 4706de348ee0SWang Shilong * we must make sure there is not any read request to 4707de348ee0SWang Shilong * submit after we stopping all workers. 4708de348ee0SWang Shilong */ 4709de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 471096192499SJosef Bacik btrfs_stop_all_workers(fs_info); 471196192499SJosef Bacik 47120a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 471336c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 47140a31daa4SFilipe Manana 4715afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 47164273eaffSAnand Jain free_root_pointers(fs_info, true); 47178c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 47189ad6b7bcSChris Mason 47194e19443dSJosef Bacik /* 47204e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 47214e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 47224e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 47234e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 47244e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 47254e19443dSJosef Bacik */ 47264e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 47274e19443dSJosef Bacik 472813e6c37bSJosef Bacik iput(fs_info->btree_inode); 4729d6bfde87SChris Mason 473021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 47310b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 47322ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 473321adbd5cSStefan Behrens #endif 473421adbd5cSStefan Behrens 47350b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 473668c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4737eb60ceacSChris Mason } 4738eb60ceacSChris Mason 4739b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4740b9fab919SChris Mason int atomic) 4741ccd467d6SChris Mason { 47421259ab75SChris Mason int ret; 4743727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 47441259ab75SChris Mason 47450b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 47461259ab75SChris Mason if (!ret) 47471259ab75SChris Mason return ret; 47481259ab75SChris Mason 47491259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4750b9fab919SChris Mason parent_transid, atomic); 4751b9fab919SChris Mason if (ret == -EAGAIN) 4752b9fab919SChris Mason return ret; 47531259ab75SChris Mason return !ret; 47545f39d397SChris Mason } 47556702ed49SChris Mason 47565f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 47575f39d397SChris Mason { 47582f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 47595f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4760b9473439SChris Mason int was_dirty; 4761b4ce94deSChris Mason 476206ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 476306ea65a3SJosef Bacik /* 476406ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 476552042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 476606ea65a3SJosef Bacik * outside of the sanity tests. 476706ea65a3SJosef Bacik */ 4768b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 476906ea65a3SJosef Bacik return; 477006ea65a3SJosef Bacik #endif 477149d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 47720b246afaSJeff Mahoney if (transid != fs_info->generation) 47735d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 47740b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 47750b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4776e2d84521SMiao Xie if (!was_dirty) 4777104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4778e2d84521SMiao Xie buf->len, 47790b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 47801f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 478169fc6cbbSQu Wenruo /* 478269fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 478369fc6cbbSQu Wenruo * but item data not updated. 478469fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 478569fc6cbbSQu Wenruo */ 478669fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4787cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4788a4f78750SDavid Sterba btrfs_print_leaf(buf); 47891f21ef0aSFilipe Manana ASSERT(0); 47901f21ef0aSFilipe Manana } 47911f21ef0aSFilipe Manana #endif 4792eb60ceacSChris Mason } 4793eb60ceacSChris Mason 47942ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4795b53d3f5dSLiu Bo int flush_delayed) 479635b7e476SChris Mason { 4797188de649SChris Mason /* 4798188de649SChris Mason * looks as though older kernels can get into trouble with 4799188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4800188de649SChris Mason */ 4801e2d84521SMiao Xie int ret; 4802d6bfde87SChris Mason 48036933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4804d6bfde87SChris Mason return; 4805d6bfde87SChris Mason 4806b53d3f5dSLiu Bo if (flush_delayed) 48072ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 480816cdcec7SMiao Xie 4809d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4810d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4811d814a491SEthan Lien fs_info->dirty_metadata_batch); 4812e2d84521SMiao Xie if (ret > 0) { 48130b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 481416cdcec7SMiao Xie } 481516cdcec7SMiao Xie } 481616cdcec7SMiao Xie 48172ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 481816cdcec7SMiao Xie { 48192ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 482035b7e476SChris Mason } 4821b53d3f5dSLiu Bo 48222ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4823b53d3f5dSLiu Bo { 48242ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4825d6bfde87SChris Mason } 48266b80053dSChris Mason 48272ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4828acce952bSliubo { 4829fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4830fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4831fe816d0fSNikolay Borisov 48320b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 48332ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 48340b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4835acce952bSliubo 48360b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 48370b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4838acce952bSliubo } 4839acce952bSliubo 4840ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4841ef67963dSJosef Bacik { 4842fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4843fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4844fc7cbcd4SDavid Sterba int ret; 4845ef67963dSJosef Bacik 4846fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4847fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4848fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4849fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4850fc7cbcd4SDavid Sterba int i; 4851ef67963dSJosef Bacik 4852fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4853fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4854fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4855fc7cbcd4SDavid Sterba 4856fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4857fc7cbcd4SDavid Sterba if (!gang[i]) 4858ef67963dSJosef Bacik continue; 4859fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4860fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4861fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4862ef67963dSJosef Bacik } 4863fc7cbcd4SDavid Sterba root_objectid++; 4864fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4865ef67963dSJosef Bacik } 4866fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4867ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4868ef67963dSJosef Bacik } 4869ef67963dSJosef Bacik 4870143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4871acce952bSliubo { 4872acce952bSliubo struct btrfs_ordered_extent *ordered; 4873acce952bSliubo 4874199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4875779880efSJosef Bacik /* 4876779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4877779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4878779880efSJosef Bacik */ 4879199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4880779880efSJosef Bacik root_extent_list) 4881779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4882199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4883199c2a9cSMiao Xie } 4884199c2a9cSMiao Xie 4885199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4886199c2a9cSMiao Xie { 4887199c2a9cSMiao Xie struct btrfs_root *root; 4888199c2a9cSMiao Xie struct list_head splice; 4889199c2a9cSMiao Xie 4890199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4891199c2a9cSMiao Xie 4892199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4893199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4894199c2a9cSMiao Xie while (!list_empty(&splice)) { 4895199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4896199c2a9cSMiao Xie ordered_root); 48971de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 48981de2cfdeSJosef Bacik &fs_info->ordered_roots); 4899199c2a9cSMiao Xie 49002a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4901199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4902199c2a9cSMiao Xie 49032a85d9caSLiu Bo cond_resched(); 49042a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4905199c2a9cSMiao Xie } 4906199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 490774d5d229SJosef Bacik 490874d5d229SJosef Bacik /* 490974d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 491074d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 491174d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 491274d5d229SJosef Bacik * actually get run and error out properly. 491374d5d229SJosef Bacik */ 491474d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4915acce952bSliubo } 4916acce952bSliubo 491735a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 49182ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4919acce952bSliubo { 4920acce952bSliubo struct rb_node *node; 4921acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4922acce952bSliubo struct btrfs_delayed_ref_node *ref; 4923acce952bSliubo int ret = 0; 4924acce952bSliubo 4925acce952bSliubo delayed_refs = &trans->delayed_refs; 4926acce952bSliubo 4927acce952bSliubo spin_lock(&delayed_refs->lock); 4928d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4929cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4930b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4931acce952bSliubo return ret; 4932acce952bSliubo } 4933acce952bSliubo 49345c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4935d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 49360e0adbcfSJosef Bacik struct rb_node *n; 4937e78417d1SJosef Bacik bool pin_bytes = false; 4938acce952bSliubo 4939d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4940d7df2c79SJosef Bacik href_node); 49413069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4942b939d1abSJosef Bacik continue; 49433069bd26SJosef Bacik 4944d7df2c79SJosef Bacik spin_lock(&head->lock); 4945e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 49460e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 49470e0adbcfSJosef Bacik ref_node); 4948d7df2c79SJosef Bacik ref->in_tree = 0; 4949e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 49500e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 49511d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 49521d57ee94SWang Xiaoguang list_del(&ref->add_list); 4953d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4954d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4955d7df2c79SJosef Bacik } 495654067ae9SJosef Bacik if (head->must_insert_reserved) 4957e78417d1SJosef Bacik pin_bytes = true; 495878a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4959fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4960d7df2c79SJosef Bacik spin_unlock(&head->lock); 4961acce952bSliubo spin_unlock(&delayed_refs->lock); 4962e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4963acce952bSliubo 4964f603bb94SNikolay Borisov if (pin_bytes) { 4965f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4966f603bb94SNikolay Borisov 4967f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4968f603bb94SNikolay Borisov BUG_ON(!cache); 4969f603bb94SNikolay Borisov 4970f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4971f603bb94SNikolay Borisov spin_lock(&cache->lock); 4972f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4973f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4974f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4975f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4976f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4977f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4978f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4979f603bb94SNikolay Borisov 4980f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4981f603bb94SNikolay Borisov 4982f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4983f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4984f603bb94SNikolay Borisov } 498531890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4986d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4987acce952bSliubo cond_resched(); 4988acce952bSliubo spin_lock(&delayed_refs->lock); 4989acce952bSliubo } 499081f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4991acce952bSliubo 4992acce952bSliubo spin_unlock(&delayed_refs->lock); 4993acce952bSliubo 4994acce952bSliubo return ret; 4995acce952bSliubo } 4996acce952bSliubo 4997143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4998acce952bSliubo { 4999acce952bSliubo struct btrfs_inode *btrfs_inode; 5000acce952bSliubo struct list_head splice; 5001acce952bSliubo 5002acce952bSliubo INIT_LIST_HEAD(&splice); 5003acce952bSliubo 5004eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5005eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 5006acce952bSliubo 5007acce952bSliubo while (!list_empty(&splice)) { 5008fe816d0fSNikolay Borisov struct inode *inode = NULL; 5009eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 5010acce952bSliubo delalloc_inodes); 5011fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 5012eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5013acce952bSliubo 5014fe816d0fSNikolay Borisov /* 5015fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 5016fe816d0fSNikolay Borisov * meanwhile. 5017fe816d0fSNikolay Borisov */ 5018fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 5019fe816d0fSNikolay Borisov if (inode) { 5020fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 5021fe816d0fSNikolay Borisov iput(inode); 5022fe816d0fSNikolay Borisov } 5023eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5024acce952bSliubo } 5025eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5026eb73c1b7SMiao Xie } 5027eb73c1b7SMiao Xie 5028eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 5029eb73c1b7SMiao Xie { 5030eb73c1b7SMiao Xie struct btrfs_root *root; 5031eb73c1b7SMiao Xie struct list_head splice; 5032eb73c1b7SMiao Xie 5033eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 5034eb73c1b7SMiao Xie 5035eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5036eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 5037eb73c1b7SMiao Xie while (!list_empty(&splice)) { 5038eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 5039eb73c1b7SMiao Xie delalloc_root); 504000246528SJosef Bacik root = btrfs_grab_root(root); 5041eb73c1b7SMiao Xie BUG_ON(!root); 5042eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5043eb73c1b7SMiao Xie 5044eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 504500246528SJosef Bacik btrfs_put_root(root); 5046eb73c1b7SMiao Xie 5047eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5048eb73c1b7SMiao Xie } 5049eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5050acce952bSliubo } 5051acce952bSliubo 50522ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 5053acce952bSliubo struct extent_io_tree *dirty_pages, 5054acce952bSliubo int mark) 5055acce952bSliubo { 5056acce952bSliubo int ret; 5057acce952bSliubo struct extent_buffer *eb; 5058acce952bSliubo u64 start = 0; 5059acce952bSliubo u64 end; 5060acce952bSliubo 5061acce952bSliubo while (1) { 5062acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 5063e6138876SJosef Bacik mark, NULL); 5064acce952bSliubo if (ret) 5065acce952bSliubo break; 5066acce952bSliubo 506791166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 5068acce952bSliubo while (start <= end) { 50690b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 50700b246afaSJeff Mahoney start += fs_info->nodesize; 5071fd8b2b61SJosef Bacik if (!eb) 5072acce952bSliubo continue; 5073fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 5074acce952bSliubo 5075fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 5076fd8b2b61SJosef Bacik &eb->bflags)) 5077fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 5078fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 5079acce952bSliubo } 5080acce952bSliubo } 5081acce952bSliubo 5082acce952bSliubo return ret; 5083acce952bSliubo } 5084acce952bSliubo 50852ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 5086fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 5087acce952bSliubo { 5088acce952bSliubo u64 start; 5089acce952bSliubo u64 end; 5090acce952bSliubo int ret; 5091acce952bSliubo 5092acce952bSliubo while (1) { 50930e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 50940e6ec385SFilipe Manana 5095fcd5e742SLu Fengqi /* 5096fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 5097fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 5098fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 5099fcd5e742SLu Fengqi * the same extent range. 5100fcd5e742SLu Fengqi */ 5101fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 5102acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 51030e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 5104fcd5e742SLu Fengqi if (ret) { 5105fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5106acce952bSliubo break; 5107fcd5e742SLu Fengqi } 5108acce952bSliubo 51090e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 51100e6ec385SFilipe Manana free_extent_state(cached_state); 51112ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 5112fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5113acce952bSliubo cond_resched(); 5114acce952bSliubo } 5115acce952bSliubo 5116acce952bSliubo return 0; 5117acce952bSliubo } 5118acce952bSliubo 511932da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 5120c79a1751SLiu Bo { 5121c79a1751SLiu Bo struct inode *inode; 5122c79a1751SLiu Bo 5123c79a1751SLiu Bo inode = cache->io_ctl.inode; 5124c79a1751SLiu Bo if (inode) { 5125c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 5126c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 5127c79a1751SLiu Bo cache->io_ctl.inode = NULL; 5128c79a1751SLiu Bo iput(inode); 5129c79a1751SLiu Bo } 5130bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 5131c79a1751SLiu Bo btrfs_put_block_group(cache); 5132c79a1751SLiu Bo } 5133c79a1751SLiu Bo 5134c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 51352ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 5136c79a1751SLiu Bo { 513732da5386SDavid Sterba struct btrfs_block_group *cache; 5138c79a1751SLiu Bo 5139c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5140c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 5141c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 514232da5386SDavid Sterba struct btrfs_block_group, 5143c79a1751SLiu Bo dirty_list); 5144c79a1751SLiu Bo 5145c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 5146c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5147c79a1751SLiu Bo list_del_init(&cache->io_list); 5148c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5149c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5150c79a1751SLiu Bo } 5151c79a1751SLiu Bo 5152c79a1751SLiu Bo list_del_init(&cache->dirty_list); 5153c79a1751SLiu Bo spin_lock(&cache->lock); 5154c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5155c79a1751SLiu Bo spin_unlock(&cache->lock); 5156c79a1751SLiu Bo 5157c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5158c79a1751SLiu Bo btrfs_put_block_group(cache); 5159ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 5160c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5161c79a1751SLiu Bo } 5162c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5163c79a1751SLiu Bo 516445ae2c18SNikolay Borisov /* 516545ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 516645ae2c18SNikolay Borisov * to use it without any locking 516745ae2c18SNikolay Borisov */ 5168c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 5169c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 517032da5386SDavid Sterba struct btrfs_block_group, 5171c79a1751SLiu Bo io_list); 5172c79a1751SLiu Bo 5173c79a1751SLiu Bo list_del_init(&cache->io_list); 5174c79a1751SLiu Bo spin_lock(&cache->lock); 5175c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5176c79a1751SLiu Bo spin_unlock(&cache->lock); 5177c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5178c79a1751SLiu Bo } 5179c79a1751SLiu Bo } 5180c79a1751SLiu Bo 518149b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 51822ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 518349b25e05SJeff Mahoney { 5184bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 5185bbbf7243SNikolay Borisov 51862ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 5187c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 5188c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 5189c79a1751SLiu Bo 5190bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 5191bbbf7243SNikolay Borisov post_commit_list) { 5192bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 5193bbbf7243SNikolay Borisov } 5194bbbf7243SNikolay Borisov 51952ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 519649b25e05SJeff Mahoney 51974a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 51980b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 519949b25e05SJeff Mahoney 52004a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 52010b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 520249b25e05SJeff Mahoney 5203ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 520467cde344SMiao Xie 52052ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 520649b25e05SJeff Mahoney EXTENT_DIRTY); 5207fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 520849b25e05SJeff Mahoney 5209d3575156SNaohiro Aota btrfs_free_redirty_list(cur_trans); 5210d3575156SNaohiro Aota 52114a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 52124a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 521349b25e05SJeff Mahoney } 521449b25e05SJeff Mahoney 52152ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 5216acce952bSliubo { 5217acce952bSliubo struct btrfs_transaction *t; 5218acce952bSliubo 52190b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 5220acce952bSliubo 52210b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 52220b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 52230b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 5224724e2315SJosef Bacik struct btrfs_transaction, list); 5225724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 52269b64f57dSElena Reshetova refcount_inc(&t->use_count); 52270b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 52282ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 5229724e2315SJosef Bacik btrfs_put_transaction(t); 52300b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5231724e2315SJosef Bacik continue; 5232724e2315SJosef Bacik } 52330b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 5234724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 52350b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5236724e2315SJosef Bacik /* 5237724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 5238724e2315SJosef Bacik * handle open currently for this transaction. 5239724e2315SJosef Bacik */ 5240724e2315SJosef Bacik wait_event(t->writer_wait, 5241724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 5242724e2315SJosef Bacik } else { 52430b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5244724e2315SJosef Bacik } 52452ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 5246724e2315SJosef Bacik 52470b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 52480b246afaSJeff Mahoney if (t == fs_info->running_transaction) 52490b246afaSJeff Mahoney fs_info->running_transaction = NULL; 5250724e2315SJosef Bacik list_del_init(&t->list); 52510b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5252a4abeea4SJosef Bacik 5253724e2315SJosef Bacik btrfs_put_transaction(t); 52542e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 52550b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5256724e2315SJosef Bacik } 52570b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 52580b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 5259ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 5260ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 52610b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 5262ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 52630b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 5264acce952bSliubo 5265acce952bSliubo return 0; 5266acce952bSliubo } 5267ec7d6dfdSNikolay Borisov 5268453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 5269ec7d6dfdSNikolay Borisov { 5270ec7d6dfdSNikolay Borisov struct btrfs_path *path; 5271ec7d6dfdSNikolay Borisov int ret; 5272ec7d6dfdSNikolay Borisov struct extent_buffer *l; 5273ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 5274ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 5275ec7d6dfdSNikolay Borisov int slot; 5276ec7d6dfdSNikolay Borisov 5277ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 5278ec7d6dfdSNikolay Borisov if (!path) 5279ec7d6dfdSNikolay Borisov return -ENOMEM; 5280ec7d6dfdSNikolay Borisov 5281ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 5282ec7d6dfdSNikolay Borisov search_key.type = -1; 5283ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 5284ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 5285ec7d6dfdSNikolay Borisov if (ret < 0) 5286ec7d6dfdSNikolay Borisov goto error; 5287ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 5288ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 5289ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 5290ec7d6dfdSNikolay Borisov l = path->nodes[0]; 5291ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 529223125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 529323125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 5294ec7d6dfdSNikolay Borisov } else { 529523125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 5296ec7d6dfdSNikolay Borisov } 5297ec7d6dfdSNikolay Borisov ret = 0; 5298ec7d6dfdSNikolay Borisov error: 5299ec7d6dfdSNikolay Borisov btrfs_free_path(path); 5300ec7d6dfdSNikolay Borisov return ret; 5301ec7d6dfdSNikolay Borisov } 5302ec7d6dfdSNikolay Borisov 5303543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 5304ec7d6dfdSNikolay Borisov { 5305ec7d6dfdSNikolay Borisov int ret; 5306ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5307ec7d6dfdSNikolay Borisov 53086b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5309ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5310ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5311ec7d6dfdSNikolay Borisov root->root_key.objectid); 5312ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5313ec7d6dfdSNikolay Borisov goto out; 5314ec7d6dfdSNikolay Borisov } 5315ec7d6dfdSNikolay Borisov 531623125104SNikolay Borisov *objectid = root->free_objectid++; 5317ec7d6dfdSNikolay Borisov ret = 0; 5318ec7d6dfdSNikolay Borisov out: 5319ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5320ec7d6dfdSNikolay Borisov return ret; 5321ec7d6dfdSNikolay Borisov } 5322