1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 26cbd5570SChris Mason /* 36cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 46cbd5570SChris Mason */ 56cbd5570SChris Mason 6e20d96d6SChris Mason #include <linux/fs.h> 7d98237b3SChris Mason #include <linux/blkdev.h> 8fc7cbcd4SDavid Sterba #include <linux/radix-tree.h> 935b7e476SChris Mason #include <linux/writeback.h> 10ce9adaa5SChris Mason #include <linux/workqueue.h> 11a74a4b97SChris Mason #include <linux/kthread.h> 125a0e3ad6STejun Heo #include <linux/slab.h> 13784b4e29SChris Mason #include <linux/migrate.h> 147a36ddecSDavid Sterba #include <linux/ratelimit.h> 156463fe58SStefan Behrens #include <linux/uuid.h> 16803b2f54SStefan Behrens #include <linux/semaphore.h> 17540adea3SMasami Hiramatsu #include <linux/error-injection.h> 189678c543SNikolay Borisov #include <linux/crc32c.h> 19b89f6d1fSFilipe Manana #include <linux/sched/mm.h> 207e75bf3fSDavid Sterba #include <asm/unaligned.h> 216d97c6e3SJohannes Thumshirn #include <crypto/hash.h> 22eb60ceacSChris Mason #include "ctree.h" 23eb60ceacSChris Mason #include "disk-io.h" 24e089f05cSChris Mason #include "transaction.h" 250f7d52f4SChris Mason #include "btrfs_inode.h" 26103c1972SChristoph Hellwig #include "bio.h" 27db94535dSChris Mason #include "print-tree.h" 28925baeddSChris Mason #include "locking.h" 29e02119d5SChris Mason #include "tree-log.h" 30fa9c0d79SChris Mason #include "free-space-cache.h" 3170f6d82eSOmar Sandoval #include "free-space-tree.h" 3221adbd5cSStefan Behrens #include "check-integrity.h" 33606686eeSJosef Bacik #include "rcu-string.h" 348dabb742SStefan Behrens #include "dev-replace.h" 3553b381b3SDavid Woodhouse #include "raid56.h" 365ac1d209SJeff Mahoney #include "sysfs.h" 37fcebe456SJosef Bacik #include "qgroup.h" 38ebb8765bSAnand Jain #include "compression.h" 39557ea5ddSQu Wenruo #include "tree-checker.h" 40fd708b81SJosef Bacik #include "ref-verify.h" 41aac0023cSJosef Bacik #include "block-group.h" 42b0643e59SDennis Zhou #include "discard.h" 43f603bb94SNikolay Borisov #include "space-info.h" 44b70f5097SNaohiro Aota #include "zoned.h" 45139e8cd3SQu Wenruo #include "subpage.h" 46c7f13d42SJosef Bacik #include "fs.h" 4707e81dc9SJosef Bacik #include "accessors.h" 48a0231804SJosef Bacik #include "extent-tree.h" 4945c40c8fSJosef Bacik #include "root-tree.h" 5059b818e0SJosef Bacik #include "defrag.h" 51c7a03b52SJosef Bacik #include "uuid-tree.h" 5267707479SJosef Bacik #include "relocation.h" 532fc6822cSJosef Bacik #include "scrub.h" 54c03b2207SJosef Bacik #include "super.h" 55eb60ceacSChris Mason 56319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 57319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 58319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 59319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 60e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 61e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 62319e4d06SQu Wenruo 63143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 64acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 652ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 66143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 672ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 68acce952bSliubo struct extent_io_tree *dirty_pages, 69acce952bSliubo int mark); 702ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 71acce952bSliubo struct extent_io_tree *pinned_extents); 722ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 732ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 74ce9adaa5SChris Mason 75141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 76141386e1SJosef Bacik { 77141386e1SJosef Bacik if (fs_info->csum_shash) 78141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 79141386e1SJosef Bacik } 80141386e1SJosef Bacik 81d352ac68SChris Mason /* 82d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 83d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 84d352ac68SChris Mason * just before they are sent down the IO stack. 85d352ac68SChris Mason */ 8644b8bd7eSChris Mason struct async_submit_bio { 87da67daabSDavid Sterba struct btrfs_inode *inode; 8844b8bd7eSChris Mason struct bio *bio; 89ab2072b2SDavid Sterba enum btrfs_wq_submit_cmd submit_cmd; 9044b8bd7eSChris Mason int mirror_num; 911941b64bSQu Wenruo 92ab2072b2SDavid Sterba /* Optional parameter for used by direct io */ 931941b64bSQu Wenruo u64 dio_file_offset; 948b712842SChris Mason struct btrfs_work work; 954e4cbee9SChristoph Hellwig blk_status_t status; 9644b8bd7eSChris Mason }; 9744b8bd7eSChris Mason 9885d4e461SChris Mason /* 992996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 100d352ac68SChris Mason */ 101c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 10219c00ddcSChris Mason { 103d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 1047280305eSDavid Sterba const int num_pages = num_extent_pages(buf); 105a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 106d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 10719c00ddcSChris Mason char *kaddr; 108e9be5a30SDavid Sterba int i; 109d5178578SJohannes Thumshirn 110d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 111d5178578SJohannes Thumshirn crypto_shash_init(shash); 112a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 113e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 114a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 11519c00ddcSChris Mason 116e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 117e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 118e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 11919c00ddcSChris Mason } 12071a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 121d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 12219c00ddcSChris Mason } 12319c00ddcSChris Mason 124d352ac68SChris Mason /* 125d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 126d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 127d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 128d352ac68SChris Mason * in the wrong place. 129d352ac68SChris Mason */ 1301259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 131b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 132b9fab919SChris Mason int atomic) 1331259ab75SChris Mason { 1342ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 1351259ab75SChris Mason int ret; 1361259ab75SChris Mason 1371259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 1381259ab75SChris Mason return 0; 1391259ab75SChris Mason 140b9fab919SChris Mason if (atomic) 141b9fab919SChris Mason return -EAGAIN; 142b9fab919SChris Mason 143570eb97bSJosef Bacik lock_extent(io_tree, eb->start, eb->start + eb->len - 1, &cached_state); 1440b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 1451259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 1461259ab75SChris Mason ret = 0; 1471259ab75SChris Mason goto out; 1481259ab75SChris Mason } 14994647322SDavid Sterba btrfs_err_rl(eb->fs_info, 1508f0ed7d4SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 1518f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 15229549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 1531259ab75SChris Mason ret = 1; 1540b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 15533958dc6SChris Mason out: 156570eb97bSJosef Bacik unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, 157e43bbe5eSDavid Sterba &cached_state); 1581259ab75SChris Mason return ret; 1591259ab75SChris Mason } 1601259ab75SChris Mason 161e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 162e7e16f48SJohannes Thumshirn { 163e7e16f48SJohannes Thumshirn switch (csum_type) { 164e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 1653951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 1663831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 167352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 168e7e16f48SJohannes Thumshirn return true; 169e7e16f48SJohannes Thumshirn default: 170e7e16f48SJohannes Thumshirn return false; 171e7e16f48SJohannes Thumshirn } 172e7e16f48SJohannes Thumshirn } 173e7e16f48SJohannes Thumshirn 174d352ac68SChris Mason /* 1751104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 1761104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 1771104a885SDavid Sterba */ 1783d17adeaSQu Wenruo int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 1793d17adeaSQu Wenruo const struct btrfs_super_block *disk_sb) 1801104a885SDavid Sterba { 18151bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 182d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 183d5178578SJohannes Thumshirn 184d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 1851104a885SDavid Sterba 1861104a885SDavid Sterba /* 1871104a885SDavid Sterba * The super_block structure does not span the whole 18851bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 18951bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 1901104a885SDavid Sterba */ 1913d17adeaSQu Wenruo crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, 192fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 1931104a885SDavid Sterba 19455fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 195e7e16f48SJohannes Thumshirn return 1; 1961104a885SDavid Sterba 197e7e16f48SJohannes Thumshirn return 0; 1981104a885SDavid Sterba } 1991104a885SDavid Sterba 200e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 201ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 202581c1760SQu Wenruo { 203e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 204581c1760SQu Wenruo int found_level; 205581c1760SQu Wenruo struct btrfs_key found_key; 206581c1760SQu Wenruo int ret; 207581c1760SQu Wenruo 208581c1760SQu Wenruo found_level = btrfs_header_level(eb); 209581c1760SQu Wenruo if (found_level != level) { 21063489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 21163489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 212581c1760SQu Wenruo btrfs_err(fs_info, 213581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 214581c1760SQu Wenruo eb->start, level, found_level); 215581c1760SQu Wenruo return -EIO; 216581c1760SQu Wenruo } 217581c1760SQu Wenruo 218581c1760SQu Wenruo if (!first_key) 219581c1760SQu Wenruo return 0; 220581c1760SQu Wenruo 2215d41be6fSQu Wenruo /* 2225d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 2235d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 2245d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 2255d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 2265d41be6fSQu Wenruo */ 2275d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 2285d41be6fSQu Wenruo return 0; 22962fdaa52SQu Wenruo 23062fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 23162fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 23262fdaa52SQu Wenruo btrfs_err(fs_info, 23362fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 23462fdaa52SQu Wenruo eb->start); 23562fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 23662fdaa52SQu Wenruo return -EUCLEAN; 23762fdaa52SQu Wenruo } 23862fdaa52SQu Wenruo 239581c1760SQu Wenruo if (found_level) 240581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 241581c1760SQu Wenruo else 242581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 243581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 244581c1760SQu Wenruo 245581c1760SQu Wenruo if (ret) { 24663489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 24763489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 248581c1760SQu Wenruo btrfs_err(fs_info, 249ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 250ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 251ff76a864SLiu Bo first_key->type, first_key->offset, 252ff76a864SLiu Bo found_key.objectid, found_key.type, 253ff76a864SLiu Bo found_key.offset); 254581c1760SQu Wenruo } 255581c1760SQu Wenruo return ret; 256581c1760SQu Wenruo } 257581c1760SQu Wenruo 258bacf60e5SChristoph Hellwig static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, 259bacf60e5SChristoph Hellwig int mirror_num) 260bacf60e5SChristoph Hellwig { 261bacf60e5SChristoph Hellwig struct btrfs_fs_info *fs_info = eb->fs_info; 262bacf60e5SChristoph Hellwig u64 start = eb->start; 263bacf60e5SChristoph Hellwig int i, num_pages = num_extent_pages(eb); 264bacf60e5SChristoph Hellwig int ret = 0; 265bacf60e5SChristoph Hellwig 266bacf60e5SChristoph Hellwig if (sb_rdonly(fs_info->sb)) 267bacf60e5SChristoph Hellwig return -EROFS; 268bacf60e5SChristoph Hellwig 269bacf60e5SChristoph Hellwig for (i = 0; i < num_pages; i++) { 270bacf60e5SChristoph Hellwig struct page *p = eb->pages[i]; 271bacf60e5SChristoph Hellwig 272bacf60e5SChristoph Hellwig ret = btrfs_repair_io_failure(fs_info, 0, start, PAGE_SIZE, 273bacf60e5SChristoph Hellwig start, p, start - page_offset(p), mirror_num); 274bacf60e5SChristoph Hellwig if (ret) 275bacf60e5SChristoph Hellwig break; 276bacf60e5SChristoph Hellwig start += PAGE_SIZE; 277bacf60e5SChristoph Hellwig } 278bacf60e5SChristoph Hellwig 279bacf60e5SChristoph Hellwig return ret; 280bacf60e5SChristoph Hellwig } 281bacf60e5SChristoph Hellwig 2821104a885SDavid Sterba /* 283d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 284d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 285581c1760SQu Wenruo * 286789d6a3aSQu Wenruo * @check: expected tree parentness check, see the comments of the 287789d6a3aSQu Wenruo * structure for details. 288d352ac68SChris Mason */ 2896a2e9dc4SFilipe Manana int btrfs_read_extent_buffer(struct extent_buffer *eb, 290789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 291f188591eSChris Mason { 2925ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 293ea466794SJosef Bacik int failed = 0; 294f188591eSChris Mason int ret; 295f188591eSChris Mason int num_copies = 0; 296f188591eSChris Mason int mirror_num = 0; 297ea466794SJosef Bacik int failed_mirror = 0; 298f188591eSChris Mason 299789d6a3aSQu Wenruo ASSERT(check); 300789d6a3aSQu Wenruo 301f188591eSChris Mason while (1) { 302f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 303947a6299SQu Wenruo ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num, check); 304947a6299SQu Wenruo if (!ret) 305581c1760SQu Wenruo break; 306d397712bSChris Mason 3070b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 308f188591eSChris Mason eb->start, eb->len); 3094235298eSChris Mason if (num_copies == 1) 310ea466794SJosef Bacik break; 3114235298eSChris Mason 3125cf1ab56SJosef Bacik if (!failed_mirror) { 3135cf1ab56SJosef Bacik failed = 1; 3145cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 3155cf1ab56SJosef Bacik } 3165cf1ab56SJosef Bacik 317f188591eSChris Mason mirror_num++; 318ea466794SJosef Bacik if (mirror_num == failed_mirror) 319ea466794SJosef Bacik mirror_num++; 320ea466794SJosef Bacik 3214235298eSChris Mason if (mirror_num > num_copies) 322ea466794SJosef Bacik break; 323f188591eSChris Mason } 324ea466794SJosef Bacik 325c0901581SStefan Behrens if (failed && !ret && failed_mirror) 32620a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 327ea466794SJosef Bacik 328ea466794SJosef Bacik return ret; 329f188591eSChris Mason } 33019c00ddcSChris Mason 331eca0f6f6SQu Wenruo static int csum_one_extent_buffer(struct extent_buffer *eb) 332eca0f6f6SQu Wenruo { 333eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 334eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 335eca0f6f6SQu Wenruo int ret; 336eca0f6f6SQu Wenruo 337eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 338eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 339eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 340eca0f6f6SQu Wenruo csum_tree_block(eb, result); 341eca0f6f6SQu Wenruo 342eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 343eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 344eca0f6f6SQu Wenruo else 345eca0f6f6SQu Wenruo ret = btrfs_check_leaf_full(eb); 346eca0f6f6SQu Wenruo 3473777369fSQu Wenruo if (ret < 0) 3483777369fSQu Wenruo goto error; 3493777369fSQu Wenruo 3503777369fSQu Wenruo /* 3513777369fSQu Wenruo * Also check the generation, the eb reached here must be newer than 3523777369fSQu Wenruo * last committed. Or something seriously wrong happened. 3533777369fSQu Wenruo */ 3543777369fSQu Wenruo if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { 3553777369fSQu Wenruo ret = -EUCLEAN; 356eca0f6f6SQu Wenruo btrfs_err(fs_info, 3573777369fSQu Wenruo "block=%llu bad generation, have %llu expect > %llu", 3583777369fSQu Wenruo eb->start, btrfs_header_generation(eb), 3593777369fSQu Wenruo fs_info->last_trans_committed); 3603777369fSQu Wenruo goto error; 361eca0f6f6SQu Wenruo } 362eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 363eca0f6f6SQu Wenruo 364eca0f6f6SQu Wenruo return 0; 3653777369fSQu Wenruo 3663777369fSQu Wenruo error: 3673777369fSQu Wenruo btrfs_print_tree(eb, 0); 3683777369fSQu Wenruo btrfs_err(fs_info, "block=%llu write time tree block corruption detected", 3693777369fSQu Wenruo eb->start); 3703777369fSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 3713777369fSQu Wenruo return ret; 372eca0f6f6SQu Wenruo } 373eca0f6f6SQu Wenruo 374eca0f6f6SQu Wenruo /* Checksum all dirty extent buffers in one bio_vec */ 375eca0f6f6SQu Wenruo static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, 376eca0f6f6SQu Wenruo struct bio_vec *bvec) 377eca0f6f6SQu Wenruo { 378eca0f6f6SQu Wenruo struct page *page = bvec->bv_page; 379eca0f6f6SQu Wenruo u64 bvec_start = page_offset(page) + bvec->bv_offset; 380eca0f6f6SQu Wenruo u64 cur; 381eca0f6f6SQu Wenruo int ret = 0; 382eca0f6f6SQu Wenruo 383eca0f6f6SQu Wenruo for (cur = bvec_start; cur < bvec_start + bvec->bv_len; 384eca0f6f6SQu Wenruo cur += fs_info->nodesize) { 385eca0f6f6SQu Wenruo struct extent_buffer *eb; 386eca0f6f6SQu Wenruo bool uptodate; 387eca0f6f6SQu Wenruo 388eca0f6f6SQu Wenruo eb = find_extent_buffer(fs_info, cur); 389eca0f6f6SQu Wenruo uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, 390eca0f6f6SQu Wenruo fs_info->nodesize); 391eca0f6f6SQu Wenruo 39201cd3909SDavid Sterba /* A dirty eb shouldn't disappear from buffer_radix */ 393eca0f6f6SQu Wenruo if (WARN_ON(!eb)) 394eca0f6f6SQu Wenruo return -EUCLEAN; 395eca0f6f6SQu Wenruo 396eca0f6f6SQu Wenruo if (WARN_ON(cur != btrfs_header_bytenr(eb))) { 397eca0f6f6SQu Wenruo free_extent_buffer(eb); 398eca0f6f6SQu Wenruo return -EUCLEAN; 399eca0f6f6SQu Wenruo } 400eca0f6f6SQu Wenruo if (WARN_ON(!uptodate)) { 401eca0f6f6SQu Wenruo free_extent_buffer(eb); 402eca0f6f6SQu Wenruo return -EUCLEAN; 403eca0f6f6SQu Wenruo } 404eca0f6f6SQu Wenruo 405eca0f6f6SQu Wenruo ret = csum_one_extent_buffer(eb); 406eca0f6f6SQu Wenruo free_extent_buffer(eb); 407eca0f6f6SQu Wenruo if (ret < 0) 408eca0f6f6SQu Wenruo return ret; 409eca0f6f6SQu Wenruo } 410eca0f6f6SQu Wenruo return ret; 411eca0f6f6SQu Wenruo } 412eca0f6f6SQu Wenruo 413d352ac68SChris Mason /* 414ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 415ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 416ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 417d352ac68SChris Mason */ 418ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 41919c00ddcSChris Mason { 420ac303b69SQu Wenruo struct page *page = bvec->bv_page; 4214eee4fa4SMiao Xie u64 start = page_offset(page); 42219c00ddcSChris Mason u64 found_start; 42319c00ddcSChris Mason struct extent_buffer *eb; 424eca0f6f6SQu Wenruo 425fbca46ebSQu Wenruo if (fs_info->nodesize < PAGE_SIZE) 426eca0f6f6SQu Wenruo return csum_dirty_subpage_buffers(fs_info, bvec); 427f188591eSChris Mason 4284f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4294f2de97aSJosef Bacik if (page != eb->pages[0]) 4304f2de97aSJosef Bacik return 0; 4310f805531SAlex Lyakas 43219c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 433d3575156SNaohiro Aota 434d3575156SNaohiro Aota if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 435d3575156SNaohiro Aota WARN_ON(found_start != 0); 436d3575156SNaohiro Aota return 0; 437d3575156SNaohiro Aota } 438d3575156SNaohiro Aota 4390f805531SAlex Lyakas /* 4400f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 4410f805531SAlex Lyakas * It is useful to know what went wrong. 4420f805531SAlex Lyakas */ 4430f805531SAlex Lyakas if (WARN_ON(found_start != start)) 4440f805531SAlex Lyakas return -EUCLEAN; 4450f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 4460f805531SAlex Lyakas return -EUCLEAN; 4470f805531SAlex Lyakas 448eca0f6f6SQu Wenruo return csum_one_extent_buffer(eb); 44919c00ddcSChris Mason } 45019c00ddcSChris Mason 451b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 4522b82032cSYan Zheng { 453b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 454944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 45544880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 456944d3f9fSNikolay Borisov u8 *metadata_uuid; 4572b82032cSYan Zheng 4589a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 4599a8658e3SDavid Sterba BTRFS_FSID_SIZE); 4607239ff4bSNikolay Borisov /* 461944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 462944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 463944d3f9fSNikolay Borisov * ->fsid in this case is fine 4647239ff4bSNikolay Borisov */ 465944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 4667239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 4677239ff4bSNikolay Borisov else 4687239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 4697239ff4bSNikolay Borisov 470944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 471944d3f9fSNikolay Borisov return 0; 472944d3f9fSNikolay Borisov 473944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 474944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 475944d3f9fSNikolay Borisov return 0; 476944d3f9fSNikolay Borisov 477944d3f9fSNikolay Borisov return 1; 4782b82032cSYan Zheng } 4792b82032cSYan Zheng 48077bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 481947a6299SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb, 482947a6299SQu Wenruo struct btrfs_tree_parent_check *check) 483ce9adaa5SChris Mason { 48477bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 485ce9adaa5SChris Mason u64 found_start; 48677bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 48777bf40a2SQu Wenruo u8 found_level; 4882996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 489dfd29eedSDavid Sterba const u8 *header_csum; 49077bf40a2SQu Wenruo int ret = 0; 491ea466794SJosef Bacik 492947a6299SQu Wenruo ASSERT(check); 493947a6299SQu Wenruo 494ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 495727011e0SChris Mason if (found_start != eb->start) { 4968f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, 4978f0ed7d4SQu Wenruo "bad tree block start, mirror %u want %llu have %llu", 4988f0ed7d4SQu Wenruo eb->read_mirror, eb->start, found_start); 499f188591eSChris Mason ret = -EIO; 50077bf40a2SQu Wenruo goto out; 501ce9adaa5SChris Mason } 502b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 5038f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", 5048f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 5051259ab75SChris Mason ret = -EIO; 50677bf40a2SQu Wenruo goto out; 5071259ab75SChris Mason } 508ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 5091c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 5108f0ed7d4SQu Wenruo btrfs_err(fs_info, 5118f0ed7d4SQu Wenruo "bad tree block level, mirror %u level %d on logical %llu", 5128f0ed7d4SQu Wenruo eb->read_mirror, btrfs_header_level(eb), eb->start); 5131c24c3ceSJosef Bacik ret = -EIO; 51477bf40a2SQu Wenruo goto out; 5151c24c3ceSJosef Bacik } 516ce9adaa5SChris Mason 517c67b3892SDavid Sterba csum_tree_block(eb, result); 518dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 519dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 520a826d6dcSJosef Bacik 521dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 5222996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 5238f0ed7d4SQu Wenruo "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", 5248f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 525dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 52635be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 52735be8851SJohannes Thumshirn btrfs_header_level(eb)); 5282996e1f8SJohannes Thumshirn ret = -EUCLEAN; 52977bf40a2SQu Wenruo goto out; 5302996e1f8SJohannes Thumshirn } 5312996e1f8SJohannes Thumshirn 532947a6299SQu Wenruo if (found_level != check->level) { 53377177ed1SQu Wenruo btrfs_err(fs_info, 53477177ed1SQu Wenruo "level verify failed on logical %llu mirror %u wanted %u found %u", 53577177ed1SQu Wenruo eb->start, eb->read_mirror, check->level, found_level); 536947a6299SQu Wenruo ret = -EIO; 537947a6299SQu Wenruo goto out; 538947a6299SQu Wenruo } 539947a6299SQu Wenruo if (unlikely(check->transid && 540947a6299SQu Wenruo btrfs_header_generation(eb) != check->transid)) { 541947a6299SQu Wenruo btrfs_err_rl(eb->fs_info, 542947a6299SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 543947a6299SQu Wenruo eb->start, eb->read_mirror, check->transid, 544947a6299SQu Wenruo btrfs_header_generation(eb)); 545947a6299SQu Wenruo ret = -EIO; 546947a6299SQu Wenruo goto out; 547947a6299SQu Wenruo } 548947a6299SQu Wenruo if (check->has_first_key) { 549947a6299SQu Wenruo struct btrfs_key *expect_key = &check->first_key; 550947a6299SQu Wenruo struct btrfs_key found_key; 551947a6299SQu Wenruo 552947a6299SQu Wenruo if (found_level) 553947a6299SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 554947a6299SQu Wenruo else 555947a6299SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 556947a6299SQu Wenruo if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { 557947a6299SQu Wenruo btrfs_err(fs_info, 558947a6299SQu Wenruo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 559947a6299SQu Wenruo eb->start, check->transid, 560947a6299SQu Wenruo expect_key->objectid, 561947a6299SQu Wenruo expect_key->type, expect_key->offset, 562947a6299SQu Wenruo found_key.objectid, found_key.type, 563947a6299SQu Wenruo found_key.offset); 564947a6299SQu Wenruo ret = -EUCLEAN; 565947a6299SQu Wenruo goto out; 566947a6299SQu Wenruo } 567947a6299SQu Wenruo } 568947a6299SQu Wenruo if (check->owner_root) { 569947a6299SQu Wenruo ret = btrfs_check_eb_owner(eb, check->owner_root); 570947a6299SQu Wenruo if (ret < 0) 571947a6299SQu Wenruo goto out; 572947a6299SQu Wenruo } 573947a6299SQu Wenruo 574a826d6dcSJosef Bacik /* 575a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 576a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 577a826d6dcSJosef Bacik * return -EIO. 578a826d6dcSJosef Bacik */ 5791c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 580a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 581a826d6dcSJosef Bacik ret = -EIO; 582a826d6dcSJosef Bacik } 583ce9adaa5SChris Mason 584813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 585053ab70fSLiu Bo ret = -EIO; 586053ab70fSLiu Bo 5870b32f4bbSJosef Bacik if (!ret) 5880b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 58975391f0dSQu Wenruo else 59075391f0dSQu Wenruo btrfs_err(fs_info, 5918f0ed7d4SQu Wenruo "read time tree block corruption detected on logical %llu mirror %u", 5928f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 59377bf40a2SQu Wenruo out: 59477bf40a2SQu Wenruo return ret; 59577bf40a2SQu Wenruo } 59677bf40a2SQu Wenruo 597371cdc07SQu Wenruo static int validate_subpage_buffer(struct page *page, u64 start, u64 end, 598947a6299SQu Wenruo int mirror, struct btrfs_tree_parent_check *check) 599371cdc07SQu Wenruo { 600371cdc07SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 601371cdc07SQu Wenruo struct extent_buffer *eb; 602371cdc07SQu Wenruo bool reads_done; 603371cdc07SQu Wenruo int ret = 0; 604371cdc07SQu Wenruo 605947a6299SQu Wenruo ASSERT(check); 606947a6299SQu Wenruo 607371cdc07SQu Wenruo /* 608371cdc07SQu Wenruo * We don't allow bio merge for subpage metadata read, so we should 609371cdc07SQu Wenruo * only get one eb for each endio hook. 610371cdc07SQu Wenruo */ 611371cdc07SQu Wenruo ASSERT(end == start + fs_info->nodesize - 1); 612371cdc07SQu Wenruo ASSERT(PagePrivate(page)); 613371cdc07SQu Wenruo 614371cdc07SQu Wenruo eb = find_extent_buffer(fs_info, start); 615371cdc07SQu Wenruo /* 616371cdc07SQu Wenruo * When we are reading one tree block, eb must have been inserted into 617371cdc07SQu Wenruo * the radix tree. If not, something is wrong. 618371cdc07SQu Wenruo */ 619371cdc07SQu Wenruo ASSERT(eb); 620371cdc07SQu Wenruo 621371cdc07SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 622371cdc07SQu Wenruo /* Subpage read must finish in page read */ 623371cdc07SQu Wenruo ASSERT(reads_done); 624371cdc07SQu Wenruo 625371cdc07SQu Wenruo eb->read_mirror = mirror; 626371cdc07SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 627371cdc07SQu Wenruo ret = -EIO; 628371cdc07SQu Wenruo goto err; 629371cdc07SQu Wenruo } 630947a6299SQu Wenruo ret = validate_extent_buffer(eb, check); 631371cdc07SQu Wenruo if (ret < 0) 632371cdc07SQu Wenruo goto err; 633371cdc07SQu Wenruo 634371cdc07SQu Wenruo set_extent_buffer_uptodate(eb); 635371cdc07SQu Wenruo 636371cdc07SQu Wenruo free_extent_buffer(eb); 637371cdc07SQu Wenruo return ret; 638371cdc07SQu Wenruo err: 639371cdc07SQu Wenruo /* 640371cdc07SQu Wenruo * end_bio_extent_readpage decrements io_pages in case of error, 641371cdc07SQu Wenruo * make sure it has something to decrement. 642371cdc07SQu Wenruo */ 643371cdc07SQu Wenruo atomic_inc(&eb->io_pages); 644371cdc07SQu Wenruo clear_extent_buffer_uptodate(eb); 645371cdc07SQu Wenruo free_extent_buffer(eb); 646371cdc07SQu Wenruo return ret; 647371cdc07SQu Wenruo } 648371cdc07SQu Wenruo 649c3a3b19bSQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, 65077bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 65177bf40a2SQu Wenruo int mirror) 65277bf40a2SQu Wenruo { 65377bf40a2SQu Wenruo struct extent_buffer *eb; 65477bf40a2SQu Wenruo int ret = 0; 65577bf40a2SQu Wenruo int reads_done; 65677bf40a2SQu Wenruo 65777bf40a2SQu Wenruo ASSERT(page->private); 658371cdc07SQu Wenruo 659fbca46ebSQu Wenruo if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) 660947a6299SQu Wenruo return validate_subpage_buffer(page, start, end, mirror, 661947a6299SQu Wenruo &bbio->parent_check); 662371cdc07SQu Wenruo 66377bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 66477bf40a2SQu Wenruo 66577bf40a2SQu Wenruo /* 66677bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 66777bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 66877bf40a2SQu Wenruo */ 66977bf40a2SQu Wenruo atomic_inc(&eb->refs); 67077bf40a2SQu Wenruo 67177bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 67277bf40a2SQu Wenruo if (!reads_done) 67377bf40a2SQu Wenruo goto err; 67477bf40a2SQu Wenruo 67577bf40a2SQu Wenruo eb->read_mirror = mirror; 67677bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 67777bf40a2SQu Wenruo ret = -EIO; 67877bf40a2SQu Wenruo goto err; 67977bf40a2SQu Wenruo } 680947a6299SQu Wenruo ret = validate_extent_buffer(eb, &bbio->parent_check); 681ce9adaa5SChris Mason err: 68253b381b3SDavid Woodhouse if (ret) { 68353b381b3SDavid Woodhouse /* 68453b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 68553b381b3SDavid Woodhouse * again, we have to make sure it has something 68653b381b3SDavid Woodhouse * to decrement 68753b381b3SDavid Woodhouse */ 68853b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6890b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 69053b381b3SDavid Woodhouse } 6910b32f4bbSJosef Bacik free_extent_buffer(eb); 69277bf40a2SQu Wenruo 693f188591eSChris Mason return ret; 694ce9adaa5SChris Mason } 695ce9adaa5SChris Mason 6964a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 6974a69a410SChris Mason { 6984a69a410SChris Mason struct async_submit_bio *async; 6994e4cbee9SChristoph Hellwig blk_status_t ret; 7004a69a410SChris Mason 7014a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 702ab2072b2SDavid Sterba switch (async->submit_cmd) { 703ab2072b2SDavid Sterba case WQ_SUBMIT_METADATA: 704ad65ecf3SDavid Sterba ret = btree_submit_bio_start(async->bio); 705ab2072b2SDavid Sterba break; 706ab2072b2SDavid Sterba case WQ_SUBMIT_DATA: 707882681acSDavid Sterba ret = btrfs_submit_bio_start(async->inode, async->bio); 708ab2072b2SDavid Sterba break; 709ab2072b2SDavid Sterba case WQ_SUBMIT_DATA_DIO: 710bfa17066SDavid Sterba ret = btrfs_submit_bio_start_direct_io(async->inode, 711da67daabSDavid Sterba async->bio, async->dio_file_offset); 712ab2072b2SDavid Sterba break; 713ab2072b2SDavid Sterba } 71479787eaaSJeff Mahoney if (ret) 7154e4cbee9SChristoph Hellwig async->status = ret; 7164a69a410SChris Mason } 7174a69a410SChris Mason 71806ea01b1SDavid Sterba /* 71906ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 72006ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 72106ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 72206ea01b1SDavid Sterba * 72306ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 72406ea01b1SDavid Sterba * inserted into the tree. 72506ea01b1SDavid Sterba */ 7264a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7278b712842SChris Mason { 728917f32a2SChristoph Hellwig struct async_submit_bio *async = 729917f32a2SChristoph Hellwig container_of(work, struct async_submit_bio, work); 730da67daabSDavid Sterba struct btrfs_inode *inode = async->inode; 731917f32a2SChristoph Hellwig struct btrfs_bio *bbio = btrfs_bio(async->bio); 7324854ddd0SChris Mason 733bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 7344e4cbee9SChristoph Hellwig if (async->status) { 735917f32a2SChristoph Hellwig btrfs_bio_end_io(bbio, async->status); 73679787eaaSJeff Mahoney return; 73779787eaaSJeff Mahoney } 73879787eaaSJeff Mahoney 739ec39f769SChris Mason /* 740ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 741ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 742ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 743ec39f769SChris Mason */ 744ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 745da67daabSDavid Sterba btrfs_submit_bio(inode->root->fs_info, async->bio, async->mirror_num); 7464a69a410SChris Mason } 7474a69a410SChris Mason 7484a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7494a69a410SChris Mason { 7504a69a410SChris Mason struct async_submit_bio *async; 7514a69a410SChris Mason 7524a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7538b712842SChris Mason kfree(async); 7548b712842SChris Mason } 7558b712842SChris Mason 756ea1f0cedSChristoph Hellwig /* 757ea1f0cedSChristoph Hellwig * Submit bio to an async queue. 758ea1f0cedSChristoph Hellwig * 759ea1f0cedSChristoph Hellwig * Retrun: 760ea1f0cedSChristoph Hellwig * - true if the work has been succesfuly submitted 761ea1f0cedSChristoph Hellwig * - false in case of error 762ea1f0cedSChristoph Hellwig */ 7635fcdadc2SDavid Sterba bool btrfs_wq_submit_bio(struct btrfs_inode *inode, struct bio *bio, int mirror_num, 764ab2072b2SDavid Sterba u64 dio_file_offset, enum btrfs_wq_submit_cmd cmd) 76544b8bd7eSChris Mason { 7665fcdadc2SDavid Sterba struct btrfs_fs_info *fs_info = inode->root->fs_info; 76744b8bd7eSChris Mason struct async_submit_bio *async; 76844b8bd7eSChris Mason 76944b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 77044b8bd7eSChris Mason if (!async) 771ea1f0cedSChristoph Hellwig return false; 77244b8bd7eSChris Mason 7735fcdadc2SDavid Sterba async->inode = inode; 77444b8bd7eSChris Mason async->bio = bio; 77544b8bd7eSChris Mason async->mirror_num = mirror_num; 776ab2072b2SDavid Sterba async->submit_cmd = cmd; 7774a69a410SChris Mason 778a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 779a0cac0ecSOmar Sandoval run_one_async_free); 7804a69a410SChris Mason 7811941b64bSQu Wenruo async->dio_file_offset = dio_file_offset; 7828c8bee1dSChris Mason 7834e4cbee9SChristoph Hellwig async->status = 0; 78479787eaaSJeff Mahoney 78567f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 786a31b4a43SChristoph Hellwig btrfs_queue_work(fs_info->hipri_workers, &async->work); 787a31b4a43SChristoph Hellwig else 7885cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 789ea1f0cedSChristoph Hellwig return true; 79044b8bd7eSChris Mason } 79144b8bd7eSChris Mason 7924e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 793ce3ed71aSChris Mason { 7942c30c71bSKent Overstreet struct bio_vec *bvec; 795ce3ed71aSChris Mason struct btrfs_root *root; 7962b070cfeSChristoph Hellwig int ret = 0; 7976dc4f100SMing Lei struct bvec_iter_all iter_all; 798ce3ed71aSChris Mason 799c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 8002b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 801ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 802ac303b69SQu Wenruo ret = csum_dirty_buffer(root->fs_info, bvec); 80379787eaaSJeff Mahoney if (ret) 80479787eaaSJeff Mahoney break; 805ce3ed71aSChris Mason } 8062c30c71bSKent Overstreet 8074e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 808ce3ed71aSChris Mason } 809ce3ed71aSChris Mason 810ad65ecf3SDavid Sterba blk_status_t btree_submit_bio_start(struct bio *bio) 81122c59948SChris Mason { 8128b712842SChris Mason /* 8138b712842SChris Mason * when we're called for a write, we're already in the async 8141a722d8fSChristoph Hellwig * submission context. Just jump into btrfs_submit_bio. 8158b712842SChris Mason */ 81679787eaaSJeff Mahoney return btree_csum_one_bio(bio); 81722c59948SChris Mason } 81822c59948SChris Mason 819f4dcfb30SJohannes Thumshirn static bool should_async_write(struct btrfs_fs_info *fs_info, 8209b4e675aSDavid Sterba struct btrfs_inode *bi) 821de0022b9SJosef Bacik { 8224eef29efSNaohiro Aota if (btrfs_is_zoned(fs_info)) 823f4dcfb30SJohannes Thumshirn return false; 8246300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 825f4dcfb30SJohannes Thumshirn return false; 8269b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 827f4dcfb30SJohannes Thumshirn return false; 828f4dcfb30SJohannes Thumshirn return true; 829de0022b9SJosef Bacik } 830de0022b9SJosef Bacik 831644094fdSDavid Sterba void btrfs_submit_metadata_bio(struct btrfs_inode *inode, struct bio *bio, int mirror_num) 83244b8bd7eSChris Mason { 833644094fdSDavid Sterba struct btrfs_fs_info *fs_info = inode->root->fs_info; 834917f32a2SChristoph Hellwig struct btrfs_bio *bbio = btrfs_bio(bio); 8354e4cbee9SChristoph Hellwig blk_status_t ret; 836cad321adSChris Mason 83708a6f464SChristoph Hellwig bio->bi_opf |= REQ_META; 838947a6299SQu Wenruo bbio->is_metadata = 1; 83908a6f464SChristoph Hellwig 840cfe94440SNaohiro Aota if (btrfs_op(bio) != BTRFS_MAP_WRITE) { 8411a722d8fSChristoph Hellwig btrfs_submit_bio(fs_info, bio, mirror_num); 8421a722d8fSChristoph Hellwig return; 84344b8bd7eSChris Mason } 84444b8bd7eSChris Mason 845ea1f0cedSChristoph Hellwig /* 846ea1f0cedSChristoph Hellwig * Kthread helpers are used to submit writes so that checksumming can 847ea1f0cedSChristoph Hellwig * happen in parallel across all CPUs. 848ea1f0cedSChristoph Hellwig */ 849644094fdSDavid Sterba if (should_async_write(fs_info, inode) && 850644094fdSDavid Sterba btrfs_wq_submit_bio(inode, bio, mirror_num, 0, WQ_SUBMIT_METADATA)) 851ea1f0cedSChristoph Hellwig return; 852ea1f0cedSChristoph Hellwig 853ea1f0cedSChristoph Hellwig ret = btree_csum_one_bio(bio); 85494d9e11bSChristoph Hellwig if (ret) { 855917f32a2SChristoph Hellwig btrfs_bio_end_io(bbio, ret); 856ea1f0cedSChristoph Hellwig return; 85794d9e11bSChristoph Hellwig } 858ea1f0cedSChristoph Hellwig 859ea1f0cedSChristoph Hellwig btrfs_submit_bio(fs_info, bio, mirror_num); 86061891923SStefan Behrens } 86161891923SStefan Behrens 8623dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 8638958b551SMatthew Wilcox (Oracle) static int btree_migrate_folio(struct address_space *mapping, 8648958b551SMatthew Wilcox (Oracle) struct folio *dst, struct folio *src, enum migrate_mode mode) 865784b4e29SChris Mason { 866784b4e29SChris Mason /* 867784b4e29SChris Mason * we can't safely write a btree page from here, 868784b4e29SChris Mason * we haven't done the locking hook 869784b4e29SChris Mason */ 8708958b551SMatthew Wilcox (Oracle) if (folio_test_dirty(src)) 871784b4e29SChris Mason return -EAGAIN; 872784b4e29SChris Mason /* 873784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 874784b4e29SChris Mason * We must have no buffers or drop them. 875784b4e29SChris Mason */ 8768958b551SMatthew Wilcox (Oracle) if (folio_get_private(src) && 8778958b551SMatthew Wilcox (Oracle) !filemap_release_folio(src, GFP_KERNEL)) 878784b4e29SChris Mason return -EAGAIN; 87954184650SMatthew Wilcox (Oracle) return migrate_folio(mapping, dst, src, mode); 880784b4e29SChris Mason } 8818958b551SMatthew Wilcox (Oracle) #else 8828958b551SMatthew Wilcox (Oracle) #define btree_migrate_folio NULL 8833dd1462eSJan Beulich #endif 884784b4e29SChris Mason 8850da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 8860da5468fSChris Mason struct writeback_control *wbc) 8870da5468fSChris Mason { 888e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 889e2d84521SMiao Xie int ret; 890e2d84521SMiao Xie 891d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 892448d640bSChris Mason 893448d640bSChris Mason if (wbc->for_kupdate) 894448d640bSChris Mason return 0; 895448d640bSChris Mason 896e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 897b9473439SChris Mason /* this is a bit racy, but that's ok */ 898d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 899d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 900d814a491SEthan Lien fs_info->dirty_metadata_batch); 901e2d84521SMiao Xie if (ret < 0) 902793955bcSChris Mason return 0; 903793955bcSChris Mason } 9040b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9050da5468fSChris Mason } 9060da5468fSChris Mason 907f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 9085f39d397SChris Mason { 909f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 910f913cff3SMatthew Wilcox (Oracle) return false; 9110c4e538bSDavid Sterba 912f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 913d98237b3SChris Mason } 914d98237b3SChris Mason 915895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 916895586ebSMatthew Wilcox (Oracle) size_t length) 917d98237b3SChris Mason { 918d1310b2eSChris Mason struct extent_io_tree *tree; 919895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 920895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 921f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 922895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 923895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 924895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 925895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 926895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 9279ad6b7bcSChris Mason } 928d98237b3SChris Mason } 929d98237b3SChris Mason 930bb146eb2SJosef Bacik #ifdef DEBUG 9310079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 9320079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 9330079c3b1SMatthew Wilcox (Oracle) { 9340079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 935139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 9360b32f4bbSJosef Bacik struct extent_buffer *eb; 937139e8cd3SQu Wenruo int cur_bit = 0; 9380079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 9390b32f4bbSJosef Bacik 940139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 9410079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 9420b32f4bbSJosef Bacik BUG_ON(!eb); 9430b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 9440b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 94549d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 9460079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 947139e8cd3SQu Wenruo } 9480079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 949139e8cd3SQu Wenruo 950139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 951139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 952139e8cd3SQu Wenruo unsigned long flags; 953139e8cd3SQu Wenruo u64 cur; 954139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 955139e8cd3SQu Wenruo 956139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 957139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 958139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 959139e8cd3SQu Wenruo cur_bit++; 960139e8cd3SQu Wenruo continue; 961139e8cd3SQu Wenruo } 962139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 963139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 964139e8cd3SQu Wenruo 965139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 966139e8cd3SQu Wenruo ASSERT(eb); 967139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 968139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 96949d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 970139e8cd3SQu Wenruo free_extent_buffer(eb); 971139e8cd3SQu Wenruo 972139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 973139e8cd3SQu Wenruo } 9740079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 9750b32f4bbSJosef Bacik } 9760079c3b1SMatthew Wilcox (Oracle) #else 9770079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 9780079c3b1SMatthew Wilcox (Oracle) #endif 9790b32f4bbSJosef Bacik 9807f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 9810da5468fSChris Mason .writepages = btree_writepages, 982f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 983895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 9848958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 9850079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 986d98237b3SChris Mason }; 987123abc88SChris Mason 9882ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 9892ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 9903fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 9913fbaf258SJosef Bacik int level) 9920999df54SChris Mason { 9930b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 9940b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 9953fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 9960999df54SChris Mason } 9970999df54SChris Mason 998581c1760SQu Wenruo /* 999581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 1000581c1760SQu Wenruo * verification. 1001581c1760SQu Wenruo * 1002789d6a3aSQu Wenruo * @check: expected tree parentness check, see comments of the 1003789d6a3aSQu Wenruo * structure for details. 1004581c1760SQu Wenruo */ 10052ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 1006789d6a3aSQu Wenruo struct btrfs_tree_parent_check *check) 1007e20d96d6SChris Mason { 10085f39d397SChris Mason struct extent_buffer *buf = NULL; 100919c00ddcSChris Mason int ret; 101019c00ddcSChris Mason 1011789d6a3aSQu Wenruo ASSERT(check); 1012789d6a3aSQu Wenruo 1013789d6a3aSQu Wenruo buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, 1014789d6a3aSQu Wenruo check->level); 1015c871b0f2SLiu Bo if (IS_ERR(buf)) 1016c871b0f2SLiu Bo return buf; 1017e4204dedSChris Mason 1018789d6a3aSQu Wenruo ret = btrfs_read_extent_buffer(buf, check); 10190f0fe8f7SFilipe David Borba Manana if (ret) { 1020537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 102164c043deSLiu Bo return ERR_PTR(ret); 10220f0fe8f7SFilipe David Borba Manana } 1023789d6a3aSQu Wenruo if (btrfs_check_eb_owner(buf, check->owner_root)) { 102488c602abSQu Wenruo free_extent_buffer_stale(buf); 102588c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 102688c602abSQu Wenruo } 10275f39d397SChris Mason return buf; 1028ce9adaa5SChris Mason 1029eb60ceacSChris Mason } 1030eb60ceacSChris Mason 10316a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 1032ed2ff2cbSChris Mason { 10336a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 103455c69072SChris Mason if (btrfs_header_generation(buf) == 1035e2d84521SMiao Xie fs_info->running_transaction->transid) { 103649d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 1037b4ce94deSChris Mason 1038b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1039104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1040e2d84521SMiao Xie -buf->len, 1041e2d84521SMiao Xie fs_info->dirty_metadata_batch); 10420b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1043925baeddSChris Mason } 10445f39d397SChris Mason } 1045ed7b63ebSJosef Bacik } 10465f39d397SChris Mason 1047da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1048e20d96d6SChris Mason u64 objectid) 1049d97e63b6SChris Mason { 10507c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 10512e608bd1SJosef Bacik 10522e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 10532e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 10542e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 105596dfcb46SJosef Bacik root->fs_info = fs_info; 10562e608bd1SJosef Bacik root->root_key.objectid = objectid; 1057cfaa7295SChris Mason root->node = NULL; 1058a28ec197SChris Mason root->commit_root = NULL; 105927cdeb70SMiao Xie root->state = 0; 1060abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 10610b86a832SChris Mason 10620f7d52f4SChris Mason root->last_trans = 0; 10636b8fad57SNikolay Borisov root->free_objectid = 0; 1064eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1065199c2a9cSMiao Xie root->nr_ordered_extents = 0; 10666bef4d31SEric Paris root->inode_tree = RB_ROOT; 1067088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 10682e608bd1SJosef Bacik 10692e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 10700b86a832SChris Mason 10710b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 10725d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1073eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1074eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1075199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1076199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1077d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 10782ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 10792ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 10805d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1081eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1082199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1083f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 10842ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 10852ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 10868287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1087a2135011SChris Mason mutex_init(&root->objectid_mutex); 1088e02119d5SChris Mason mutex_init(&root->log_mutex); 108931f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1090573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1091c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 10927237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 10937237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 10947237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 10958b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 10968b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 10977237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 10987237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 10997237f183SYan Zheng atomic_set(&root->log_writers, 0); 11002ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 11010700cea7SElena Reshetova refcount_set(&root->refs, 1); 11028ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1103eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 11047237f183SYan Zheng root->log_transid = 0; 1105d1433debSMiao Xie root->log_transid_committed = -1; 1106257c62e1SChris Mason root->last_log_commit = 0; 11072e608bd1SJosef Bacik root->anon_dev = 0; 1108e289f03eSFilipe Manana if (!dummy) { 110943eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 111035da5a7eSDavid Sterba IO_TREE_ROOT_DIRTY_LOG_PAGES); 1111e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 111235da5a7eSDavid Sterba IO_TREE_LOG_CSUM_RANGE); 1113e289f03eSFilipe Manana } 1114017e5369SChris Mason 11155f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1116370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1117bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1118bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1119fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1120bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1121fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1122bd647ce3SJosef Bacik #endif 11233768f368SChris Mason } 11243768f368SChris Mason 112574e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 112696dfcb46SJosef Bacik u64 objectid, gfp_t flags) 11276f07e42eSAl Viro { 112874e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 11296f07e42eSAl Viro if (root) 113096dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 11316f07e42eSAl Viro return root; 11326f07e42eSAl Viro } 11336f07e42eSAl Viro 113406ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 113506ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1136da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 113706ea65a3SJosef Bacik { 113806ea65a3SJosef Bacik struct btrfs_root *root; 113906ea65a3SJosef Bacik 11407c0260eeSJeff Mahoney if (!fs_info) 11417c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 11427c0260eeSJeff Mahoney 114396dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 114406ea65a3SJosef Bacik if (!root) 114506ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1146da17066cSJeff Mahoney 1147b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1148faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 114906ea65a3SJosef Bacik 115006ea65a3SJosef Bacik return root; 115106ea65a3SJosef Bacik } 115206ea65a3SJosef Bacik #endif 115306ea65a3SJosef Bacik 1154abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 1155abed4aaaSJosef Bacik { 1156abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 1157abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 1158abed4aaaSJosef Bacik 1159abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 1160abed4aaaSJosef Bacik } 1161abed4aaaSJosef Bacik 1162abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 1163abed4aaaSJosef Bacik { 1164abed4aaaSJosef Bacik const struct btrfs_key *key = k; 1165abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 1166abed4aaaSJosef Bacik 1167abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 1168abed4aaaSJosef Bacik } 1169abed4aaaSJosef Bacik 1170abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 1171abed4aaaSJosef Bacik { 1172abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1173abed4aaaSJosef Bacik struct rb_node *tmp; 1174abed4aaaSJosef Bacik 1175abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1176abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 1177abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1178abed4aaaSJosef Bacik ASSERT(!tmp); 1179abed4aaaSJosef Bacik 1180abed4aaaSJosef Bacik return tmp ? -EEXIST : 0; 1181abed4aaaSJosef Bacik } 1182abed4aaaSJosef Bacik 1183abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 1184abed4aaaSJosef Bacik { 1185abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1186abed4aaaSJosef Bacik 1187abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1188abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1189abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1190abed4aaaSJosef Bacik } 1191abed4aaaSJosef Bacik 1192abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 1193abed4aaaSJosef Bacik struct btrfs_key *key) 1194abed4aaaSJosef Bacik { 1195abed4aaaSJosef Bacik struct rb_node *node; 1196abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 1197abed4aaaSJosef Bacik 1198abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 1199abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 1200abed4aaaSJosef Bacik if (node) 1201abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 1202abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 1203abed4aaaSJosef Bacik 1204abed4aaaSJosef Bacik return root; 1205abed4aaaSJosef Bacik } 1206abed4aaaSJosef Bacik 1207f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 1208f7238e50SJosef Bacik { 1209f7238e50SJosef Bacik struct btrfs_block_group *block_group; 1210f7238e50SJosef Bacik u64 ret; 1211f7238e50SJosef Bacik 1212f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 1213f7238e50SJosef Bacik return 0; 1214f7238e50SJosef Bacik 1215f7238e50SJosef Bacik if (bytenr) 1216f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 1217f7238e50SJosef Bacik else 1218f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 1219f7238e50SJosef Bacik ASSERT(block_group); 1220f7238e50SJosef Bacik if (!block_group) 1221f7238e50SJosef Bacik return 0; 1222f7238e50SJosef Bacik ret = block_group->global_root_id; 1223f7238e50SJosef Bacik btrfs_put_block_group(block_group); 1224f7238e50SJosef Bacik 1225f7238e50SJosef Bacik return ret; 1226f7238e50SJosef Bacik } 1227f7238e50SJosef Bacik 1228abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1229abed4aaaSJosef Bacik { 1230abed4aaaSJosef Bacik struct btrfs_key key = { 1231abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 1232abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1233f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1234abed4aaaSJosef Bacik }; 1235abed4aaaSJosef Bacik 1236abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1237abed4aaaSJosef Bacik } 1238abed4aaaSJosef Bacik 1239abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1240abed4aaaSJosef Bacik { 1241abed4aaaSJosef Bacik struct btrfs_key key = { 1242abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 1243abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1244f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1245abed4aaaSJosef Bacik }; 1246abed4aaaSJosef Bacik 1247abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1248abed4aaaSJosef Bacik } 1249abed4aaaSJosef Bacik 125051129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 125151129b33SJosef Bacik { 125251129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 125351129b33SJosef Bacik return fs_info->block_group_root; 125451129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 125551129b33SJosef Bacik } 125651129b33SJosef Bacik 125720897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 125820897f5cSArne Jansen u64 objectid) 125920897f5cSArne Jansen { 12609b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 126120897f5cSArne Jansen struct extent_buffer *leaf; 126220897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 126320897f5cSArne Jansen struct btrfs_root *root; 126420897f5cSArne Jansen struct btrfs_key key; 1265b89f6d1fSFilipe Manana unsigned int nofs_flag; 126620897f5cSArne Jansen int ret = 0; 126720897f5cSArne Jansen 1268b89f6d1fSFilipe Manana /* 1269b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1270b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1271b89f6d1fSFilipe Manana */ 1272b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 127396dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1274b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 127520897f5cSArne Jansen if (!root) 127620897f5cSArne Jansen return ERR_PTR(-ENOMEM); 127720897f5cSArne Jansen 127820897f5cSArne Jansen root->root_key.objectid = objectid; 127920897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 128020897f5cSArne Jansen root->root_key.offset = 0; 128120897f5cSArne Jansen 12829631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 12839631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 128420897f5cSArne Jansen if (IS_ERR(leaf)) { 128520897f5cSArne Jansen ret = PTR_ERR(leaf); 12861dd05682STsutomu Itoh leaf = NULL; 1287c1b07854SPeng Hao goto fail; 128820897f5cSArne Jansen } 128920897f5cSArne Jansen 129020897f5cSArne Jansen root->node = leaf; 129120897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 129220897f5cSArne Jansen 129320897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 129427cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 129520897f5cSArne Jansen 1296f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1297f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 129820897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 129920897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 130020897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 130120897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 130220897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 130320897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 130420897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 130533d85fdaSQu Wenruo if (is_fstree(objectid)) 1306807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1307807fc790SAndy Shevchenko else 1308807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1309c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 131020897f5cSArne Jansen 13118a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 13128a6a87cdSBoris Burkov 131320897f5cSArne Jansen key.objectid = objectid; 131420897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 131520897f5cSArne Jansen key.offset = 0; 131620897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 131720897f5cSArne Jansen if (ret) 131820897f5cSArne Jansen goto fail; 131920897f5cSArne Jansen 132020897f5cSArne Jansen return root; 13211dd05682STsutomu Itoh 13228a6a87cdSBoris Burkov fail: 132300246528SJosef Bacik btrfs_put_root(root); 13241dd05682STsutomu Itoh 13251dd05682STsutomu Itoh return ERR_PTR(ret); 132620897f5cSArne Jansen } 132720897f5cSArne Jansen 13287237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1329e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13300f7d52f4SChris Mason { 13310f7d52f4SChris Mason struct btrfs_root *root; 1332e02119d5SChris Mason 133396dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1334e02119d5SChris Mason if (!root) 13357237f183SYan Zheng return ERR_PTR(-ENOMEM); 1336e02119d5SChris Mason 1337e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1338e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1339e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 134027cdeb70SMiao Xie 13416ab6ebb7SNaohiro Aota return root; 13426ab6ebb7SNaohiro Aota } 13436ab6ebb7SNaohiro Aota 13446ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 13456ab6ebb7SNaohiro Aota struct btrfs_root *root) 13466ab6ebb7SNaohiro Aota { 13476ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 13486ab6ebb7SNaohiro Aota 13497237f183SYan Zheng /* 135092a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 135127cdeb70SMiao Xie * 135292a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 135392a7cc42SQu Wenruo * and they go away before a real commit is actually done. 135492a7cc42SQu Wenruo * 135592a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 135692a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 13577237f183SYan Zheng */ 1358e02119d5SChris Mason 13594d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 13609631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 13616ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 13626ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 1363e02119d5SChris Mason 13647237f183SYan Zheng root->node = leaf; 1365e02119d5SChris Mason 1366e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1367e02119d5SChris Mason btrfs_tree_unlock(root->node); 13686ab6ebb7SNaohiro Aota 13696ab6ebb7SNaohiro Aota return 0; 13707237f183SYan Zheng } 13717237f183SYan Zheng 13727237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13737237f183SYan Zheng struct btrfs_fs_info *fs_info) 13747237f183SYan Zheng { 13757237f183SYan Zheng struct btrfs_root *log_root; 13767237f183SYan Zheng 13777237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13787237f183SYan Zheng if (IS_ERR(log_root)) 13797237f183SYan Zheng return PTR_ERR(log_root); 13806ab6ebb7SNaohiro Aota 13813ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 13823ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 13833ddebf27SNaohiro Aota 13846ab6ebb7SNaohiro Aota if (ret) { 13856ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13866ab6ebb7SNaohiro Aota return ret; 13876ab6ebb7SNaohiro Aota } 13883ddebf27SNaohiro Aota } 13896ab6ebb7SNaohiro Aota 13907237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13917237f183SYan Zheng fs_info->log_root_tree = log_root; 13927237f183SYan Zheng return 0; 13937237f183SYan Zheng } 13947237f183SYan Zheng 13957237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13967237f183SYan Zheng struct btrfs_root *root) 13977237f183SYan Zheng { 13980b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13997237f183SYan Zheng struct btrfs_root *log_root; 14007237f183SYan Zheng struct btrfs_inode_item *inode_item; 14016ab6ebb7SNaohiro Aota int ret; 14027237f183SYan Zheng 14030b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 14047237f183SYan Zheng if (IS_ERR(log_root)) 14057237f183SYan Zheng return PTR_ERR(log_root); 14067237f183SYan Zheng 14076ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 14086ab6ebb7SNaohiro Aota if (ret) { 14096ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 14106ab6ebb7SNaohiro Aota return ret; 14116ab6ebb7SNaohiro Aota } 14126ab6ebb7SNaohiro Aota 14137237f183SYan Zheng log_root->last_trans = trans->transid; 14147237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14157237f183SYan Zheng 14167237f183SYan Zheng inode_item = &log_root->root_item.inode; 14173cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14183cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14193cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1420da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 14210b246afaSJeff Mahoney fs_info->nodesize); 14223cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14237237f183SYan Zheng 14245d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14257237f183SYan Zheng 14267237f183SYan Zheng WARN_ON(root->log_root); 14277237f183SYan Zheng root->log_root = log_root; 14287237f183SYan Zheng root->log_transid = 0; 1429d1433debSMiao Xie root->log_transid_committed = -1; 1430257c62e1SChris Mason root->last_log_commit = 0; 1431e02119d5SChris Mason return 0; 1432e02119d5SChris Mason } 1433e02119d5SChris Mason 143449d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 143549d11beaSJosef Bacik struct btrfs_path *path, 1436cb517eabSMiao Xie struct btrfs_key *key) 1437e02119d5SChris Mason { 1438e02119d5SChris Mason struct btrfs_root *root; 1439789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 1440e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 144184234f3aSYan Zheng u64 generation; 1442cb517eabSMiao Xie int ret; 1443581c1760SQu Wenruo int level; 1444cb517eabSMiao Xie 144596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 144649d11beaSJosef Bacik if (!root) 144749d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 14480f7d52f4SChris Mason 1449cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1450cb517eabSMiao Xie &root->root_item, &root->root_key); 14510f7d52f4SChris Mason if (ret) { 145213a8a7c8SYan, Zheng if (ret > 0) 145313a8a7c8SYan, Zheng ret = -ENOENT; 145449d11beaSJosef Bacik goto fail; 14550f7d52f4SChris Mason } 145613a8a7c8SYan, Zheng 145784234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1458581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 1459789d6a3aSQu Wenruo check.level = level; 1460789d6a3aSQu Wenruo check.transid = generation; 1461789d6a3aSQu Wenruo check.owner_root = key->objectid; 1462789d6a3aSQu Wenruo root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), 1463789d6a3aSQu Wenruo &check); 146464c043deSLiu Bo if (IS_ERR(root->node)) { 146564c043deSLiu Bo ret = PTR_ERR(root->node); 14668c38938cSJosef Bacik root->node = NULL; 146749d11beaSJosef Bacik goto fail; 14684eb150d6SQu Wenruo } 14694eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1470cb517eabSMiao Xie ret = -EIO; 147149d11beaSJosef Bacik goto fail; 1472416bc658SJosef Bacik } 147388c602abSQu Wenruo 147488c602abSQu Wenruo /* 147588c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 147688c602abSQu Wenruo * match its root node owner 147788c602abSQu Wenruo */ 147888c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 147988c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 148088c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 148188c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 148288c602abSQu Wenruo btrfs_crit(fs_info, 148388c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 148488c602abSQu Wenruo root->root_key.objectid, root->node->start, 148588c602abSQu Wenruo btrfs_header_owner(root->node), 148688c602abSQu Wenruo root->root_key.objectid); 148788c602abSQu Wenruo ret = -EUCLEAN; 148888c602abSQu Wenruo goto fail; 148988c602abSQu Wenruo } 14905d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1491cb517eabSMiao Xie return root; 149249d11beaSJosef Bacik fail: 149300246528SJosef Bacik btrfs_put_root(root); 149449d11beaSJosef Bacik return ERR_PTR(ret); 149549d11beaSJosef Bacik } 149649d11beaSJosef Bacik 149749d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 149849d11beaSJosef Bacik struct btrfs_key *key) 149949d11beaSJosef Bacik { 150049d11beaSJosef Bacik struct btrfs_root *root; 150149d11beaSJosef Bacik struct btrfs_path *path; 150249d11beaSJosef Bacik 150349d11beaSJosef Bacik path = btrfs_alloc_path(); 150449d11beaSJosef Bacik if (!path) 150549d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 150649d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 150749d11beaSJosef Bacik btrfs_free_path(path); 150849d11beaSJosef Bacik 150949d11beaSJosef Bacik return root; 1510cb517eabSMiao Xie } 1511cb517eabSMiao Xie 15122dfb1e43SQu Wenruo /* 15132dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 15142dfb1e43SQu Wenruo * 15152dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 15162dfb1e43SQu Wenruo */ 15172dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1518cb517eabSMiao Xie { 1519cb517eabSMiao Xie int ret; 1520dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1521cb517eabSMiao Xie 1522dcc3eb96SNikolay Borisov /* 1523dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1524dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1525dcc3eb96SNikolay Borisov */ 1526dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1527dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1528dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1529dcc3eb96SNikolay Borisov if (ret) 15308257b2dcSMiao Xie goto fail; 15318257b2dcSMiao Xie 1532aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 153337f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 153492a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1535f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1536f39e4571SJosef Bacik } 1537f39e4571SJosef Bacik 1538851fd730SQu Wenruo /* 1539851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1540851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1541851fd730SQu Wenruo */ 1542851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1543851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 15442dfb1e43SQu Wenruo if (!anon_dev) { 1545cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1546cb517eabSMiao Xie if (ret) 1547876d2cf1SLiu Bo goto fail; 15482dfb1e43SQu Wenruo } else { 15492dfb1e43SQu Wenruo root->anon_dev = anon_dev; 15502dfb1e43SQu Wenruo } 1551851fd730SQu Wenruo } 1552f32e48e9SChandan Rajendra 1553f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1554453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1555f32e48e9SChandan Rajendra if (ret) { 1556f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1557876d2cf1SLiu Bo goto fail; 1558f32e48e9SChandan Rajendra } 1559f32e48e9SChandan Rajendra 15606b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1561f32e48e9SChandan Rajendra 1562f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1563f32e48e9SChandan Rajendra 1564cb517eabSMiao Xie return 0; 1565cb517eabSMiao Xie fail: 156684db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1567cb517eabSMiao Xie return ret; 1568cb517eabSMiao Xie } 1569cb517eabSMiao Xie 1570a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1571cb517eabSMiao Xie u64 root_id) 1572cb517eabSMiao Xie { 1573cb517eabSMiao Xie struct btrfs_root *root; 1574cb517eabSMiao Xie 1575fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1576fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1577fc7cbcd4SDavid Sterba (unsigned long)root_id); 1578bc44d7c4SJosef Bacik if (root) 157900246528SJosef Bacik root = btrfs_grab_root(root); 1580fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1581cb517eabSMiao Xie return root; 1582cb517eabSMiao Xie } 1583cb517eabSMiao Xie 158449d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 158549d11beaSJosef Bacik u64 objectid) 158649d11beaSJosef Bacik { 1587abed4aaaSJosef Bacik struct btrfs_key key = { 1588abed4aaaSJosef Bacik .objectid = objectid, 1589abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1590abed4aaaSJosef Bacik .offset = 0, 1591abed4aaaSJosef Bacik }; 1592abed4aaaSJosef Bacik 159349d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 159449d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 159549d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 1596abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 159749d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 159849d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 159949d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 160049d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 160149d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 1602abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 160349d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 160449d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 160549d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 160649d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 160749d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 160849d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 160914033b08SQu Wenruo if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) 161014033b08SQu Wenruo return btrfs_grab_root(fs_info->block_group_root) ? 161114033b08SQu Wenruo fs_info->block_group_root : ERR_PTR(-ENOENT); 1612abed4aaaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { 1613abed4aaaSJosef Bacik struct btrfs_root *root = btrfs_global_root(fs_info, &key); 1614abed4aaaSJosef Bacik 1615abed4aaaSJosef Bacik return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); 1616abed4aaaSJosef Bacik } 161749d11beaSJosef Bacik return NULL; 161849d11beaSJosef Bacik } 161949d11beaSJosef Bacik 1620cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1621cb517eabSMiao Xie struct btrfs_root *root) 1622cb517eabSMiao Xie { 1623cb517eabSMiao Xie int ret; 1624cb517eabSMiao Xie 1625fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1626fc7cbcd4SDavid Sterba if (ret) 1627fc7cbcd4SDavid Sterba return ret; 1628fc7cbcd4SDavid Sterba 1629fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1630fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1631fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1632fc7cbcd4SDavid Sterba root); 1633af01d2e5SJosef Bacik if (ret == 0) { 163400246528SJosef Bacik btrfs_grab_root(root); 1635fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1636af01d2e5SJosef Bacik } 1637fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1638fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1639cb517eabSMiao Xie 1640cb517eabSMiao Xie return ret; 1641cb517eabSMiao Xie } 1642cb517eabSMiao Xie 1643bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1644bd647ce3SJosef Bacik { 1645bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1646bd647ce3SJosef Bacik struct btrfs_root *root; 1647bd647ce3SJosef Bacik 1648bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1649457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1650457f1864SJosef Bacik 1651bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1652bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1653457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 165471008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1655bd647ce3SJosef Bacik refcount_read(&root->refs)); 1656bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 165700246528SJosef Bacik btrfs_put_root(root); 165800246528SJosef Bacik btrfs_put_root(root); 1659bd647ce3SJosef Bacik } 1660bd647ce3SJosef Bacik #endif 1661bd647ce3SJosef Bacik } 1662bd647ce3SJosef Bacik 1663abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1664abed4aaaSJosef Bacik { 1665abed4aaaSJosef Bacik struct btrfs_root *root; 1666abed4aaaSJosef Bacik struct rb_node *node; 1667abed4aaaSJosef Bacik 1668abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1669abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1670abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1671abed4aaaSJosef Bacik btrfs_put_root(root); 1672abed4aaaSJosef Bacik } 1673abed4aaaSJosef Bacik } 1674abed4aaaSJosef Bacik 16750d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 16760d4b0463SJosef Bacik { 1677141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1678141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 16795deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1680141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1681141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1682141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1683141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 16840d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 16850d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1686abed4aaaSJosef Bacik free_global_roots(fs_info); 168700246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 168800246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 168900246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 169000246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 169100246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 169200246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1693aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 16949c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1695bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 16963fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 16970d4b0463SJosef Bacik kfree(fs_info->super_copy); 16980d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 16998481dd80SQu Wenruo kfree(fs_info->subpage_info); 17000d4b0463SJosef Bacik kvfree(fs_info); 17010d4b0463SJosef Bacik } 17020d4b0463SJosef Bacik 17030d4b0463SJosef Bacik 17042dfb1e43SQu Wenruo /* 17052dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 17062dfb1e43SQu Wenruo * 17072dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 17082dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 17092dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 17102dfb1e43SQu Wenruo * 17112dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 17122dfb1e43SQu Wenruo * 17132dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 17142dfb1e43SQu Wenruo * same root objectid. 17152dfb1e43SQu Wenruo * 17162dfb1e43SQu Wenruo * @objectid: root id 17172dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 17182dfb1e43SQu Wenruo * pass 0 for new allocation. 17192dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 17202dfb1e43SQu Wenruo * for orphan roots 17212dfb1e43SQu Wenruo */ 17222dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 17232dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 17242dfb1e43SQu Wenruo bool check_ref) 17255eda7b5eSChris Mason { 17265eda7b5eSChris Mason struct btrfs_root *root; 1727381cf658SDavid Sterba struct btrfs_path *path; 17281d4c08e0SDavid Sterba struct btrfs_key key; 17295eda7b5eSChris Mason int ret; 17305eda7b5eSChris Mason 173149d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 173249d11beaSJosef Bacik if (root) 173349d11beaSJosef Bacik return root; 17344df27c4dSYan, Zheng again: 173556e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 173648475471SStefan Behrens if (root) { 17372dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 17382dfb1e43SQu Wenruo ASSERT(!anon_dev); 1739bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 174000246528SJosef Bacik btrfs_put_root(root); 174148475471SStefan Behrens return ERR_PTR(-ENOENT); 1742bc44d7c4SJosef Bacik } 17435eda7b5eSChris Mason return root; 174448475471SStefan Behrens } 17455eda7b5eSChris Mason 174656e9357aSDavid Sterba key.objectid = objectid; 174756e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 174856e9357aSDavid Sterba key.offset = (u64)-1; 174956e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 17505eda7b5eSChris Mason if (IS_ERR(root)) 17515eda7b5eSChris Mason return root; 17523394e160SChris Mason 1753c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1754d68fc57bSYan, Zheng ret = -ENOENT; 1755d68fc57bSYan, Zheng goto fail; 1756d68fc57bSYan, Zheng } 1757d68fc57bSYan, Zheng 17582dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1759cb517eabSMiao Xie if (ret) 1760cb517eabSMiao Xie goto fail; 1761cb517eabSMiao Xie 1762381cf658SDavid Sterba path = btrfs_alloc_path(); 1763381cf658SDavid Sterba if (!path) { 1764381cf658SDavid Sterba ret = -ENOMEM; 1765381cf658SDavid Sterba goto fail; 1766381cf658SDavid Sterba } 17671d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 17681d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 176956e9357aSDavid Sterba key.offset = objectid; 17701d4c08e0SDavid Sterba 17711d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1772381cf658SDavid Sterba btrfs_free_path(path); 1773d68fc57bSYan, Zheng if (ret < 0) 1774d68fc57bSYan, Zheng goto fail; 1775d68fc57bSYan, Zheng if (ret == 0) 177627cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1777d68fc57bSYan, Zheng 1778cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 17790f7d52f4SChris Mason if (ret) { 1780168a2f77SJia-Ju Bai if (ret == -EEXIST) { 178100246528SJosef Bacik btrfs_put_root(root); 17824df27c4dSYan, Zheng goto again; 1783168a2f77SJia-Ju Bai } 17844df27c4dSYan, Zheng goto fail; 17854df27c4dSYan, Zheng } 1786edbd8d4eSChris Mason return root; 17874df27c4dSYan, Zheng fail: 178833fab972SFilipe Manana /* 178933fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1790143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 179133fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 179233fab972SFilipe Manana * and once again by our caller. 179333fab972SFilipe Manana */ 179433fab972SFilipe Manana if (anon_dev) 179533fab972SFilipe Manana root->anon_dev = 0; 17968c38938cSJosef Bacik btrfs_put_root(root); 17974df27c4dSYan, Zheng return ERR_PTR(ret); 1798edbd8d4eSChris Mason } 1799edbd8d4eSChris Mason 18002dfb1e43SQu Wenruo /* 18012dfb1e43SQu Wenruo * Get in-memory reference of a root structure 18022dfb1e43SQu Wenruo * 18032dfb1e43SQu Wenruo * @objectid: tree objectid 18042dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 18052dfb1e43SQu Wenruo * one reference 18062dfb1e43SQu Wenruo */ 18072dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 18082dfb1e43SQu Wenruo u64 objectid, bool check_ref) 18092dfb1e43SQu Wenruo { 18102dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 18112dfb1e43SQu Wenruo } 18122dfb1e43SQu Wenruo 18132dfb1e43SQu Wenruo /* 18142dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 18152dfb1e43SQu Wenruo * the anonymous block device id 18162dfb1e43SQu Wenruo * 18172dfb1e43SQu Wenruo * @objectid: tree objectid 18182dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 18192dfb1e43SQu Wenruo * parameter value 18202dfb1e43SQu Wenruo */ 18212dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 18222dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 18232dfb1e43SQu Wenruo { 18242dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 18252dfb1e43SQu Wenruo } 18262dfb1e43SQu Wenruo 18278b712842SChris Mason /* 182849d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 182949d11beaSJosef Bacik * @fs_info: the fs_info 183049d11beaSJosef Bacik * @objectid: the objectid we need to lookup 183149d11beaSJosef Bacik * 183249d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 183349d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 183449d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 183549d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 183649d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 183749d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 183849d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 183949d11beaSJosef Bacik * backref code is finished using the root. 184049d11beaSJosef Bacik */ 184149d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 184249d11beaSJosef Bacik struct btrfs_path *path, 184349d11beaSJosef Bacik u64 objectid) 184449d11beaSJosef Bacik { 184549d11beaSJosef Bacik struct btrfs_root *root; 184649d11beaSJosef Bacik struct btrfs_key key; 184749d11beaSJosef Bacik 184849d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 184949d11beaSJosef Bacik 185049d11beaSJosef Bacik /* 185149d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 185249d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 185349d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 185449d11beaSJosef Bacik * != NULL just return it. 185549d11beaSJosef Bacik */ 185649d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 185749d11beaSJosef Bacik if (root) 185849d11beaSJosef Bacik return root; 185949d11beaSJosef Bacik 186049d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 186149d11beaSJosef Bacik if (root) 186249d11beaSJosef Bacik return root; 186349d11beaSJosef Bacik 186449d11beaSJosef Bacik key.objectid = objectid; 186549d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 186649d11beaSJosef Bacik key.offset = (u64)-1; 186749d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 186849d11beaSJosef Bacik btrfs_release_path(path); 186949d11beaSJosef Bacik 187049d11beaSJosef Bacik return root; 187149d11beaSJosef Bacik } 187249d11beaSJosef Bacik 1873a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1874a74a4b97SChris Mason { 18750d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1876d0278245SMiao Xie int again; 1877a74a4b97SChris Mason 1878d6fd0ae2SOmar Sandoval while (1) { 1879d0278245SMiao Xie again = 0; 18809d1a2a3aSDavid Sterba 1881fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1882fd340d0fSJosef Bacik 1883d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 18842ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1885d0278245SMiao Xie goto sleep; 1886d0278245SMiao Xie 188790c711abSZygo Blaxell /* 188890c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 188990c711abSZygo Blaxell * before we have finished mounting the filesystem. 189090c711abSZygo Blaxell */ 18910b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 189290c711abSZygo Blaxell goto sleep; 189390c711abSZygo Blaxell 18940b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1895d0278245SMiao Xie goto sleep; 1896d0278245SMiao Xie 1897dc7f370cSMiao Xie /* 1898dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1899dc7f370cSMiao Xie * during the above check and trylock. 1900dc7f370cSMiao Xie */ 19012ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 19020b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1903dc7f370cSMiao Xie goto sleep; 1904dc7f370cSMiao Xie } 1905dc7f370cSMiao Xie 19062ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1907c2d6cb16SFilipe Manana 190833c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 19090b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1910a74a4b97SChris Mason 1911d0278245SMiao Xie /* 191205323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 191305323cd1SMiao Xie * needn't do anything special here. 1914d0278245SMiao Xie */ 19150b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 191667c5e7d4SFilipe Manana 191767c5e7d4SFilipe Manana /* 1918f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 191967c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 192067c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1921f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 192267c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 192367c5e7d4SFilipe Manana * unused block groups. 192467c5e7d4SFilipe Manana */ 19250b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 192618bb8bbfSJohannes Thumshirn 192718bb8bbfSJohannes Thumshirn /* 192818bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 192918bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 193018bb8bbfSJohannes Thumshirn * space. 193118bb8bbfSJohannes Thumshirn */ 193218bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1933d0278245SMiao Xie sleep: 1934a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1935d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1936d6fd0ae2SOmar Sandoval kthread_parkme(); 1937d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1938d6fd0ae2SOmar Sandoval return 0; 1939838fe188SJiri Kosina if (!again) { 1940a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1941a74a4b97SChris Mason schedule(); 1942a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1943a74a4b97SChris Mason } 1944da288d28SFilipe Manana } 1945a74a4b97SChris Mason } 1946a74a4b97SChris Mason 1947a74a4b97SChris Mason static int transaction_kthread(void *arg) 1948a74a4b97SChris Mason { 1949a74a4b97SChris Mason struct btrfs_root *root = arg; 19500b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1951a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1952a74a4b97SChris Mason struct btrfs_transaction *cur; 19538929ecfaSYan, Zheng u64 transid; 1954643900beSNikolay Borisov time64_t delta; 1955a74a4b97SChris Mason unsigned long delay; 1956914b2007SJan Kara bool cannot_commit; 1957a74a4b97SChris Mason 1958a74a4b97SChris Mason do { 1959914b2007SJan Kara cannot_commit = false; 1960ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 19610b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1962a74a4b97SChris Mason 19630b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 19640b246afaSJeff Mahoney cur = fs_info->running_transaction; 1965a74a4b97SChris Mason if (!cur) { 19660b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1967a74a4b97SChris Mason goto sleep; 1968a74a4b97SChris Mason } 196931153d81SYan Zheng 1970643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1971fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1972fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1973643900beSNikolay Borisov delta < fs_info->commit_interval) { 19740b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1975fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1976fb8a7e94SNikolay Borisov delay = min(delay, 1977fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1978a74a4b97SChris Mason goto sleep; 1979a74a4b97SChris Mason } 19808929ecfaSYan, Zheng transid = cur->transid; 19810b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 198256bec294SChris Mason 198379787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1984354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1985914b2007SJan Kara if (IS_ERR(trans)) { 1986354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1987914b2007SJan Kara cannot_commit = true; 198879787eaaSJeff Mahoney goto sleep; 1989914b2007SJan Kara } 19908929ecfaSYan, Zheng if (transid == trans->transid) { 19913a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 19928929ecfaSYan, Zheng } else { 19933a45bb20SJeff Mahoney btrfs_end_transaction(trans); 19948929ecfaSYan, Zheng } 1995a74a4b97SChris Mason sleep: 19960b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 19970b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1998a74a4b97SChris Mason 199984961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 20002ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 20018929ecfaSYan, Zheng if (!kthread_should_stop() && 20020b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 2003914b2007SJan Kara cannot_commit)) 2004bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 2005a74a4b97SChris Mason } while (!kthread_should_stop()); 2006a74a4b97SChris Mason return 0; 2007a74a4b97SChris Mason } 2008a74a4b97SChris Mason 2009af31f5e5SChris Mason /* 201001f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 201101f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 201201f0f9daSNikolay Borisov * anything. 2013af31f5e5SChris Mason * 2014af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 2015af31f5e5SChris Mason * generation of the latest root in the array with the generation 2016af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 2017af31f5e5SChris Mason */ 201801f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 2019af31f5e5SChris Mason { 202001f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 2021af31f5e5SChris Mason u64 cur; 2022af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2023af31f5e5SChris Mason int i; 2024af31f5e5SChris Mason 2025af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2026af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 2027af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 2028af31f5e5SChris Mason if (cur == newest_gen) 202901f0f9daSNikolay Borisov return i; 2030af31f5e5SChris Mason } 2031af31f5e5SChris Mason 203201f0f9daSNikolay Borisov return -EINVAL; 2033af31f5e5SChris Mason } 2034af31f5e5SChris Mason 2035af31f5e5SChris Mason /* 2036af31f5e5SChris Mason * copy all the root pointers into the super backup array. 2037af31f5e5SChris Mason * this will bump the backup pointer by one when it is 2038af31f5e5SChris Mason * done 2039af31f5e5SChris Mason */ 2040af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 2041af31f5e5SChris Mason { 20426ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 2043af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2044af31f5e5SChris Mason 2045af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 2046af31f5e5SChris Mason 2047af31f5e5SChris Mason /* 2048af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 2049af31f5e5SChris Mason * regardless of which ones we use today 2050af31f5e5SChris Mason */ 2051af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 2052af31f5e5SChris Mason 2053af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 2054af31f5e5SChris Mason 2055af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 2056af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 2057af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 2058af31f5e5SChris Mason 2059af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 2060af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 2061af31f5e5SChris Mason 2062af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 2063af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 2064af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 2065af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 2066af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 2067af31f5e5SChris Mason 20681c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 20699c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 2070f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 20719c54e80dSJosef Bacik 20729c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 20739c54e80dSJosef Bacik extent_root->node->start); 2074af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 207529cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 2076af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 207729cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 2078af31f5e5SChris Mason 2079f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 2080f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 2081f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 2082f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 2083f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 20849c54e80dSJosef Bacik } 2085af31f5e5SChris Mason 20867c7e82a7SChris Mason /* 20877c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 20887c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 20897c7e82a7SChris Mason */ 20907c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 20917c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 20927c7e82a7SChris Mason info->fs_root->node->start); 2093af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2094af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2095af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2096af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 20977c7e82a7SChris Mason } 2098af31f5e5SChris Mason 2099af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2100af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2101af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2102af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2103af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2104af31f5e5SChris Mason 2105af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2106af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2107af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2108af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2109af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2110af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2111af31f5e5SChris Mason 2112af31f5e5SChris Mason /* 2113af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2114af31f5e5SChris Mason * for the next commit 2115af31f5e5SChris Mason */ 2116af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2117af31f5e5SChris Mason &info->super_for_commit->super_roots, 2118af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2119af31f5e5SChris Mason } 2120af31f5e5SChris Mason 2121af31f5e5SChris Mason /* 2122bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 2123bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 2124bd2336b2SNikolay Borisov * 2125bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 2126bd2336b2SNikolay Borisov * priority - priority of backup root required 2127bd2336b2SNikolay Borisov * 2128bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 2129bd2336b2SNikolay Borisov */ 2130bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 2131bd2336b2SNikolay Borisov { 2132bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2133bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 2134bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 2135bd2336b2SNikolay Borisov 2136bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 2137bd2336b2SNikolay Borisov if (priority == 0) 2138bd2336b2SNikolay Borisov return backup_index; 2139bd2336b2SNikolay Borisov 2140bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 2141bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 2142bd2336b2SNikolay Borisov } else { 2143bd2336b2SNikolay Borisov return -EINVAL; 2144bd2336b2SNikolay Borisov } 2145bd2336b2SNikolay Borisov 2146bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 2147bd2336b2SNikolay Borisov 2148bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 2149bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 2150bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2151bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 2152bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 2153bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2154bd2336b2SNikolay Borisov 2155bd2336b2SNikolay Borisov /* 2156bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 2157bd2336b2SNikolay Borisov * need a fsck 2158bd2336b2SNikolay Borisov */ 2159bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2160bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2161bd2336b2SNikolay Borisov 2162bd2336b2SNikolay Borisov return backup_index; 2163bd2336b2SNikolay Borisov } 2164bd2336b2SNikolay Borisov 21657abadb64SLiu Bo /* helper to cleanup workers */ 21667abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 21677abadb64SLiu Bo { 2168dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2169afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 2170a31b4a43SChristoph Hellwig btrfs_destroy_workqueue(fs_info->hipri_workers); 21715cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2172d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 2173d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 2174385de0efSChristoph Hellwig if (fs_info->rmw_workers) 2175385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 2176fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 2177fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 2178fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2179fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 21805b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2181e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2182a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2183fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2184b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 2185b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 2186a9b9477dSFilipe Manana /* 2187a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2188a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2189a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2190a9b9477dSFilipe Manana */ 2191d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 2192d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 21937abadb64SLiu Bo } 21947abadb64SLiu Bo 21952e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 21962e9f5954SRashika { 21972e9f5954SRashika if (root) { 21982e9f5954SRashika free_extent_buffer(root->node); 21992e9f5954SRashika free_extent_buffer(root->commit_root); 22002e9f5954SRashika root->node = NULL; 22012e9f5954SRashika root->commit_root = NULL; 22022e9f5954SRashika } 22032e9f5954SRashika } 22042e9f5954SRashika 2205abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 2206abed4aaaSJosef Bacik { 2207abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 2208abed4aaaSJosef Bacik 2209abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 2210abed4aaaSJosef Bacik &fs_info->global_root_tree, 2211abed4aaaSJosef Bacik rb_node) 2212abed4aaaSJosef Bacik free_root_extent_buffers(root); 2213abed4aaaSJosef Bacik } 2214abed4aaaSJosef Bacik 2215af31f5e5SChris Mason /* helper to cleanup tree roots */ 22164273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2217af31f5e5SChris Mason { 22182e9f5954SRashika free_root_extent_buffers(info->tree_root); 2219655b09feSJosef Bacik 2220abed4aaaSJosef Bacik free_global_root_pointers(info); 22212e9f5954SRashika free_root_extent_buffers(info->dev_root); 22222e9f5954SRashika free_root_extent_buffers(info->quota_root); 22232e9f5954SRashika free_root_extent_buffers(info->uuid_root); 22248c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2225aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 22269c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 22274273eaffSAnand Jain if (free_chunk_root) 22282e9f5954SRashika free_root_extent_buffers(info->chunk_root); 2229af31f5e5SChris Mason } 2230af31f5e5SChris Mason 22318c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 22328c38938cSJosef Bacik { 22338c38938cSJosef Bacik if (!root) 22348c38938cSJosef Bacik return; 22358c38938cSJosef Bacik 22368c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 22378c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 22381dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 22398c38938cSJosef Bacik if (root->anon_dev) 22408c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 22418c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2242923eb523SJohannes Thumshirn free_root_extent_buffers(root); 22438c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2244fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 22458c38938cSJosef Bacik list_del_init(&root->leak_list); 2246fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 22478c38938cSJosef Bacik #endif 22488c38938cSJosef Bacik kfree(root); 22498c38938cSJosef Bacik } 22508c38938cSJosef Bacik } 22518c38938cSJosef Bacik 2252faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2253171f6537SJosef Bacik { 2254fc7cbcd4SDavid Sterba int ret; 2255fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 2256fc7cbcd4SDavid Sterba int i; 2257171f6537SJosef Bacik 2258171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2259fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 2260171f6537SJosef Bacik struct btrfs_root, root_list); 2261fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 2262171f6537SJosef Bacik 2263fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2264fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2265fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 2266171f6537SJosef Bacik } 2267171f6537SJosef Bacik 2268fc7cbcd4SDavid Sterba while (1) { 2269fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2270fc7cbcd4SDavid Sterba (void **)gang, 0, 2271fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 2272fc7cbcd4SDavid Sterba if (!ret) 2273fc7cbcd4SDavid Sterba break; 2274fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 2275fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2276171f6537SJosef Bacik } 2277171f6537SJosef Bacik } 2278af31f5e5SChris Mason 2279638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2280638aa7edSEric Sandeen { 2281638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2282638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2283638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2284638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2285638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2286638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2287ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2288638aa7edSEric Sandeen } 2289638aa7edSEric Sandeen 2290779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2291779a65a4SEric Sandeen { 2292779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2293779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2294779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2295779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2296779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2297779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2298907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 2299779a65a4SEric Sandeen } 2300779a65a4SEric Sandeen 23016bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2302f37938e0SEric Sandeen { 23032ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 2304e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 2305e256927bSJosef Bacik fs_info->tree_root); 23062ff7e61eSJeff Mahoney 23072ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 23082ff7e61eSJeff Mahoney set_nlink(inode, 1); 2309f37938e0SEric Sandeen /* 2310f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2311f37938e0SEric Sandeen * the real end of the address space is determined by all of 2312f37938e0SEric Sandeen * the devices in the system 2313f37938e0SEric Sandeen */ 23142ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 23152ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2316f37938e0SEric Sandeen 23172ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 231843eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 231935da5a7eSDavid Sterba IO_TREE_BTREE_INODE_IO); 23202ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2321f37938e0SEric Sandeen 23225c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 2323adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 2324adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 2325adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 23262ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 2327e256927bSJosef Bacik __insert_inode_hash(inode, hash); 2328f37938e0SEric Sandeen } 2329f37938e0SEric Sandeen 2330ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2331ad618368SEric Sandeen { 2332ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2333129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 23347f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2335ad618368SEric Sandeen } 2336ad618368SEric Sandeen 2337f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2338f9e92e40SEric Sandeen { 2339f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2340f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2341f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2342f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2343f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2344f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2345d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2346011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 2347f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2348f9e92e40SEric Sandeen } 2349f9e92e40SEric Sandeen 2350d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 23512a458198SEric Sandeen { 2352f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 23536f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 23542a458198SEric Sandeen 23552a458198SEric Sandeen fs_info->workers = 2356a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 2357a31b4a43SChristoph Hellwig fs_info->hipri_workers = 2358a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker-high", 2359cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 23602a458198SEric Sandeen 23612a458198SEric Sandeen fs_info->delalloc_workers = 2362cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2363cb001095SJeff Mahoney flags, max_active, 2); 23642a458198SEric Sandeen 23652a458198SEric Sandeen fs_info->flush_workers = 2366cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2367cb001095SJeff Mahoney flags, max_active, 0); 23682a458198SEric Sandeen 23692a458198SEric Sandeen fs_info->caching_workers = 2370cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 23712a458198SEric Sandeen 23722a458198SEric Sandeen fs_info->fixup_workers = 2373cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 23742a458198SEric Sandeen 23752a458198SEric Sandeen fs_info->endio_workers = 2376d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 23772a458198SEric Sandeen fs_info->endio_meta_workers = 2378d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 2379385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 23802a458198SEric Sandeen fs_info->endio_write_workers = 2381cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2382cb001095SJeff Mahoney max_active, 2); 2383fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 2384fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 23852a458198SEric Sandeen fs_info->endio_freespace_worker = 2386cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2387cb001095SJeff Mahoney max_active, 0); 23882a458198SEric Sandeen fs_info->delayed_workers = 2389cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2390cb001095SJeff Mahoney max_active, 0); 23912a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2392cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2393b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2394b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 23952a458198SEric Sandeen 2396a31b4a43SChristoph Hellwig if (!(fs_info->workers && fs_info->hipri_workers && 2397a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 23982a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 2399fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 24001a1a2851SQu Wenruo fs_info->endio_write_workers && 24012a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 2402f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 2403f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 2404b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 24052a458198SEric Sandeen return -ENOMEM; 24062a458198SEric Sandeen } 24072a458198SEric Sandeen 24082a458198SEric Sandeen return 0; 24092a458198SEric Sandeen } 24102a458198SEric Sandeen 24116d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 24126d97c6e3SJohannes Thumshirn { 24136d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2414b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 24156d97c6e3SJohannes Thumshirn 2416b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 24176d97c6e3SJohannes Thumshirn 24186d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 24196d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2420b4e967beSDavid Sterba csum_driver); 24216d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 24226d97c6e3SJohannes Thumshirn } 24236d97c6e3SJohannes Thumshirn 24246d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 24256d97c6e3SJohannes Thumshirn 2426c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2427c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2428c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 24296d97c6e3SJohannes Thumshirn return 0; 24306d97c6e3SJohannes Thumshirn } 24316d97c6e3SJohannes Thumshirn 243263443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 243363443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 243463443bf5SEric Sandeen { 243563443bf5SEric Sandeen int ret; 2436789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 0 }; 243763443bf5SEric Sandeen struct btrfs_root *log_tree_root; 243863443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 243963443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2440581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 244163443bf5SEric Sandeen 244263443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2443f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 244463443bf5SEric Sandeen return -EIO; 244563443bf5SEric Sandeen } 244663443bf5SEric Sandeen 244796dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 244896dfcb46SJosef Bacik GFP_KERNEL); 244963443bf5SEric Sandeen if (!log_tree_root) 245063443bf5SEric Sandeen return -ENOMEM; 245163443bf5SEric Sandeen 2452789d6a3aSQu Wenruo check.level = level; 2453789d6a3aSQu Wenruo check.transid = fs_info->generation + 1; 2454789d6a3aSQu Wenruo check.owner_root = BTRFS_TREE_LOG_OBJECTID; 2455789d6a3aSQu Wenruo log_tree_root->node = read_tree_block(fs_info, bytenr, &check); 245664c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2457f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 24580eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 24598c38938cSJosef Bacik log_tree_root->node = NULL; 246000246528SJosef Bacik btrfs_put_root(log_tree_root); 24610eeff236SLiu Bo return ret; 24624eb150d6SQu Wenruo } 24634eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2464f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 246500246528SJosef Bacik btrfs_put_root(log_tree_root); 246663443bf5SEric Sandeen return -EIO; 246763443bf5SEric Sandeen } 24684eb150d6SQu Wenruo 246963443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 247063443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 247163443bf5SEric Sandeen if (ret) { 24720b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 247363443bf5SEric Sandeen "Failed to recover log tree"); 247400246528SJosef Bacik btrfs_put_root(log_tree_root); 247563443bf5SEric Sandeen return ret; 247663443bf5SEric Sandeen } 247763443bf5SEric Sandeen 2478bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 24796bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 248063443bf5SEric Sandeen if (ret) 248163443bf5SEric Sandeen return ret; 248263443bf5SEric Sandeen } 248363443bf5SEric Sandeen 248463443bf5SEric Sandeen return 0; 248563443bf5SEric Sandeen } 248663443bf5SEric Sandeen 2487abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2488abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2489abed4aaaSJosef Bacik const char *name) 2490abed4aaaSJosef Bacik { 2491abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2492abed4aaaSJosef Bacik struct btrfs_root *root; 2493f7238e50SJosef Bacik u64 max_global_id = 0; 2494abed4aaaSJosef Bacik int ret; 2495abed4aaaSJosef Bacik struct btrfs_key key = { 2496abed4aaaSJosef Bacik .objectid = objectid, 2497abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2498abed4aaaSJosef Bacik .offset = 0, 2499abed4aaaSJosef Bacik }; 2500abed4aaaSJosef Bacik bool found = false; 2501abed4aaaSJosef Bacik 2502abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2503abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2504abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2505abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2506abed4aaaSJosef Bacik return 0; 2507abed4aaaSJosef Bacik } 2508abed4aaaSJosef Bacik 2509abed4aaaSJosef Bacik while (1) { 2510abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2511abed4aaaSJosef Bacik if (ret < 0) 2512abed4aaaSJosef Bacik break; 2513abed4aaaSJosef Bacik 2514abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2515abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2516abed4aaaSJosef Bacik if (ret) { 2517abed4aaaSJosef Bacik if (ret > 0) 2518abed4aaaSJosef Bacik ret = 0; 2519abed4aaaSJosef Bacik break; 2520abed4aaaSJosef Bacik } 2521abed4aaaSJosef Bacik } 2522abed4aaaSJosef Bacik ret = 0; 2523abed4aaaSJosef Bacik 2524abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2525abed4aaaSJosef Bacik if (key.objectid != objectid) 2526abed4aaaSJosef Bacik break; 2527abed4aaaSJosef Bacik btrfs_release_path(path); 2528abed4aaaSJosef Bacik 2529f7238e50SJosef Bacik /* 2530f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2531f7238e50SJosef Bacik * everybody. 2532f7238e50SJosef Bacik */ 2533f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2534f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2535f7238e50SJosef Bacik 2536abed4aaaSJosef Bacik found = true; 2537abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2538abed4aaaSJosef Bacik if (IS_ERR(root)) { 2539abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2540abed4aaaSJosef Bacik ret = PTR_ERR(root); 2541abed4aaaSJosef Bacik break; 2542abed4aaaSJosef Bacik } 2543abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2544abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2545abed4aaaSJosef Bacik if (ret) { 2546abed4aaaSJosef Bacik btrfs_put_root(root); 2547abed4aaaSJosef Bacik break; 2548abed4aaaSJosef Bacik } 2549abed4aaaSJosef Bacik key.offset++; 2550abed4aaaSJosef Bacik } 2551abed4aaaSJosef Bacik btrfs_release_path(path); 2552abed4aaaSJosef Bacik 2553f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2554f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2555f7238e50SJosef Bacik 2556abed4aaaSJosef Bacik if (!found || ret) { 2557abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2558abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2559abed4aaaSJosef Bacik 2560abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2561abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2562abed4aaaSJosef Bacik else 2563abed4aaaSJosef Bacik ret = 0; 2564abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2565abed4aaaSJosef Bacik } 2566abed4aaaSJosef Bacik return ret; 2567abed4aaaSJosef Bacik } 2568abed4aaaSJosef Bacik 2569abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2570abed4aaaSJosef Bacik { 2571abed4aaaSJosef Bacik struct btrfs_path *path; 2572abed4aaaSJosef Bacik int ret = 0; 2573abed4aaaSJosef Bacik 2574abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2575abed4aaaSJosef Bacik if (!path) 2576abed4aaaSJosef Bacik return -ENOMEM; 2577abed4aaaSJosef Bacik 2578abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2579abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2580abed4aaaSJosef Bacik if (ret) 2581abed4aaaSJosef Bacik goto out; 2582abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2583abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2584abed4aaaSJosef Bacik if (ret) 2585abed4aaaSJosef Bacik goto out; 2586abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2587abed4aaaSJosef Bacik goto out; 2588abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2589abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2590abed4aaaSJosef Bacik "free space"); 2591abed4aaaSJosef Bacik out: 2592abed4aaaSJosef Bacik btrfs_free_path(path); 2593abed4aaaSJosef Bacik return ret; 2594abed4aaaSJosef Bacik } 2595abed4aaaSJosef Bacik 25966bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 25974bbcaa64SEric Sandeen { 25986bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2599a4f3d2c4SDavid Sterba struct btrfs_root *root; 26004bbcaa64SEric Sandeen struct btrfs_key location; 26014bbcaa64SEric Sandeen int ret; 26024bbcaa64SEric Sandeen 26036bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 26046bccf3abSJeff Mahoney 2605abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2606abed4aaaSJosef Bacik if (ret) 2607abed4aaaSJosef Bacik return ret; 2608abed4aaaSJosef Bacik 26094bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 26104bbcaa64SEric Sandeen location.offset = 0; 26114bbcaa64SEric Sandeen 26121c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 261314033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 261414033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 261514033b08SQu Wenruo if (IS_ERR(root)) { 261614033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 261714033b08SQu Wenruo ret = PTR_ERR(root); 261814033b08SQu Wenruo goto out; 261914033b08SQu Wenruo } 262014033b08SQu Wenruo } else { 262114033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 262214033b08SQu Wenruo fs_info->block_group_root = root; 262314033b08SQu Wenruo } 262414033b08SQu Wenruo } 262514033b08SQu Wenruo 262614033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2627a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2628f50f4353SLiu Bo if (IS_ERR(root)) { 262942437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2630f50f4353SLiu Bo ret = PTR_ERR(root); 2631f50f4353SLiu Bo goto out; 2632f50f4353SLiu Bo } 263342437a63SJosef Bacik } else { 2634a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2635a4f3d2c4SDavid Sterba fs_info->dev_root = root; 263642437a63SJosef Bacik } 2637820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2638a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2639a8d1b164SJohannes Thumshirn if (ret) 2640a8d1b164SJohannes Thumshirn goto out; 26414bbcaa64SEric Sandeen 2642aeb935a4SQu Wenruo /* 2643aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2644aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2645aeb935a4SQu Wenruo */ 264656e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 264756e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2648aeb935a4SQu Wenruo if (IS_ERR(root)) { 264942437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2650aeb935a4SQu Wenruo ret = PTR_ERR(root); 2651aeb935a4SQu Wenruo goto out; 2652aeb935a4SQu Wenruo } 265342437a63SJosef Bacik } else { 2654aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2655aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 265642437a63SJosef Bacik } 2657aeb935a4SQu Wenruo 26584bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2659a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2660a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2661a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2662afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2663a4f3d2c4SDavid Sterba fs_info->quota_root = root; 26644bbcaa64SEric Sandeen } 26654bbcaa64SEric Sandeen 26664bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2667a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2668a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 266942437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2670a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 26714bbcaa64SEric Sandeen if (ret != -ENOENT) 2672f50f4353SLiu Bo goto out; 267342437a63SJosef Bacik } 26744bbcaa64SEric Sandeen } else { 2675a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2676a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 26774bbcaa64SEric Sandeen } 26784bbcaa64SEric Sandeen 26794bbcaa64SEric Sandeen return 0; 2680f50f4353SLiu Bo out: 2681f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2682f50f4353SLiu Bo location.objectid, ret); 2683f50f4353SLiu Bo return ret; 26844bbcaa64SEric Sandeen } 26854bbcaa64SEric Sandeen 2686069ec957SQu Wenruo /* 2687069ec957SQu Wenruo * Real super block validation 2688069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2689069ec957SQu Wenruo * 2690069ec957SQu Wenruo * @sb: super block to check 2691069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2692069ec957SQu Wenruo * 0 the primary (1st) sb 2693069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2694069ec957SQu Wenruo * -1 skip bytenr check 2695069ec957SQu Wenruo */ 2696a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2697069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 269821a852b0SQu Wenruo { 269921a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 270021a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 270121a852b0SQu Wenruo int ret = 0; 270221a852b0SQu Wenruo 270321a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 270421a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 270521a852b0SQu Wenruo ret = -EINVAL; 270621a852b0SQu Wenruo } 270721a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 270821a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 270921a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 271021a852b0SQu Wenruo ret = -EINVAL; 271121a852b0SQu Wenruo } 271221a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 271321a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 271421a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 271521a852b0SQu Wenruo ret = -EINVAL; 271621a852b0SQu Wenruo } 271721a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 271821a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 271921a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 272021a852b0SQu Wenruo ret = -EINVAL; 272121a852b0SQu Wenruo } 272221a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 272321a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 272421a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 272521a852b0SQu Wenruo ret = -EINVAL; 272621a852b0SQu Wenruo } 272721a852b0SQu Wenruo 272821a852b0SQu Wenruo /* 272921a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 273021a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 273121a852b0SQu Wenruo */ 273221a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 273321a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 273421a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 273521a852b0SQu Wenruo ret = -EINVAL; 273621a852b0SQu Wenruo } 27370bb3eb3eSQu Wenruo 27380bb3eb3eSQu Wenruo /* 27391a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 27401a42daabSQu Wenruo * 27411a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 27421a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 27431a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 27441a42daabSQu Wenruo * beginning. 27450bb3eb3eSQu Wenruo */ 27461a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 274721a852b0SQu Wenruo btrfs_err(fs_info, 27480bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 274921a852b0SQu Wenruo sectorsize, PAGE_SIZE); 275021a852b0SQu Wenruo ret = -EINVAL; 275121a852b0SQu Wenruo } 27520bb3eb3eSQu Wenruo 275321a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 275421a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 275521a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 275621a852b0SQu Wenruo ret = -EINVAL; 275721a852b0SQu Wenruo } 275821a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 275921a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 276021a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 276121a852b0SQu Wenruo ret = -EINVAL; 276221a852b0SQu Wenruo } 276321a852b0SQu Wenruo 276421a852b0SQu Wenruo /* Root alignment check */ 276521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 276621a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 276721a852b0SQu Wenruo btrfs_super_root(sb)); 276821a852b0SQu Wenruo ret = -EINVAL; 276921a852b0SQu Wenruo } 277021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 277121a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 277221a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 277321a852b0SQu Wenruo ret = -EINVAL; 277421a852b0SQu Wenruo } 277521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 277621a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 277721a852b0SQu Wenruo btrfs_super_log_root(sb)); 277821a852b0SQu Wenruo ret = -EINVAL; 277921a852b0SQu Wenruo } 278021a852b0SQu Wenruo 2781aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2782aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2783aefd7f70SNikolay Borisov btrfs_err(fs_info, 2784aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2785aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2786aefd7f70SNikolay Borisov ret = -EINVAL; 2787aefd7f70SNikolay Borisov } 2788aefd7f70SNikolay Borisov 2789aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2790aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2791aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2792aefd7f70SNikolay Borisov btrfs_err(fs_info, 2793aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2794aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2795aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2796aefd7f70SNikolay Borisov ret = -EINVAL; 2797aefd7f70SNikolay Borisov } 2798aefd7f70SNikolay Borisov 27991c56ab99SQu Wenruo /* 28001c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 28011c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 28021c56ab99SQu Wenruo */ 28031c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 28041c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 28051c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 28061c56ab99SQu Wenruo btrfs_err(fs_info, 28071c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 28081c56ab99SQu Wenruo ret = -EINVAL; 28091c56ab99SQu Wenruo } 28101c56ab99SQu Wenruo 2811de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 28127239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 281321a852b0SQu Wenruo btrfs_err(fs_info, 28147239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2815de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 281621a852b0SQu Wenruo ret = -EINVAL; 281721a852b0SQu Wenruo } 281821a852b0SQu Wenruo 281921a852b0SQu Wenruo /* 282021a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 282121a852b0SQu Wenruo * done later 282221a852b0SQu Wenruo */ 282321a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 282421a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 282521a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 282621a852b0SQu Wenruo ret = -EINVAL; 282721a852b0SQu Wenruo } 282821a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 282921a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 283021a852b0SQu Wenruo btrfs_super_stripesize(sb)); 283121a852b0SQu Wenruo ret = -EINVAL; 283221a852b0SQu Wenruo } 283321a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 283421a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 283521a852b0SQu Wenruo btrfs_super_num_devices(sb)); 283621a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 283721a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 283821a852b0SQu Wenruo ret = -EINVAL; 283921a852b0SQu Wenruo } 284021a852b0SQu Wenruo 2841069ec957SQu Wenruo if (mirror_num >= 0 && 2842069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 284321a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 284421a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 284521a852b0SQu Wenruo ret = -EINVAL; 284621a852b0SQu Wenruo } 284721a852b0SQu Wenruo 284821a852b0SQu Wenruo /* 284921a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 285021a852b0SQu Wenruo * and one chunk 285121a852b0SQu Wenruo */ 285221a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 285321a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 285421a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 285521a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 285621a852b0SQu Wenruo ret = -EINVAL; 285721a852b0SQu Wenruo } 285821a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 285921a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 286021a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 286121a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 286221a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 286321a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 286421a852b0SQu Wenruo ret = -EINVAL; 286521a852b0SQu Wenruo } 286621a852b0SQu Wenruo 286721a852b0SQu Wenruo /* 286821a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 286921a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 287021a852b0SQu Wenruo */ 287121a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 287221a852b0SQu Wenruo btrfs_warn(fs_info, 287321a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 287421a852b0SQu Wenruo btrfs_super_generation(sb), 287521a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 287621a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 287721a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 287821a852b0SQu Wenruo btrfs_warn(fs_info, 287921a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 288021a852b0SQu Wenruo btrfs_super_generation(sb), 288121a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 288221a852b0SQu Wenruo 288321a852b0SQu Wenruo return ret; 288421a852b0SQu Wenruo } 288521a852b0SQu Wenruo 2886069ec957SQu Wenruo /* 2887069ec957SQu Wenruo * Validation of super block at mount time. 2888069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2889069ec957SQu Wenruo * flags will be skipped. 2890069ec957SQu Wenruo */ 2891069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2892069ec957SQu Wenruo { 2893a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2894069ec957SQu Wenruo } 2895069ec957SQu Wenruo 289675cb857dSQu Wenruo /* 289775cb857dSQu Wenruo * Validation of super block at write time. 289875cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 289975cb857dSQu Wenruo * overwritten soon. 290075cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 290175cb857dSQu Wenruo */ 290275cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 290375cb857dSQu Wenruo struct btrfs_super_block *sb) 290475cb857dSQu Wenruo { 290575cb857dSQu Wenruo int ret; 290675cb857dSQu Wenruo 2907a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 290875cb857dSQu Wenruo if (ret < 0) 290975cb857dSQu Wenruo goto out; 2910e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 291175cb857dSQu Wenruo ret = -EUCLEAN; 291275cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 291375cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 291475cb857dSQu Wenruo goto out; 291575cb857dSQu Wenruo } 291675cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 291775cb857dSQu Wenruo ret = -EUCLEAN; 291875cb857dSQu Wenruo btrfs_err(fs_info, 291975cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 292075cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 292175cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 292275cb857dSQu Wenruo goto out; 292375cb857dSQu Wenruo } 292475cb857dSQu Wenruo out: 292575cb857dSQu Wenruo if (ret < 0) 292675cb857dSQu Wenruo btrfs_err(fs_info, 292775cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 292875cb857dSQu Wenruo return ret; 292975cb857dSQu Wenruo } 293075cb857dSQu Wenruo 2931bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2932bd676446SJosef Bacik { 2933789d6a3aSQu Wenruo struct btrfs_tree_parent_check check = { 2934789d6a3aSQu Wenruo .level = level, 2935789d6a3aSQu Wenruo .transid = gen, 2936789d6a3aSQu Wenruo .owner_root = root->root_key.objectid 2937789d6a3aSQu Wenruo }; 2938bd676446SJosef Bacik int ret = 0; 2939bd676446SJosef Bacik 2940789d6a3aSQu Wenruo root->node = read_tree_block(root->fs_info, bytenr, &check); 2941bd676446SJosef Bacik if (IS_ERR(root->node)) { 2942bd676446SJosef Bacik ret = PTR_ERR(root->node); 2943bd676446SJosef Bacik root->node = NULL; 29444eb150d6SQu Wenruo return ret; 29454eb150d6SQu Wenruo } 29464eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2947bd676446SJosef Bacik free_extent_buffer(root->node); 2948bd676446SJosef Bacik root->node = NULL; 29494eb150d6SQu Wenruo return -EIO; 2950bd676446SJosef Bacik } 2951bd676446SJosef Bacik 2952bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2953bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2954bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2955bd676446SJosef Bacik return ret; 2956bd676446SJosef Bacik } 2957bd676446SJosef Bacik 2958bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2959bd676446SJosef Bacik { 2960bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2961bd676446SJosef Bacik u64 gen, bytenr; 2962bd676446SJosef Bacik int level, ret; 2963bd676446SJosef Bacik 2964bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2965bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2966bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2967bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 29689c54e80dSJosef Bacik if (ret) { 2969bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2970bd676446SJosef Bacik return ret; 2971bd676446SJosef Bacik } 29729c54e80dSJosef Bacik return 0; 29739c54e80dSJosef Bacik } 29749c54e80dSJosef Bacik 29756ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2976b8522a1eSNikolay Borisov { 29776ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2978b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2979b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2980b8522a1eSNikolay Borisov bool handle_error = false; 2981b8522a1eSNikolay Borisov int ret = 0; 2982b8522a1eSNikolay Borisov int i; 2983b8522a1eSNikolay Borisov 2984b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2985b8522a1eSNikolay Borisov if (handle_error) { 2986b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2987b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2988b8522a1eSNikolay Borisov tree_root->node = NULL; 2989b8522a1eSNikolay Borisov 2990b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2991b8522a1eSNikolay Borisov break; 2992b8522a1eSNikolay Borisov 2993b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2994b8522a1eSNikolay Borisov 2995b8522a1eSNikolay Borisov /* 2996b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2997b8522a1eSNikolay Borisov * valid 2998b8522a1eSNikolay Borisov */ 2999b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 3000b8522a1eSNikolay Borisov 3001b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 3002b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3003b8522a1eSNikolay Borisov 3004b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 30056ef108ddSNikolay Borisov backup_index = ret; 3006b8522a1eSNikolay Borisov if (ret < 0) 3007b8522a1eSNikolay Borisov return ret; 3008b8522a1eSNikolay Borisov } 3009b8522a1eSNikolay Borisov 3010bd676446SJosef Bacik ret = load_important_roots(fs_info); 3011bd676446SJosef Bacik if (ret) { 3012217f5004SNikolay Borisov handle_error = true; 3013b8522a1eSNikolay Borisov continue; 3014b8522a1eSNikolay Borisov } 3015b8522a1eSNikolay Borisov 3016336a0d8dSNikolay Borisov /* 3017336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 3018336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 3019336a0d8dSNikolay Borisov */ 3020453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 3021b8522a1eSNikolay Borisov if (ret < 0) { 3022b8522a1eSNikolay Borisov handle_error = true; 3023b8522a1eSNikolay Borisov continue; 3024b8522a1eSNikolay Borisov } 3025b8522a1eSNikolay Borisov 30266b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 3027b8522a1eSNikolay Borisov 3028b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 3029b8522a1eSNikolay Borisov if (ret < 0) { 3030b8522a1eSNikolay Borisov handle_error = true; 3031b8522a1eSNikolay Borisov continue; 3032b8522a1eSNikolay Borisov } 3033b8522a1eSNikolay Borisov 3034b8522a1eSNikolay Borisov /* All successful */ 3035bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 3036bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 3037d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 30386ef108ddSNikolay Borisov 30396ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 30406ef108ddSNikolay Borisov if (backup_index < 0) { 30416ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 30426ef108ddSNikolay Borisov } else { 30436ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 30446ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 30456ef108ddSNikolay Borisov } 3046b8522a1eSNikolay Borisov break; 3047b8522a1eSNikolay Borisov } 3048b8522a1eSNikolay Borisov 3049b8522a1eSNikolay Borisov return ret; 3050b8522a1eSNikolay Borisov } 3051b8522a1eSNikolay Borisov 30528260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 3053eb60ceacSChris Mason { 3054fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 305501cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 30568fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 3057facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 305824bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 3059eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 306011833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 3061eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 3062a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 3063fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 306424bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 30654cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 3066ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 3067f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 306847ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 306940ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 3070afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 3071c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 3072f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 3073abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 3074d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 3075f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 30767585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 3077573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 30780bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 30795f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 3080de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 308119c00ddcSChris Mason 3082e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 30835a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 30848b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 30855f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 30863e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 30873e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 30883e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 30893e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 30903e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 30913e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 30923e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 30933e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 3094e1489b4fSIoannis Angelakopoulos 30950b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 30966324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 3097f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 309847ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 309918bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 3100afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 3101bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 3102bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 31033fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 31043fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 3105bd647ce3SJosef Bacik #endif 3106c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 310766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 310866d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 310966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 311066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 311166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 311266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 311366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 3114ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 3115ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 3116ba2c4d4eSJosef Bacik 3117771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 31184cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 3119034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 3120fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 3121abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 312295ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 31239ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 31244cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 3125a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 3126f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 31278b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 3128f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 3129fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 3130c8b97818SChris Mason 3131b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 3132b34b086cSChris Mason num_online_cpus() + 2, 8); 31330afbaf8cSChris Mason 3134199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 3135199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 313669fe2d75SJosef Bacik 3137638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 313821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 313921adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 314021adbd5cSStefan Behrens #endif 3141779a65a4SEric Sandeen btrfs_init_balance(fs_info); 314257056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 3143a2de733cSArne Jansen 314416b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 314508dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 31460f9dd46cSJosef Bacik 3147fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 314835da5a7eSDavid Sterba IO_TREE_FS_EXCLUDED_EXTENTS); 314939279cc3SChris Mason 31505a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 3151e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 3152925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 3153a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 3154a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 31551bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 31569e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 3157c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 315876dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 3159803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 3160fa9c0d79SChris Mason 3161ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 3162f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 3163b0643e59SDennis Zhou btrfs_discard_init(fs_info); 3164416ac51dSArne Jansen 3165fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 3166fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 3167fa9c0d79SChris Mason 3168e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 3169f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 3170bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 31714854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 3172034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 31733768f368SChris Mason 3174da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 3175da17066cSJeff Mahoney fs_info->nodesize = 4096; 3176da17066cSJeff Mahoney fs_info->sectorsize = 4096; 3177ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 3178da17066cSJeff Mahoney fs_info->stripesize = 4096; 3179da17066cSJeff Mahoney 3180f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 3181f7b12a62SNaohiro Aota 3182eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 3183eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 3184eede2bf3SOmar Sandoval 318518bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 318618bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 31878260edbaSJosef Bacik } 31888260edbaSJosef Bacik 31898260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 31908260edbaSJosef Bacik { 31918260edbaSJosef Bacik int ret; 31928260edbaSJosef Bacik 31938260edbaSJosef Bacik fs_info->sb = sb; 31948260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 31958260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 31969e967495SFilipe Manana 31975deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 3198ae18c37aSJosef Bacik if (ret) 3199ae18c37aSJosef Bacik return ret; 3200ae18c37aSJosef Bacik 3201ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 3202ae18c37aSJosef Bacik if (ret) 3203c75e8394SJosef Bacik return ret; 3204ae18c37aSJosef Bacik 3205ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 3206ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 3207ae18c37aSJosef Bacik 3208ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 3209ae18c37aSJosef Bacik if (ret) 3210c75e8394SJosef Bacik return ret; 3211ae18c37aSJosef Bacik 3212ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 3213ae18c37aSJosef Bacik GFP_KERNEL); 3214ae18c37aSJosef Bacik if (ret) 3215c75e8394SJosef Bacik return ret; 3216ae18c37aSJosef Bacik 3217ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 3218ae18c37aSJosef Bacik GFP_KERNEL); 3219c75e8394SJosef Bacik if (!fs_info->delayed_root) 3220c75e8394SJosef Bacik return -ENOMEM; 3221ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 3222ae18c37aSJosef Bacik 3223a0a1db70SFilipe Manana if (sb_rdonly(sb)) 3224a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 3225a0a1db70SFilipe Manana 3226c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 3227ae18c37aSJosef Bacik } 3228ae18c37aSJosef Bacik 322997f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 323097f4dd09SNikolay Borisov { 32310d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 323297f4dd09SNikolay Borisov int ret; 323397f4dd09SNikolay Borisov 323497f4dd09SNikolay Borisov /* 323597f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 323697f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 323797f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 323897f4dd09SNikolay Borisov */ 323997f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 324097f4dd09SNikolay Borisov if (ret < 0) { 3241c94bec2cSJosef Bacik if (ret != -EINTR) 3242c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 3243c94bec2cSJosef Bacik ret); 324497f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 324597f4dd09SNikolay Borisov return ret; 324697f4dd09SNikolay Borisov } 324797f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 324897f4dd09SNikolay Borisov } 324997f4dd09SNikolay Borisov 325097f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 325197f4dd09SNikolay Borisov { 325297f4dd09SNikolay Borisov struct task_struct *task; 325397f4dd09SNikolay Borisov 325497f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 325597f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 325697f4dd09SNikolay Borisov if (IS_ERR(task)) { 325797f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 325897f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 325997f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 326097f4dd09SNikolay Borisov return PTR_ERR(task); 326197f4dd09SNikolay Borisov } 326297f4dd09SNikolay Borisov 326397f4dd09SNikolay Borisov return 0; 326497f4dd09SNikolay Borisov } 326597f4dd09SNikolay Borisov 326644c0ca21SBoris Burkov /* 32678cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 32688cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 32698cd29088SBoris Burkov * code paths. 32708cd29088SBoris Burkov */ 32718cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 32728cd29088SBoris Burkov { 32738cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 32748b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 32758cd29088SBoris Burkov } 32768cd29088SBoris Burkov 32778cd29088SBoris Burkov /* 327844c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 327944c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 328044c0ca21SBoris Burkov */ 328144c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 328244c0ca21SBoris Burkov { 328344c0ca21SBoris Burkov int ret; 328494846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 32858b228324SBoris Burkov bool clear_free_space_tree = false; 32868b228324SBoris Burkov 32878b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 32888b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 32898b228324SBoris Burkov clear_free_space_tree = true; 32908b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 32918b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 32928b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 32938b228324SBoris Burkov clear_free_space_tree = true; 32948b228324SBoris Burkov } 32958b228324SBoris Burkov 32968b228324SBoris Burkov if (clear_free_space_tree) { 32978b228324SBoris Burkov btrfs_info(fs_info, "clearing free space tree"); 32988b228324SBoris Burkov ret = btrfs_clear_free_space_tree(fs_info); 32998b228324SBoris Burkov if (ret) { 33008b228324SBoris Burkov btrfs_warn(fs_info, 33018b228324SBoris Burkov "failed to clear free space tree: %d", ret); 33028b228324SBoris Burkov goto out; 33038b228324SBoris Burkov } 33048b228324SBoris Burkov } 330544c0ca21SBoris Burkov 33068d488a8cSFilipe Manana /* 33078d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 33088d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 3309fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 33108d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 33118d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 33128d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 33138d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 33148d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 33158d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 33168d488a8cSFilipe Manana */ 33178d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 33188d488a8cSFilipe Manana if (ret) 33198d488a8cSFilipe Manana goto out; 33208d488a8cSFilipe Manana 332144c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 332244c0ca21SBoris Burkov if (ret) 332344c0ca21SBoris Burkov goto out; 332444c0ca21SBoris Burkov 33258f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 33268f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 33278f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 33288f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 33298f1c21d7SBoris Burkov goto out; 33308f1c21d7SBoris Burkov } 33318f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 33328f1c21d7SBoris Burkov 333344c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 33347eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 333544c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 333644c0ca21SBoris Burkov if (ret < 0) { 333744c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 333844c0ca21SBoris Burkov goto out; 333944c0ca21SBoris Burkov } 334044c0ca21SBoris Burkov 33415011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 33425011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 33435011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 33445011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 33455011139aSBoris Burkov if (ret) { 33465011139aSBoris Burkov btrfs_warn(fs_info, 33475011139aSBoris Burkov "failed to create free space tree: %d", ret); 33485011139aSBoris Burkov goto out; 33495011139aSBoris Burkov } 33505011139aSBoris Burkov } 33515011139aSBoris Burkov 335294846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 335394846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 335494846229SBoris Burkov if (ret) 335594846229SBoris Burkov goto out; 335694846229SBoris Burkov } 335794846229SBoris Burkov 335844c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 335944c0ca21SBoris Burkov if (ret) 336044c0ca21SBoris Burkov goto out; 336144c0ca21SBoris Burkov 336244c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 336344c0ca21SBoris Burkov if (ret) { 336444c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 336544c0ca21SBoris Burkov goto out; 336644c0ca21SBoris Burkov } 336744c0ca21SBoris Burkov 336844c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 336944c0ca21SBoris Burkov 337044c0ca21SBoris Burkov if (!fs_info->uuid_root) { 337144c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 337244c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 337344c0ca21SBoris Burkov if (ret) { 337444c0ca21SBoris Burkov btrfs_warn(fs_info, 337544c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 337644c0ca21SBoris Burkov goto out; 337744c0ca21SBoris Burkov } 337844c0ca21SBoris Burkov } 337944c0ca21SBoris Burkov 338044c0ca21SBoris Burkov out: 338144c0ca21SBoris Burkov return ret; 338244c0ca21SBoris Burkov } 338344c0ca21SBoris Burkov 3384d7f67ac9SQu Wenruo /* 3385d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3386d7f67ac9SQu Wenruo * 3387*2ba48b20SQu Wenruo * @is_rw_mount: If the mount is read-write. 3388*2ba48b20SQu Wenruo * 3389d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3390d7f67ac9SQu Wenruo * feature dependencies). 3391d7f67ac9SQu Wenruo * 3392d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3393d7f67ac9SQu Wenruo * btrfs_validate_super(). 3394d7f67ac9SQu Wenruo * 3395d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3396d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3397d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3398d7f67ac9SQu Wenruo */ 3399*2ba48b20SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) 3400d7f67ac9SQu Wenruo { 3401d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3402d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3403d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3404d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3405d7f67ac9SQu Wenruo 3406d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3407d7f67ac9SQu Wenruo btrfs_err(fs_info, 3408d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3409d7f67ac9SQu Wenruo incompat); 3410d7f67ac9SQu Wenruo return -EINVAL; 3411d7f67ac9SQu Wenruo } 3412d7f67ac9SQu Wenruo 3413d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3414d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3415d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3416d7f67ac9SQu Wenruo btrfs_err(fs_info, 3417d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3418d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3419d7f67ac9SQu Wenruo return -EINVAL; 3420d7f67ac9SQu Wenruo } 3421d7f67ac9SQu Wenruo 3422d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3423d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3424d7f67ac9SQu Wenruo 3425d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3426d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3427d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3428d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3429d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3430d7f67ac9SQu Wenruo 3431d7f67ac9SQu Wenruo /* 3432d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3433d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3434d7f67ac9SQu Wenruo */ 3435d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3436d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3437d7f67ac9SQu Wenruo 3438*2ba48b20SQu Wenruo if (compat_ro_unsupp && is_rw_mount) { 3439d7f67ac9SQu Wenruo btrfs_err(fs_info, 3440d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3441d7f67ac9SQu Wenruo compat_ro); 3442d7f67ac9SQu Wenruo return -EINVAL; 3443d7f67ac9SQu Wenruo } 3444d7f67ac9SQu Wenruo 3445d7f67ac9SQu Wenruo /* 3446d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3447d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3448d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3449d7f67ac9SQu Wenruo */ 3450d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3451d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3452d7f67ac9SQu Wenruo btrfs_err(fs_info, 3453d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3454d7f67ac9SQu Wenruo compat_ro); 3455d7f67ac9SQu Wenruo return -EINVAL; 3456d7f67ac9SQu Wenruo } 3457d7f67ac9SQu Wenruo 3458d7f67ac9SQu Wenruo /* 3459d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3460d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3461d7f67ac9SQu Wenruo */ 3462d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3463d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3464d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3465d7f67ac9SQu Wenruo btrfs_err(fs_info, 3466d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3467d7f67ac9SQu Wenruo return -EINVAL; 3468d7f67ac9SQu Wenruo } 3469d7f67ac9SQu Wenruo 3470d7f67ac9SQu Wenruo /* 3471d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3472d7f67ac9SQu Wenruo * 3473d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3474d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3475d7f67ac9SQu Wenruo * going to be deprecated anyway. 3476d7f67ac9SQu Wenruo */ 3477d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3478d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3479d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3480d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3481d7f67ac9SQu Wenruo return -EINVAL; 3482d7f67ac9SQu Wenruo } 3483d7f67ac9SQu Wenruo 3484d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3485d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3486d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3487d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3488d7f67ac9SQu Wenruo 3489d7f67ac9SQu Wenruo return 0; 3490d7f67ac9SQu Wenruo } 3491d7f67ac9SQu Wenruo 3492ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3493ae18c37aSJosef Bacik char *options) 3494ae18c37aSJosef Bacik { 3495ae18c37aSJosef Bacik u32 sectorsize; 3496ae18c37aSJosef Bacik u32 nodesize; 3497ae18c37aSJosef Bacik u32 stripesize; 3498ae18c37aSJosef Bacik u64 generation; 3499ae18c37aSJosef Bacik u64 features; 3500ae18c37aSJosef Bacik u16 csum_type; 3501ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3502ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3503ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3504ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3505ae18c37aSJosef Bacik int ret; 3506ae18c37aSJosef Bacik int err = -EINVAL; 3507ae18c37aSJosef Bacik int level; 3508ae18c37aSJosef Bacik 35098260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 351053b381b3SDavid Woodhouse if (ret) { 351183c8266aSDavid Sterba err = ret; 3512ae18c37aSJosef Bacik goto fail; 351353b381b3SDavid Woodhouse } 351453b381b3SDavid Woodhouse 3515ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3516ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3517ae18c37aSJosef Bacik GFP_KERNEL); 3518ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3519ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3520ae18c37aSJosef Bacik GFP_KERNEL); 3521ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3522ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 3523ae18c37aSJosef Bacik err = -ENOMEM; 3524c75e8394SJosef Bacik goto fail; 3525ae18c37aSJosef Bacik } 3526ae18c37aSJosef Bacik 3527ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 3528ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 3529ae18c37aSJosef Bacik err = -ENOMEM; 3530c75e8394SJosef Bacik goto fail; 3531ae18c37aSJosef Bacik } 3532ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 3533ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 3534ae18c37aSJosef Bacik 3535d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 35361104a885SDavid Sterba 35371104a885SDavid Sterba /* 35381104a885SDavid Sterba * Read super block and check the signature bytes only 35391104a885SDavid Sterba */ 3540d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 35418f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 35428f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 354316cdcec7SMiao Xie goto fail_alloc; 354420b45077SDave Young } 354539279cc3SChris Mason 35461104a885SDavid Sterba /* 3547260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 35488dc3f22cSJohannes Thumshirn * corrupted, we'll find out 35498dc3f22cSJohannes Thumshirn */ 35508f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 355151bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 35528dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 355351bce6c9SJohannes Thumshirn csum_type); 35548dc3f22cSJohannes Thumshirn err = -EINVAL; 35558f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35568dc3f22cSJohannes Thumshirn goto fail_alloc; 35578dc3f22cSJohannes Thumshirn } 35588dc3f22cSJohannes Thumshirn 355983c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 356083c68bbcSSu Yue 35616d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 35626d97c6e3SJohannes Thumshirn if (ret) { 35636d97c6e3SJohannes Thumshirn err = ret; 35648f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35656d97c6e3SJohannes Thumshirn goto fail_alloc; 35666d97c6e3SJohannes Thumshirn } 35676d97c6e3SJohannes Thumshirn 35688dc3f22cSJohannes Thumshirn /* 35691104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 35701104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 35711104a885SDavid Sterba */ 35723d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 357305135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 35741104a885SDavid Sterba err = -EINVAL; 35758f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3576141386e1SJosef Bacik goto fail_alloc; 35771104a885SDavid Sterba } 35781104a885SDavid Sterba 35791104a885SDavid Sterba /* 35801104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 35811104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 35821104a885SDavid Sterba * the whole block of INFO_SIZE 35831104a885SDavid Sterba */ 35848f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 35858f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35865f39d397SChris Mason 3587fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3588fbc6feaeSNikolay Borisov 35890b86a832SChris Mason 3590fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3591fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3592fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3593fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3594fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3595fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3596fbc6feaeSNikolay Borisov } 3597fbc6feaeSNikolay Borisov 3598fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3599fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3600de37aa51SNikolay Borisov 3601069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 36021104a885SDavid Sterba if (ret) { 360305135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 36041104a885SDavid Sterba err = -EINVAL; 3605141386e1SJosef Bacik goto fail_alloc; 36061104a885SDavid Sterba } 36071104a885SDavid Sterba 36080f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3609141386e1SJosef Bacik goto fail_alloc; 36100f7d52f4SChris Mason 3611acce952bSliubo /* check FS state, whether FS is broken. */ 361287533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 361387533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3614acce952bSliubo 361575e7cb7fSLiu Bo /* 361675e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 361775e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 361875e7cb7fSLiu Bo */ 361975e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 362075e7cb7fSLiu Bo 36216f93e834SAnand Jain 36226f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 36236f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 36246f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 36256f93e834SAnand Jain stripesize = sectorsize; 36266f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 36276f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 36286f93e834SAnand Jain 36296f93e834SAnand Jain fs_info->nodesize = nodesize; 36306f93e834SAnand Jain fs_info->sectorsize = sectorsize; 36316f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 36326f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 36336f93e834SAnand Jain fs_info->stripesize = stripesize; 36346f93e834SAnand Jain 36352ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 36362b82032cSYan Zheng if (ret) { 36372b82032cSYan Zheng err = ret; 3638141386e1SJosef Bacik goto fail_alloc; 36392b82032cSYan Zheng } 3640dfe25020SChris Mason 3641*2ba48b20SQu Wenruo ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); 3642d7f67ac9SQu Wenruo if (ret < 0) { 3643d7f67ac9SQu Wenruo err = ret; 3644141386e1SJosef Bacik goto fail_alloc; 3645f2b636e8SJosef Bacik } 3646f2b636e8SJosef Bacik 36478481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 36488481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 36498481dd80SQu Wenruo 36509f73f1aeSQu Wenruo /* 36519f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 36529f73f1aeSQu Wenruo * going to be deprecated. 36539f73f1aeSQu Wenruo * 36549f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 36559f73f1aeSQu Wenruo */ 36569f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 36579f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 36589f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 36599f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 36609f73f1aeSQu Wenruo 366195ea0486SQu Wenruo btrfs_warn(fs_info, 366295ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 36630bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 36648481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 36658481dd80SQu Wenruo if (!subpage_info) 36668481dd80SQu Wenruo goto fail_alloc; 36678481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 36688481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3669c8050b3bSQu Wenruo } 36700bb3eb3eSQu Wenruo 3671d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 36722a458198SEric Sandeen if (ret) { 36732a458198SEric Sandeen err = ret; 36740dc3b84aSJosef Bacik goto fail_sb_buffer; 36750dc3b84aSJosef Bacik } 36764543df7eSChris Mason 36779e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 36789e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 36794575c9ccSChris Mason 3680a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3681a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3682de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3683db94535dSChris Mason 3684925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 36856bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3686925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 368784eed90fSChris Mason if (ret) { 368805135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 36895d4f98a2SYan Zheng goto fail_sb_buffer; 369084eed90fSChris Mason } 36910b86a832SChris Mason 369284234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3693581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3694bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3695bd676446SJosef Bacik generation, level); 3696bd676446SJosef Bacik if (ret) { 369705135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3698af31f5e5SChris Mason goto fail_tree_roots; 369983121942SDavid Woodhouse } 37000b86a832SChris Mason 3701e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3702c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3703c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3704e17cade2SChris Mason 37055b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 37062b82032cSYan Zheng if (ret) { 370705135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3708af31f5e5SChris Mason goto fail_tree_roots; 37092b82032cSYan Zheng } 37100b86a832SChris Mason 37118dabb742SStefan Behrens /* 3712bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3713bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3714bacce86aSAnand Jain * target is required. So free devices that are not part of this 37151a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3716bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 37178dabb742SStefan Behrens */ 3718bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3719d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 372005135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3721a6b0d5c8SChris Mason goto fail_tree_roots; 3722a6b0d5c8SChris Mason } 3723a6b0d5c8SChris Mason 3724b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 37254bbcaa64SEric Sandeen if (ret) 3726b8522a1eSNikolay Borisov goto fail_tree_roots; 37278929ecfaSYan, Zheng 372875ec1db8SJosef Bacik /* 372973651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 373073651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 373173651042SNaohiro Aota * zoned incompat feature flag set. 373273651042SNaohiro Aota */ 373373651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 373473651042SNaohiro Aota if (ret) { 373573651042SNaohiro Aota btrfs_err(fs_info, 373673651042SNaohiro Aota "zoned: failed to read device zone info: %d", 373773651042SNaohiro Aota ret); 373873651042SNaohiro Aota goto fail_block_groups; 373973651042SNaohiro Aota } 374073651042SNaohiro Aota 374173651042SNaohiro Aota /* 374275ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 374375ec1db8SJosef Bacik * check the generation here so we can set the 374475ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 374575ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 374675ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 374775ec1db8SJosef Bacik * even though it was perfectly fine. 374875ec1db8SJosef Bacik */ 374975ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 375075ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 375175ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 375275ec1db8SJosef Bacik 3753cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3754cf90d884SQu Wenruo if (ret) { 3755cf90d884SQu Wenruo btrfs_err(fs_info, 3756cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3757cf90d884SQu Wenruo ret); 3758cf90d884SQu Wenruo goto fail_block_groups; 3759cf90d884SQu Wenruo } 376068310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 376168310a5eSIlya Dryomov if (ret) { 376205135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 376368310a5eSIlya Dryomov goto fail_block_groups; 376468310a5eSIlya Dryomov } 376568310a5eSIlya Dryomov 3766733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3767733f4fbbSStefan Behrens if (ret) { 376805135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3769733f4fbbSStefan Behrens goto fail_block_groups; 3770733f4fbbSStefan Behrens } 3771733f4fbbSStefan Behrens 37728dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 37738dabb742SStefan Behrens if (ret) { 377405135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 37758dabb742SStefan Behrens goto fail_block_groups; 37768dabb742SStefan Behrens } 37778dabb742SStefan Behrens 3778b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3779b70f5097SNaohiro Aota if (ret) { 3780b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3781b70f5097SNaohiro Aota ret); 3782b70f5097SNaohiro Aota goto fail_block_groups; 3783b70f5097SNaohiro Aota } 3784b70f5097SNaohiro Aota 3785c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3786b7c35e81SAnand Jain if (ret) { 378705135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 378805135f59SDavid Sterba ret); 3789b7c35e81SAnand Jain goto fail_block_groups; 3790b7c35e81SAnand Jain } 3791b7c35e81SAnand Jain 379296f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 37935ac1d209SJeff Mahoney if (ret) { 379405135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3795b7c35e81SAnand Jain goto fail_fsdev_sysfs; 37965ac1d209SJeff Mahoney } 37975ac1d209SJeff Mahoney 3798c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3799c59021f8Sliubo if (ret) { 380005135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 38012365dd3cSAnand Jain goto fail_sysfs; 3802c59021f8Sliubo } 3803c59021f8Sliubo 38045b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 38051b1d1f66SJosef Bacik if (ret) { 380605135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 38072365dd3cSAnand Jain goto fail_sysfs; 38081b1d1f66SJosef Bacik } 38094330e183SQu Wenruo 381016beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 381116beac87SNaohiro Aota 38125c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 38135c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 381405135f59SDavid Sterba btrfs_warn(fs_info, 381552042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 38162365dd3cSAnand Jain goto fail_sysfs; 3817292fd7fcSStefan Behrens } 38189078a3e1SChris Mason 381933c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3820a74a4b97SChris Mason "btrfs-cleaner"); 382157506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 38222365dd3cSAnand Jain goto fail_sysfs; 3823a74a4b97SChris Mason 3824a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3825a74a4b97SChris Mason tree_root, 3826a74a4b97SChris Mason "btrfs-transaction"); 382757506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 38283f157a2fSChris Mason goto fail_cleaner; 3829a74a4b97SChris Mason 3830583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3831c289811cSChris Mason !fs_info->fs_devices->rotating) { 3832583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3833c289811cSChris Mason } 3834c289811cSChris Mason 3835572d9ab7SDavid Sterba /* 383663a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 383763a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 383863a7cb13SDavid Sterba * nodiscard for the same mount. 383963a7cb13SDavid Sterba */ 384063a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 384163a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 384263a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 384363a7cb13SDavid Sterba fs_info->fs_devices->discardable) { 384463a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 384563a7cb13SDavid Sterba "auto enabling async discard"); 384663a7cb13SDavid Sterba btrfs_clear_opt(fs_info->mount_opt, NODISCARD); 384763a7cb13SDavid Sterba } 384863a7cb13SDavid Sterba 384921adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 38500b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 38512ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 38520b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3853cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 385421adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 385521adbd5cSStefan Behrens if (ret) 385605135f59SDavid Sterba btrfs_warn(fs_info, 385705135f59SDavid Sterba "failed to initialize integrity check module: %d", 385805135f59SDavid Sterba ret); 385921adbd5cSStefan Behrens } 386021adbd5cSStefan Behrens #endif 3861bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3862bcef60f2SArne Jansen if (ret) 3863bcef60f2SArne Jansen goto fail_trans_kthread; 386421adbd5cSStefan Behrens 3865fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3866fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3867fd708b81SJosef Bacik 386896da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 386996da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 38700b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3871e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 387263443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 387379787eaaSJeff Mahoney if (ret) { 387463443bf5SEric Sandeen err = ret; 387528c16cbbSWang Shilong goto fail_qgroup; 3876e556ce2cSYan Zheng } 3877e02119d5SChris Mason } 38781a40e23bSZheng Yan 387956e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 38803140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 38813140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3882f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3883315bf8efSJosef Bacik fs_info->fs_root = NULL; 3884bcef60f2SArne Jansen goto fail_qgroup; 38853140c9a3SDan Carpenter } 3886c289811cSChris Mason 3887bc98a42cSDavid Howells if (sb_rdonly(sb)) 38888cd29088SBoris Burkov goto clear_oneshot; 38892b6ba629SIlya Dryomov 389044c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 38912b6ba629SIlya Dryomov if (ret) { 38926bccf3abSJeff Mahoney close_ctree(fs_info); 38932b6ba629SIlya Dryomov return ret; 3894e3acc2a6SJosef Bacik } 3895b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3896b382a324SJan Schmidt 389744c0ca21SBoris Burkov if (fs_info->uuid_root && 389844c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 389944c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 390005135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 390170f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 390270f80175SStefan Behrens if (ret) { 390305135f59SDavid Sterba btrfs_warn(fs_info, 390405135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 39056bccf3abSJeff Mahoney close_ctree(fs_info); 390670f80175SStefan Behrens return ret; 390770f80175SStefan Behrens } 3908f7a81ea4SStefan Behrens } 390994846229SBoris Burkov 3910afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 391147ab2a6cSJosef Bacik 3912b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3913b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3914b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3915b4be6aefSJosef Bacik 39168cd29088SBoris Burkov clear_oneshot: 39178cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3918ad2b2c80SAl Viro return 0; 391939279cc3SChris Mason 3920bcef60f2SArne Jansen fail_qgroup: 3921bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 39227c2ca468SChris Mason fail_trans_kthread: 39237c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 39242ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3925faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 39263f157a2fSChris Mason fail_cleaner: 3927a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 39287c2ca468SChris Mason 39297c2ca468SChris Mason /* 39307c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 39317c2ca468SChris Mason * kthreads 39327c2ca468SChris Mason */ 39337c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 39347c2ca468SChris Mason 39352365dd3cSAnand Jain fail_sysfs: 39366618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 39372365dd3cSAnand Jain 3938b7c35e81SAnand Jain fail_fsdev_sysfs: 3939b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3940b7c35e81SAnand Jain 39411b1d1f66SJosef Bacik fail_block_groups: 394254067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3943af31f5e5SChris Mason 3944af31f5e5SChris Mason fail_tree_roots: 39459e3aa805SJosef Bacik if (fs_info->data_reloc_root) 39469e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 39474273eaffSAnand Jain free_root_pointers(fs_info, true); 39482b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3949af31f5e5SChris Mason 395039279cc3SChris Mason fail_sb_buffer: 39517abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 39525cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 395316cdcec7SMiao Xie fail_alloc: 3954586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3955586e46e2SIlya Dryomov 39564543df7eSChris Mason iput(fs_info->btree_inode); 39577e662854SQinghuang Feng fail: 3958586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3959ad2b2c80SAl Viro return err; 3960eb60ceacSChris Mason } 3961663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3962eb60ceacSChris Mason 3963314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3964f2984462SChris Mason { 3965314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3966314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3967314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3968314b6dd0SJohannes Thumshirn struct page *page; 3969442a4f63SStefan Behrens 3970314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3971314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3972314b6dd0SJohannes Thumshirn 3973314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3974fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3975314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3976cb3e217bSQu Wenruo btrfs_dev_name(device), 3977314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3978314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3979314b6dd0SJohannes Thumshirn SetPageError(page); 3980314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3981314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3982314b6dd0SJohannes Thumshirn } else { 3983314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3984f2984462SChris Mason } 3985314b6dd0SJohannes Thumshirn 3986314b6dd0SJohannes Thumshirn put_page(page); 3987314b6dd0SJohannes Thumshirn unlock_page(page); 3988314b6dd0SJohannes Thumshirn } 3989314b6dd0SJohannes Thumshirn 3990314b6dd0SJohannes Thumshirn bio_put(bio); 3991f2984462SChris Mason } 3992f2984462SChris Mason 39938f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3994a05d3c91SQu Wenruo int copy_num, bool drop_cache) 399529c36d72SAnand Jain { 399629c36d72SAnand Jain struct btrfs_super_block *super; 39978f32380dSJohannes Thumshirn struct page *page; 399812659251SNaohiro Aota u64 bytenr, bytenr_orig; 39998f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 400012659251SNaohiro Aota int ret; 400129c36d72SAnand Jain 400212659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 400312659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 400412659251SNaohiro Aota if (ret == -ENOENT) 400512659251SNaohiro Aota return ERR_PTR(-EINVAL); 400612659251SNaohiro Aota else if (ret) 400712659251SNaohiro Aota return ERR_PTR(ret); 400812659251SNaohiro Aota 4009cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 40108f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 401129c36d72SAnand Jain 4012a05d3c91SQu Wenruo if (drop_cache) { 4013a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 4014a05d3c91SQu Wenruo ASSERT(copy_num == 0); 4015a05d3c91SQu Wenruo 4016a05d3c91SQu Wenruo /* 4017a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 4018a05d3c91SQu Wenruo * always read from the device. 4019a05d3c91SQu Wenruo */ 4020a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 4021a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 4022a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 4023a05d3c91SQu Wenruo } 4024a05d3c91SQu Wenruo 40258f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 40268f32380dSJohannes Thumshirn if (IS_ERR(page)) 40278f32380dSJohannes Thumshirn return ERR_CAST(page); 402829c36d72SAnand Jain 40298f32380dSJohannes Thumshirn super = page_address(page); 403096c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 403196c2e067SAnand Jain btrfs_release_disk_super(super); 403296c2e067SAnand Jain return ERR_PTR(-ENODATA); 403396c2e067SAnand Jain } 403496c2e067SAnand Jain 403512659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 40368f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 40378f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 403829c36d72SAnand Jain } 403929c36d72SAnand Jain 40408f32380dSJohannes Thumshirn return super; 404129c36d72SAnand Jain } 404229c36d72SAnand Jain 404329c36d72SAnand Jain 40448f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 4045a512bbf8SYan Zheng { 40468f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 4047a512bbf8SYan Zheng int i; 4048a512bbf8SYan Zheng u64 transid = 0; 4049a512bbf8SYan Zheng 4050a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 4051a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 4052a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 4053a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 4054a512bbf8SYan Zheng */ 4055a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 4056a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 40578f32380dSJohannes Thumshirn if (IS_ERR(super)) 4058a512bbf8SYan Zheng continue; 4059a512bbf8SYan Zheng 4060a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 40618f32380dSJohannes Thumshirn if (latest) 40628f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 40638f32380dSJohannes Thumshirn 40648f32380dSJohannes Thumshirn latest = super; 4065a512bbf8SYan Zheng transid = btrfs_super_generation(super); 4066a512bbf8SYan Zheng } 4067a512bbf8SYan Zheng } 406892fc03fbSAnand Jain 40698f32380dSJohannes Thumshirn return super; 4070a512bbf8SYan Zheng } 4071a512bbf8SYan Zheng 40724eedeb75SHisashi Hifumi /* 4073abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 4074314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 40754eedeb75SHisashi Hifumi * 4076abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 4077abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 4078abbb3b8eSDavid Sterba * same for write and wait phases. 40794eedeb75SHisashi Hifumi * 4080314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 40814eedeb75SHisashi Hifumi */ 4082a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 4083abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 4084a512bbf8SYan Zheng { 4085d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 4086314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 4087d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 4088a512bbf8SYan Zheng int i; 4089a512bbf8SYan Zheng int errors = 0; 409012659251SNaohiro Aota int ret; 409112659251SNaohiro Aota u64 bytenr, bytenr_orig; 4092a512bbf8SYan Zheng 4093a512bbf8SYan Zheng if (max_mirrors == 0) 4094a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4095a512bbf8SYan Zheng 4096d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 4097d5178578SJohannes Thumshirn 4098a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 4099314b6dd0SJohannes Thumshirn struct page *page; 4100314b6dd0SJohannes Thumshirn struct bio *bio; 4101314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 4102314b6dd0SJohannes Thumshirn 410312659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 410412659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 410512659251SNaohiro Aota if (ret == -ENOENT) { 410612659251SNaohiro Aota continue; 410712659251SNaohiro Aota } else if (ret < 0) { 410812659251SNaohiro Aota btrfs_err(device->fs_info, 410912659251SNaohiro Aota "couldn't get super block location for mirror %d", 411012659251SNaohiro Aota i); 411112659251SNaohiro Aota errors++; 411212659251SNaohiro Aota continue; 411312659251SNaohiro Aota } 4114935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4115935e5cc9SMiao Xie device->commit_total_bytes) 4116a512bbf8SYan Zheng break; 4117a512bbf8SYan Zheng 411812659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 4119a512bbf8SYan Zheng 4120fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 4121fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 4122fd08001fSEric Biggers sb->csum); 4123a512bbf8SYan Zheng 4124314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 4125314b6dd0SJohannes Thumshirn GFP_NOFS); 4126314b6dd0SJohannes Thumshirn if (!page) { 4127fb456252SJeff Mahoney btrfs_err(device->fs_info, 4128314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 4129f14d104dSDavid Sterba bytenr); 4130634554dcSJosef Bacik errors++; 4131634554dcSJosef Bacik continue; 4132634554dcSJosef Bacik } 4133634554dcSJosef Bacik 4134314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 4135314b6dd0SJohannes Thumshirn get_page(page); 4136a512bbf8SYan Zheng 4137314b6dd0SJohannes Thumshirn disk_super = page_address(page); 4138314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 4139a512bbf8SYan Zheng 4140387125fcSChris Mason /* 4141314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 4142314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 4143314b6dd0SJohannes Thumshirn * checking. 4144387125fcSChris Mason */ 414507888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 414607888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 414707888c66SChristoph Hellwig GFP_NOFS); 4148314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 4149314b6dd0SJohannes Thumshirn bio->bi_private = device; 4150314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 4151314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 4152314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 4153314b6dd0SJohannes Thumshirn 4154314b6dd0SJohannes Thumshirn /* 4155314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 4156314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 4157314b6dd0SJohannes Thumshirn * copies might still contain the older version. 4158314b6dd0SJohannes Thumshirn */ 41591b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 4160314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 4161314b6dd0SJohannes Thumshirn 416258ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 416358ff51f1SChristoph Hellwig submit_bio(bio); 41648376d9e1SNaohiro Aota 41658376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 41668376d9e1SNaohiro Aota errors++; 4167a512bbf8SYan Zheng } 4168a512bbf8SYan Zheng return errors < i ? 0 : -1; 4169a512bbf8SYan Zheng } 4170a512bbf8SYan Zheng 4171387125fcSChris Mason /* 4172abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 4173abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 4174abbb3b8eSDavid Sterba * 4175314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 4176abbb3b8eSDavid Sterba * date. 4177abbb3b8eSDavid Sterba */ 4178abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 4179abbb3b8eSDavid Sterba { 4180abbb3b8eSDavid Sterba int i; 4181abbb3b8eSDavid Sterba int errors = 0; 4182b6a535faSHoward McLauchlan bool primary_failed = false; 418312659251SNaohiro Aota int ret; 4184abbb3b8eSDavid Sterba u64 bytenr; 4185abbb3b8eSDavid Sterba 4186abbb3b8eSDavid Sterba if (max_mirrors == 0) 4187abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4188abbb3b8eSDavid Sterba 4189abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 4190314b6dd0SJohannes Thumshirn struct page *page; 4191314b6dd0SJohannes Thumshirn 419212659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 419312659251SNaohiro Aota if (ret == -ENOENT) { 419412659251SNaohiro Aota break; 419512659251SNaohiro Aota } else if (ret < 0) { 419612659251SNaohiro Aota errors++; 419712659251SNaohiro Aota if (i == 0) 419812659251SNaohiro Aota primary_failed = true; 419912659251SNaohiro Aota continue; 420012659251SNaohiro Aota } 4201abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4202abbb3b8eSDavid Sterba device->commit_total_bytes) 4203abbb3b8eSDavid Sterba break; 4204abbb3b8eSDavid Sterba 4205314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 4206314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 4207314b6dd0SJohannes Thumshirn if (!page) { 4208abbb3b8eSDavid Sterba errors++; 4209b6a535faSHoward McLauchlan if (i == 0) 4210b6a535faSHoward McLauchlan primary_failed = true; 4211abbb3b8eSDavid Sterba continue; 4212abbb3b8eSDavid Sterba } 4213314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 4214314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 4215314b6dd0SJohannes Thumshirn if (PageError(page)) { 4216abbb3b8eSDavid Sterba errors++; 4217b6a535faSHoward McLauchlan if (i == 0) 4218b6a535faSHoward McLauchlan primary_failed = true; 4219b6a535faSHoward McLauchlan } 4220abbb3b8eSDavid Sterba 4221314b6dd0SJohannes Thumshirn /* Drop our reference */ 4222314b6dd0SJohannes Thumshirn put_page(page); 4223abbb3b8eSDavid Sterba 4224314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 4225314b6dd0SJohannes Thumshirn put_page(page); 4226abbb3b8eSDavid Sterba } 4227abbb3b8eSDavid Sterba 4228b6a535faSHoward McLauchlan /* log error, force error return */ 4229b6a535faSHoward McLauchlan if (primary_failed) { 4230b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 4231b6a535faSHoward McLauchlan device->devid); 4232b6a535faSHoward McLauchlan return -1; 4233b6a535faSHoward McLauchlan } 4234b6a535faSHoward McLauchlan 4235abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 4236abbb3b8eSDavid Sterba } 4237abbb3b8eSDavid Sterba 4238abbb3b8eSDavid Sterba /* 4239387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 4240387125fcSChris Mason * for the barrier when it is done 4241387125fcSChris Mason */ 42424246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 4243387125fcSChris Mason { 4244f9e69aa9SChristoph Hellwig bio_uninit(bio); 4245387125fcSChris Mason complete(bio->bi_private); 4246387125fcSChris Mason } 4247387125fcSChris Mason 4248387125fcSChris Mason /* 42494fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 42504fc6441aSAnand Jain * done in the waiting counterpart. 4251387125fcSChris Mason */ 42524fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 4253387125fcSChris Mason { 4254f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 4255387125fcSChris Mason 4256a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 4257a91cf0ffSWang Yugui /* 4258a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 4259a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 4260a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 4261a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 4262a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 4263a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 4264a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 4265a91cf0ffSWang Yugui * non-debug builds. 4266a91cf0ffSWang Yugui */ 426708e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 42684fc6441aSAnand Jain return; 4269a91cf0ffSWang Yugui #endif 4270387125fcSChris Mason 4271f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 4272f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 4273387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 4274387125fcSChris Mason init_completion(&device->flush_wait); 4275387125fcSChris Mason bio->bi_private = &device->flush_wait; 4276387125fcSChris Mason 427758ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 427858ff51f1SChristoph Hellwig submit_bio(bio); 42791c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 42804fc6441aSAnand Jain } 4281387125fcSChris Mason 42824fc6441aSAnand Jain /* 42834fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 42844fc6441aSAnand Jain */ 42858c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 42864fc6441aSAnand Jain { 4287f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 42884fc6441aSAnand Jain 42891c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 429058efbc9fSOmar Sandoval return BLK_STS_OK; 42914fc6441aSAnand Jain 42921c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 42932980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 42944fc6441aSAnand Jain 42958c27cb35SLinus Torvalds return bio->bi_status; 4296387125fcSChris Mason } 4297387125fcSChris Mason 4298d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 4299401b41e5SAnand Jain { 43006528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 4301401b41e5SAnand Jain return -EIO; 4302387125fcSChris Mason return 0; 4303387125fcSChris Mason } 4304387125fcSChris Mason 4305387125fcSChris Mason /* 4306387125fcSChris Mason * send an empty flush down to each device in parallel, 4307387125fcSChris Mason * then wait for them 4308387125fcSChris Mason */ 4309387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 4310387125fcSChris Mason { 4311387125fcSChris Mason struct list_head *head; 4312387125fcSChris Mason struct btrfs_device *dev; 43135af3e8ccSStefan Behrens int errors_wait = 0; 43144e4cbee9SChristoph Hellwig blk_status_t ret; 4315387125fcSChris Mason 43161538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 4317387125fcSChris Mason /* send down all the barriers */ 4318387125fcSChris Mason head = &info->fs_devices->devices; 43191538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4320e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4321f88ba6a2SHidetoshi Seto continue; 4322cea7c8bfSAnand Jain if (!dev->bdev) 4323387125fcSChris Mason continue; 4324e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4325ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4326387125fcSChris Mason continue; 4327387125fcSChris Mason 43284fc6441aSAnand Jain write_dev_flush(dev); 432958efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 4330387125fcSChris Mason } 4331387125fcSChris Mason 4332387125fcSChris Mason /* wait for all the barriers */ 43331538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4334e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4335f88ba6a2SHidetoshi Seto continue; 4336387125fcSChris Mason if (!dev->bdev) { 43375af3e8ccSStefan Behrens errors_wait++; 4338387125fcSChris Mason continue; 4339387125fcSChris Mason } 4340e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4341ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4342387125fcSChris Mason continue; 4343387125fcSChris Mason 43444fc6441aSAnand Jain ret = wait_dev_flush(dev); 4345401b41e5SAnand Jain if (ret) { 4346401b41e5SAnand Jain dev->last_flush_error = ret; 434766b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 434866b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 43495af3e8ccSStefan Behrens errors_wait++; 4350387125fcSChris Mason } 4351401b41e5SAnand Jain } 4352401b41e5SAnand Jain 4353cea7c8bfSAnand Jain if (errors_wait) { 4354401b41e5SAnand Jain /* 4355401b41e5SAnand Jain * At some point we need the status of all disks 4356401b41e5SAnand Jain * to arrive at the volume status. So error checking 4357401b41e5SAnand Jain * is being pushed to a separate loop. 4358401b41e5SAnand Jain */ 4359d10b82feSQu Wenruo return check_barrier_error(info); 4360401b41e5SAnand Jain } 4361387125fcSChris Mason return 0; 4362387125fcSChris Mason } 4363387125fcSChris Mason 4364943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4365943c6e99SZhao Lei { 43668789f4feSZhao Lei int raid_type; 43678789f4feSZhao Lei int min_tolerated = INT_MAX; 4368943c6e99SZhao Lei 43698789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 43708789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 43718c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 43728789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 43738789f4feSZhao Lei tolerated_failures); 4374943c6e99SZhao Lei 43758789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 43768789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 43778789f4feSZhao Lei continue; 437841a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 43798789f4feSZhao Lei continue; 43808c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 43818789f4feSZhao Lei btrfs_raid_array[raid_type]. 43828789f4feSZhao Lei tolerated_failures); 43838789f4feSZhao Lei } 4384943c6e99SZhao Lei 43858789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4386ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 43878789f4feSZhao Lei min_tolerated = 0; 43888789f4feSZhao Lei } 43898789f4feSZhao Lei 43908789f4feSZhao Lei return min_tolerated; 4391943c6e99SZhao Lei } 4392943c6e99SZhao Lei 4393eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4394f2984462SChris Mason { 4395e5e9a520SChris Mason struct list_head *head; 4396f2984462SChris Mason struct btrfs_device *dev; 4397a061fc8dSChris Mason struct btrfs_super_block *sb; 4398f2984462SChris Mason struct btrfs_dev_item *dev_item; 4399f2984462SChris Mason int ret; 4400f2984462SChris Mason int do_barriers; 4401a236aed1SChris Mason int max_errors; 4402a236aed1SChris Mason int total_errors = 0; 4403a061fc8dSChris Mason u64 flags; 4404f2984462SChris Mason 44050b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4406fed3b381SLiu Bo 4407fed3b381SLiu Bo /* 4408fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4409fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4410fed3b381SLiu Bo * been consistent on disk. 4411fed3b381SLiu Bo */ 4412fed3b381SLiu Bo if (max_mirrors == 0) 44130b246afaSJeff Mahoney backup_super_roots(fs_info); 4414f2984462SChris Mason 44150b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4416a061fc8dSChris Mason dev_item = &sb->dev_item; 4417e5e9a520SChris Mason 44180b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 44190b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 44200b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4421387125fcSChris Mason 44225af3e8ccSStefan Behrens if (do_barriers) { 44230b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 44245af3e8ccSStefan Behrens if (ret) { 44255af3e8ccSStefan Behrens mutex_unlock( 44260b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 44270b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 44285af3e8ccSStefan Behrens "errors while submitting device barriers."); 44295af3e8ccSStefan Behrens return ret; 44305af3e8ccSStefan Behrens } 44315af3e8ccSStefan Behrens } 4432387125fcSChris Mason 44331538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4434dfe25020SChris Mason if (!dev->bdev) { 4435dfe25020SChris Mason total_errors++; 4436dfe25020SChris Mason continue; 4437dfe25020SChris Mason } 4438e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4439ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4440dfe25020SChris Mason continue; 4441dfe25020SChris Mason 44422b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4443a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4444a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 44457df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4446935e5cc9SMiao Xie dev->commit_total_bytes); 4447ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4448ce7213c7SMiao Xie dev->commit_bytes_used); 4449a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4450a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4451a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4452a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 44537239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 44547239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4455a512bbf8SYan Zheng 4456a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4457a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4458f2984462SChris Mason 445975cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 446075cb857dSQu Wenruo if (ret < 0) { 446175cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 446275cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 446375cb857dSQu Wenruo "unexpected superblock corruption detected"); 446475cb857dSQu Wenruo return -EUCLEAN; 446575cb857dSQu Wenruo } 446675cb857dSQu Wenruo 4467abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4468a236aed1SChris Mason if (ret) 4469a236aed1SChris Mason total_errors++; 4470f2984462SChris Mason } 4471a236aed1SChris Mason if (total_errors > max_errors) { 44720b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4473d397712bSChris Mason total_errors); 44740b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 447579787eaaSJeff Mahoney 44769d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 44770b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 44780b246afaSJeff Mahoney "%d errors while writing supers", 44790b246afaSJeff Mahoney total_errors); 44809d565ba4SStefan Behrens return -EIO; 4481a236aed1SChris Mason } 4482f2984462SChris Mason 4483a512bbf8SYan Zheng total_errors = 0; 44841538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4485dfe25020SChris Mason if (!dev->bdev) 4486dfe25020SChris Mason continue; 4487e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4488ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4489dfe25020SChris Mason continue; 4490dfe25020SChris Mason 4491abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4492a512bbf8SYan Zheng if (ret) 44931259ab75SChris Mason total_errors++; 4494f2984462SChris Mason } 44950b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4496a236aed1SChris Mason if (total_errors > max_errors) { 44970b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 44980b246afaSJeff Mahoney "%d errors while writing supers", 44990b246afaSJeff Mahoney total_errors); 450079787eaaSJeff Mahoney return -EIO; 4501a236aed1SChris Mason } 4502f2984462SChris Mason return 0; 4503f2984462SChris Mason } 4504f2984462SChris Mason 4505cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4506cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4507cb517eabSMiao Xie struct btrfs_root *root) 45082619ba1fSChris Mason { 45094785e24fSJosef Bacik bool drop_ref = false; 45104785e24fSJosef Bacik 4511fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4512fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4513fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4514fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 45154785e24fSJosef Bacik drop_ref = true; 4516fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 451776dda93cSYan, Zheng 451884961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4519ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 45201c1ea4f7SLiu Bo if (root->reloc_root) { 452100246528SJosef Bacik btrfs_put_root(root->reloc_root); 45221c1ea4f7SLiu Bo root->reloc_root = NULL; 45231c1ea4f7SLiu Bo } 45241c1ea4f7SLiu Bo } 45253321719eSLiu Bo 45264785e24fSJosef Bacik if (drop_ref) 452700246528SJosef Bacik btrfs_put_root(root); 45282619ba1fSChris Mason } 45292619ba1fSChris Mason 4530c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4531c146afadSYan Zheng { 4532fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4533fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4534fc7cbcd4SDavid Sterba int i = 0; 453565d33fd7SQu Wenruo int err = 0; 4536fc7cbcd4SDavid Sterba unsigned int ret = 0; 4537c146afadSYan Zheng 4538c146afadSYan Zheng while (1) { 4539fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4540fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4541fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4542fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4543fc7cbcd4SDavid Sterba if (!ret) { 4544fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4545c146afadSYan Zheng break; 454665d33fd7SQu Wenruo } 4547fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 454866b4ffd1SJosef Bacik 4549fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4550fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4551fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4552fc7cbcd4SDavid Sterba gang[i] = NULL; 455365d33fd7SQu Wenruo continue; 4554c146afadSYan Zheng } 4555fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4556fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4557fc7cbcd4SDavid Sterba } 4558fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4559fc7cbcd4SDavid Sterba 4560fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4561fc7cbcd4SDavid Sterba if (!gang[i]) 4562fc7cbcd4SDavid Sterba continue; 4563fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4564fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4565fc7cbcd4SDavid Sterba if (err) 4566fc7cbcd4SDavid Sterba break; 4567fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4568fc7cbcd4SDavid Sterba } 4569fc7cbcd4SDavid Sterba root_objectid++; 4570c146afadSYan Zheng } 457165d33fd7SQu Wenruo 4572fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4573fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4574fc7cbcd4SDavid Sterba if (gang[i]) 4575fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 457665d33fd7SQu Wenruo } 457765d33fd7SQu Wenruo return err; 4578c146afadSYan Zheng } 4579c146afadSYan Zheng 45806bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4581c146afadSYan Zheng { 45826bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4583c146afadSYan Zheng struct btrfs_trans_handle *trans; 4584c146afadSYan Zheng 45850b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45862ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45870b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 45880b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4589c71bf099SYan, Zheng 4590c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 45910b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45920b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4593c71bf099SYan, Zheng 45947a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 45953612b495STsutomu Itoh if (IS_ERR(trans)) 45963612b495STsutomu Itoh return PTR_ERR(trans); 45973a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4598c146afadSYan Zheng } 4599c146afadSYan Zheng 460036c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 460136c86a9eSQu Wenruo { 460236c86a9eSQu Wenruo struct btrfs_transaction *trans; 460336c86a9eSQu Wenruo struct btrfs_transaction *tmp; 460436c86a9eSQu Wenruo bool found = false; 460536c86a9eSQu Wenruo 460636c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 460736c86a9eSQu Wenruo return; 460836c86a9eSQu Wenruo 460936c86a9eSQu Wenruo /* 461036c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 461136c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 461236c86a9eSQu Wenruo */ 461336c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 461436c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 461536c86a9eSQu Wenruo struct extent_state *cached = NULL; 461636c86a9eSQu Wenruo u64 dirty_bytes = 0; 461736c86a9eSQu Wenruo u64 cur = 0; 461836c86a9eSQu Wenruo u64 found_start; 461936c86a9eSQu Wenruo u64 found_end; 462036c86a9eSQu Wenruo 462136c86a9eSQu Wenruo found = true; 462236c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 462336c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 462436c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 462536c86a9eSQu Wenruo cur = found_end + 1; 462636c86a9eSQu Wenruo } 462736c86a9eSQu Wenruo btrfs_warn(fs_info, 462836c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 462936c86a9eSQu Wenruo trans->transid, dirty_bytes); 463036c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 463136c86a9eSQu Wenruo 463236c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 463336c86a9eSQu Wenruo fs_info->running_transaction = NULL; 463436c86a9eSQu Wenruo list_del_init(&trans->list); 463536c86a9eSQu Wenruo 463636c86a9eSQu Wenruo btrfs_put_transaction(trans); 463736c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 463836c86a9eSQu Wenruo } 463936c86a9eSQu Wenruo ASSERT(!found); 464036c86a9eSQu Wenruo } 464136c86a9eSQu Wenruo 4642b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4643eb60ceacSChris Mason { 4644c146afadSYan Zheng int ret; 4645e089f05cSChris Mason 4646afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 464731e70e52SFilipe Manana 464831e70e52SFilipe Manana /* 46498a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 46508a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 46518a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 46528a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 46538a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 46548a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 46558a1f1e3dSFilipe Manana * return 1. 46568a1f1e3dSFilipe Manana */ 46578a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 46588a1f1e3dSFilipe Manana 46598a1f1e3dSFilipe Manana /* 466031e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 466131e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 466231e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 466331e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 466431e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 466531e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 466631e70e52SFilipe Manana * trying to stop the async reclaim task. 466731e70e52SFilipe Manana */ 466831e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4669d6fd0ae2SOmar Sandoval /* 4670d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4671d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4672d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4673d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4674d6fd0ae2SOmar Sandoval */ 4675d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4676a2135011SChris Mason 46777343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4678d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 46797343dd61SJustin Maggard 4680803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4681803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4682803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4683803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4684803b2f54SStefan Behrens 4685837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4686aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4687837d5b6eSIlya Dryomov 46888dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 46898dabb742SStefan Behrens 4690aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 46914cb5300bSChris Mason 46924cb5300bSChris Mason /* wait for any defraggers to finish */ 46934cb5300bSChris Mason wait_event(fs_info->transaction_wait, 46944cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 46954cb5300bSChris Mason 46964cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 469726176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 46984cb5300bSChris Mason 4699a362bb86SFilipe Manana /* 4700a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4701a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4702a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4703a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4704a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4705a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4706a362bb86SFilipe Manana * no one else to run iputs. 4707a362bb86SFilipe Manana * 4708a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4709a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4710a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4711a362bb86SFilipe Manana * through some other means. 4712a362bb86SFilipe Manana * 4713a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4714a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4715a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4716a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4717a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4718a362bb86SFilipe Manana */ 4719a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4720a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4721a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4722a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4723a362bb86SFilipe Manana 472421c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 472557056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4726576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 472721c7e756SMiao Xie 4728b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4729b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4730b0643e59SDennis Zhou 4731bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4732e44163e1SJeff Mahoney /* 4733d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4734d6fd0ae2SOmar Sandoval * unused block groups. 4735e44163e1SJeff Mahoney */ 47360b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4737e44163e1SJeff Mahoney 4738f0cc2cd7SFilipe Manana /* 4739f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4740f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4741f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4742f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4743f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4744f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4745f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4746f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4747f0cc2cd7SFilipe Manana * This is a very rare case. 4748f0cc2cd7SFilipe Manana */ 4749f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4750f0cc2cd7SFilipe Manana 47516bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4752d397712bSChris Mason if (ret) 475304892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4754c146afadSYan Zheng } 4755ed2ff2cbSChris Mason 475684961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 47572ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4758acce952bSliubo 4759e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4760e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 47618929ecfaSYan, Zheng 4762e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4763afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4764f25784b3SYan Zheng 47655958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 47665958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 47675958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 47685958253cSQu Wenruo } 47695958253cSQu Wenruo 477004892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4771fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4772bcef60f2SArne Jansen 4773963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 477404892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4775963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4776b0c68f8bSChris Mason } 477731153d81SYan Zheng 47785deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 47794297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 47805deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 47814297ff84SJosef Bacik 47826618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4783b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 47845ac1d209SJeff Mahoney 47851a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 47861a4319ccSLiu Bo 4787de348ee0SWang Shilong /* 4788de348ee0SWang Shilong * we must make sure there is not any read request to 4789de348ee0SWang Shilong * submit after we stopping all workers. 4790de348ee0SWang Shilong */ 4791de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 479296192499SJosef Bacik btrfs_stop_all_workers(fs_info); 479396192499SJosef Bacik 47940a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 479536c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 47960a31daa4SFilipe Manana 4797afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 47984273eaffSAnand Jain free_root_pointers(fs_info, true); 47998c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 48009ad6b7bcSChris Mason 48014e19443dSJosef Bacik /* 48024e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 48034e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 48044e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 48054e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 48064e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 48074e19443dSJosef Bacik */ 48084e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 48094e19443dSJosef Bacik 481013e6c37bSJosef Bacik iput(fs_info->btree_inode); 4811d6bfde87SChris Mason 481221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 48130b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 48142ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 481521adbd5cSStefan Behrens #endif 481621adbd5cSStefan Behrens 48170b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 481868c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4819eb60ceacSChris Mason } 4820eb60ceacSChris Mason 4821b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4822b9fab919SChris Mason int atomic) 4823ccd467d6SChris Mason { 48241259ab75SChris Mason int ret; 4825727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 48261259ab75SChris Mason 48270b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 48281259ab75SChris Mason if (!ret) 48291259ab75SChris Mason return ret; 48301259ab75SChris Mason 48311259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4832b9fab919SChris Mason parent_transid, atomic); 4833b9fab919SChris Mason if (ret == -EAGAIN) 4834b9fab919SChris Mason return ret; 48351259ab75SChris Mason return !ret; 48365f39d397SChris Mason } 48376702ed49SChris Mason 48385f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 48395f39d397SChris Mason { 48402f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 48415f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4842b9473439SChris Mason int was_dirty; 4843b4ce94deSChris Mason 484406ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 484506ea65a3SJosef Bacik /* 484606ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 484752042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 484806ea65a3SJosef Bacik * outside of the sanity tests. 484906ea65a3SJosef Bacik */ 4850b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 485106ea65a3SJosef Bacik return; 485206ea65a3SJosef Bacik #endif 485349d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 48540b246afaSJeff Mahoney if (transid != fs_info->generation) 48555d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 48560b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 48570b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4858e2d84521SMiao Xie if (!was_dirty) 4859104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4860e2d84521SMiao Xie buf->len, 48610b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 48621f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 486369fc6cbbSQu Wenruo /* 486469fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 486569fc6cbbSQu Wenruo * but item data not updated. 486669fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 486769fc6cbbSQu Wenruo */ 486869fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4869cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4870a4f78750SDavid Sterba btrfs_print_leaf(buf); 48711f21ef0aSFilipe Manana ASSERT(0); 48721f21ef0aSFilipe Manana } 48731f21ef0aSFilipe Manana #endif 4874eb60ceacSChris Mason } 4875eb60ceacSChris Mason 48762ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4877b53d3f5dSLiu Bo int flush_delayed) 487835b7e476SChris Mason { 4879188de649SChris Mason /* 4880188de649SChris Mason * looks as though older kernels can get into trouble with 4881188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4882188de649SChris Mason */ 4883e2d84521SMiao Xie int ret; 4884d6bfde87SChris Mason 48856933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4886d6bfde87SChris Mason return; 4887d6bfde87SChris Mason 4888b53d3f5dSLiu Bo if (flush_delayed) 48892ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 489016cdcec7SMiao Xie 4891d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4892d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4893d814a491SEthan Lien fs_info->dirty_metadata_batch); 4894e2d84521SMiao Xie if (ret > 0) { 48950b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 489616cdcec7SMiao Xie } 489716cdcec7SMiao Xie } 489816cdcec7SMiao Xie 48992ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 490016cdcec7SMiao Xie { 49012ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 490235b7e476SChris Mason } 4903b53d3f5dSLiu Bo 49042ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4905b53d3f5dSLiu Bo { 49062ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4907d6bfde87SChris Mason } 49086b80053dSChris Mason 49092ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4910acce952bSliubo { 4911fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4912fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4913fe816d0fSNikolay Borisov 49140b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 49152ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 49160b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4917acce952bSliubo 49180b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 49190b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4920acce952bSliubo } 4921acce952bSliubo 4922ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4923ef67963dSJosef Bacik { 4924fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4925fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4926fc7cbcd4SDavid Sterba int ret; 4927ef67963dSJosef Bacik 4928fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4929fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4930fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4931fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4932fc7cbcd4SDavid Sterba int i; 4933ef67963dSJosef Bacik 4934fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4935fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4936fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4937fc7cbcd4SDavid Sterba 4938fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4939fc7cbcd4SDavid Sterba if (!gang[i]) 4940ef67963dSJosef Bacik continue; 4941fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4942fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4943fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4944ef67963dSJosef Bacik } 4945fc7cbcd4SDavid Sterba root_objectid++; 4946fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4947ef67963dSJosef Bacik } 4948fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4949ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4950ef67963dSJosef Bacik } 4951ef67963dSJosef Bacik 4952143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4953acce952bSliubo { 4954acce952bSliubo struct btrfs_ordered_extent *ordered; 4955acce952bSliubo 4956199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4957779880efSJosef Bacik /* 4958779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4959779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4960779880efSJosef Bacik */ 4961199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4962779880efSJosef Bacik root_extent_list) 4963779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4964199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4965199c2a9cSMiao Xie } 4966199c2a9cSMiao Xie 4967199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4968199c2a9cSMiao Xie { 4969199c2a9cSMiao Xie struct btrfs_root *root; 4970199c2a9cSMiao Xie struct list_head splice; 4971199c2a9cSMiao Xie 4972199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4973199c2a9cSMiao Xie 4974199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4975199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4976199c2a9cSMiao Xie while (!list_empty(&splice)) { 4977199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4978199c2a9cSMiao Xie ordered_root); 49791de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 49801de2cfdeSJosef Bacik &fs_info->ordered_roots); 4981199c2a9cSMiao Xie 49822a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4983199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4984199c2a9cSMiao Xie 49852a85d9caSLiu Bo cond_resched(); 49862a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4987199c2a9cSMiao Xie } 4988199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 498974d5d229SJosef Bacik 499074d5d229SJosef Bacik /* 499174d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 499274d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 499374d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 499474d5d229SJosef Bacik * actually get run and error out properly. 499574d5d229SJosef Bacik */ 499674d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4997acce952bSliubo } 4998acce952bSliubo 499935a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 50002ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 5001acce952bSliubo { 5002acce952bSliubo struct rb_node *node; 5003acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 5004acce952bSliubo struct btrfs_delayed_ref_node *ref; 5005acce952bSliubo int ret = 0; 5006acce952bSliubo 5007acce952bSliubo delayed_refs = &trans->delayed_refs; 5008acce952bSliubo 5009acce952bSliubo spin_lock(&delayed_refs->lock); 5010d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 5011cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 5012b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 5013acce952bSliubo return ret; 5014acce952bSliubo } 5015acce952bSliubo 50165c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 5017d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 50180e0adbcfSJosef Bacik struct rb_node *n; 5019e78417d1SJosef Bacik bool pin_bytes = false; 5020acce952bSliubo 5021d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 5022d7df2c79SJosef Bacik href_node); 50233069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 5024b939d1abSJosef Bacik continue; 50253069bd26SJosef Bacik 5026d7df2c79SJosef Bacik spin_lock(&head->lock); 5027e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 50280e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 50290e0adbcfSJosef Bacik ref_node); 5030d7df2c79SJosef Bacik ref->in_tree = 0; 5031e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 50320e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 50331d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 50341d57ee94SWang Xiaoguang list_del(&ref->add_list); 5035d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 5036d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 5037d7df2c79SJosef Bacik } 503854067ae9SJosef Bacik if (head->must_insert_reserved) 5039e78417d1SJosef Bacik pin_bytes = true; 504078a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 5041fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 5042d7df2c79SJosef Bacik spin_unlock(&head->lock); 5043acce952bSliubo spin_unlock(&delayed_refs->lock); 5044e78417d1SJosef Bacik mutex_unlock(&head->mutex); 5045acce952bSliubo 5046f603bb94SNikolay Borisov if (pin_bytes) { 5047f603bb94SNikolay Borisov struct btrfs_block_group *cache; 5048f603bb94SNikolay Borisov 5049f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 5050f603bb94SNikolay Borisov BUG_ON(!cache); 5051f603bb94SNikolay Borisov 5052f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 5053f603bb94SNikolay Borisov spin_lock(&cache->lock); 5054f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 5055f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 5056f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 5057f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 5058f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 5059f603bb94SNikolay Borisov spin_unlock(&cache->lock); 5060f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 5061f603bb94SNikolay Borisov 5062f603bb94SNikolay Borisov btrfs_put_block_group(cache); 5063f603bb94SNikolay Borisov 5064f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 5065f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 5066f603bb94SNikolay Borisov } 506731890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 5068d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 5069acce952bSliubo cond_resched(); 5070acce952bSliubo spin_lock(&delayed_refs->lock); 5071acce952bSliubo } 507281f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 5073acce952bSliubo 5074acce952bSliubo spin_unlock(&delayed_refs->lock); 5075acce952bSliubo 5076acce952bSliubo return ret; 5077acce952bSliubo } 5078acce952bSliubo 5079143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 5080acce952bSliubo { 5081acce952bSliubo struct btrfs_inode *btrfs_inode; 5082acce952bSliubo struct list_head splice; 5083acce952bSliubo 5084acce952bSliubo INIT_LIST_HEAD(&splice); 5085acce952bSliubo 5086eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5087eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 5088acce952bSliubo 5089acce952bSliubo while (!list_empty(&splice)) { 5090fe816d0fSNikolay Borisov struct inode *inode = NULL; 5091eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 5092acce952bSliubo delalloc_inodes); 5093fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 5094eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5095acce952bSliubo 5096fe816d0fSNikolay Borisov /* 5097fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 5098fe816d0fSNikolay Borisov * meanwhile. 5099fe816d0fSNikolay Borisov */ 5100fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 5101fe816d0fSNikolay Borisov if (inode) { 5102fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 5103fe816d0fSNikolay Borisov iput(inode); 5104fe816d0fSNikolay Borisov } 5105eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5106acce952bSliubo } 5107eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5108eb73c1b7SMiao Xie } 5109eb73c1b7SMiao Xie 5110eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 5111eb73c1b7SMiao Xie { 5112eb73c1b7SMiao Xie struct btrfs_root *root; 5113eb73c1b7SMiao Xie struct list_head splice; 5114eb73c1b7SMiao Xie 5115eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 5116eb73c1b7SMiao Xie 5117eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5118eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 5119eb73c1b7SMiao Xie while (!list_empty(&splice)) { 5120eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 5121eb73c1b7SMiao Xie delalloc_root); 512200246528SJosef Bacik root = btrfs_grab_root(root); 5123eb73c1b7SMiao Xie BUG_ON(!root); 5124eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5125eb73c1b7SMiao Xie 5126eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 512700246528SJosef Bacik btrfs_put_root(root); 5128eb73c1b7SMiao Xie 5129eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5130eb73c1b7SMiao Xie } 5131eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5132acce952bSliubo } 5133acce952bSliubo 51342ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 5135acce952bSliubo struct extent_io_tree *dirty_pages, 5136acce952bSliubo int mark) 5137acce952bSliubo { 5138acce952bSliubo int ret; 5139acce952bSliubo struct extent_buffer *eb; 5140acce952bSliubo u64 start = 0; 5141acce952bSliubo u64 end; 5142acce952bSliubo 5143acce952bSliubo while (1) { 5144acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 5145e6138876SJosef Bacik mark, NULL); 5146acce952bSliubo if (ret) 5147acce952bSliubo break; 5148acce952bSliubo 514991166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 5150acce952bSliubo while (start <= end) { 51510b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 51520b246afaSJeff Mahoney start += fs_info->nodesize; 5153fd8b2b61SJosef Bacik if (!eb) 5154acce952bSliubo continue; 5155fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 5156acce952bSliubo 5157fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 5158fd8b2b61SJosef Bacik &eb->bflags)) 5159fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 5160fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 5161acce952bSliubo } 5162acce952bSliubo } 5163acce952bSliubo 5164acce952bSliubo return ret; 5165acce952bSliubo } 5166acce952bSliubo 51672ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 5168fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 5169acce952bSliubo { 5170acce952bSliubo u64 start; 5171acce952bSliubo u64 end; 5172acce952bSliubo int ret; 5173acce952bSliubo 5174acce952bSliubo while (1) { 51750e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 51760e6ec385SFilipe Manana 5177fcd5e742SLu Fengqi /* 5178fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 5179fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 5180fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 5181fcd5e742SLu Fengqi * the same extent range. 5182fcd5e742SLu Fengqi */ 5183fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 5184acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 51850e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 5186fcd5e742SLu Fengqi if (ret) { 5187fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5188acce952bSliubo break; 5189fcd5e742SLu Fengqi } 5190acce952bSliubo 51910e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 51920e6ec385SFilipe Manana free_extent_state(cached_state); 51932ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 5194fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5195acce952bSliubo cond_resched(); 5196acce952bSliubo } 5197acce952bSliubo 5198acce952bSliubo return 0; 5199acce952bSliubo } 5200acce952bSliubo 520132da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 5202c79a1751SLiu Bo { 5203c79a1751SLiu Bo struct inode *inode; 5204c79a1751SLiu Bo 5205c79a1751SLiu Bo inode = cache->io_ctl.inode; 5206c79a1751SLiu Bo if (inode) { 5207c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 5208c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 5209c79a1751SLiu Bo cache->io_ctl.inode = NULL; 5210c79a1751SLiu Bo iput(inode); 5211c79a1751SLiu Bo } 5212bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 5213c79a1751SLiu Bo btrfs_put_block_group(cache); 5214c79a1751SLiu Bo } 5215c79a1751SLiu Bo 5216c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 52172ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 5218c79a1751SLiu Bo { 521932da5386SDavid Sterba struct btrfs_block_group *cache; 5220c79a1751SLiu Bo 5221c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5222c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 5223c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 522432da5386SDavid Sterba struct btrfs_block_group, 5225c79a1751SLiu Bo dirty_list); 5226c79a1751SLiu Bo 5227c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 5228c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5229c79a1751SLiu Bo list_del_init(&cache->io_list); 5230c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5231c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5232c79a1751SLiu Bo } 5233c79a1751SLiu Bo 5234c79a1751SLiu Bo list_del_init(&cache->dirty_list); 5235c79a1751SLiu Bo spin_lock(&cache->lock); 5236c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5237c79a1751SLiu Bo spin_unlock(&cache->lock); 5238c79a1751SLiu Bo 5239c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5240c79a1751SLiu Bo btrfs_put_block_group(cache); 5241ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 5242c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5243c79a1751SLiu Bo } 5244c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5245c79a1751SLiu Bo 524645ae2c18SNikolay Borisov /* 524745ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 524845ae2c18SNikolay Borisov * to use it without any locking 524945ae2c18SNikolay Borisov */ 5250c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 5251c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 525232da5386SDavid Sterba struct btrfs_block_group, 5253c79a1751SLiu Bo io_list); 5254c79a1751SLiu Bo 5255c79a1751SLiu Bo list_del_init(&cache->io_list); 5256c79a1751SLiu Bo spin_lock(&cache->lock); 5257c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5258c79a1751SLiu Bo spin_unlock(&cache->lock); 5259c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5260c79a1751SLiu Bo } 5261c79a1751SLiu Bo } 5262c79a1751SLiu Bo 526349b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 52642ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 526549b25e05SJeff Mahoney { 5266bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 5267bbbf7243SNikolay Borisov 52682ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 5269c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 5270c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 5271c79a1751SLiu Bo 5272bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 5273bbbf7243SNikolay Borisov post_commit_list) { 5274bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 5275bbbf7243SNikolay Borisov } 5276bbbf7243SNikolay Borisov 52772ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 527849b25e05SJeff Mahoney 52794a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 52800b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 528149b25e05SJeff Mahoney 52824a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 52830b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 528449b25e05SJeff Mahoney 5285ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 528667cde344SMiao Xie 52872ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 528849b25e05SJeff Mahoney EXTENT_DIRTY); 5289fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 529049b25e05SJeff Mahoney 5291d3575156SNaohiro Aota btrfs_free_redirty_list(cur_trans); 5292d3575156SNaohiro Aota 52934a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 52944a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 529549b25e05SJeff Mahoney } 529649b25e05SJeff Mahoney 52972ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 5298acce952bSliubo { 5299acce952bSliubo struct btrfs_transaction *t; 5300acce952bSliubo 53010b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 5302acce952bSliubo 53030b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 53040b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 53050b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 5306724e2315SJosef Bacik struct btrfs_transaction, list); 5307724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 53089b64f57dSElena Reshetova refcount_inc(&t->use_count); 53090b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 53102ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 5311724e2315SJosef Bacik btrfs_put_transaction(t); 53120b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5313724e2315SJosef Bacik continue; 5314724e2315SJosef Bacik } 53150b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 5316724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 53170b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5318724e2315SJosef Bacik /* 5319724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 5320724e2315SJosef Bacik * handle open currently for this transaction. 5321724e2315SJosef Bacik */ 5322724e2315SJosef Bacik wait_event(t->writer_wait, 5323724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 5324724e2315SJosef Bacik } else { 53250b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5326724e2315SJosef Bacik } 53272ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 5328724e2315SJosef Bacik 53290b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 53300b246afaSJeff Mahoney if (t == fs_info->running_transaction) 53310b246afaSJeff Mahoney fs_info->running_transaction = NULL; 5332724e2315SJosef Bacik list_del_init(&t->list); 53330b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5334a4abeea4SJosef Bacik 5335724e2315SJosef Bacik btrfs_put_transaction(t); 53362e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 53370b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5338724e2315SJosef Bacik } 53390b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 53400b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 5341ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 5342ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 53430b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 5344ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 53450b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 5346acce952bSliubo 5347acce952bSliubo return 0; 5348acce952bSliubo } 5349ec7d6dfdSNikolay Borisov 5350453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 5351ec7d6dfdSNikolay Borisov { 5352ec7d6dfdSNikolay Borisov struct btrfs_path *path; 5353ec7d6dfdSNikolay Borisov int ret; 5354ec7d6dfdSNikolay Borisov struct extent_buffer *l; 5355ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 5356ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 5357ec7d6dfdSNikolay Borisov int slot; 5358ec7d6dfdSNikolay Borisov 5359ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 5360ec7d6dfdSNikolay Borisov if (!path) 5361ec7d6dfdSNikolay Borisov return -ENOMEM; 5362ec7d6dfdSNikolay Borisov 5363ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 5364ec7d6dfdSNikolay Borisov search_key.type = -1; 5365ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 5366ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 5367ec7d6dfdSNikolay Borisov if (ret < 0) 5368ec7d6dfdSNikolay Borisov goto error; 5369ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 5370ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 5371ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 5372ec7d6dfdSNikolay Borisov l = path->nodes[0]; 5373ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 537423125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 537523125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 5376ec7d6dfdSNikolay Borisov } else { 537723125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 5378ec7d6dfdSNikolay Borisov } 5379ec7d6dfdSNikolay Borisov ret = 0; 5380ec7d6dfdSNikolay Borisov error: 5381ec7d6dfdSNikolay Borisov btrfs_free_path(path); 5382ec7d6dfdSNikolay Borisov return ret; 5383ec7d6dfdSNikolay Borisov } 5384ec7d6dfdSNikolay Borisov 5385543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 5386ec7d6dfdSNikolay Borisov { 5387ec7d6dfdSNikolay Borisov int ret; 5388ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5389ec7d6dfdSNikolay Borisov 53906b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5391ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5392ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5393ec7d6dfdSNikolay Borisov root->root_key.objectid); 5394ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5395ec7d6dfdSNikolay Borisov goto out; 5396ec7d6dfdSNikolay Borisov } 5397ec7d6dfdSNikolay Borisov 539823125104SNikolay Borisov *objectid = root->free_objectid++; 5399ec7d6dfdSNikolay Borisov ret = 0; 5400ec7d6dfdSNikolay Borisov out: 5401ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5402ec7d6dfdSNikolay Borisov return ret; 5403ec7d6dfdSNikolay Borisov } 5404