1c1d7c514SDavid Sterba // SPDX-License-Identifier: GPL-2.0 26cbd5570SChris Mason /* 36cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 46cbd5570SChris Mason */ 56cbd5570SChris Mason 6e20d96d6SChris Mason #include <linux/fs.h> 7d98237b3SChris Mason #include <linux/blkdev.h> 8fc7cbcd4SDavid Sterba #include <linux/radix-tree.h> 935b7e476SChris Mason #include <linux/writeback.h> 10ce9adaa5SChris Mason #include <linux/workqueue.h> 11a74a4b97SChris Mason #include <linux/kthread.h> 125a0e3ad6STejun Heo #include <linux/slab.h> 13784b4e29SChris Mason #include <linux/migrate.h> 147a36ddecSDavid Sterba #include <linux/ratelimit.h> 156463fe58SStefan Behrens #include <linux/uuid.h> 16803b2f54SStefan Behrens #include <linux/semaphore.h> 17540adea3SMasami Hiramatsu #include <linux/error-injection.h> 189678c543SNikolay Borisov #include <linux/crc32c.h> 19b89f6d1fSFilipe Manana #include <linux/sched/mm.h> 207e75bf3fSDavid Sterba #include <asm/unaligned.h> 216d97c6e3SJohannes Thumshirn #include <crypto/hash.h> 22eb60ceacSChris Mason #include "ctree.h" 23eb60ceacSChris Mason #include "disk-io.h" 24e089f05cSChris Mason #include "transaction.h" 250f7d52f4SChris Mason #include "btrfs_inode.h" 260b86a832SChris Mason #include "volumes.h" 27db94535dSChris Mason #include "print-tree.h" 28925baeddSChris Mason #include "locking.h" 29e02119d5SChris Mason #include "tree-log.h" 30fa9c0d79SChris Mason #include "free-space-cache.h" 3170f6d82eSOmar Sandoval #include "free-space-tree.h" 3221adbd5cSStefan Behrens #include "check-integrity.h" 33606686eeSJosef Bacik #include "rcu-string.h" 348dabb742SStefan Behrens #include "dev-replace.h" 3553b381b3SDavid Woodhouse #include "raid56.h" 365ac1d209SJeff Mahoney #include "sysfs.h" 37fcebe456SJosef Bacik #include "qgroup.h" 38ebb8765bSAnand Jain #include "compression.h" 39557ea5ddSQu Wenruo #include "tree-checker.h" 40fd708b81SJosef Bacik #include "ref-verify.h" 41aac0023cSJosef Bacik #include "block-group.h" 42b0643e59SDennis Zhou #include "discard.h" 43f603bb94SNikolay Borisov #include "space-info.h" 44b70f5097SNaohiro Aota #include "zoned.h" 45139e8cd3SQu Wenruo #include "subpage.h" 46c7f13d42SJosef Bacik #include "fs.h" 4707e81dc9SJosef Bacik #include "accessors.h" 48a0231804SJosef Bacik #include "extent-tree.h" 4945c40c8fSJosef Bacik #include "root-tree.h" 5059b818e0SJosef Bacik #include "defrag.h" 51c7a03b52SJosef Bacik #include "uuid-tree.h" 5267707479SJosef Bacik #include "relocation.h" 53*2fc6822cSJosef Bacik #include "scrub.h" 54eb60ceacSChris Mason 55319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 56319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 57319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 58319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 59e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 60e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 61319e4d06SQu Wenruo 62143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 63acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 642ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 65143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 662ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 67acce952bSliubo struct extent_io_tree *dirty_pages, 68acce952bSliubo int mark); 692ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 70acce952bSliubo struct extent_io_tree *pinned_extents); 712ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 722ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 73ce9adaa5SChris Mason 74141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 75141386e1SJosef Bacik { 76141386e1SJosef Bacik if (fs_info->csum_shash) 77141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 78141386e1SJosef Bacik } 79141386e1SJosef Bacik 80d352ac68SChris Mason /* 81d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 82d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 83d352ac68SChris Mason * just before they are sent down the IO stack. 84d352ac68SChris Mason */ 8544b8bd7eSChris Mason struct async_submit_bio { 868896a08dSQu Wenruo struct inode *inode; 8744b8bd7eSChris Mason struct bio *bio; 88a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 8944b8bd7eSChris Mason int mirror_num; 901941b64bSQu Wenruo 911941b64bSQu Wenruo /* Optional parameter for submit_bio_start used by direct io */ 921941b64bSQu Wenruo u64 dio_file_offset; 938b712842SChris Mason struct btrfs_work work; 944e4cbee9SChristoph Hellwig blk_status_t status; 9544b8bd7eSChris Mason }; 9644b8bd7eSChris Mason 9785d4e461SChris Mason /* 982996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 99d352ac68SChris Mason */ 100c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 10119c00ddcSChris Mason { 102d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 1037280305eSDavid Sterba const int num_pages = num_extent_pages(buf); 104a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 105d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 10619c00ddcSChris Mason char *kaddr; 107e9be5a30SDavid Sterba int i; 108d5178578SJohannes Thumshirn 109d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 110d5178578SJohannes Thumshirn crypto_shash_init(shash); 111a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 112e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 113a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 11419c00ddcSChris Mason 115e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 116e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 117e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 11819c00ddcSChris Mason } 11971a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 120d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 12119c00ddcSChris Mason } 12219c00ddcSChris Mason 123d352ac68SChris Mason /* 124d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 125d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 126d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 127d352ac68SChris Mason * in the wrong place. 128d352ac68SChris Mason */ 1291259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 130b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 131b9fab919SChris Mason int atomic) 1321259ab75SChris Mason { 1332ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 1341259ab75SChris Mason int ret; 1351259ab75SChris Mason 1361259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 1371259ab75SChris Mason return 0; 1381259ab75SChris Mason 139b9fab919SChris Mason if (atomic) 140b9fab919SChris Mason return -EAGAIN; 141b9fab919SChris Mason 142570eb97bSJosef Bacik lock_extent(io_tree, eb->start, eb->start + eb->len - 1, &cached_state); 1430b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 1441259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 1451259ab75SChris Mason ret = 0; 1461259ab75SChris Mason goto out; 1471259ab75SChris Mason } 14894647322SDavid Sterba btrfs_err_rl(eb->fs_info, 1498f0ed7d4SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 1508f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 15129549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 1521259ab75SChris Mason ret = 1; 1530b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 15433958dc6SChris Mason out: 155570eb97bSJosef Bacik unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, 156e43bbe5eSDavid Sterba &cached_state); 1571259ab75SChris Mason return ret; 1581259ab75SChris Mason } 1591259ab75SChris Mason 160e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 161e7e16f48SJohannes Thumshirn { 162e7e16f48SJohannes Thumshirn switch (csum_type) { 163e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 1643951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 1653831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 166352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 167e7e16f48SJohannes Thumshirn return true; 168e7e16f48SJohannes Thumshirn default: 169e7e16f48SJohannes Thumshirn return false; 170e7e16f48SJohannes Thumshirn } 171e7e16f48SJohannes Thumshirn } 172e7e16f48SJohannes Thumshirn 173d352ac68SChris Mason /* 1741104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 1751104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 1761104a885SDavid Sterba */ 1773d17adeaSQu Wenruo int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 1783d17adeaSQu Wenruo const struct btrfs_super_block *disk_sb) 1791104a885SDavid Sterba { 18051bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 181d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 182d5178578SJohannes Thumshirn 183d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 1841104a885SDavid Sterba 1851104a885SDavid Sterba /* 1861104a885SDavid Sterba * The super_block structure does not span the whole 18751bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 18851bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 1891104a885SDavid Sterba */ 1903d17adeaSQu Wenruo crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, 191fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 1921104a885SDavid Sterba 19355fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 194e7e16f48SJohannes Thumshirn return 1; 1951104a885SDavid Sterba 196e7e16f48SJohannes Thumshirn return 0; 1971104a885SDavid Sterba } 1981104a885SDavid Sterba 199e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 200ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 201581c1760SQu Wenruo { 202e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 203581c1760SQu Wenruo int found_level; 204581c1760SQu Wenruo struct btrfs_key found_key; 205581c1760SQu Wenruo int ret; 206581c1760SQu Wenruo 207581c1760SQu Wenruo found_level = btrfs_header_level(eb); 208581c1760SQu Wenruo if (found_level != level) { 20963489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 21063489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 211581c1760SQu Wenruo btrfs_err(fs_info, 212581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 213581c1760SQu Wenruo eb->start, level, found_level); 214581c1760SQu Wenruo return -EIO; 215581c1760SQu Wenruo } 216581c1760SQu Wenruo 217581c1760SQu Wenruo if (!first_key) 218581c1760SQu Wenruo return 0; 219581c1760SQu Wenruo 2205d41be6fSQu Wenruo /* 2215d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 2225d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 2235d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 2245d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 2255d41be6fSQu Wenruo */ 2265d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 2275d41be6fSQu Wenruo return 0; 22862fdaa52SQu Wenruo 22962fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 23062fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 23162fdaa52SQu Wenruo btrfs_err(fs_info, 23262fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 23362fdaa52SQu Wenruo eb->start); 23462fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 23562fdaa52SQu Wenruo return -EUCLEAN; 23662fdaa52SQu Wenruo } 23762fdaa52SQu Wenruo 238581c1760SQu Wenruo if (found_level) 239581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 240581c1760SQu Wenruo else 241581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 242581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 243581c1760SQu Wenruo 244581c1760SQu Wenruo if (ret) { 24563489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 24663489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 247581c1760SQu Wenruo btrfs_err(fs_info, 248ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 249ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 250ff76a864SLiu Bo first_key->type, first_key->offset, 251ff76a864SLiu Bo found_key.objectid, found_key.type, 252ff76a864SLiu Bo found_key.offset); 253581c1760SQu Wenruo } 254581c1760SQu Wenruo return ret; 255581c1760SQu Wenruo } 256581c1760SQu Wenruo 2571104a885SDavid Sterba /* 258d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 259d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 260581c1760SQu Wenruo * 261581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 262581c1760SQu Wenruo * @level: expected level, mandatory check 263581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 264d352ac68SChris Mason */ 2656a2e9dc4SFilipe Manana int btrfs_read_extent_buffer(struct extent_buffer *eb, 266581c1760SQu Wenruo u64 parent_transid, int level, 267581c1760SQu Wenruo struct btrfs_key *first_key) 268f188591eSChris Mason { 2695ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 270f188591eSChris Mason struct extent_io_tree *io_tree; 271ea466794SJosef Bacik int failed = 0; 272f188591eSChris Mason int ret; 273f188591eSChris Mason int num_copies = 0; 274f188591eSChris Mason int mirror_num = 0; 275ea466794SJosef Bacik int failed_mirror = 0; 276f188591eSChris Mason 2770b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 278f188591eSChris Mason while (1) { 279f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 280c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); 281256dd1bbSStefan Behrens if (!ret) { 282581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 283b9fab919SChris Mason parent_transid, 0)) 284256dd1bbSStefan Behrens ret = -EIO; 285e064d5e9SDavid Sterba else if (btrfs_verify_level_key(eb, level, 286ff76a864SLiu Bo first_key, parent_transid)) 287581c1760SQu Wenruo ret = -EUCLEAN; 288581c1760SQu Wenruo else 289581c1760SQu Wenruo break; 290256dd1bbSStefan Behrens } 291d397712bSChris Mason 2920b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 293f188591eSChris Mason eb->start, eb->len); 2944235298eSChris Mason if (num_copies == 1) 295ea466794SJosef Bacik break; 2964235298eSChris Mason 2975cf1ab56SJosef Bacik if (!failed_mirror) { 2985cf1ab56SJosef Bacik failed = 1; 2995cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 3005cf1ab56SJosef Bacik } 3015cf1ab56SJosef Bacik 302f188591eSChris Mason mirror_num++; 303ea466794SJosef Bacik if (mirror_num == failed_mirror) 304ea466794SJosef Bacik mirror_num++; 305ea466794SJosef Bacik 3064235298eSChris Mason if (mirror_num > num_copies) 307ea466794SJosef Bacik break; 308f188591eSChris Mason } 309ea466794SJosef Bacik 310c0901581SStefan Behrens if (failed && !ret && failed_mirror) 31120a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 312ea466794SJosef Bacik 313ea466794SJosef Bacik return ret; 314f188591eSChris Mason } 31519c00ddcSChris Mason 316eca0f6f6SQu Wenruo static int csum_one_extent_buffer(struct extent_buffer *eb) 317eca0f6f6SQu Wenruo { 318eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 319eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 320eca0f6f6SQu Wenruo int ret; 321eca0f6f6SQu Wenruo 322eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 323eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 324eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 325eca0f6f6SQu Wenruo csum_tree_block(eb, result); 326eca0f6f6SQu Wenruo 327eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 328eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 329eca0f6f6SQu Wenruo else 330eca0f6f6SQu Wenruo ret = btrfs_check_leaf_full(eb); 331eca0f6f6SQu Wenruo 3323777369fSQu Wenruo if (ret < 0) 3333777369fSQu Wenruo goto error; 3343777369fSQu Wenruo 3353777369fSQu Wenruo /* 3363777369fSQu Wenruo * Also check the generation, the eb reached here must be newer than 3373777369fSQu Wenruo * last committed. Or something seriously wrong happened. 3383777369fSQu Wenruo */ 3393777369fSQu Wenruo if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { 3403777369fSQu Wenruo ret = -EUCLEAN; 341eca0f6f6SQu Wenruo btrfs_err(fs_info, 3423777369fSQu Wenruo "block=%llu bad generation, have %llu expect > %llu", 3433777369fSQu Wenruo eb->start, btrfs_header_generation(eb), 3443777369fSQu Wenruo fs_info->last_trans_committed); 3453777369fSQu Wenruo goto error; 346eca0f6f6SQu Wenruo } 347eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 348eca0f6f6SQu Wenruo 349eca0f6f6SQu Wenruo return 0; 3503777369fSQu Wenruo 3513777369fSQu Wenruo error: 3523777369fSQu Wenruo btrfs_print_tree(eb, 0); 3533777369fSQu Wenruo btrfs_err(fs_info, "block=%llu write time tree block corruption detected", 3543777369fSQu Wenruo eb->start); 3553777369fSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 3563777369fSQu Wenruo return ret; 357eca0f6f6SQu Wenruo } 358eca0f6f6SQu Wenruo 359eca0f6f6SQu Wenruo /* Checksum all dirty extent buffers in one bio_vec */ 360eca0f6f6SQu Wenruo static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, 361eca0f6f6SQu Wenruo struct bio_vec *bvec) 362eca0f6f6SQu Wenruo { 363eca0f6f6SQu Wenruo struct page *page = bvec->bv_page; 364eca0f6f6SQu Wenruo u64 bvec_start = page_offset(page) + bvec->bv_offset; 365eca0f6f6SQu Wenruo u64 cur; 366eca0f6f6SQu Wenruo int ret = 0; 367eca0f6f6SQu Wenruo 368eca0f6f6SQu Wenruo for (cur = bvec_start; cur < bvec_start + bvec->bv_len; 369eca0f6f6SQu Wenruo cur += fs_info->nodesize) { 370eca0f6f6SQu Wenruo struct extent_buffer *eb; 371eca0f6f6SQu Wenruo bool uptodate; 372eca0f6f6SQu Wenruo 373eca0f6f6SQu Wenruo eb = find_extent_buffer(fs_info, cur); 374eca0f6f6SQu Wenruo uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, 375eca0f6f6SQu Wenruo fs_info->nodesize); 376eca0f6f6SQu Wenruo 37701cd3909SDavid Sterba /* A dirty eb shouldn't disappear from buffer_radix */ 378eca0f6f6SQu Wenruo if (WARN_ON(!eb)) 379eca0f6f6SQu Wenruo return -EUCLEAN; 380eca0f6f6SQu Wenruo 381eca0f6f6SQu Wenruo if (WARN_ON(cur != btrfs_header_bytenr(eb))) { 382eca0f6f6SQu Wenruo free_extent_buffer(eb); 383eca0f6f6SQu Wenruo return -EUCLEAN; 384eca0f6f6SQu Wenruo } 385eca0f6f6SQu Wenruo if (WARN_ON(!uptodate)) { 386eca0f6f6SQu Wenruo free_extent_buffer(eb); 387eca0f6f6SQu Wenruo return -EUCLEAN; 388eca0f6f6SQu Wenruo } 389eca0f6f6SQu Wenruo 390eca0f6f6SQu Wenruo ret = csum_one_extent_buffer(eb); 391eca0f6f6SQu Wenruo free_extent_buffer(eb); 392eca0f6f6SQu Wenruo if (ret < 0) 393eca0f6f6SQu Wenruo return ret; 394eca0f6f6SQu Wenruo } 395eca0f6f6SQu Wenruo return ret; 396eca0f6f6SQu Wenruo } 397eca0f6f6SQu Wenruo 398d352ac68SChris Mason /* 399ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 400ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 401ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 402d352ac68SChris Mason */ 403ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 40419c00ddcSChris Mason { 405ac303b69SQu Wenruo struct page *page = bvec->bv_page; 4064eee4fa4SMiao Xie u64 start = page_offset(page); 40719c00ddcSChris Mason u64 found_start; 40819c00ddcSChris Mason struct extent_buffer *eb; 409eca0f6f6SQu Wenruo 410fbca46ebSQu Wenruo if (fs_info->nodesize < PAGE_SIZE) 411eca0f6f6SQu Wenruo return csum_dirty_subpage_buffers(fs_info, bvec); 412f188591eSChris Mason 4134f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4144f2de97aSJosef Bacik if (page != eb->pages[0]) 4154f2de97aSJosef Bacik return 0; 4160f805531SAlex Lyakas 41719c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 418d3575156SNaohiro Aota 419d3575156SNaohiro Aota if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 420d3575156SNaohiro Aota WARN_ON(found_start != 0); 421d3575156SNaohiro Aota return 0; 422d3575156SNaohiro Aota } 423d3575156SNaohiro Aota 4240f805531SAlex Lyakas /* 4250f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 4260f805531SAlex Lyakas * It is useful to know what went wrong. 4270f805531SAlex Lyakas */ 4280f805531SAlex Lyakas if (WARN_ON(found_start != start)) 4290f805531SAlex Lyakas return -EUCLEAN; 4300f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 4310f805531SAlex Lyakas return -EUCLEAN; 4320f805531SAlex Lyakas 433eca0f6f6SQu Wenruo return csum_one_extent_buffer(eb); 43419c00ddcSChris Mason } 43519c00ddcSChris Mason 436b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 4372b82032cSYan Zheng { 438b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 439944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 44044880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 441944d3f9fSNikolay Borisov u8 *metadata_uuid; 4422b82032cSYan Zheng 4439a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 4449a8658e3SDavid Sterba BTRFS_FSID_SIZE); 4457239ff4bSNikolay Borisov /* 446944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 447944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 448944d3f9fSNikolay Borisov * ->fsid in this case is fine 4497239ff4bSNikolay Borisov */ 450944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 4517239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 4527239ff4bSNikolay Borisov else 4537239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 4547239ff4bSNikolay Borisov 455944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 456944d3f9fSNikolay Borisov return 0; 457944d3f9fSNikolay Borisov 458944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 459944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 460944d3f9fSNikolay Borisov return 0; 461944d3f9fSNikolay Borisov 462944d3f9fSNikolay Borisov return 1; 4632b82032cSYan Zheng } 4642b82032cSYan Zheng 46577bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 46677bf40a2SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb) 467ce9adaa5SChris Mason { 46877bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 469ce9adaa5SChris Mason u64 found_start; 47077bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 47177bf40a2SQu Wenruo u8 found_level; 4722996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 473dfd29eedSDavid Sterba const u8 *header_csum; 47477bf40a2SQu Wenruo int ret = 0; 475ea466794SJosef Bacik 476ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 477727011e0SChris Mason if (found_start != eb->start) { 4788f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, 4798f0ed7d4SQu Wenruo "bad tree block start, mirror %u want %llu have %llu", 4808f0ed7d4SQu Wenruo eb->read_mirror, eb->start, found_start); 481f188591eSChris Mason ret = -EIO; 48277bf40a2SQu Wenruo goto out; 483ce9adaa5SChris Mason } 484b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 4858f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", 4868f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 4871259ab75SChris Mason ret = -EIO; 48877bf40a2SQu Wenruo goto out; 4891259ab75SChris Mason } 490ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 4911c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 4928f0ed7d4SQu Wenruo btrfs_err(fs_info, 4938f0ed7d4SQu Wenruo "bad tree block level, mirror %u level %d on logical %llu", 4948f0ed7d4SQu Wenruo eb->read_mirror, btrfs_header_level(eb), eb->start); 4951c24c3ceSJosef Bacik ret = -EIO; 49677bf40a2SQu Wenruo goto out; 4971c24c3ceSJosef Bacik } 498ce9adaa5SChris Mason 499c67b3892SDavid Sterba csum_tree_block(eb, result); 500dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 501dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 502a826d6dcSJosef Bacik 503dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 5042996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 5058f0ed7d4SQu Wenruo "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", 5068f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 507dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 50835be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 50935be8851SJohannes Thumshirn btrfs_header_level(eb)); 5102996e1f8SJohannes Thumshirn ret = -EUCLEAN; 51177bf40a2SQu Wenruo goto out; 5122996e1f8SJohannes Thumshirn } 5132996e1f8SJohannes Thumshirn 514a826d6dcSJosef Bacik /* 515a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 516a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 517a826d6dcSJosef Bacik * return -EIO. 518a826d6dcSJosef Bacik */ 5191c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 520a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 521a826d6dcSJosef Bacik ret = -EIO; 522a826d6dcSJosef Bacik } 523ce9adaa5SChris Mason 524813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 525053ab70fSLiu Bo ret = -EIO; 526053ab70fSLiu Bo 5270b32f4bbSJosef Bacik if (!ret) 5280b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 52975391f0dSQu Wenruo else 53075391f0dSQu Wenruo btrfs_err(fs_info, 5318f0ed7d4SQu Wenruo "read time tree block corruption detected on logical %llu mirror %u", 5328f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 53377bf40a2SQu Wenruo out: 53477bf40a2SQu Wenruo return ret; 53577bf40a2SQu Wenruo } 53677bf40a2SQu Wenruo 537371cdc07SQu Wenruo static int validate_subpage_buffer(struct page *page, u64 start, u64 end, 538371cdc07SQu Wenruo int mirror) 539371cdc07SQu Wenruo { 540371cdc07SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 541371cdc07SQu Wenruo struct extent_buffer *eb; 542371cdc07SQu Wenruo bool reads_done; 543371cdc07SQu Wenruo int ret = 0; 544371cdc07SQu Wenruo 545371cdc07SQu Wenruo /* 546371cdc07SQu Wenruo * We don't allow bio merge for subpage metadata read, so we should 547371cdc07SQu Wenruo * only get one eb for each endio hook. 548371cdc07SQu Wenruo */ 549371cdc07SQu Wenruo ASSERT(end == start + fs_info->nodesize - 1); 550371cdc07SQu Wenruo ASSERT(PagePrivate(page)); 551371cdc07SQu Wenruo 552371cdc07SQu Wenruo eb = find_extent_buffer(fs_info, start); 553371cdc07SQu Wenruo /* 554371cdc07SQu Wenruo * When we are reading one tree block, eb must have been inserted into 555371cdc07SQu Wenruo * the radix tree. If not, something is wrong. 556371cdc07SQu Wenruo */ 557371cdc07SQu Wenruo ASSERT(eb); 558371cdc07SQu Wenruo 559371cdc07SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 560371cdc07SQu Wenruo /* Subpage read must finish in page read */ 561371cdc07SQu Wenruo ASSERT(reads_done); 562371cdc07SQu Wenruo 563371cdc07SQu Wenruo eb->read_mirror = mirror; 564371cdc07SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 565371cdc07SQu Wenruo ret = -EIO; 566371cdc07SQu Wenruo goto err; 567371cdc07SQu Wenruo } 568371cdc07SQu Wenruo ret = validate_extent_buffer(eb); 569371cdc07SQu Wenruo if (ret < 0) 570371cdc07SQu Wenruo goto err; 571371cdc07SQu Wenruo 572371cdc07SQu Wenruo set_extent_buffer_uptodate(eb); 573371cdc07SQu Wenruo 574371cdc07SQu Wenruo free_extent_buffer(eb); 575371cdc07SQu Wenruo return ret; 576371cdc07SQu Wenruo err: 577371cdc07SQu Wenruo /* 578371cdc07SQu Wenruo * end_bio_extent_readpage decrements io_pages in case of error, 579371cdc07SQu Wenruo * make sure it has something to decrement. 580371cdc07SQu Wenruo */ 581371cdc07SQu Wenruo atomic_inc(&eb->io_pages); 582371cdc07SQu Wenruo clear_extent_buffer_uptodate(eb); 583371cdc07SQu Wenruo free_extent_buffer(eb); 584371cdc07SQu Wenruo return ret; 585371cdc07SQu Wenruo } 586371cdc07SQu Wenruo 587c3a3b19bSQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, 58877bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 58977bf40a2SQu Wenruo int mirror) 59077bf40a2SQu Wenruo { 59177bf40a2SQu Wenruo struct extent_buffer *eb; 59277bf40a2SQu Wenruo int ret = 0; 59377bf40a2SQu Wenruo int reads_done; 59477bf40a2SQu Wenruo 59577bf40a2SQu Wenruo ASSERT(page->private); 596371cdc07SQu Wenruo 597fbca46ebSQu Wenruo if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) 598371cdc07SQu Wenruo return validate_subpage_buffer(page, start, end, mirror); 599371cdc07SQu Wenruo 60077bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 60177bf40a2SQu Wenruo 60277bf40a2SQu Wenruo /* 60377bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 60477bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 60577bf40a2SQu Wenruo */ 60677bf40a2SQu Wenruo atomic_inc(&eb->refs); 60777bf40a2SQu Wenruo 60877bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 60977bf40a2SQu Wenruo if (!reads_done) 61077bf40a2SQu Wenruo goto err; 61177bf40a2SQu Wenruo 61277bf40a2SQu Wenruo eb->read_mirror = mirror; 61377bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 61477bf40a2SQu Wenruo ret = -EIO; 61577bf40a2SQu Wenruo goto err; 61677bf40a2SQu Wenruo } 61777bf40a2SQu Wenruo ret = validate_extent_buffer(eb); 618ce9adaa5SChris Mason err: 61953b381b3SDavid Woodhouse if (ret) { 62053b381b3SDavid Woodhouse /* 62153b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 62253b381b3SDavid Woodhouse * again, we have to make sure it has something 62353b381b3SDavid Woodhouse * to decrement 62453b381b3SDavid Woodhouse */ 62553b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6260b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 62753b381b3SDavid Woodhouse } 6280b32f4bbSJosef Bacik free_extent_buffer(eb); 62977bf40a2SQu Wenruo 630f188591eSChris Mason return ret; 631ce9adaa5SChris Mason } 632ce9adaa5SChris Mason 6334a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 6344a69a410SChris Mason { 6354a69a410SChris Mason struct async_submit_bio *async; 6364e4cbee9SChristoph Hellwig blk_status_t ret; 6374a69a410SChris Mason 6384a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 6391941b64bSQu Wenruo ret = async->submit_bio_start(async->inode, async->bio, 6401941b64bSQu Wenruo async->dio_file_offset); 64179787eaaSJeff Mahoney if (ret) 6424e4cbee9SChristoph Hellwig async->status = ret; 6434a69a410SChris Mason } 6444a69a410SChris Mason 64506ea01b1SDavid Sterba /* 64606ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 64706ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 64806ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 64906ea01b1SDavid Sterba * 65006ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 65106ea01b1SDavid Sterba * inserted into the tree. 65206ea01b1SDavid Sterba */ 6534a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 6548b712842SChris Mason { 655917f32a2SChristoph Hellwig struct async_submit_bio *async = 656917f32a2SChristoph Hellwig container_of(work, struct async_submit_bio, work); 657917f32a2SChristoph Hellwig struct inode *inode = async->inode; 658917f32a2SChristoph Hellwig struct btrfs_bio *bbio = btrfs_bio(async->bio); 6594854ddd0SChris Mason 660bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 6614e4cbee9SChristoph Hellwig if (async->status) { 662917f32a2SChristoph Hellwig btrfs_bio_end_io(bbio, async->status); 66379787eaaSJeff Mahoney return; 66479787eaaSJeff Mahoney } 66579787eaaSJeff Mahoney 666ec39f769SChris Mason /* 667ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 668ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 669ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 670ec39f769SChris Mason */ 671ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 6721a722d8fSChristoph Hellwig btrfs_submit_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 6734a69a410SChris Mason } 6744a69a410SChris Mason 6754a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 6764a69a410SChris Mason { 6774a69a410SChris Mason struct async_submit_bio *async; 6784a69a410SChris Mason 6794a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 6808b712842SChris Mason kfree(async); 6818b712842SChris Mason } 6828b712842SChris Mason 683ea1f0cedSChristoph Hellwig /* 684ea1f0cedSChristoph Hellwig * Submit bio to an async queue. 685ea1f0cedSChristoph Hellwig * 686ea1f0cedSChristoph Hellwig * Retrun: 687ea1f0cedSChristoph Hellwig * - true if the work has been succesfuly submitted 688ea1f0cedSChristoph Hellwig * - false in case of error 689ea1f0cedSChristoph Hellwig */ 690ea1f0cedSChristoph Hellwig bool btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, int mirror_num, 691ea1f0cedSChristoph Hellwig u64 dio_file_offset, 692e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 69344b8bd7eSChris Mason { 6948896a08dSQu Wenruo struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 69544b8bd7eSChris Mason struct async_submit_bio *async; 69644b8bd7eSChris Mason 69744b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 69844b8bd7eSChris Mason if (!async) 699ea1f0cedSChristoph Hellwig return false; 70044b8bd7eSChris Mason 7018896a08dSQu Wenruo async->inode = inode; 70244b8bd7eSChris Mason async->bio = bio; 70344b8bd7eSChris Mason async->mirror_num = mirror_num; 7044a69a410SChris Mason async->submit_bio_start = submit_bio_start; 7054a69a410SChris Mason 706a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 707a0cac0ecSOmar Sandoval run_one_async_free); 7084a69a410SChris Mason 7091941b64bSQu Wenruo async->dio_file_offset = dio_file_offset; 7108c8bee1dSChris Mason 7114e4cbee9SChristoph Hellwig async->status = 0; 71279787eaaSJeff Mahoney 71367f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 714a31b4a43SChristoph Hellwig btrfs_queue_work(fs_info->hipri_workers, &async->work); 715a31b4a43SChristoph Hellwig else 7165cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 717ea1f0cedSChristoph Hellwig return true; 71844b8bd7eSChris Mason } 71944b8bd7eSChris Mason 7204e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 721ce3ed71aSChris Mason { 7222c30c71bSKent Overstreet struct bio_vec *bvec; 723ce3ed71aSChris Mason struct btrfs_root *root; 7242b070cfeSChristoph Hellwig int ret = 0; 7256dc4f100SMing Lei struct bvec_iter_all iter_all; 726ce3ed71aSChris Mason 727c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 7282b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 729ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 730ac303b69SQu Wenruo ret = csum_dirty_buffer(root->fs_info, bvec); 73179787eaaSJeff Mahoney if (ret) 73279787eaaSJeff Mahoney break; 733ce3ed71aSChris Mason } 7342c30c71bSKent Overstreet 7354e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 736ce3ed71aSChris Mason } 737ce3ed71aSChris Mason 7388896a08dSQu Wenruo static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, 7391941b64bSQu Wenruo u64 dio_file_offset) 74022c59948SChris Mason { 7418b712842SChris Mason /* 7428b712842SChris Mason * when we're called for a write, we're already in the async 7431a722d8fSChristoph Hellwig * submission context. Just jump into btrfs_submit_bio. 7448b712842SChris Mason */ 74579787eaaSJeff Mahoney return btree_csum_one_bio(bio); 74622c59948SChris Mason } 74722c59948SChris Mason 748f4dcfb30SJohannes Thumshirn static bool should_async_write(struct btrfs_fs_info *fs_info, 7499b4e675aSDavid Sterba struct btrfs_inode *bi) 750de0022b9SJosef Bacik { 7514eef29efSNaohiro Aota if (btrfs_is_zoned(fs_info)) 752f4dcfb30SJohannes Thumshirn return false; 7536300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 754f4dcfb30SJohannes Thumshirn return false; 7559b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 756f4dcfb30SJohannes Thumshirn return false; 757f4dcfb30SJohannes Thumshirn return true; 758de0022b9SJosef Bacik } 759de0022b9SJosef Bacik 76094d9e11bSChristoph Hellwig void btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, int mirror_num) 76144b8bd7eSChris Mason { 7620b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 763917f32a2SChristoph Hellwig struct btrfs_bio *bbio = btrfs_bio(bio); 7644e4cbee9SChristoph Hellwig blk_status_t ret; 765cad321adSChris Mason 76608a6f464SChristoph Hellwig bio->bi_opf |= REQ_META; 76708a6f464SChristoph Hellwig 768cfe94440SNaohiro Aota if (btrfs_op(bio) != BTRFS_MAP_WRITE) { 7691a722d8fSChristoph Hellwig btrfs_submit_bio(fs_info, bio, mirror_num); 7701a722d8fSChristoph Hellwig return; 77144b8bd7eSChris Mason } 77244b8bd7eSChris Mason 773ea1f0cedSChristoph Hellwig /* 774ea1f0cedSChristoph Hellwig * Kthread helpers are used to submit writes so that checksumming can 775ea1f0cedSChristoph Hellwig * happen in parallel across all CPUs. 776ea1f0cedSChristoph Hellwig */ 777ea1f0cedSChristoph Hellwig if (should_async_write(fs_info, BTRFS_I(inode)) && 778ea1f0cedSChristoph Hellwig btrfs_wq_submit_bio(inode, bio, mirror_num, 0, btree_submit_bio_start)) 779ea1f0cedSChristoph Hellwig return; 780ea1f0cedSChristoph Hellwig 781ea1f0cedSChristoph Hellwig ret = btree_csum_one_bio(bio); 78294d9e11bSChristoph Hellwig if (ret) { 783917f32a2SChristoph Hellwig btrfs_bio_end_io(bbio, ret); 784ea1f0cedSChristoph Hellwig return; 78594d9e11bSChristoph Hellwig } 786ea1f0cedSChristoph Hellwig 787ea1f0cedSChristoph Hellwig btrfs_submit_bio(fs_info, bio, mirror_num); 78861891923SStefan Behrens } 78961891923SStefan Behrens 7903dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 7918958b551SMatthew Wilcox (Oracle) static int btree_migrate_folio(struct address_space *mapping, 7928958b551SMatthew Wilcox (Oracle) struct folio *dst, struct folio *src, enum migrate_mode mode) 793784b4e29SChris Mason { 794784b4e29SChris Mason /* 795784b4e29SChris Mason * we can't safely write a btree page from here, 796784b4e29SChris Mason * we haven't done the locking hook 797784b4e29SChris Mason */ 7988958b551SMatthew Wilcox (Oracle) if (folio_test_dirty(src)) 799784b4e29SChris Mason return -EAGAIN; 800784b4e29SChris Mason /* 801784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 802784b4e29SChris Mason * We must have no buffers or drop them. 803784b4e29SChris Mason */ 8048958b551SMatthew Wilcox (Oracle) if (folio_get_private(src) && 8058958b551SMatthew Wilcox (Oracle) !filemap_release_folio(src, GFP_KERNEL)) 806784b4e29SChris Mason return -EAGAIN; 80754184650SMatthew Wilcox (Oracle) return migrate_folio(mapping, dst, src, mode); 808784b4e29SChris Mason } 8098958b551SMatthew Wilcox (Oracle) #else 8108958b551SMatthew Wilcox (Oracle) #define btree_migrate_folio NULL 8113dd1462eSJan Beulich #endif 812784b4e29SChris Mason 8130da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 8140da5468fSChris Mason struct writeback_control *wbc) 8150da5468fSChris Mason { 816e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 817e2d84521SMiao Xie int ret; 818e2d84521SMiao Xie 819d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 820448d640bSChris Mason 821448d640bSChris Mason if (wbc->for_kupdate) 822448d640bSChris Mason return 0; 823448d640bSChris Mason 824e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 825b9473439SChris Mason /* this is a bit racy, but that's ok */ 826d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 827d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 828d814a491SEthan Lien fs_info->dirty_metadata_batch); 829e2d84521SMiao Xie if (ret < 0) 830793955bcSChris Mason return 0; 831793955bcSChris Mason } 8320b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 8330da5468fSChris Mason } 8340da5468fSChris Mason 835f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 8365f39d397SChris Mason { 837f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 838f913cff3SMatthew Wilcox (Oracle) return false; 8390c4e538bSDavid Sterba 840f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 841d98237b3SChris Mason } 842d98237b3SChris Mason 843895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 844895586ebSMatthew Wilcox (Oracle) size_t length) 845d98237b3SChris Mason { 846d1310b2eSChris Mason struct extent_io_tree *tree; 847895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 848895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 849f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 850895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 851895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 852895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 853895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 854895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 8559ad6b7bcSChris Mason } 856d98237b3SChris Mason } 857d98237b3SChris Mason 858bb146eb2SJosef Bacik #ifdef DEBUG 8590079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 8600079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 8610079c3b1SMatthew Wilcox (Oracle) { 8620079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 863139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 8640b32f4bbSJosef Bacik struct extent_buffer *eb; 865139e8cd3SQu Wenruo int cur_bit = 0; 8660079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 8670b32f4bbSJosef Bacik 868139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 8690079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 8700b32f4bbSJosef Bacik BUG_ON(!eb); 8710b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 8720b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 87349d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 8740079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 875139e8cd3SQu Wenruo } 8760079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 877139e8cd3SQu Wenruo 878139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 879139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 880139e8cd3SQu Wenruo unsigned long flags; 881139e8cd3SQu Wenruo u64 cur; 882139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 883139e8cd3SQu Wenruo 884139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 885139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 886139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 887139e8cd3SQu Wenruo cur_bit++; 888139e8cd3SQu Wenruo continue; 889139e8cd3SQu Wenruo } 890139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 891139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 892139e8cd3SQu Wenruo 893139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 894139e8cd3SQu Wenruo ASSERT(eb); 895139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 896139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 89749d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 898139e8cd3SQu Wenruo free_extent_buffer(eb); 899139e8cd3SQu Wenruo 900139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 901139e8cd3SQu Wenruo } 9020079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 9030b32f4bbSJosef Bacik } 9040079c3b1SMatthew Wilcox (Oracle) #else 9050079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 9060079c3b1SMatthew Wilcox (Oracle) #endif 9070b32f4bbSJosef Bacik 9087f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 9090da5468fSChris Mason .writepages = btree_writepages, 910f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 911895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 9128958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 9130079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 914d98237b3SChris Mason }; 915123abc88SChris Mason 9162ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 9172ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 9183fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 9193fbaf258SJosef Bacik int level) 9200999df54SChris Mason { 9210b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 9220b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 9233fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 9240999df54SChris Mason } 9250999df54SChris Mason 926581c1760SQu Wenruo /* 927581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 928581c1760SQu Wenruo * verification. 929581c1760SQu Wenruo * 9301b7ec85eSJosef Bacik * @owner_root: the objectid of the root owner for this block. 931581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 932581c1760SQu Wenruo * @level: expected level, mandatory check 933581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 934581c1760SQu Wenruo */ 9352ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 9361b7ec85eSJosef Bacik u64 owner_root, u64 parent_transid, 9371b7ec85eSJosef Bacik int level, struct btrfs_key *first_key) 938e20d96d6SChris Mason { 9395f39d397SChris Mason struct extent_buffer *buf = NULL; 94019c00ddcSChris Mason int ret; 94119c00ddcSChris Mason 9423fbaf258SJosef Bacik buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); 943c871b0f2SLiu Bo if (IS_ERR(buf)) 944c871b0f2SLiu Bo return buf; 945e4204dedSChris Mason 9466a2e9dc4SFilipe Manana ret = btrfs_read_extent_buffer(buf, parent_transid, level, first_key); 9470f0fe8f7SFilipe David Borba Manana if (ret) { 948537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 94964c043deSLiu Bo return ERR_PTR(ret); 9500f0fe8f7SFilipe David Borba Manana } 95188c602abSQu Wenruo if (btrfs_check_eb_owner(buf, owner_root)) { 95288c602abSQu Wenruo free_extent_buffer_stale(buf); 95388c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 95488c602abSQu Wenruo } 9555f39d397SChris Mason return buf; 956ce9adaa5SChris Mason 957eb60ceacSChris Mason } 958eb60ceacSChris Mason 9596a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 960ed2ff2cbSChris Mason { 9616a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 96255c69072SChris Mason if (btrfs_header_generation(buf) == 963e2d84521SMiao Xie fs_info->running_transaction->transid) { 96449d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 965b4ce94deSChris Mason 966b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 967104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 968e2d84521SMiao Xie -buf->len, 969e2d84521SMiao Xie fs_info->dirty_metadata_batch); 9700b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 971925baeddSChris Mason } 9725f39d397SChris Mason } 973ed7b63ebSJosef Bacik } 9745f39d397SChris Mason 975da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 976e20d96d6SChris Mason u64 objectid) 977d97e63b6SChris Mason { 9787c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 9792e608bd1SJosef Bacik 9802e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 9812e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 9822e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 98396dfcb46SJosef Bacik root->fs_info = fs_info; 9842e608bd1SJosef Bacik root->root_key.objectid = objectid; 985cfaa7295SChris Mason root->node = NULL; 986a28ec197SChris Mason root->commit_root = NULL; 98727cdeb70SMiao Xie root->state = 0; 988abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 9890b86a832SChris Mason 9900f7d52f4SChris Mason root->last_trans = 0; 9916b8fad57SNikolay Borisov root->free_objectid = 0; 992eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 993199c2a9cSMiao Xie root->nr_ordered_extents = 0; 9946bef4d31SEric Paris root->inode_tree = RB_ROOT; 995088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 9962e608bd1SJosef Bacik 9972e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 9980b86a832SChris Mason 9990b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 10005d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1001eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1002eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1003199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1004199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1005d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 10062ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 10072ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 10085d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1009eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1010199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1011f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 10122ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 10132ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 10148287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1015a2135011SChris Mason mutex_init(&root->objectid_mutex); 1016e02119d5SChris Mason mutex_init(&root->log_mutex); 101731f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1018573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1019c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 10207237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 10217237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 10227237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 10238b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 10248b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 10257237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 10267237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 10277237f183SYan Zheng atomic_set(&root->log_writers, 0); 10282ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 10290700cea7SElena Reshetova refcount_set(&root->refs, 1); 10308ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1031eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 10327237f183SYan Zheng root->log_transid = 0; 1033d1433debSMiao Xie root->log_transid_committed = -1; 1034257c62e1SChris Mason root->last_log_commit = 0; 10352e608bd1SJosef Bacik root->anon_dev = 0; 1036e289f03eSFilipe Manana if (!dummy) { 103743eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 103843eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1039e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1040e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1041e289f03eSFilipe Manana } 1042017e5369SChris Mason 10435f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1044370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1045bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1046bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1047fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1048bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1049fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1050bd647ce3SJosef Bacik #endif 10513768f368SChris Mason } 10523768f368SChris Mason 105374e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 105496dfcb46SJosef Bacik u64 objectid, gfp_t flags) 10556f07e42eSAl Viro { 105674e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 10576f07e42eSAl Viro if (root) 105896dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 10596f07e42eSAl Viro return root; 10606f07e42eSAl Viro } 10616f07e42eSAl Viro 106206ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 106306ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1064da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 106506ea65a3SJosef Bacik { 106606ea65a3SJosef Bacik struct btrfs_root *root; 106706ea65a3SJosef Bacik 10687c0260eeSJeff Mahoney if (!fs_info) 10697c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 10707c0260eeSJeff Mahoney 107196dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 107206ea65a3SJosef Bacik if (!root) 107306ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1074da17066cSJeff Mahoney 1075b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1076faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 107706ea65a3SJosef Bacik 107806ea65a3SJosef Bacik return root; 107906ea65a3SJosef Bacik } 108006ea65a3SJosef Bacik #endif 108106ea65a3SJosef Bacik 1082abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 1083abed4aaaSJosef Bacik { 1084abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 1085abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 1086abed4aaaSJosef Bacik 1087abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 1088abed4aaaSJosef Bacik } 1089abed4aaaSJosef Bacik 1090abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 1091abed4aaaSJosef Bacik { 1092abed4aaaSJosef Bacik const struct btrfs_key *key = k; 1093abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 1094abed4aaaSJosef Bacik 1095abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 1096abed4aaaSJosef Bacik } 1097abed4aaaSJosef Bacik 1098abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 1099abed4aaaSJosef Bacik { 1100abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1101abed4aaaSJosef Bacik struct rb_node *tmp; 1102abed4aaaSJosef Bacik 1103abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1104abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 1105abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1106abed4aaaSJosef Bacik ASSERT(!tmp); 1107abed4aaaSJosef Bacik 1108abed4aaaSJosef Bacik return tmp ? -EEXIST : 0; 1109abed4aaaSJosef Bacik } 1110abed4aaaSJosef Bacik 1111abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 1112abed4aaaSJosef Bacik { 1113abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1114abed4aaaSJosef Bacik 1115abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1116abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1117abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1118abed4aaaSJosef Bacik } 1119abed4aaaSJosef Bacik 1120abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 1121abed4aaaSJosef Bacik struct btrfs_key *key) 1122abed4aaaSJosef Bacik { 1123abed4aaaSJosef Bacik struct rb_node *node; 1124abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 1125abed4aaaSJosef Bacik 1126abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 1127abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 1128abed4aaaSJosef Bacik if (node) 1129abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 1130abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 1131abed4aaaSJosef Bacik 1132abed4aaaSJosef Bacik return root; 1133abed4aaaSJosef Bacik } 1134abed4aaaSJosef Bacik 1135f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 1136f7238e50SJosef Bacik { 1137f7238e50SJosef Bacik struct btrfs_block_group *block_group; 1138f7238e50SJosef Bacik u64 ret; 1139f7238e50SJosef Bacik 1140f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 1141f7238e50SJosef Bacik return 0; 1142f7238e50SJosef Bacik 1143f7238e50SJosef Bacik if (bytenr) 1144f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 1145f7238e50SJosef Bacik else 1146f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 1147f7238e50SJosef Bacik ASSERT(block_group); 1148f7238e50SJosef Bacik if (!block_group) 1149f7238e50SJosef Bacik return 0; 1150f7238e50SJosef Bacik ret = block_group->global_root_id; 1151f7238e50SJosef Bacik btrfs_put_block_group(block_group); 1152f7238e50SJosef Bacik 1153f7238e50SJosef Bacik return ret; 1154f7238e50SJosef Bacik } 1155f7238e50SJosef Bacik 1156abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1157abed4aaaSJosef Bacik { 1158abed4aaaSJosef Bacik struct btrfs_key key = { 1159abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 1160abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1161f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1162abed4aaaSJosef Bacik }; 1163abed4aaaSJosef Bacik 1164abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1165abed4aaaSJosef Bacik } 1166abed4aaaSJosef Bacik 1167abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1168abed4aaaSJosef Bacik { 1169abed4aaaSJosef Bacik struct btrfs_key key = { 1170abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 1171abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1172f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1173abed4aaaSJosef Bacik }; 1174abed4aaaSJosef Bacik 1175abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1176abed4aaaSJosef Bacik } 1177abed4aaaSJosef Bacik 117851129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 117951129b33SJosef Bacik { 118051129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 118151129b33SJosef Bacik return fs_info->block_group_root; 118251129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 118351129b33SJosef Bacik } 118451129b33SJosef Bacik 118520897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 118620897f5cSArne Jansen u64 objectid) 118720897f5cSArne Jansen { 11889b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 118920897f5cSArne Jansen struct extent_buffer *leaf; 119020897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 119120897f5cSArne Jansen struct btrfs_root *root; 119220897f5cSArne Jansen struct btrfs_key key; 1193b89f6d1fSFilipe Manana unsigned int nofs_flag; 119420897f5cSArne Jansen int ret = 0; 119520897f5cSArne Jansen 1196b89f6d1fSFilipe Manana /* 1197b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1198b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1199b89f6d1fSFilipe Manana */ 1200b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 120196dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1202b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 120320897f5cSArne Jansen if (!root) 120420897f5cSArne Jansen return ERR_PTR(-ENOMEM); 120520897f5cSArne Jansen 120620897f5cSArne Jansen root->root_key.objectid = objectid; 120720897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 120820897f5cSArne Jansen root->root_key.offset = 0; 120920897f5cSArne Jansen 12109631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 12119631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 121220897f5cSArne Jansen if (IS_ERR(leaf)) { 121320897f5cSArne Jansen ret = PTR_ERR(leaf); 12141dd05682STsutomu Itoh leaf = NULL; 1215c1b07854SPeng Hao goto fail; 121620897f5cSArne Jansen } 121720897f5cSArne Jansen 121820897f5cSArne Jansen root->node = leaf; 121920897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 122020897f5cSArne Jansen 122120897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 122227cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 122320897f5cSArne Jansen 1224f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1225f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 122620897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 122720897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 122820897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 122920897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 123020897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 123120897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 123220897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 123333d85fdaSQu Wenruo if (is_fstree(objectid)) 1234807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1235807fc790SAndy Shevchenko else 1236807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1237c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 123820897f5cSArne Jansen 12398a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 12408a6a87cdSBoris Burkov 124120897f5cSArne Jansen key.objectid = objectid; 124220897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 124320897f5cSArne Jansen key.offset = 0; 124420897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 124520897f5cSArne Jansen if (ret) 124620897f5cSArne Jansen goto fail; 124720897f5cSArne Jansen 124820897f5cSArne Jansen return root; 12491dd05682STsutomu Itoh 12508a6a87cdSBoris Burkov fail: 125100246528SJosef Bacik btrfs_put_root(root); 12521dd05682STsutomu Itoh 12531dd05682STsutomu Itoh return ERR_PTR(ret); 125420897f5cSArne Jansen } 125520897f5cSArne Jansen 12567237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1257e02119d5SChris Mason struct btrfs_fs_info *fs_info) 12580f7d52f4SChris Mason { 12590f7d52f4SChris Mason struct btrfs_root *root; 1260e02119d5SChris Mason 126196dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1262e02119d5SChris Mason if (!root) 12637237f183SYan Zheng return ERR_PTR(-ENOMEM); 1264e02119d5SChris Mason 1265e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1266e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1267e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 126827cdeb70SMiao Xie 12696ab6ebb7SNaohiro Aota return root; 12706ab6ebb7SNaohiro Aota } 12716ab6ebb7SNaohiro Aota 12726ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 12736ab6ebb7SNaohiro Aota struct btrfs_root *root) 12746ab6ebb7SNaohiro Aota { 12756ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 12766ab6ebb7SNaohiro Aota 12777237f183SYan Zheng /* 127892a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 127927cdeb70SMiao Xie * 128092a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 128192a7cc42SQu Wenruo * and they go away before a real commit is actually done. 128292a7cc42SQu Wenruo * 128392a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 128492a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 12857237f183SYan Zheng */ 1286e02119d5SChris Mason 12874d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 12889631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 12896ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 12906ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 1291e02119d5SChris Mason 12927237f183SYan Zheng root->node = leaf; 1293e02119d5SChris Mason 1294e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1295e02119d5SChris Mason btrfs_tree_unlock(root->node); 12966ab6ebb7SNaohiro Aota 12976ab6ebb7SNaohiro Aota return 0; 12987237f183SYan Zheng } 12997237f183SYan Zheng 13007237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13017237f183SYan Zheng struct btrfs_fs_info *fs_info) 13027237f183SYan Zheng { 13037237f183SYan Zheng struct btrfs_root *log_root; 13047237f183SYan Zheng 13057237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13067237f183SYan Zheng if (IS_ERR(log_root)) 13077237f183SYan Zheng return PTR_ERR(log_root); 13086ab6ebb7SNaohiro Aota 13093ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 13103ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 13113ddebf27SNaohiro Aota 13126ab6ebb7SNaohiro Aota if (ret) { 13136ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13146ab6ebb7SNaohiro Aota return ret; 13156ab6ebb7SNaohiro Aota } 13163ddebf27SNaohiro Aota } 13176ab6ebb7SNaohiro Aota 13187237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13197237f183SYan Zheng fs_info->log_root_tree = log_root; 13207237f183SYan Zheng return 0; 13217237f183SYan Zheng } 13227237f183SYan Zheng 13237237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13247237f183SYan Zheng struct btrfs_root *root) 13257237f183SYan Zheng { 13260b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13277237f183SYan Zheng struct btrfs_root *log_root; 13287237f183SYan Zheng struct btrfs_inode_item *inode_item; 13296ab6ebb7SNaohiro Aota int ret; 13307237f183SYan Zheng 13310b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13327237f183SYan Zheng if (IS_ERR(log_root)) 13337237f183SYan Zheng return PTR_ERR(log_root); 13347237f183SYan Zheng 13356ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 13366ab6ebb7SNaohiro Aota if (ret) { 13376ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13386ab6ebb7SNaohiro Aota return ret; 13396ab6ebb7SNaohiro Aota } 13406ab6ebb7SNaohiro Aota 13417237f183SYan Zheng log_root->last_trans = trans->transid; 13427237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 13437237f183SYan Zheng 13447237f183SYan Zheng inode_item = &log_root->root_item.inode; 13453cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 13463cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 13473cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1348da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 13490b246afaSJeff Mahoney fs_info->nodesize); 13503cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 13517237f183SYan Zheng 13525d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 13537237f183SYan Zheng 13547237f183SYan Zheng WARN_ON(root->log_root); 13557237f183SYan Zheng root->log_root = log_root; 13567237f183SYan Zheng root->log_transid = 0; 1357d1433debSMiao Xie root->log_transid_committed = -1; 1358257c62e1SChris Mason root->last_log_commit = 0; 1359e02119d5SChris Mason return 0; 1360e02119d5SChris Mason } 1361e02119d5SChris Mason 136249d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 136349d11beaSJosef Bacik struct btrfs_path *path, 1364cb517eabSMiao Xie struct btrfs_key *key) 1365e02119d5SChris Mason { 1366e02119d5SChris Mason struct btrfs_root *root; 1367e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 136884234f3aSYan Zheng u64 generation; 1369cb517eabSMiao Xie int ret; 1370581c1760SQu Wenruo int level; 1371cb517eabSMiao Xie 137296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 137349d11beaSJosef Bacik if (!root) 137449d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 13750f7d52f4SChris Mason 1376cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1377cb517eabSMiao Xie &root->root_item, &root->root_key); 13780f7d52f4SChris Mason if (ret) { 137913a8a7c8SYan, Zheng if (ret > 0) 138013a8a7c8SYan, Zheng ret = -ENOENT; 138149d11beaSJosef Bacik goto fail; 13820f7d52f4SChris Mason } 138313a8a7c8SYan, Zheng 138484234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1385581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 13862ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 13872ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 13881b7ec85eSJosef Bacik key->objectid, generation, level, NULL); 138964c043deSLiu Bo if (IS_ERR(root->node)) { 139064c043deSLiu Bo ret = PTR_ERR(root->node); 13918c38938cSJosef Bacik root->node = NULL; 139249d11beaSJosef Bacik goto fail; 13934eb150d6SQu Wenruo } 13944eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1395cb517eabSMiao Xie ret = -EIO; 139649d11beaSJosef Bacik goto fail; 1397416bc658SJosef Bacik } 139888c602abSQu Wenruo 139988c602abSQu Wenruo /* 140088c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 140188c602abSQu Wenruo * match its root node owner 140288c602abSQu Wenruo */ 140388c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 140488c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 140588c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 140688c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 140788c602abSQu Wenruo btrfs_crit(fs_info, 140888c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 140988c602abSQu Wenruo root->root_key.objectid, root->node->start, 141088c602abSQu Wenruo btrfs_header_owner(root->node), 141188c602abSQu Wenruo root->root_key.objectid); 141288c602abSQu Wenruo ret = -EUCLEAN; 141388c602abSQu Wenruo goto fail; 141488c602abSQu Wenruo } 14155d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1416cb517eabSMiao Xie return root; 141749d11beaSJosef Bacik fail: 141800246528SJosef Bacik btrfs_put_root(root); 141949d11beaSJosef Bacik return ERR_PTR(ret); 142049d11beaSJosef Bacik } 142149d11beaSJosef Bacik 142249d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 142349d11beaSJosef Bacik struct btrfs_key *key) 142449d11beaSJosef Bacik { 142549d11beaSJosef Bacik struct btrfs_root *root; 142649d11beaSJosef Bacik struct btrfs_path *path; 142749d11beaSJosef Bacik 142849d11beaSJosef Bacik path = btrfs_alloc_path(); 142949d11beaSJosef Bacik if (!path) 143049d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 143149d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 143249d11beaSJosef Bacik btrfs_free_path(path); 143349d11beaSJosef Bacik 143449d11beaSJosef Bacik return root; 1435cb517eabSMiao Xie } 1436cb517eabSMiao Xie 14372dfb1e43SQu Wenruo /* 14382dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 14392dfb1e43SQu Wenruo * 14402dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 14412dfb1e43SQu Wenruo */ 14422dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1443cb517eabSMiao Xie { 1444cb517eabSMiao Xie int ret; 1445dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1446cb517eabSMiao Xie 1447dcc3eb96SNikolay Borisov /* 1448dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1449dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1450dcc3eb96SNikolay Borisov */ 1451dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1452dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1453dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1454dcc3eb96SNikolay Borisov if (ret) 14558257b2dcSMiao Xie goto fail; 14568257b2dcSMiao Xie 1457aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 145837f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 145992a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1460f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1461f39e4571SJosef Bacik } 1462f39e4571SJosef Bacik 1463851fd730SQu Wenruo /* 1464851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1465851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1466851fd730SQu Wenruo */ 1467851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1468851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 14692dfb1e43SQu Wenruo if (!anon_dev) { 1470cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1471cb517eabSMiao Xie if (ret) 1472876d2cf1SLiu Bo goto fail; 14732dfb1e43SQu Wenruo } else { 14742dfb1e43SQu Wenruo root->anon_dev = anon_dev; 14752dfb1e43SQu Wenruo } 1476851fd730SQu Wenruo } 1477f32e48e9SChandan Rajendra 1478f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1479453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1480f32e48e9SChandan Rajendra if (ret) { 1481f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1482876d2cf1SLiu Bo goto fail; 1483f32e48e9SChandan Rajendra } 1484f32e48e9SChandan Rajendra 14856b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1486f32e48e9SChandan Rajendra 1487f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1488f32e48e9SChandan Rajendra 1489cb517eabSMiao Xie return 0; 1490cb517eabSMiao Xie fail: 149184db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1492cb517eabSMiao Xie return ret; 1493cb517eabSMiao Xie } 1494cb517eabSMiao Xie 1495a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1496cb517eabSMiao Xie u64 root_id) 1497cb517eabSMiao Xie { 1498cb517eabSMiao Xie struct btrfs_root *root; 1499cb517eabSMiao Xie 1500fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1501fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1502fc7cbcd4SDavid Sterba (unsigned long)root_id); 1503bc44d7c4SJosef Bacik if (root) 150400246528SJosef Bacik root = btrfs_grab_root(root); 1505fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1506cb517eabSMiao Xie return root; 1507cb517eabSMiao Xie } 1508cb517eabSMiao Xie 150949d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 151049d11beaSJosef Bacik u64 objectid) 151149d11beaSJosef Bacik { 1512abed4aaaSJosef Bacik struct btrfs_key key = { 1513abed4aaaSJosef Bacik .objectid = objectid, 1514abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1515abed4aaaSJosef Bacik .offset = 0, 1516abed4aaaSJosef Bacik }; 1517abed4aaaSJosef Bacik 151849d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 151949d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 152049d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 1521abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 152249d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 152349d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 152449d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 152549d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 152649d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 1527abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 152849d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 152949d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 153049d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 153149d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 153249d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 153349d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 153414033b08SQu Wenruo if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) 153514033b08SQu Wenruo return btrfs_grab_root(fs_info->block_group_root) ? 153614033b08SQu Wenruo fs_info->block_group_root : ERR_PTR(-ENOENT); 1537abed4aaaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { 1538abed4aaaSJosef Bacik struct btrfs_root *root = btrfs_global_root(fs_info, &key); 1539abed4aaaSJosef Bacik 1540abed4aaaSJosef Bacik return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); 1541abed4aaaSJosef Bacik } 154249d11beaSJosef Bacik return NULL; 154349d11beaSJosef Bacik } 154449d11beaSJosef Bacik 1545cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1546cb517eabSMiao Xie struct btrfs_root *root) 1547cb517eabSMiao Xie { 1548cb517eabSMiao Xie int ret; 1549cb517eabSMiao Xie 1550fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1551fc7cbcd4SDavid Sterba if (ret) 1552fc7cbcd4SDavid Sterba return ret; 1553fc7cbcd4SDavid Sterba 1554fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1555fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1556fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1557fc7cbcd4SDavid Sterba root); 1558af01d2e5SJosef Bacik if (ret == 0) { 155900246528SJosef Bacik btrfs_grab_root(root); 1560fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1561af01d2e5SJosef Bacik } 1562fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1563fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1564cb517eabSMiao Xie 1565cb517eabSMiao Xie return ret; 1566cb517eabSMiao Xie } 1567cb517eabSMiao Xie 1568bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1569bd647ce3SJosef Bacik { 1570bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1571bd647ce3SJosef Bacik struct btrfs_root *root; 1572bd647ce3SJosef Bacik 1573bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1574457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1575457f1864SJosef Bacik 1576bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1577bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1578457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 157971008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1580bd647ce3SJosef Bacik refcount_read(&root->refs)); 1581bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 158200246528SJosef Bacik btrfs_put_root(root); 158300246528SJosef Bacik btrfs_put_root(root); 1584bd647ce3SJosef Bacik } 1585bd647ce3SJosef Bacik #endif 1586bd647ce3SJosef Bacik } 1587bd647ce3SJosef Bacik 1588abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1589abed4aaaSJosef Bacik { 1590abed4aaaSJosef Bacik struct btrfs_root *root; 1591abed4aaaSJosef Bacik struct rb_node *node; 1592abed4aaaSJosef Bacik 1593abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1594abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1595abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1596abed4aaaSJosef Bacik btrfs_put_root(root); 1597abed4aaaSJosef Bacik } 1598abed4aaaSJosef Bacik } 1599abed4aaaSJosef Bacik 16000d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 16010d4b0463SJosef Bacik { 1602141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1603141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 16045deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1605141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1606141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1607141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1608141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 16090d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 16100d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1611abed4aaaSJosef Bacik free_global_roots(fs_info); 161200246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 161300246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 161400246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 161500246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 161600246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 161700246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1618aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 16199c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1620bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 16213fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 16220d4b0463SJosef Bacik kfree(fs_info->super_copy); 16230d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 16248481dd80SQu Wenruo kfree(fs_info->subpage_info); 16250d4b0463SJosef Bacik kvfree(fs_info); 16260d4b0463SJosef Bacik } 16270d4b0463SJosef Bacik 16280d4b0463SJosef Bacik 16292dfb1e43SQu Wenruo /* 16302dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 16312dfb1e43SQu Wenruo * 16322dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 16332dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 16342dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 16352dfb1e43SQu Wenruo * 16362dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 16372dfb1e43SQu Wenruo * 16382dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 16392dfb1e43SQu Wenruo * same root objectid. 16402dfb1e43SQu Wenruo * 16412dfb1e43SQu Wenruo * @objectid: root id 16422dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 16432dfb1e43SQu Wenruo * pass 0 for new allocation. 16442dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 16452dfb1e43SQu Wenruo * for orphan roots 16462dfb1e43SQu Wenruo */ 16472dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 16482dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 16492dfb1e43SQu Wenruo bool check_ref) 16505eda7b5eSChris Mason { 16515eda7b5eSChris Mason struct btrfs_root *root; 1652381cf658SDavid Sterba struct btrfs_path *path; 16531d4c08e0SDavid Sterba struct btrfs_key key; 16545eda7b5eSChris Mason int ret; 16555eda7b5eSChris Mason 165649d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 165749d11beaSJosef Bacik if (root) 165849d11beaSJosef Bacik return root; 16594df27c4dSYan, Zheng again: 166056e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 166148475471SStefan Behrens if (root) { 16622dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 16632dfb1e43SQu Wenruo ASSERT(!anon_dev); 1664bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 166500246528SJosef Bacik btrfs_put_root(root); 166648475471SStefan Behrens return ERR_PTR(-ENOENT); 1667bc44d7c4SJosef Bacik } 16685eda7b5eSChris Mason return root; 166948475471SStefan Behrens } 16705eda7b5eSChris Mason 167156e9357aSDavid Sterba key.objectid = objectid; 167256e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 167356e9357aSDavid Sterba key.offset = (u64)-1; 167456e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 16755eda7b5eSChris Mason if (IS_ERR(root)) 16765eda7b5eSChris Mason return root; 16773394e160SChris Mason 1678c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1679d68fc57bSYan, Zheng ret = -ENOENT; 1680d68fc57bSYan, Zheng goto fail; 1681d68fc57bSYan, Zheng } 1682d68fc57bSYan, Zheng 16832dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1684cb517eabSMiao Xie if (ret) 1685cb517eabSMiao Xie goto fail; 1686cb517eabSMiao Xie 1687381cf658SDavid Sterba path = btrfs_alloc_path(); 1688381cf658SDavid Sterba if (!path) { 1689381cf658SDavid Sterba ret = -ENOMEM; 1690381cf658SDavid Sterba goto fail; 1691381cf658SDavid Sterba } 16921d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 16931d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 169456e9357aSDavid Sterba key.offset = objectid; 16951d4c08e0SDavid Sterba 16961d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1697381cf658SDavid Sterba btrfs_free_path(path); 1698d68fc57bSYan, Zheng if (ret < 0) 1699d68fc57bSYan, Zheng goto fail; 1700d68fc57bSYan, Zheng if (ret == 0) 170127cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1702d68fc57bSYan, Zheng 1703cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 17040f7d52f4SChris Mason if (ret) { 1705168a2f77SJia-Ju Bai if (ret == -EEXIST) { 170600246528SJosef Bacik btrfs_put_root(root); 17074df27c4dSYan, Zheng goto again; 1708168a2f77SJia-Ju Bai } 17094df27c4dSYan, Zheng goto fail; 17104df27c4dSYan, Zheng } 1711edbd8d4eSChris Mason return root; 17124df27c4dSYan, Zheng fail: 171333fab972SFilipe Manana /* 171433fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1715143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 171633fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 171733fab972SFilipe Manana * and once again by our caller. 171833fab972SFilipe Manana */ 171933fab972SFilipe Manana if (anon_dev) 172033fab972SFilipe Manana root->anon_dev = 0; 17218c38938cSJosef Bacik btrfs_put_root(root); 17224df27c4dSYan, Zheng return ERR_PTR(ret); 1723edbd8d4eSChris Mason } 1724edbd8d4eSChris Mason 17252dfb1e43SQu Wenruo /* 17262dfb1e43SQu Wenruo * Get in-memory reference of a root structure 17272dfb1e43SQu Wenruo * 17282dfb1e43SQu Wenruo * @objectid: tree objectid 17292dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 17302dfb1e43SQu Wenruo * one reference 17312dfb1e43SQu Wenruo */ 17322dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 17332dfb1e43SQu Wenruo u64 objectid, bool check_ref) 17342dfb1e43SQu Wenruo { 17352dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 17362dfb1e43SQu Wenruo } 17372dfb1e43SQu Wenruo 17382dfb1e43SQu Wenruo /* 17392dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 17402dfb1e43SQu Wenruo * the anonymous block device id 17412dfb1e43SQu Wenruo * 17422dfb1e43SQu Wenruo * @objectid: tree objectid 17432dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 17442dfb1e43SQu Wenruo * parameter value 17452dfb1e43SQu Wenruo */ 17462dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 17472dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 17482dfb1e43SQu Wenruo { 17492dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 17502dfb1e43SQu Wenruo } 17512dfb1e43SQu Wenruo 17528b712842SChris Mason /* 175349d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 175449d11beaSJosef Bacik * @fs_info: the fs_info 175549d11beaSJosef Bacik * @objectid: the objectid we need to lookup 175649d11beaSJosef Bacik * 175749d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 175849d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 175949d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 176049d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 176149d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 176249d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 176349d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 176449d11beaSJosef Bacik * backref code is finished using the root. 176549d11beaSJosef Bacik */ 176649d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 176749d11beaSJosef Bacik struct btrfs_path *path, 176849d11beaSJosef Bacik u64 objectid) 176949d11beaSJosef Bacik { 177049d11beaSJosef Bacik struct btrfs_root *root; 177149d11beaSJosef Bacik struct btrfs_key key; 177249d11beaSJosef Bacik 177349d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 177449d11beaSJosef Bacik 177549d11beaSJosef Bacik /* 177649d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 177749d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 177849d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 177949d11beaSJosef Bacik * != NULL just return it. 178049d11beaSJosef Bacik */ 178149d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 178249d11beaSJosef Bacik if (root) 178349d11beaSJosef Bacik return root; 178449d11beaSJosef Bacik 178549d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 178649d11beaSJosef Bacik if (root) 178749d11beaSJosef Bacik return root; 178849d11beaSJosef Bacik 178949d11beaSJosef Bacik key.objectid = objectid; 179049d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 179149d11beaSJosef Bacik key.offset = (u64)-1; 179249d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 179349d11beaSJosef Bacik btrfs_release_path(path); 179449d11beaSJosef Bacik 179549d11beaSJosef Bacik return root; 179649d11beaSJosef Bacik } 179749d11beaSJosef Bacik 1798a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1799a74a4b97SChris Mason { 18000d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1801d0278245SMiao Xie int again; 1802a74a4b97SChris Mason 1803d6fd0ae2SOmar Sandoval while (1) { 1804d0278245SMiao Xie again = 0; 18059d1a2a3aSDavid Sterba 1806fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1807fd340d0fSJosef Bacik 1808d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 18092ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1810d0278245SMiao Xie goto sleep; 1811d0278245SMiao Xie 181290c711abSZygo Blaxell /* 181390c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 181490c711abSZygo Blaxell * before we have finished mounting the filesystem. 181590c711abSZygo Blaxell */ 18160b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 181790c711abSZygo Blaxell goto sleep; 181890c711abSZygo Blaxell 18190b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1820d0278245SMiao Xie goto sleep; 1821d0278245SMiao Xie 1822dc7f370cSMiao Xie /* 1823dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1824dc7f370cSMiao Xie * during the above check and trylock. 1825dc7f370cSMiao Xie */ 18262ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 18270b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1828dc7f370cSMiao Xie goto sleep; 1829dc7f370cSMiao Xie } 1830dc7f370cSMiao Xie 18312ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1832c2d6cb16SFilipe Manana 183333c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 18340b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1835a74a4b97SChris Mason 1836d0278245SMiao Xie /* 183705323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 183805323cd1SMiao Xie * needn't do anything special here. 1839d0278245SMiao Xie */ 18400b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 184167c5e7d4SFilipe Manana 184267c5e7d4SFilipe Manana /* 1843f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 184467c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 184567c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1846f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 184767c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 184867c5e7d4SFilipe Manana * unused block groups. 184967c5e7d4SFilipe Manana */ 18500b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 185118bb8bbfSJohannes Thumshirn 185218bb8bbfSJohannes Thumshirn /* 185318bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 185418bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 185518bb8bbfSJohannes Thumshirn * space. 185618bb8bbfSJohannes Thumshirn */ 185718bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1858d0278245SMiao Xie sleep: 1859a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1860d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1861d6fd0ae2SOmar Sandoval kthread_parkme(); 1862d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1863d6fd0ae2SOmar Sandoval return 0; 1864838fe188SJiri Kosina if (!again) { 1865a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1866a74a4b97SChris Mason schedule(); 1867a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1868a74a4b97SChris Mason } 1869da288d28SFilipe Manana } 1870a74a4b97SChris Mason } 1871a74a4b97SChris Mason 1872a74a4b97SChris Mason static int transaction_kthread(void *arg) 1873a74a4b97SChris Mason { 1874a74a4b97SChris Mason struct btrfs_root *root = arg; 18750b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1876a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1877a74a4b97SChris Mason struct btrfs_transaction *cur; 18788929ecfaSYan, Zheng u64 transid; 1879643900beSNikolay Borisov time64_t delta; 1880a74a4b97SChris Mason unsigned long delay; 1881914b2007SJan Kara bool cannot_commit; 1882a74a4b97SChris Mason 1883a74a4b97SChris Mason do { 1884914b2007SJan Kara cannot_commit = false; 1885ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 18860b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1887a74a4b97SChris Mason 18880b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 18890b246afaSJeff Mahoney cur = fs_info->running_transaction; 1890a74a4b97SChris Mason if (!cur) { 18910b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1892a74a4b97SChris Mason goto sleep; 1893a74a4b97SChris Mason } 189431153d81SYan Zheng 1895643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1896fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1897fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1898643900beSNikolay Borisov delta < fs_info->commit_interval) { 18990b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1900fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1901fb8a7e94SNikolay Borisov delay = min(delay, 1902fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1903a74a4b97SChris Mason goto sleep; 1904a74a4b97SChris Mason } 19058929ecfaSYan, Zheng transid = cur->transid; 19060b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 190756bec294SChris Mason 190879787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1909354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1910914b2007SJan Kara if (IS_ERR(trans)) { 1911354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1912914b2007SJan Kara cannot_commit = true; 191379787eaaSJeff Mahoney goto sleep; 1914914b2007SJan Kara } 19158929ecfaSYan, Zheng if (transid == trans->transid) { 19163a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 19178929ecfaSYan, Zheng } else { 19183a45bb20SJeff Mahoney btrfs_end_transaction(trans); 19198929ecfaSYan, Zheng } 1920a74a4b97SChris Mason sleep: 19210b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 19220b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1923a74a4b97SChris Mason 192484961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 19252ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 19268929ecfaSYan, Zheng if (!kthread_should_stop() && 19270b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1928914b2007SJan Kara cannot_commit)) 1929bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1930a74a4b97SChris Mason } while (!kthread_should_stop()); 1931a74a4b97SChris Mason return 0; 1932a74a4b97SChris Mason } 1933a74a4b97SChris Mason 1934af31f5e5SChris Mason /* 193501f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 193601f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 193701f0f9daSNikolay Borisov * anything. 1938af31f5e5SChris Mason * 1939af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1940af31f5e5SChris Mason * generation of the latest root in the array with the generation 1941af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1942af31f5e5SChris Mason */ 194301f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1944af31f5e5SChris Mason { 194501f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1946af31f5e5SChris Mason u64 cur; 1947af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1948af31f5e5SChris Mason int i; 1949af31f5e5SChris Mason 1950af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1951af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1952af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1953af31f5e5SChris Mason if (cur == newest_gen) 195401f0f9daSNikolay Borisov return i; 1955af31f5e5SChris Mason } 1956af31f5e5SChris Mason 195701f0f9daSNikolay Borisov return -EINVAL; 1958af31f5e5SChris Mason } 1959af31f5e5SChris Mason 1960af31f5e5SChris Mason /* 1961af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1962af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1963af31f5e5SChris Mason * done 1964af31f5e5SChris Mason */ 1965af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1966af31f5e5SChris Mason { 19676ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1968af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1969af31f5e5SChris Mason 1970af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1971af31f5e5SChris Mason 1972af31f5e5SChris Mason /* 1973af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1974af31f5e5SChris Mason * regardless of which ones we use today 1975af31f5e5SChris Mason */ 1976af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1977af31f5e5SChris Mason 1978af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1979af31f5e5SChris Mason 1980af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1981af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1982af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1983af31f5e5SChris Mason 1984af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1985af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1986af31f5e5SChris Mason 1987af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1988af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1989af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1990af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1991af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1992af31f5e5SChris Mason 19931c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 19949c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1995f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 19969c54e80dSJosef Bacik 19979c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 19989c54e80dSJosef Bacik extent_root->node->start); 1999af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 200029cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 2001af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 200229cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 2003af31f5e5SChris Mason 2004f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 2005f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 2006f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 2007f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 2008f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 20099c54e80dSJosef Bacik } 2010af31f5e5SChris Mason 20117c7e82a7SChris Mason /* 20127c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 20137c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 20147c7e82a7SChris Mason */ 20157c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 20167c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 20177c7e82a7SChris Mason info->fs_root->node->start); 2018af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2019af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2020af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2021af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 20227c7e82a7SChris Mason } 2023af31f5e5SChris Mason 2024af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2025af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2026af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2027af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2028af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2029af31f5e5SChris Mason 2030af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2031af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2032af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2033af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2034af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2035af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2036af31f5e5SChris Mason 2037af31f5e5SChris Mason /* 2038af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2039af31f5e5SChris Mason * for the next commit 2040af31f5e5SChris Mason */ 2041af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2042af31f5e5SChris Mason &info->super_for_commit->super_roots, 2043af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2044af31f5e5SChris Mason } 2045af31f5e5SChris Mason 2046af31f5e5SChris Mason /* 2047bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 2048bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 2049bd2336b2SNikolay Borisov * 2050bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 2051bd2336b2SNikolay Borisov * priority - priority of backup root required 2052bd2336b2SNikolay Borisov * 2053bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 2054bd2336b2SNikolay Borisov */ 2055bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 2056bd2336b2SNikolay Borisov { 2057bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2058bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 2059bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 2060bd2336b2SNikolay Borisov 2061bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 2062bd2336b2SNikolay Borisov if (priority == 0) 2063bd2336b2SNikolay Borisov return backup_index; 2064bd2336b2SNikolay Borisov 2065bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 2066bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 2067bd2336b2SNikolay Borisov } else { 2068bd2336b2SNikolay Borisov return -EINVAL; 2069bd2336b2SNikolay Borisov } 2070bd2336b2SNikolay Borisov 2071bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 2072bd2336b2SNikolay Borisov 2073bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 2074bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 2075bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2076bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 2077bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 2078bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2079bd2336b2SNikolay Borisov 2080bd2336b2SNikolay Borisov /* 2081bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 2082bd2336b2SNikolay Borisov * need a fsck 2083bd2336b2SNikolay Borisov */ 2084bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2085bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2086bd2336b2SNikolay Borisov 2087bd2336b2SNikolay Borisov return backup_index; 2088bd2336b2SNikolay Borisov } 2089bd2336b2SNikolay Borisov 20907abadb64SLiu Bo /* helper to cleanup workers */ 20917abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20927abadb64SLiu Bo { 2093dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2094afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 2095a31b4a43SChristoph Hellwig btrfs_destroy_workqueue(fs_info->hipri_workers); 20965cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2097d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 2098d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 2099d34e123dSChristoph Hellwig if (fs_info->endio_raid56_workers) 2100d34e123dSChristoph Hellwig destroy_workqueue(fs_info->endio_raid56_workers); 2101385de0efSChristoph Hellwig if (fs_info->rmw_workers) 2102385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 2103fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 2104fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 2105fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2106fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 21075b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2108e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2109a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2110fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2111b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 2112b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 2113a9b9477dSFilipe Manana /* 2114a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2115a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2116a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2117a9b9477dSFilipe Manana */ 2118d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 2119d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 21207abadb64SLiu Bo } 21217abadb64SLiu Bo 21222e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 21232e9f5954SRashika { 21242e9f5954SRashika if (root) { 21252e9f5954SRashika free_extent_buffer(root->node); 21262e9f5954SRashika free_extent_buffer(root->commit_root); 21272e9f5954SRashika root->node = NULL; 21282e9f5954SRashika root->commit_root = NULL; 21292e9f5954SRashika } 21302e9f5954SRashika } 21312e9f5954SRashika 2132abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 2133abed4aaaSJosef Bacik { 2134abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 2135abed4aaaSJosef Bacik 2136abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 2137abed4aaaSJosef Bacik &fs_info->global_root_tree, 2138abed4aaaSJosef Bacik rb_node) 2139abed4aaaSJosef Bacik free_root_extent_buffers(root); 2140abed4aaaSJosef Bacik } 2141abed4aaaSJosef Bacik 2142af31f5e5SChris Mason /* helper to cleanup tree roots */ 21434273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2144af31f5e5SChris Mason { 21452e9f5954SRashika free_root_extent_buffers(info->tree_root); 2146655b09feSJosef Bacik 2147abed4aaaSJosef Bacik free_global_root_pointers(info); 21482e9f5954SRashika free_root_extent_buffers(info->dev_root); 21492e9f5954SRashika free_root_extent_buffers(info->quota_root); 21502e9f5954SRashika free_root_extent_buffers(info->uuid_root); 21518c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2152aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 21539c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 21544273eaffSAnand Jain if (free_chunk_root) 21552e9f5954SRashika free_root_extent_buffers(info->chunk_root); 2156af31f5e5SChris Mason } 2157af31f5e5SChris Mason 21588c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 21598c38938cSJosef Bacik { 21608c38938cSJosef Bacik if (!root) 21618c38938cSJosef Bacik return; 21628c38938cSJosef Bacik 21638c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 21648c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 21651dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 21668c38938cSJosef Bacik if (root->anon_dev) 21678c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 21688c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2169923eb523SJohannes Thumshirn free_root_extent_buffers(root); 21708c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2171fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 21728c38938cSJosef Bacik list_del_init(&root->leak_list); 2173fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 21748c38938cSJosef Bacik #endif 21758c38938cSJosef Bacik kfree(root); 21768c38938cSJosef Bacik } 21778c38938cSJosef Bacik } 21788c38938cSJosef Bacik 2179faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2180171f6537SJosef Bacik { 2181fc7cbcd4SDavid Sterba int ret; 2182fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 2183fc7cbcd4SDavid Sterba int i; 2184171f6537SJosef Bacik 2185171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2186fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 2187171f6537SJosef Bacik struct btrfs_root, root_list); 2188fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 2189171f6537SJosef Bacik 2190fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2191fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2192fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 2193171f6537SJosef Bacik } 2194171f6537SJosef Bacik 2195fc7cbcd4SDavid Sterba while (1) { 2196fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2197fc7cbcd4SDavid Sterba (void **)gang, 0, 2198fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 2199fc7cbcd4SDavid Sterba if (!ret) 2200fc7cbcd4SDavid Sterba break; 2201fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 2202fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2203171f6537SJosef Bacik } 2204171f6537SJosef Bacik } 2205af31f5e5SChris Mason 2206638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2207638aa7edSEric Sandeen { 2208638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2209638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2210638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2211638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2212638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2213638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2214ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2215638aa7edSEric Sandeen } 2216638aa7edSEric Sandeen 2217779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2218779a65a4SEric Sandeen { 2219779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2220779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2221779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2222779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2223779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2224779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2225907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 2226779a65a4SEric Sandeen } 2227779a65a4SEric Sandeen 22286bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2229f37938e0SEric Sandeen { 22302ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 2231e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 2232e256927bSJosef Bacik fs_info->tree_root); 22332ff7e61eSJeff Mahoney 22342ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 22352ff7e61eSJeff Mahoney set_nlink(inode, 1); 2236f37938e0SEric Sandeen /* 2237f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2238f37938e0SEric Sandeen * the real end of the address space is determined by all of 2239f37938e0SEric Sandeen * the devices in the system 2240f37938e0SEric Sandeen */ 22412ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 22422ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2243f37938e0SEric Sandeen 22442ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 224543eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 2246efb0645bSJosef Bacik IO_TREE_BTREE_INODE_IO, NULL); 22472ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2248f37938e0SEric Sandeen 22495c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 2250adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 2251adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 2252adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 22532ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 2254e256927bSJosef Bacik __insert_inode_hash(inode, hash); 2255f37938e0SEric Sandeen } 2256f37938e0SEric Sandeen 2257ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2258ad618368SEric Sandeen { 2259ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2260129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 22617f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2262ad618368SEric Sandeen } 2263ad618368SEric Sandeen 2264f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2265f9e92e40SEric Sandeen { 2266f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2267f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2268f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2269f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2270f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2271f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2272d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2273011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 2274f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2275f9e92e40SEric Sandeen } 2276f9e92e40SEric Sandeen 2277d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 22782a458198SEric Sandeen { 2279f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 22806f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 22812a458198SEric Sandeen 22822a458198SEric Sandeen fs_info->workers = 2283a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 2284a31b4a43SChristoph Hellwig fs_info->hipri_workers = 2285a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker-high", 2286cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 22872a458198SEric Sandeen 22882a458198SEric Sandeen fs_info->delalloc_workers = 2289cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2290cb001095SJeff Mahoney flags, max_active, 2); 22912a458198SEric Sandeen 22922a458198SEric Sandeen fs_info->flush_workers = 2293cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2294cb001095SJeff Mahoney flags, max_active, 0); 22952a458198SEric Sandeen 22962a458198SEric Sandeen fs_info->caching_workers = 2297cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 22982a458198SEric Sandeen 22992a458198SEric Sandeen fs_info->fixup_workers = 2300cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 23012a458198SEric Sandeen 23022a458198SEric Sandeen fs_info->endio_workers = 2303d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 23042a458198SEric Sandeen fs_info->endio_meta_workers = 2305d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 23062a458198SEric Sandeen fs_info->endio_raid56_workers = 2307d34e123dSChristoph Hellwig alloc_workqueue("btrfs-endio-raid56", flags, max_active); 2308385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 23092a458198SEric Sandeen fs_info->endio_write_workers = 2310cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2311cb001095SJeff Mahoney max_active, 2); 2312fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 2313fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 23142a458198SEric Sandeen fs_info->endio_freespace_worker = 2315cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2316cb001095SJeff Mahoney max_active, 0); 23172a458198SEric Sandeen fs_info->delayed_workers = 2318cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2319cb001095SJeff Mahoney max_active, 0); 23202a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2321cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2322b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2323b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 23242a458198SEric Sandeen 2325a31b4a43SChristoph Hellwig if (!(fs_info->workers && fs_info->hipri_workers && 2326a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 23272a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 2328fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 23292a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 23302a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 2331f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 2332f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 2333b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 23342a458198SEric Sandeen return -ENOMEM; 23352a458198SEric Sandeen } 23362a458198SEric Sandeen 23372a458198SEric Sandeen return 0; 23382a458198SEric Sandeen } 23392a458198SEric Sandeen 23406d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 23416d97c6e3SJohannes Thumshirn { 23426d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2343b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 23446d97c6e3SJohannes Thumshirn 2345b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 23466d97c6e3SJohannes Thumshirn 23476d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 23486d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2349b4e967beSDavid Sterba csum_driver); 23506d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 23516d97c6e3SJohannes Thumshirn } 23526d97c6e3SJohannes Thumshirn 23536d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 23546d97c6e3SJohannes Thumshirn 2355c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2356c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2357c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 23586d97c6e3SJohannes Thumshirn return 0; 23596d97c6e3SJohannes Thumshirn } 23606d97c6e3SJohannes Thumshirn 236163443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 236263443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 236363443bf5SEric Sandeen { 236463443bf5SEric Sandeen int ret; 236563443bf5SEric Sandeen struct btrfs_root *log_tree_root; 236663443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 236763443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2368581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 236963443bf5SEric Sandeen 237063443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2371f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 237263443bf5SEric Sandeen return -EIO; 237363443bf5SEric Sandeen } 237463443bf5SEric Sandeen 237596dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 237696dfcb46SJosef Bacik GFP_KERNEL); 237763443bf5SEric Sandeen if (!log_tree_root) 237863443bf5SEric Sandeen return -ENOMEM; 237963443bf5SEric Sandeen 23802ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 23811b7ec85eSJosef Bacik BTRFS_TREE_LOG_OBJECTID, 23821b7ec85eSJosef Bacik fs_info->generation + 1, level, 23831b7ec85eSJosef Bacik NULL); 238464c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2385f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 23860eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 23878c38938cSJosef Bacik log_tree_root->node = NULL; 238800246528SJosef Bacik btrfs_put_root(log_tree_root); 23890eeff236SLiu Bo return ret; 23904eb150d6SQu Wenruo } 23914eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2392f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 239300246528SJosef Bacik btrfs_put_root(log_tree_root); 239463443bf5SEric Sandeen return -EIO; 239563443bf5SEric Sandeen } 23964eb150d6SQu Wenruo 239763443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 239863443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 239963443bf5SEric Sandeen if (ret) { 24000b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 240163443bf5SEric Sandeen "Failed to recover log tree"); 240200246528SJosef Bacik btrfs_put_root(log_tree_root); 240363443bf5SEric Sandeen return ret; 240463443bf5SEric Sandeen } 240563443bf5SEric Sandeen 2406bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 24076bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 240863443bf5SEric Sandeen if (ret) 240963443bf5SEric Sandeen return ret; 241063443bf5SEric Sandeen } 241163443bf5SEric Sandeen 241263443bf5SEric Sandeen return 0; 241363443bf5SEric Sandeen } 241463443bf5SEric Sandeen 2415abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2416abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2417abed4aaaSJosef Bacik const char *name) 2418abed4aaaSJosef Bacik { 2419abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2420abed4aaaSJosef Bacik struct btrfs_root *root; 2421f7238e50SJosef Bacik u64 max_global_id = 0; 2422abed4aaaSJosef Bacik int ret; 2423abed4aaaSJosef Bacik struct btrfs_key key = { 2424abed4aaaSJosef Bacik .objectid = objectid, 2425abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2426abed4aaaSJosef Bacik .offset = 0, 2427abed4aaaSJosef Bacik }; 2428abed4aaaSJosef Bacik bool found = false; 2429abed4aaaSJosef Bacik 2430abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2431abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2432abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2433abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2434abed4aaaSJosef Bacik return 0; 2435abed4aaaSJosef Bacik } 2436abed4aaaSJosef Bacik 2437abed4aaaSJosef Bacik while (1) { 2438abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2439abed4aaaSJosef Bacik if (ret < 0) 2440abed4aaaSJosef Bacik break; 2441abed4aaaSJosef Bacik 2442abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2443abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2444abed4aaaSJosef Bacik if (ret) { 2445abed4aaaSJosef Bacik if (ret > 0) 2446abed4aaaSJosef Bacik ret = 0; 2447abed4aaaSJosef Bacik break; 2448abed4aaaSJosef Bacik } 2449abed4aaaSJosef Bacik } 2450abed4aaaSJosef Bacik ret = 0; 2451abed4aaaSJosef Bacik 2452abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2453abed4aaaSJosef Bacik if (key.objectid != objectid) 2454abed4aaaSJosef Bacik break; 2455abed4aaaSJosef Bacik btrfs_release_path(path); 2456abed4aaaSJosef Bacik 2457f7238e50SJosef Bacik /* 2458f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2459f7238e50SJosef Bacik * everybody. 2460f7238e50SJosef Bacik */ 2461f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2462f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2463f7238e50SJosef Bacik 2464abed4aaaSJosef Bacik found = true; 2465abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2466abed4aaaSJosef Bacik if (IS_ERR(root)) { 2467abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2468abed4aaaSJosef Bacik ret = PTR_ERR(root); 2469abed4aaaSJosef Bacik break; 2470abed4aaaSJosef Bacik } 2471abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2472abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2473abed4aaaSJosef Bacik if (ret) { 2474abed4aaaSJosef Bacik btrfs_put_root(root); 2475abed4aaaSJosef Bacik break; 2476abed4aaaSJosef Bacik } 2477abed4aaaSJosef Bacik key.offset++; 2478abed4aaaSJosef Bacik } 2479abed4aaaSJosef Bacik btrfs_release_path(path); 2480abed4aaaSJosef Bacik 2481f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2482f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2483f7238e50SJosef Bacik 2484abed4aaaSJosef Bacik if (!found || ret) { 2485abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2486abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2487abed4aaaSJosef Bacik 2488abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2489abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2490abed4aaaSJosef Bacik else 2491abed4aaaSJosef Bacik ret = 0; 2492abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2493abed4aaaSJosef Bacik } 2494abed4aaaSJosef Bacik return ret; 2495abed4aaaSJosef Bacik } 2496abed4aaaSJosef Bacik 2497abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2498abed4aaaSJosef Bacik { 2499abed4aaaSJosef Bacik struct btrfs_path *path; 2500abed4aaaSJosef Bacik int ret = 0; 2501abed4aaaSJosef Bacik 2502abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2503abed4aaaSJosef Bacik if (!path) 2504abed4aaaSJosef Bacik return -ENOMEM; 2505abed4aaaSJosef Bacik 2506abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2507abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2508abed4aaaSJosef Bacik if (ret) 2509abed4aaaSJosef Bacik goto out; 2510abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2511abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2512abed4aaaSJosef Bacik if (ret) 2513abed4aaaSJosef Bacik goto out; 2514abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2515abed4aaaSJosef Bacik goto out; 2516abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2517abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2518abed4aaaSJosef Bacik "free space"); 2519abed4aaaSJosef Bacik out: 2520abed4aaaSJosef Bacik btrfs_free_path(path); 2521abed4aaaSJosef Bacik return ret; 2522abed4aaaSJosef Bacik } 2523abed4aaaSJosef Bacik 25246bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 25254bbcaa64SEric Sandeen { 25266bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2527a4f3d2c4SDavid Sterba struct btrfs_root *root; 25284bbcaa64SEric Sandeen struct btrfs_key location; 25294bbcaa64SEric Sandeen int ret; 25304bbcaa64SEric Sandeen 25316bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 25326bccf3abSJeff Mahoney 2533abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2534abed4aaaSJosef Bacik if (ret) 2535abed4aaaSJosef Bacik return ret; 2536abed4aaaSJosef Bacik 25374bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 25384bbcaa64SEric Sandeen location.offset = 0; 25394bbcaa64SEric Sandeen 25401c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 254114033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 254214033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 254314033b08SQu Wenruo if (IS_ERR(root)) { 254414033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 254514033b08SQu Wenruo ret = PTR_ERR(root); 254614033b08SQu Wenruo goto out; 254714033b08SQu Wenruo } 254814033b08SQu Wenruo } else { 254914033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 255014033b08SQu Wenruo fs_info->block_group_root = root; 255114033b08SQu Wenruo } 255214033b08SQu Wenruo } 255314033b08SQu Wenruo 255414033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2555a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2556f50f4353SLiu Bo if (IS_ERR(root)) { 255742437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2558f50f4353SLiu Bo ret = PTR_ERR(root); 2559f50f4353SLiu Bo goto out; 2560f50f4353SLiu Bo } 256142437a63SJosef Bacik } else { 2562a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2563a4f3d2c4SDavid Sterba fs_info->dev_root = root; 256442437a63SJosef Bacik } 2565820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2566a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2567a8d1b164SJohannes Thumshirn if (ret) 2568a8d1b164SJohannes Thumshirn goto out; 25694bbcaa64SEric Sandeen 2570aeb935a4SQu Wenruo /* 2571aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2572aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2573aeb935a4SQu Wenruo */ 257456e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 257556e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2576aeb935a4SQu Wenruo if (IS_ERR(root)) { 257742437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2578aeb935a4SQu Wenruo ret = PTR_ERR(root); 2579aeb935a4SQu Wenruo goto out; 2580aeb935a4SQu Wenruo } 258142437a63SJosef Bacik } else { 2582aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2583aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 258442437a63SJosef Bacik } 2585aeb935a4SQu Wenruo 25864bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2587a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2588a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2589a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2590afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2591a4f3d2c4SDavid Sterba fs_info->quota_root = root; 25924bbcaa64SEric Sandeen } 25934bbcaa64SEric Sandeen 25944bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2595a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2596a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 259742437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2598a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 25994bbcaa64SEric Sandeen if (ret != -ENOENT) 2600f50f4353SLiu Bo goto out; 260142437a63SJosef Bacik } 26024bbcaa64SEric Sandeen } else { 2603a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2604a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 26054bbcaa64SEric Sandeen } 26064bbcaa64SEric Sandeen 26074bbcaa64SEric Sandeen return 0; 2608f50f4353SLiu Bo out: 2609f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2610f50f4353SLiu Bo location.objectid, ret); 2611f50f4353SLiu Bo return ret; 26124bbcaa64SEric Sandeen } 26134bbcaa64SEric Sandeen 2614069ec957SQu Wenruo /* 2615069ec957SQu Wenruo * Real super block validation 2616069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2617069ec957SQu Wenruo * 2618069ec957SQu Wenruo * @sb: super block to check 2619069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2620069ec957SQu Wenruo * 0 the primary (1st) sb 2621069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2622069ec957SQu Wenruo * -1 skip bytenr check 2623069ec957SQu Wenruo */ 2624a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2625069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 262621a852b0SQu Wenruo { 262721a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 262821a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 262921a852b0SQu Wenruo int ret = 0; 263021a852b0SQu Wenruo 263121a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 263221a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 263321a852b0SQu Wenruo ret = -EINVAL; 263421a852b0SQu Wenruo } 263521a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 263621a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 263721a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 263821a852b0SQu Wenruo ret = -EINVAL; 263921a852b0SQu Wenruo } 264021a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 264121a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 264221a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 264321a852b0SQu Wenruo ret = -EINVAL; 264421a852b0SQu Wenruo } 264521a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 264621a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 264721a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 264821a852b0SQu Wenruo ret = -EINVAL; 264921a852b0SQu Wenruo } 265021a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 265121a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 265221a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 265321a852b0SQu Wenruo ret = -EINVAL; 265421a852b0SQu Wenruo } 265521a852b0SQu Wenruo 265621a852b0SQu Wenruo /* 265721a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 265821a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 265921a852b0SQu Wenruo */ 266021a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 266121a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 266221a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 266321a852b0SQu Wenruo ret = -EINVAL; 266421a852b0SQu Wenruo } 26650bb3eb3eSQu Wenruo 26660bb3eb3eSQu Wenruo /* 26671a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 26681a42daabSQu Wenruo * 26691a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 26701a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 26711a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 26721a42daabSQu Wenruo * beginning. 26730bb3eb3eSQu Wenruo */ 26741a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 267521a852b0SQu Wenruo btrfs_err(fs_info, 26760bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 267721a852b0SQu Wenruo sectorsize, PAGE_SIZE); 267821a852b0SQu Wenruo ret = -EINVAL; 267921a852b0SQu Wenruo } 26800bb3eb3eSQu Wenruo 268121a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 268221a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 268321a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 268421a852b0SQu Wenruo ret = -EINVAL; 268521a852b0SQu Wenruo } 268621a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 268721a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 268821a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 268921a852b0SQu Wenruo ret = -EINVAL; 269021a852b0SQu Wenruo } 269121a852b0SQu Wenruo 269221a852b0SQu Wenruo /* Root alignment check */ 269321a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 269421a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 269521a852b0SQu Wenruo btrfs_super_root(sb)); 269621a852b0SQu Wenruo ret = -EINVAL; 269721a852b0SQu Wenruo } 269821a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 269921a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 270021a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 270121a852b0SQu Wenruo ret = -EINVAL; 270221a852b0SQu Wenruo } 270321a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 270421a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 270521a852b0SQu Wenruo btrfs_super_log_root(sb)); 270621a852b0SQu Wenruo ret = -EINVAL; 270721a852b0SQu Wenruo } 270821a852b0SQu Wenruo 2709aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2710aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2711aefd7f70SNikolay Borisov btrfs_err(fs_info, 2712aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2713aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2714aefd7f70SNikolay Borisov ret = -EINVAL; 2715aefd7f70SNikolay Borisov } 2716aefd7f70SNikolay Borisov 2717aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2718aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2719aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2720aefd7f70SNikolay Borisov btrfs_err(fs_info, 2721aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2722aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2723aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2724aefd7f70SNikolay Borisov ret = -EINVAL; 2725aefd7f70SNikolay Borisov } 2726aefd7f70SNikolay Borisov 27271c56ab99SQu Wenruo /* 27281c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 27291c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 27301c56ab99SQu Wenruo */ 27311c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 27321c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 27331c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 27341c56ab99SQu Wenruo btrfs_err(fs_info, 27351c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 27361c56ab99SQu Wenruo ret = -EINVAL; 27371c56ab99SQu Wenruo } 27381c56ab99SQu Wenruo 2739de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 27407239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 274121a852b0SQu Wenruo btrfs_err(fs_info, 27427239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2743de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 274421a852b0SQu Wenruo ret = -EINVAL; 274521a852b0SQu Wenruo } 274621a852b0SQu Wenruo 274721a852b0SQu Wenruo /* 274821a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 274921a852b0SQu Wenruo * done later 275021a852b0SQu Wenruo */ 275121a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 275221a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 275321a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 275421a852b0SQu Wenruo ret = -EINVAL; 275521a852b0SQu Wenruo } 275621a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 275721a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 275821a852b0SQu Wenruo btrfs_super_stripesize(sb)); 275921a852b0SQu Wenruo ret = -EINVAL; 276021a852b0SQu Wenruo } 276121a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 276221a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 276321a852b0SQu Wenruo btrfs_super_num_devices(sb)); 276421a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 276521a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 276621a852b0SQu Wenruo ret = -EINVAL; 276721a852b0SQu Wenruo } 276821a852b0SQu Wenruo 2769069ec957SQu Wenruo if (mirror_num >= 0 && 2770069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 277121a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 277221a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 277321a852b0SQu Wenruo ret = -EINVAL; 277421a852b0SQu Wenruo } 277521a852b0SQu Wenruo 277621a852b0SQu Wenruo /* 277721a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 277821a852b0SQu Wenruo * and one chunk 277921a852b0SQu Wenruo */ 278021a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 278121a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 278221a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 278321a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 278421a852b0SQu Wenruo ret = -EINVAL; 278521a852b0SQu Wenruo } 278621a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 278721a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 278821a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 278921a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 279021a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 279121a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 279221a852b0SQu Wenruo ret = -EINVAL; 279321a852b0SQu Wenruo } 279421a852b0SQu Wenruo 279521a852b0SQu Wenruo /* 279621a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 279721a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 279821a852b0SQu Wenruo */ 279921a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 280021a852b0SQu Wenruo btrfs_warn(fs_info, 280121a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 280221a852b0SQu Wenruo btrfs_super_generation(sb), 280321a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 280421a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 280521a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 280621a852b0SQu Wenruo btrfs_warn(fs_info, 280721a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 280821a852b0SQu Wenruo btrfs_super_generation(sb), 280921a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 281021a852b0SQu Wenruo 281121a852b0SQu Wenruo return ret; 281221a852b0SQu Wenruo } 281321a852b0SQu Wenruo 2814069ec957SQu Wenruo /* 2815069ec957SQu Wenruo * Validation of super block at mount time. 2816069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2817069ec957SQu Wenruo * flags will be skipped. 2818069ec957SQu Wenruo */ 2819069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2820069ec957SQu Wenruo { 2821a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2822069ec957SQu Wenruo } 2823069ec957SQu Wenruo 282475cb857dSQu Wenruo /* 282575cb857dSQu Wenruo * Validation of super block at write time. 282675cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 282775cb857dSQu Wenruo * overwritten soon. 282875cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 282975cb857dSQu Wenruo */ 283075cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 283175cb857dSQu Wenruo struct btrfs_super_block *sb) 283275cb857dSQu Wenruo { 283375cb857dSQu Wenruo int ret; 283475cb857dSQu Wenruo 2835a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 283675cb857dSQu Wenruo if (ret < 0) 283775cb857dSQu Wenruo goto out; 2838e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 283975cb857dSQu Wenruo ret = -EUCLEAN; 284075cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 284175cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 284275cb857dSQu Wenruo goto out; 284375cb857dSQu Wenruo } 284475cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 284575cb857dSQu Wenruo ret = -EUCLEAN; 284675cb857dSQu Wenruo btrfs_err(fs_info, 284775cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 284875cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 284975cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 285075cb857dSQu Wenruo goto out; 285175cb857dSQu Wenruo } 285275cb857dSQu Wenruo out: 285375cb857dSQu Wenruo if (ret < 0) 285475cb857dSQu Wenruo btrfs_err(fs_info, 285575cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 285675cb857dSQu Wenruo return ret; 285775cb857dSQu Wenruo } 285875cb857dSQu Wenruo 2859bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2860bd676446SJosef Bacik { 2861bd676446SJosef Bacik int ret = 0; 2862bd676446SJosef Bacik 2863bd676446SJosef Bacik root->node = read_tree_block(root->fs_info, bytenr, 2864bd676446SJosef Bacik root->root_key.objectid, gen, level, NULL); 2865bd676446SJosef Bacik if (IS_ERR(root->node)) { 2866bd676446SJosef Bacik ret = PTR_ERR(root->node); 2867bd676446SJosef Bacik root->node = NULL; 28684eb150d6SQu Wenruo return ret; 28694eb150d6SQu Wenruo } 28704eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2871bd676446SJosef Bacik free_extent_buffer(root->node); 2872bd676446SJosef Bacik root->node = NULL; 28734eb150d6SQu Wenruo return -EIO; 2874bd676446SJosef Bacik } 2875bd676446SJosef Bacik 2876bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2877bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2878bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2879bd676446SJosef Bacik return ret; 2880bd676446SJosef Bacik } 2881bd676446SJosef Bacik 2882bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2883bd676446SJosef Bacik { 2884bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2885bd676446SJosef Bacik u64 gen, bytenr; 2886bd676446SJosef Bacik int level, ret; 2887bd676446SJosef Bacik 2888bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2889bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2890bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2891bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 28929c54e80dSJosef Bacik if (ret) { 2893bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2894bd676446SJosef Bacik return ret; 2895bd676446SJosef Bacik } 28969c54e80dSJosef Bacik return 0; 28979c54e80dSJosef Bacik } 28989c54e80dSJosef Bacik 28996ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2900b8522a1eSNikolay Borisov { 29016ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2902b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2903b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2904b8522a1eSNikolay Borisov bool handle_error = false; 2905b8522a1eSNikolay Borisov int ret = 0; 2906b8522a1eSNikolay Borisov int i; 2907b8522a1eSNikolay Borisov 2908b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2909b8522a1eSNikolay Borisov if (handle_error) { 2910b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2911b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2912b8522a1eSNikolay Borisov tree_root->node = NULL; 2913b8522a1eSNikolay Borisov 2914b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2915b8522a1eSNikolay Borisov break; 2916b8522a1eSNikolay Borisov 2917b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2918b8522a1eSNikolay Borisov 2919b8522a1eSNikolay Borisov /* 2920b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2921b8522a1eSNikolay Borisov * valid 2922b8522a1eSNikolay Borisov */ 2923b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2924b8522a1eSNikolay Borisov 2925b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2926b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2927b8522a1eSNikolay Borisov 2928b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 29296ef108ddSNikolay Borisov backup_index = ret; 2930b8522a1eSNikolay Borisov if (ret < 0) 2931b8522a1eSNikolay Borisov return ret; 2932b8522a1eSNikolay Borisov } 2933b8522a1eSNikolay Borisov 2934bd676446SJosef Bacik ret = load_important_roots(fs_info); 2935bd676446SJosef Bacik if (ret) { 2936217f5004SNikolay Borisov handle_error = true; 2937b8522a1eSNikolay Borisov continue; 2938b8522a1eSNikolay Borisov } 2939b8522a1eSNikolay Borisov 2940336a0d8dSNikolay Borisov /* 2941336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2942336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2943336a0d8dSNikolay Borisov */ 2944453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2945b8522a1eSNikolay Borisov if (ret < 0) { 2946b8522a1eSNikolay Borisov handle_error = true; 2947b8522a1eSNikolay Borisov continue; 2948b8522a1eSNikolay Borisov } 2949b8522a1eSNikolay Borisov 29506b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2951b8522a1eSNikolay Borisov 2952b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2953b8522a1eSNikolay Borisov if (ret < 0) { 2954b8522a1eSNikolay Borisov handle_error = true; 2955b8522a1eSNikolay Borisov continue; 2956b8522a1eSNikolay Borisov } 2957b8522a1eSNikolay Borisov 2958b8522a1eSNikolay Borisov /* All successful */ 2959bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2960bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2961d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 29626ef108ddSNikolay Borisov 29636ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 29646ef108ddSNikolay Borisov if (backup_index < 0) { 29656ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 29666ef108ddSNikolay Borisov } else { 29676ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 29686ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 29696ef108ddSNikolay Borisov } 2970b8522a1eSNikolay Borisov break; 2971b8522a1eSNikolay Borisov } 2972b8522a1eSNikolay Borisov 2973b8522a1eSNikolay Borisov return ret; 2974b8522a1eSNikolay Borisov } 2975b8522a1eSNikolay Borisov 29768260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2977eb60ceacSChris Mason { 2978fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 297901cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 29808fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2981facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 298224bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2983eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 298411833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2985eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2986a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2987fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 298824bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 29894cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2990ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2991f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 299247ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 299340ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2994afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2995c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2996f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2997abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2998d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2999f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 30007585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 3001573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 30020bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 30035f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 3004de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 300519c00ddcSChris Mason 3006e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 30075a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 30088b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 30095f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 30103e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 30113e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 30123e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 30133e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 30143e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 30153e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 30163e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 30173e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 3018e1489b4fSIoannis Angelakopoulos 30190b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 30206324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 3021f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 302247ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 302318bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 3024afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 3025bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 3026bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 30273fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 30283fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 3029bd647ce3SJosef Bacik #endif 3030c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 303166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 303266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 303366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 303466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 303566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 303666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 303766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 3038ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 3039ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 3040ba2c4d4eSJosef Bacik 3041771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 30424cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 3043034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 3044fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 3045abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 304695ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 30479ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 30484cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 3049a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 3050f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 30518b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 3052f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 3053fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 3054c8b97818SChris Mason 3055b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 3056b34b086cSChris Mason num_online_cpus() + 2, 8); 30570afbaf8cSChris Mason 3058199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 3059199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 306069fe2d75SJosef Bacik 3061638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 306221adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 306321adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 306421adbd5cSStefan Behrens #endif 3065779a65a4SEric Sandeen btrfs_init_balance(fs_info); 306657056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 3067a2de733cSArne Jansen 306816b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 306908dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 30700f9dd46cSJosef Bacik 3071fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 3072fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 307339279cc3SChris Mason 30745a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 3075e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 3076925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 3077a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 3078a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 30791bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 30809e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 3081c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 308276dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 3083803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 3084fa9c0d79SChris Mason 3085ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 3086f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 3087b0643e59SDennis Zhou btrfs_discard_init(fs_info); 3088416ac51dSArne Jansen 3089fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 3090fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 3091fa9c0d79SChris Mason 3092e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 3093f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 3094bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 30954854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 3096034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 30973768f368SChris Mason 3098da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 3099da17066cSJeff Mahoney fs_info->nodesize = 4096; 3100da17066cSJeff Mahoney fs_info->sectorsize = 4096; 3101ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 3102da17066cSJeff Mahoney fs_info->stripesize = 4096; 3103da17066cSJeff Mahoney 3104f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 3105f7b12a62SNaohiro Aota 3106eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 3107eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 3108eede2bf3SOmar Sandoval 310918bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 311018bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 31118260edbaSJosef Bacik } 31128260edbaSJosef Bacik 31138260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 31148260edbaSJosef Bacik { 31158260edbaSJosef Bacik int ret; 31168260edbaSJosef Bacik 31178260edbaSJosef Bacik fs_info->sb = sb; 31188260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 31198260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 31209e967495SFilipe Manana 31215deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 3122ae18c37aSJosef Bacik if (ret) 3123ae18c37aSJosef Bacik return ret; 3124ae18c37aSJosef Bacik 3125ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 3126ae18c37aSJosef Bacik if (ret) 3127c75e8394SJosef Bacik return ret; 3128ae18c37aSJosef Bacik 3129ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 3130ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 3131ae18c37aSJosef Bacik 3132ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 3133ae18c37aSJosef Bacik if (ret) 3134c75e8394SJosef Bacik return ret; 3135ae18c37aSJosef Bacik 3136ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 3137ae18c37aSJosef Bacik GFP_KERNEL); 3138ae18c37aSJosef Bacik if (ret) 3139c75e8394SJosef Bacik return ret; 3140ae18c37aSJosef Bacik 3141ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 3142ae18c37aSJosef Bacik GFP_KERNEL); 3143c75e8394SJosef Bacik if (!fs_info->delayed_root) 3144c75e8394SJosef Bacik return -ENOMEM; 3145ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 3146ae18c37aSJosef Bacik 3147a0a1db70SFilipe Manana if (sb_rdonly(sb)) 3148a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 3149a0a1db70SFilipe Manana 3150c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 3151ae18c37aSJosef Bacik } 3152ae18c37aSJosef Bacik 315397f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 315497f4dd09SNikolay Borisov { 31550d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 315697f4dd09SNikolay Borisov int ret; 315797f4dd09SNikolay Borisov 315897f4dd09SNikolay Borisov /* 315997f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 316097f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 316197f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 316297f4dd09SNikolay Borisov */ 316397f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 316497f4dd09SNikolay Borisov if (ret < 0) { 3165c94bec2cSJosef Bacik if (ret != -EINTR) 3166c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 3167c94bec2cSJosef Bacik ret); 316897f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 316997f4dd09SNikolay Borisov return ret; 317097f4dd09SNikolay Borisov } 317197f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 317297f4dd09SNikolay Borisov } 317397f4dd09SNikolay Borisov 317497f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 317597f4dd09SNikolay Borisov { 317697f4dd09SNikolay Borisov struct task_struct *task; 317797f4dd09SNikolay Borisov 317897f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 317997f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 318097f4dd09SNikolay Borisov if (IS_ERR(task)) { 318197f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 318297f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 318397f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 318497f4dd09SNikolay Borisov return PTR_ERR(task); 318597f4dd09SNikolay Borisov } 318697f4dd09SNikolay Borisov 318797f4dd09SNikolay Borisov return 0; 318897f4dd09SNikolay Borisov } 318997f4dd09SNikolay Borisov 319044c0ca21SBoris Burkov /* 31918cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 31928cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 31938cd29088SBoris Burkov * code paths. 31948cd29088SBoris Burkov */ 31958cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 31968cd29088SBoris Burkov { 31978cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 31988b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 31998cd29088SBoris Burkov } 32008cd29088SBoris Burkov 32018cd29088SBoris Burkov /* 320244c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 320344c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 320444c0ca21SBoris Burkov */ 320544c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 320644c0ca21SBoris Burkov { 320744c0ca21SBoris Burkov int ret; 320894846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 32098b228324SBoris Burkov bool clear_free_space_tree = false; 32108b228324SBoris Burkov 32118b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 32128b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 32138b228324SBoris Burkov clear_free_space_tree = true; 32148b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 32158b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 32168b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 32178b228324SBoris Burkov clear_free_space_tree = true; 32188b228324SBoris Burkov } 32198b228324SBoris Burkov 32208b228324SBoris Burkov if (clear_free_space_tree) { 32218b228324SBoris Burkov btrfs_info(fs_info, "clearing free space tree"); 32228b228324SBoris Burkov ret = btrfs_clear_free_space_tree(fs_info); 32238b228324SBoris Burkov if (ret) { 32248b228324SBoris Burkov btrfs_warn(fs_info, 32258b228324SBoris Burkov "failed to clear free space tree: %d", ret); 32268b228324SBoris Burkov goto out; 32278b228324SBoris Burkov } 32288b228324SBoris Burkov } 322944c0ca21SBoris Burkov 32308d488a8cSFilipe Manana /* 32318d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 32328d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 3233fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 32348d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 32358d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 32368d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 32378d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 32388d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 32398d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 32408d488a8cSFilipe Manana */ 32418d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 32428d488a8cSFilipe Manana if (ret) 32438d488a8cSFilipe Manana goto out; 32448d488a8cSFilipe Manana 324544c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 324644c0ca21SBoris Burkov if (ret) 324744c0ca21SBoris Burkov goto out; 324844c0ca21SBoris Burkov 32498f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 32508f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 32518f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 32528f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 32538f1c21d7SBoris Burkov goto out; 32548f1c21d7SBoris Burkov } 32558f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 32568f1c21d7SBoris Burkov 325744c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 32587eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 325944c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 326044c0ca21SBoris Burkov if (ret < 0) { 326144c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 326244c0ca21SBoris Burkov goto out; 326344c0ca21SBoris Burkov } 326444c0ca21SBoris Burkov 32655011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 32665011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 32675011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 32685011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 32695011139aSBoris Burkov if (ret) { 32705011139aSBoris Burkov btrfs_warn(fs_info, 32715011139aSBoris Burkov "failed to create free space tree: %d", ret); 32725011139aSBoris Burkov goto out; 32735011139aSBoris Burkov } 32745011139aSBoris Burkov } 32755011139aSBoris Burkov 327694846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 327794846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 327894846229SBoris Burkov if (ret) 327994846229SBoris Burkov goto out; 328094846229SBoris Burkov } 328194846229SBoris Burkov 328244c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 328344c0ca21SBoris Burkov if (ret) 328444c0ca21SBoris Burkov goto out; 328544c0ca21SBoris Burkov 328644c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 328744c0ca21SBoris Burkov if (ret) { 328844c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 328944c0ca21SBoris Burkov goto out; 329044c0ca21SBoris Burkov } 329144c0ca21SBoris Burkov 329244c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 329344c0ca21SBoris Burkov 329444c0ca21SBoris Burkov if (!fs_info->uuid_root) { 329544c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 329644c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 329744c0ca21SBoris Burkov if (ret) { 329844c0ca21SBoris Burkov btrfs_warn(fs_info, 329944c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 330044c0ca21SBoris Burkov goto out; 330144c0ca21SBoris Burkov } 330244c0ca21SBoris Burkov } 330344c0ca21SBoris Burkov 330444c0ca21SBoris Burkov out: 330544c0ca21SBoris Burkov return ret; 330644c0ca21SBoris Burkov } 330744c0ca21SBoris Burkov 3308d7f67ac9SQu Wenruo /* 3309d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3310d7f67ac9SQu Wenruo * 3311d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3312d7f67ac9SQu Wenruo * feature dependencies). 3313d7f67ac9SQu Wenruo * 3314d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3315d7f67ac9SQu Wenruo * btrfs_validate_super(). 3316d7f67ac9SQu Wenruo * 3317d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3318d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3319d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3320d7f67ac9SQu Wenruo */ 3321d7f67ac9SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, struct super_block *sb) 3322d7f67ac9SQu Wenruo { 3323d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3324d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3325d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3326d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3327d7f67ac9SQu Wenruo 3328d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3329d7f67ac9SQu Wenruo btrfs_err(fs_info, 3330d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3331d7f67ac9SQu Wenruo incompat); 3332d7f67ac9SQu Wenruo return -EINVAL; 3333d7f67ac9SQu Wenruo } 3334d7f67ac9SQu Wenruo 3335d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3336d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3337d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3338d7f67ac9SQu Wenruo btrfs_err(fs_info, 3339d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3340d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3341d7f67ac9SQu Wenruo return -EINVAL; 3342d7f67ac9SQu Wenruo } 3343d7f67ac9SQu Wenruo 3344d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3345d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3346d7f67ac9SQu Wenruo 3347d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3348d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3349d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3350d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3351d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3352d7f67ac9SQu Wenruo 3353d7f67ac9SQu Wenruo /* 3354d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3355d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3356d7f67ac9SQu Wenruo */ 3357d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3358d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3359d7f67ac9SQu Wenruo 3360d7f67ac9SQu Wenruo if (compat_ro_unsupp && !sb_rdonly(sb)) { 3361d7f67ac9SQu Wenruo btrfs_err(fs_info, 3362d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3363d7f67ac9SQu Wenruo compat_ro); 3364d7f67ac9SQu Wenruo return -EINVAL; 3365d7f67ac9SQu Wenruo } 3366d7f67ac9SQu Wenruo 3367d7f67ac9SQu Wenruo /* 3368d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3369d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3370d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3371d7f67ac9SQu Wenruo */ 3372d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3373d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3374d7f67ac9SQu Wenruo btrfs_err(fs_info, 3375d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3376d7f67ac9SQu Wenruo compat_ro); 3377d7f67ac9SQu Wenruo return -EINVAL; 3378d7f67ac9SQu Wenruo } 3379d7f67ac9SQu Wenruo 3380d7f67ac9SQu Wenruo /* 3381d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3382d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3383d7f67ac9SQu Wenruo */ 3384d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3385d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3386d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3387d7f67ac9SQu Wenruo btrfs_err(fs_info, 3388d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3389d7f67ac9SQu Wenruo return -EINVAL; 3390d7f67ac9SQu Wenruo } 3391d7f67ac9SQu Wenruo 3392d7f67ac9SQu Wenruo /* 3393d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3394d7f67ac9SQu Wenruo * 3395d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3396d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3397d7f67ac9SQu Wenruo * going to be deprecated anyway. 3398d7f67ac9SQu Wenruo */ 3399d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3400d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3401d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3402d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3403d7f67ac9SQu Wenruo return -EINVAL; 3404d7f67ac9SQu Wenruo } 3405d7f67ac9SQu Wenruo 3406d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3407d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3408d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3409d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3410d7f67ac9SQu Wenruo 3411d7f67ac9SQu Wenruo return 0; 3412d7f67ac9SQu Wenruo } 3413d7f67ac9SQu Wenruo 3414ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3415ae18c37aSJosef Bacik char *options) 3416ae18c37aSJosef Bacik { 3417ae18c37aSJosef Bacik u32 sectorsize; 3418ae18c37aSJosef Bacik u32 nodesize; 3419ae18c37aSJosef Bacik u32 stripesize; 3420ae18c37aSJosef Bacik u64 generation; 3421ae18c37aSJosef Bacik u64 features; 3422ae18c37aSJosef Bacik u16 csum_type; 3423ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3424ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3425ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3426ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3427ae18c37aSJosef Bacik int ret; 3428ae18c37aSJosef Bacik int err = -EINVAL; 3429ae18c37aSJosef Bacik int level; 3430ae18c37aSJosef Bacik 34318260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 343253b381b3SDavid Woodhouse if (ret) { 343383c8266aSDavid Sterba err = ret; 3434ae18c37aSJosef Bacik goto fail; 343553b381b3SDavid Woodhouse } 343653b381b3SDavid Woodhouse 3437ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3438ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3439ae18c37aSJosef Bacik GFP_KERNEL); 3440ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3441ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3442ae18c37aSJosef Bacik GFP_KERNEL); 3443ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3444ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 3445ae18c37aSJosef Bacik err = -ENOMEM; 3446c75e8394SJosef Bacik goto fail; 3447ae18c37aSJosef Bacik } 3448ae18c37aSJosef Bacik 3449ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 3450ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 3451ae18c37aSJosef Bacik err = -ENOMEM; 3452c75e8394SJosef Bacik goto fail; 3453ae18c37aSJosef Bacik } 3454ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 3455ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 3456ae18c37aSJosef Bacik 3457d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 34581104a885SDavid Sterba 34591104a885SDavid Sterba /* 34601104a885SDavid Sterba * Read super block and check the signature bytes only 34611104a885SDavid Sterba */ 3462d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 34638f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 34648f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 346516cdcec7SMiao Xie goto fail_alloc; 346620b45077SDave Young } 346739279cc3SChris Mason 34681104a885SDavid Sterba /* 3469260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 34708dc3f22cSJohannes Thumshirn * corrupted, we'll find out 34718dc3f22cSJohannes Thumshirn */ 34728f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 347351bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 34748dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 347551bce6c9SJohannes Thumshirn csum_type); 34768dc3f22cSJohannes Thumshirn err = -EINVAL; 34778f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 34788dc3f22cSJohannes Thumshirn goto fail_alloc; 34798dc3f22cSJohannes Thumshirn } 34808dc3f22cSJohannes Thumshirn 348183c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 348283c68bbcSSu Yue 34836d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 34846d97c6e3SJohannes Thumshirn if (ret) { 34856d97c6e3SJohannes Thumshirn err = ret; 34868f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 34876d97c6e3SJohannes Thumshirn goto fail_alloc; 34886d97c6e3SJohannes Thumshirn } 34896d97c6e3SJohannes Thumshirn 34908dc3f22cSJohannes Thumshirn /* 34911104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 34921104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 34931104a885SDavid Sterba */ 34943d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 349505135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 34961104a885SDavid Sterba err = -EINVAL; 34978f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3498141386e1SJosef Bacik goto fail_alloc; 34991104a885SDavid Sterba } 35001104a885SDavid Sterba 35011104a885SDavid Sterba /* 35021104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 35031104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 35041104a885SDavid Sterba * the whole block of INFO_SIZE 35051104a885SDavid Sterba */ 35068f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 35078f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35085f39d397SChris Mason 3509fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3510fbc6feaeSNikolay Borisov 35110b86a832SChris Mason 3512fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3513fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3514fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3515fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3516fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3517fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3518fbc6feaeSNikolay Borisov } 3519fbc6feaeSNikolay Borisov 3520fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3521fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3522de37aa51SNikolay Borisov 3523069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 35241104a885SDavid Sterba if (ret) { 352505135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 35261104a885SDavid Sterba err = -EINVAL; 3527141386e1SJosef Bacik goto fail_alloc; 35281104a885SDavid Sterba } 35291104a885SDavid Sterba 35300f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3531141386e1SJosef Bacik goto fail_alloc; 35320f7d52f4SChris Mason 3533acce952bSliubo /* check FS state, whether FS is broken. */ 353487533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 353587533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3536acce952bSliubo 353775e7cb7fSLiu Bo /* 353875e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 353975e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 354075e7cb7fSLiu Bo */ 354175e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 354275e7cb7fSLiu Bo 35436f93e834SAnand Jain 35446f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 35456f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 35466f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 35476f93e834SAnand Jain stripesize = sectorsize; 35486f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 35496f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 35506f93e834SAnand Jain 35516f93e834SAnand Jain fs_info->nodesize = nodesize; 35526f93e834SAnand Jain fs_info->sectorsize = sectorsize; 35536f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 35546f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 35556f93e834SAnand Jain fs_info->stripesize = stripesize; 35566f93e834SAnand Jain 35572ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 35582b82032cSYan Zheng if (ret) { 35592b82032cSYan Zheng err = ret; 3560141386e1SJosef Bacik goto fail_alloc; 35612b82032cSYan Zheng } 3562dfe25020SChris Mason 3563d7f67ac9SQu Wenruo ret = btrfs_check_features(fs_info, sb); 3564d7f67ac9SQu Wenruo if (ret < 0) { 3565d7f67ac9SQu Wenruo err = ret; 3566141386e1SJosef Bacik goto fail_alloc; 3567f2b636e8SJosef Bacik } 3568f2b636e8SJosef Bacik 35698481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 35708481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 35718481dd80SQu Wenruo 35729f73f1aeSQu Wenruo /* 35739f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 35749f73f1aeSQu Wenruo * going to be deprecated. 35759f73f1aeSQu Wenruo * 35769f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 35779f73f1aeSQu Wenruo */ 35789f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 35799f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 35809f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 35819f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 35829f73f1aeSQu Wenruo 358395ea0486SQu Wenruo btrfs_warn(fs_info, 358495ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 35850bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 35868481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 35878481dd80SQu Wenruo if (!subpage_info) 35888481dd80SQu Wenruo goto fail_alloc; 35898481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 35908481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3591c8050b3bSQu Wenruo } 35920bb3eb3eSQu Wenruo 3593d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 35942a458198SEric Sandeen if (ret) { 35952a458198SEric Sandeen err = ret; 35960dc3b84aSJosef Bacik goto fail_sb_buffer; 35970dc3b84aSJosef Bacik } 35984543df7eSChris Mason 35999e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 36009e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 36014575c9ccSChris Mason 3602a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3603a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3604de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3605db94535dSChris Mason 3606925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 36076bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3608925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 360984eed90fSChris Mason if (ret) { 361005135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 36115d4f98a2SYan Zheng goto fail_sb_buffer; 361284eed90fSChris Mason } 36130b86a832SChris Mason 361484234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3615581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3616bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3617bd676446SJosef Bacik generation, level); 3618bd676446SJosef Bacik if (ret) { 361905135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3620af31f5e5SChris Mason goto fail_tree_roots; 362183121942SDavid Woodhouse } 36220b86a832SChris Mason 3623e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3624c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3625c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3626e17cade2SChris Mason 36275b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 36282b82032cSYan Zheng if (ret) { 362905135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3630af31f5e5SChris Mason goto fail_tree_roots; 36312b82032cSYan Zheng } 36320b86a832SChris Mason 36338dabb742SStefan Behrens /* 3634bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3635bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3636bacce86aSAnand Jain * target is required. So free devices that are not part of this 36371a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3638bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 36398dabb742SStefan Behrens */ 3640bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3641d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 364205135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3643a6b0d5c8SChris Mason goto fail_tree_roots; 3644a6b0d5c8SChris Mason } 3645a6b0d5c8SChris Mason 3646b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 36474bbcaa64SEric Sandeen if (ret) 3648b8522a1eSNikolay Borisov goto fail_tree_roots; 36498929ecfaSYan, Zheng 365075ec1db8SJosef Bacik /* 365173651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 365273651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 365373651042SNaohiro Aota * zoned incompat feature flag set. 365473651042SNaohiro Aota */ 365573651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 365673651042SNaohiro Aota if (ret) { 365773651042SNaohiro Aota btrfs_err(fs_info, 365873651042SNaohiro Aota "zoned: failed to read device zone info: %d", 365973651042SNaohiro Aota ret); 366073651042SNaohiro Aota goto fail_block_groups; 366173651042SNaohiro Aota } 366273651042SNaohiro Aota 366373651042SNaohiro Aota /* 366475ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 366575ec1db8SJosef Bacik * check the generation here so we can set the 366675ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 366775ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 366875ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 366975ec1db8SJosef Bacik * even though it was perfectly fine. 367075ec1db8SJosef Bacik */ 367175ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 367275ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 367375ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 367475ec1db8SJosef Bacik 3675cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3676cf90d884SQu Wenruo if (ret) { 3677cf90d884SQu Wenruo btrfs_err(fs_info, 3678cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3679cf90d884SQu Wenruo ret); 3680cf90d884SQu Wenruo goto fail_block_groups; 3681cf90d884SQu Wenruo } 368268310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 368368310a5eSIlya Dryomov if (ret) { 368405135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 368568310a5eSIlya Dryomov goto fail_block_groups; 368668310a5eSIlya Dryomov } 368768310a5eSIlya Dryomov 3688733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3689733f4fbbSStefan Behrens if (ret) { 369005135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3691733f4fbbSStefan Behrens goto fail_block_groups; 3692733f4fbbSStefan Behrens } 3693733f4fbbSStefan Behrens 36948dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 36958dabb742SStefan Behrens if (ret) { 369605135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 36978dabb742SStefan Behrens goto fail_block_groups; 36988dabb742SStefan Behrens } 36998dabb742SStefan Behrens 3700b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3701b70f5097SNaohiro Aota if (ret) { 3702b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3703b70f5097SNaohiro Aota ret); 3704b70f5097SNaohiro Aota goto fail_block_groups; 3705b70f5097SNaohiro Aota } 3706b70f5097SNaohiro Aota 3707c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3708b7c35e81SAnand Jain if (ret) { 370905135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 371005135f59SDavid Sterba ret); 3711b7c35e81SAnand Jain goto fail_block_groups; 3712b7c35e81SAnand Jain } 3713b7c35e81SAnand Jain 371496f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 37155ac1d209SJeff Mahoney if (ret) { 371605135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3717b7c35e81SAnand Jain goto fail_fsdev_sysfs; 37185ac1d209SJeff Mahoney } 37195ac1d209SJeff Mahoney 3720c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3721c59021f8Sliubo if (ret) { 372205135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 37232365dd3cSAnand Jain goto fail_sysfs; 3724c59021f8Sliubo } 3725c59021f8Sliubo 37265b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 37271b1d1f66SJosef Bacik if (ret) { 372805135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 37292365dd3cSAnand Jain goto fail_sysfs; 37301b1d1f66SJosef Bacik } 37314330e183SQu Wenruo 373216beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 373316beac87SNaohiro Aota 37345c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 37355c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 373605135f59SDavid Sterba btrfs_warn(fs_info, 373752042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 37382365dd3cSAnand Jain goto fail_sysfs; 3739292fd7fcSStefan Behrens } 37409078a3e1SChris Mason 374133c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3742a74a4b97SChris Mason "btrfs-cleaner"); 374357506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 37442365dd3cSAnand Jain goto fail_sysfs; 3745a74a4b97SChris Mason 3746a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3747a74a4b97SChris Mason tree_root, 3748a74a4b97SChris Mason "btrfs-transaction"); 374957506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 37503f157a2fSChris Mason goto fail_cleaner; 3751a74a4b97SChris Mason 3752583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3753c289811cSChris Mason !fs_info->fs_devices->rotating) { 3754583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3755c289811cSChris Mason } 3756c289811cSChris Mason 3757572d9ab7SDavid Sterba /* 375863a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 375963a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 376063a7cb13SDavid Sterba * nodiscard for the same mount. 376163a7cb13SDavid Sterba */ 376263a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 376363a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 376463a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 376563a7cb13SDavid Sterba fs_info->fs_devices->discardable) { 376663a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 376763a7cb13SDavid Sterba "auto enabling async discard"); 376863a7cb13SDavid Sterba btrfs_clear_opt(fs_info->mount_opt, NODISCARD); 376963a7cb13SDavid Sterba } 377063a7cb13SDavid Sterba 377121adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 37720b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 37732ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 37740b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3775cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 377621adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 377721adbd5cSStefan Behrens if (ret) 377805135f59SDavid Sterba btrfs_warn(fs_info, 377905135f59SDavid Sterba "failed to initialize integrity check module: %d", 378005135f59SDavid Sterba ret); 378121adbd5cSStefan Behrens } 378221adbd5cSStefan Behrens #endif 3783bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3784bcef60f2SArne Jansen if (ret) 3785bcef60f2SArne Jansen goto fail_trans_kthread; 378621adbd5cSStefan Behrens 3787fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3788fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3789fd708b81SJosef Bacik 379096da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 379196da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 37920b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3793e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 379463443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 379579787eaaSJeff Mahoney if (ret) { 379663443bf5SEric Sandeen err = ret; 379728c16cbbSWang Shilong goto fail_qgroup; 3798e556ce2cSYan Zheng } 3799e02119d5SChris Mason } 38001a40e23bSZheng Yan 380156e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 38023140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 38033140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3804f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3805315bf8efSJosef Bacik fs_info->fs_root = NULL; 3806bcef60f2SArne Jansen goto fail_qgroup; 38073140c9a3SDan Carpenter } 3808c289811cSChris Mason 3809bc98a42cSDavid Howells if (sb_rdonly(sb)) 38108cd29088SBoris Burkov goto clear_oneshot; 38112b6ba629SIlya Dryomov 381244c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 38132b6ba629SIlya Dryomov if (ret) { 38146bccf3abSJeff Mahoney close_ctree(fs_info); 38152b6ba629SIlya Dryomov return ret; 3816e3acc2a6SJosef Bacik } 3817b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3818b382a324SJan Schmidt 381944c0ca21SBoris Burkov if (fs_info->uuid_root && 382044c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 382144c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 382205135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 382370f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 382470f80175SStefan Behrens if (ret) { 382505135f59SDavid Sterba btrfs_warn(fs_info, 382605135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 38276bccf3abSJeff Mahoney close_ctree(fs_info); 382870f80175SStefan Behrens return ret; 382970f80175SStefan Behrens } 3830f7a81ea4SStefan Behrens } 383194846229SBoris Burkov 3832afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 383347ab2a6cSJosef Bacik 3834b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3835b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3836b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3837b4be6aefSJosef Bacik 38388cd29088SBoris Burkov clear_oneshot: 38398cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3840ad2b2c80SAl Viro return 0; 384139279cc3SChris Mason 3842bcef60f2SArne Jansen fail_qgroup: 3843bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 38447c2ca468SChris Mason fail_trans_kthread: 38457c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 38462ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3847faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 38483f157a2fSChris Mason fail_cleaner: 3849a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 38507c2ca468SChris Mason 38517c2ca468SChris Mason /* 38527c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 38537c2ca468SChris Mason * kthreads 38547c2ca468SChris Mason */ 38557c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 38567c2ca468SChris Mason 38572365dd3cSAnand Jain fail_sysfs: 38586618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 38592365dd3cSAnand Jain 3860b7c35e81SAnand Jain fail_fsdev_sysfs: 3861b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3862b7c35e81SAnand Jain 38631b1d1f66SJosef Bacik fail_block_groups: 386454067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3865af31f5e5SChris Mason 3866af31f5e5SChris Mason fail_tree_roots: 38679e3aa805SJosef Bacik if (fs_info->data_reloc_root) 38689e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 38694273eaffSAnand Jain free_root_pointers(fs_info, true); 38702b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3871af31f5e5SChris Mason 387239279cc3SChris Mason fail_sb_buffer: 38737abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 38745cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 387516cdcec7SMiao Xie fail_alloc: 3876586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3877586e46e2SIlya Dryomov 38784543df7eSChris Mason iput(fs_info->btree_inode); 38797e662854SQinghuang Feng fail: 3880586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3881ad2b2c80SAl Viro return err; 3882eb60ceacSChris Mason } 3883663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3884eb60ceacSChris Mason 3885314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3886f2984462SChris Mason { 3887314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3888314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3889314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3890314b6dd0SJohannes Thumshirn struct page *page; 3891442a4f63SStefan Behrens 3892314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3893314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3894314b6dd0SJohannes Thumshirn 3895314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3896fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3897314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3898314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3899314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3900314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3901314b6dd0SJohannes Thumshirn SetPageError(page); 3902314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3903314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3904314b6dd0SJohannes Thumshirn } else { 3905314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3906f2984462SChris Mason } 3907314b6dd0SJohannes Thumshirn 3908314b6dd0SJohannes Thumshirn put_page(page); 3909314b6dd0SJohannes Thumshirn unlock_page(page); 3910314b6dd0SJohannes Thumshirn } 3911314b6dd0SJohannes Thumshirn 3912314b6dd0SJohannes Thumshirn bio_put(bio); 3913f2984462SChris Mason } 3914f2984462SChris Mason 39158f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3916a05d3c91SQu Wenruo int copy_num, bool drop_cache) 391729c36d72SAnand Jain { 391829c36d72SAnand Jain struct btrfs_super_block *super; 39198f32380dSJohannes Thumshirn struct page *page; 392012659251SNaohiro Aota u64 bytenr, bytenr_orig; 39218f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 392212659251SNaohiro Aota int ret; 392329c36d72SAnand Jain 392412659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 392512659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 392612659251SNaohiro Aota if (ret == -ENOENT) 392712659251SNaohiro Aota return ERR_PTR(-EINVAL); 392812659251SNaohiro Aota else if (ret) 392912659251SNaohiro Aota return ERR_PTR(ret); 393012659251SNaohiro Aota 3931cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 39328f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 393329c36d72SAnand Jain 3934a05d3c91SQu Wenruo if (drop_cache) { 3935a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3936a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3937a05d3c91SQu Wenruo 3938a05d3c91SQu Wenruo /* 3939a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3940a05d3c91SQu Wenruo * always read from the device. 3941a05d3c91SQu Wenruo */ 3942a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3943a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3944a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3945a05d3c91SQu Wenruo } 3946a05d3c91SQu Wenruo 39478f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 39488f32380dSJohannes Thumshirn if (IS_ERR(page)) 39498f32380dSJohannes Thumshirn return ERR_CAST(page); 395029c36d72SAnand Jain 39518f32380dSJohannes Thumshirn super = page_address(page); 395296c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 395396c2e067SAnand Jain btrfs_release_disk_super(super); 395496c2e067SAnand Jain return ERR_PTR(-ENODATA); 395596c2e067SAnand Jain } 395696c2e067SAnand Jain 395712659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 39588f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 39598f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 396029c36d72SAnand Jain } 396129c36d72SAnand Jain 39628f32380dSJohannes Thumshirn return super; 396329c36d72SAnand Jain } 396429c36d72SAnand Jain 396529c36d72SAnand Jain 39668f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3967a512bbf8SYan Zheng { 39688f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3969a512bbf8SYan Zheng int i; 3970a512bbf8SYan Zheng u64 transid = 0; 3971a512bbf8SYan Zheng 3972a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3973a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3974a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3975a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3976a512bbf8SYan Zheng */ 3977a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3978a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 39798f32380dSJohannes Thumshirn if (IS_ERR(super)) 3980a512bbf8SYan Zheng continue; 3981a512bbf8SYan Zheng 3982a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 39838f32380dSJohannes Thumshirn if (latest) 39848f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 39858f32380dSJohannes Thumshirn 39868f32380dSJohannes Thumshirn latest = super; 3987a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3988a512bbf8SYan Zheng } 3989a512bbf8SYan Zheng } 399092fc03fbSAnand Jain 39918f32380dSJohannes Thumshirn return super; 3992a512bbf8SYan Zheng } 3993a512bbf8SYan Zheng 39944eedeb75SHisashi Hifumi /* 3995abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3996314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 39974eedeb75SHisashi Hifumi * 3998abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3999abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 4000abbb3b8eSDavid Sterba * same for write and wait phases. 40014eedeb75SHisashi Hifumi * 4002314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 40034eedeb75SHisashi Hifumi */ 4004a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 4005abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 4006a512bbf8SYan Zheng { 4007d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 4008314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 4009d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 4010a512bbf8SYan Zheng int i; 4011a512bbf8SYan Zheng int errors = 0; 401212659251SNaohiro Aota int ret; 401312659251SNaohiro Aota u64 bytenr, bytenr_orig; 4014a512bbf8SYan Zheng 4015a512bbf8SYan Zheng if (max_mirrors == 0) 4016a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4017a512bbf8SYan Zheng 4018d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 4019d5178578SJohannes Thumshirn 4020a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 4021314b6dd0SJohannes Thumshirn struct page *page; 4022314b6dd0SJohannes Thumshirn struct bio *bio; 4023314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 4024314b6dd0SJohannes Thumshirn 402512659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 402612659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 402712659251SNaohiro Aota if (ret == -ENOENT) { 402812659251SNaohiro Aota continue; 402912659251SNaohiro Aota } else if (ret < 0) { 403012659251SNaohiro Aota btrfs_err(device->fs_info, 403112659251SNaohiro Aota "couldn't get super block location for mirror %d", 403212659251SNaohiro Aota i); 403312659251SNaohiro Aota errors++; 403412659251SNaohiro Aota continue; 403512659251SNaohiro Aota } 4036935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4037935e5cc9SMiao Xie device->commit_total_bytes) 4038a512bbf8SYan Zheng break; 4039a512bbf8SYan Zheng 404012659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 4041a512bbf8SYan Zheng 4042fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 4043fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 4044fd08001fSEric Biggers sb->csum); 4045a512bbf8SYan Zheng 4046314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 4047314b6dd0SJohannes Thumshirn GFP_NOFS); 4048314b6dd0SJohannes Thumshirn if (!page) { 4049fb456252SJeff Mahoney btrfs_err(device->fs_info, 4050314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 4051f14d104dSDavid Sterba bytenr); 4052634554dcSJosef Bacik errors++; 4053634554dcSJosef Bacik continue; 4054634554dcSJosef Bacik } 4055634554dcSJosef Bacik 4056314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 4057314b6dd0SJohannes Thumshirn get_page(page); 4058a512bbf8SYan Zheng 4059314b6dd0SJohannes Thumshirn disk_super = page_address(page); 4060314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 4061a512bbf8SYan Zheng 4062387125fcSChris Mason /* 4063314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 4064314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 4065314b6dd0SJohannes Thumshirn * checking. 4066387125fcSChris Mason */ 406707888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 406807888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 406907888c66SChristoph Hellwig GFP_NOFS); 4070314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 4071314b6dd0SJohannes Thumshirn bio->bi_private = device; 4072314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 4073314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 4074314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 4075314b6dd0SJohannes Thumshirn 4076314b6dd0SJohannes Thumshirn /* 4077314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 4078314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 4079314b6dd0SJohannes Thumshirn * copies might still contain the older version. 4080314b6dd0SJohannes Thumshirn */ 40811b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 4082314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 4083314b6dd0SJohannes Thumshirn 408458ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 408558ff51f1SChristoph Hellwig submit_bio(bio); 40868376d9e1SNaohiro Aota 40878376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 40888376d9e1SNaohiro Aota errors++; 4089a512bbf8SYan Zheng } 4090a512bbf8SYan Zheng return errors < i ? 0 : -1; 4091a512bbf8SYan Zheng } 4092a512bbf8SYan Zheng 4093387125fcSChris Mason /* 4094abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 4095abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 4096abbb3b8eSDavid Sterba * 4097314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 4098abbb3b8eSDavid Sterba * date. 4099abbb3b8eSDavid Sterba */ 4100abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 4101abbb3b8eSDavid Sterba { 4102abbb3b8eSDavid Sterba int i; 4103abbb3b8eSDavid Sterba int errors = 0; 4104b6a535faSHoward McLauchlan bool primary_failed = false; 410512659251SNaohiro Aota int ret; 4106abbb3b8eSDavid Sterba u64 bytenr; 4107abbb3b8eSDavid Sterba 4108abbb3b8eSDavid Sterba if (max_mirrors == 0) 4109abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4110abbb3b8eSDavid Sterba 4111abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 4112314b6dd0SJohannes Thumshirn struct page *page; 4113314b6dd0SJohannes Thumshirn 411412659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 411512659251SNaohiro Aota if (ret == -ENOENT) { 411612659251SNaohiro Aota break; 411712659251SNaohiro Aota } else if (ret < 0) { 411812659251SNaohiro Aota errors++; 411912659251SNaohiro Aota if (i == 0) 412012659251SNaohiro Aota primary_failed = true; 412112659251SNaohiro Aota continue; 412212659251SNaohiro Aota } 4123abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4124abbb3b8eSDavid Sterba device->commit_total_bytes) 4125abbb3b8eSDavid Sterba break; 4126abbb3b8eSDavid Sterba 4127314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 4128314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 4129314b6dd0SJohannes Thumshirn if (!page) { 4130abbb3b8eSDavid Sterba errors++; 4131b6a535faSHoward McLauchlan if (i == 0) 4132b6a535faSHoward McLauchlan primary_failed = true; 4133abbb3b8eSDavid Sterba continue; 4134abbb3b8eSDavid Sterba } 4135314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 4136314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 4137314b6dd0SJohannes Thumshirn if (PageError(page)) { 4138abbb3b8eSDavid Sterba errors++; 4139b6a535faSHoward McLauchlan if (i == 0) 4140b6a535faSHoward McLauchlan primary_failed = true; 4141b6a535faSHoward McLauchlan } 4142abbb3b8eSDavid Sterba 4143314b6dd0SJohannes Thumshirn /* Drop our reference */ 4144314b6dd0SJohannes Thumshirn put_page(page); 4145abbb3b8eSDavid Sterba 4146314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 4147314b6dd0SJohannes Thumshirn put_page(page); 4148abbb3b8eSDavid Sterba } 4149abbb3b8eSDavid Sterba 4150b6a535faSHoward McLauchlan /* log error, force error return */ 4151b6a535faSHoward McLauchlan if (primary_failed) { 4152b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 4153b6a535faSHoward McLauchlan device->devid); 4154b6a535faSHoward McLauchlan return -1; 4155b6a535faSHoward McLauchlan } 4156b6a535faSHoward McLauchlan 4157abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 4158abbb3b8eSDavid Sterba } 4159abbb3b8eSDavid Sterba 4160abbb3b8eSDavid Sterba /* 4161387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 4162387125fcSChris Mason * for the barrier when it is done 4163387125fcSChris Mason */ 41644246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 4165387125fcSChris Mason { 4166f9e69aa9SChristoph Hellwig bio_uninit(bio); 4167387125fcSChris Mason complete(bio->bi_private); 4168387125fcSChris Mason } 4169387125fcSChris Mason 4170387125fcSChris Mason /* 41714fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 41724fc6441aSAnand Jain * done in the waiting counterpart. 4173387125fcSChris Mason */ 41744fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 4175387125fcSChris Mason { 4176f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 4177387125fcSChris Mason 4178a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 4179a91cf0ffSWang Yugui /* 4180a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 4181a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 4182a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 4183a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 4184a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 4185a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 4186a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 4187a91cf0ffSWang Yugui * non-debug builds. 4188a91cf0ffSWang Yugui */ 418908e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 41904fc6441aSAnand Jain return; 4191a91cf0ffSWang Yugui #endif 4192387125fcSChris Mason 4193f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 4194f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 4195387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 4196387125fcSChris Mason init_completion(&device->flush_wait); 4197387125fcSChris Mason bio->bi_private = &device->flush_wait; 4198387125fcSChris Mason 419958ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 420058ff51f1SChristoph Hellwig submit_bio(bio); 42011c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 42024fc6441aSAnand Jain } 4203387125fcSChris Mason 42044fc6441aSAnand Jain /* 42054fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 42064fc6441aSAnand Jain */ 42078c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 42084fc6441aSAnand Jain { 4209f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 42104fc6441aSAnand Jain 42111c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 421258efbc9fSOmar Sandoval return BLK_STS_OK; 42134fc6441aSAnand Jain 42141c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 42152980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 42164fc6441aSAnand Jain 42178c27cb35SLinus Torvalds return bio->bi_status; 4218387125fcSChris Mason } 4219387125fcSChris Mason 4220d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 4221401b41e5SAnand Jain { 42226528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 4223401b41e5SAnand Jain return -EIO; 4224387125fcSChris Mason return 0; 4225387125fcSChris Mason } 4226387125fcSChris Mason 4227387125fcSChris Mason /* 4228387125fcSChris Mason * send an empty flush down to each device in parallel, 4229387125fcSChris Mason * then wait for them 4230387125fcSChris Mason */ 4231387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 4232387125fcSChris Mason { 4233387125fcSChris Mason struct list_head *head; 4234387125fcSChris Mason struct btrfs_device *dev; 42355af3e8ccSStefan Behrens int errors_wait = 0; 42364e4cbee9SChristoph Hellwig blk_status_t ret; 4237387125fcSChris Mason 42381538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 4239387125fcSChris Mason /* send down all the barriers */ 4240387125fcSChris Mason head = &info->fs_devices->devices; 42411538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4242e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4243f88ba6a2SHidetoshi Seto continue; 4244cea7c8bfSAnand Jain if (!dev->bdev) 4245387125fcSChris Mason continue; 4246e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4247ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4248387125fcSChris Mason continue; 4249387125fcSChris Mason 42504fc6441aSAnand Jain write_dev_flush(dev); 425158efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 4252387125fcSChris Mason } 4253387125fcSChris Mason 4254387125fcSChris Mason /* wait for all the barriers */ 42551538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4256e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4257f88ba6a2SHidetoshi Seto continue; 4258387125fcSChris Mason if (!dev->bdev) { 42595af3e8ccSStefan Behrens errors_wait++; 4260387125fcSChris Mason continue; 4261387125fcSChris Mason } 4262e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4263ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4264387125fcSChris Mason continue; 4265387125fcSChris Mason 42664fc6441aSAnand Jain ret = wait_dev_flush(dev); 4267401b41e5SAnand Jain if (ret) { 4268401b41e5SAnand Jain dev->last_flush_error = ret; 426966b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 427066b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 42715af3e8ccSStefan Behrens errors_wait++; 4272387125fcSChris Mason } 4273401b41e5SAnand Jain } 4274401b41e5SAnand Jain 4275cea7c8bfSAnand Jain if (errors_wait) { 4276401b41e5SAnand Jain /* 4277401b41e5SAnand Jain * At some point we need the status of all disks 4278401b41e5SAnand Jain * to arrive at the volume status. So error checking 4279401b41e5SAnand Jain * is being pushed to a separate loop. 4280401b41e5SAnand Jain */ 4281d10b82feSQu Wenruo return check_barrier_error(info); 4282401b41e5SAnand Jain } 4283387125fcSChris Mason return 0; 4284387125fcSChris Mason } 4285387125fcSChris Mason 4286943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4287943c6e99SZhao Lei { 42888789f4feSZhao Lei int raid_type; 42898789f4feSZhao Lei int min_tolerated = INT_MAX; 4290943c6e99SZhao Lei 42918789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 42928789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 42938c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 42948789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 42958789f4feSZhao Lei tolerated_failures); 4296943c6e99SZhao Lei 42978789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 42988789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 42998789f4feSZhao Lei continue; 430041a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 43018789f4feSZhao Lei continue; 43028c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 43038789f4feSZhao Lei btrfs_raid_array[raid_type]. 43048789f4feSZhao Lei tolerated_failures); 43058789f4feSZhao Lei } 4306943c6e99SZhao Lei 43078789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4308ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 43098789f4feSZhao Lei min_tolerated = 0; 43108789f4feSZhao Lei } 43118789f4feSZhao Lei 43128789f4feSZhao Lei return min_tolerated; 4313943c6e99SZhao Lei } 4314943c6e99SZhao Lei 4315eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4316f2984462SChris Mason { 4317e5e9a520SChris Mason struct list_head *head; 4318f2984462SChris Mason struct btrfs_device *dev; 4319a061fc8dSChris Mason struct btrfs_super_block *sb; 4320f2984462SChris Mason struct btrfs_dev_item *dev_item; 4321f2984462SChris Mason int ret; 4322f2984462SChris Mason int do_barriers; 4323a236aed1SChris Mason int max_errors; 4324a236aed1SChris Mason int total_errors = 0; 4325a061fc8dSChris Mason u64 flags; 4326f2984462SChris Mason 43270b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4328fed3b381SLiu Bo 4329fed3b381SLiu Bo /* 4330fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4331fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4332fed3b381SLiu Bo * been consistent on disk. 4333fed3b381SLiu Bo */ 4334fed3b381SLiu Bo if (max_mirrors == 0) 43350b246afaSJeff Mahoney backup_super_roots(fs_info); 4336f2984462SChris Mason 43370b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4338a061fc8dSChris Mason dev_item = &sb->dev_item; 4339e5e9a520SChris Mason 43400b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 43410b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 43420b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4343387125fcSChris Mason 43445af3e8ccSStefan Behrens if (do_barriers) { 43450b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 43465af3e8ccSStefan Behrens if (ret) { 43475af3e8ccSStefan Behrens mutex_unlock( 43480b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 43490b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 43505af3e8ccSStefan Behrens "errors while submitting device barriers."); 43515af3e8ccSStefan Behrens return ret; 43525af3e8ccSStefan Behrens } 43535af3e8ccSStefan Behrens } 4354387125fcSChris Mason 43551538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4356dfe25020SChris Mason if (!dev->bdev) { 4357dfe25020SChris Mason total_errors++; 4358dfe25020SChris Mason continue; 4359dfe25020SChris Mason } 4360e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4361ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4362dfe25020SChris Mason continue; 4363dfe25020SChris Mason 43642b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4365a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4366a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 43677df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4368935e5cc9SMiao Xie dev->commit_total_bytes); 4369ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4370ce7213c7SMiao Xie dev->commit_bytes_used); 4371a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4372a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4373a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4374a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 43757239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 43767239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4377a512bbf8SYan Zheng 4378a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4379a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4380f2984462SChris Mason 438175cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 438275cb857dSQu Wenruo if (ret < 0) { 438375cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 438475cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 438575cb857dSQu Wenruo "unexpected superblock corruption detected"); 438675cb857dSQu Wenruo return -EUCLEAN; 438775cb857dSQu Wenruo } 438875cb857dSQu Wenruo 4389abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4390a236aed1SChris Mason if (ret) 4391a236aed1SChris Mason total_errors++; 4392f2984462SChris Mason } 4393a236aed1SChris Mason if (total_errors > max_errors) { 43940b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4395d397712bSChris Mason total_errors); 43960b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 439779787eaaSJeff Mahoney 43989d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 43990b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 44000b246afaSJeff Mahoney "%d errors while writing supers", 44010b246afaSJeff Mahoney total_errors); 44029d565ba4SStefan Behrens return -EIO; 4403a236aed1SChris Mason } 4404f2984462SChris Mason 4405a512bbf8SYan Zheng total_errors = 0; 44061538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4407dfe25020SChris Mason if (!dev->bdev) 4408dfe25020SChris Mason continue; 4409e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4410ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4411dfe25020SChris Mason continue; 4412dfe25020SChris Mason 4413abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4414a512bbf8SYan Zheng if (ret) 44151259ab75SChris Mason total_errors++; 4416f2984462SChris Mason } 44170b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4418a236aed1SChris Mason if (total_errors > max_errors) { 44190b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 44200b246afaSJeff Mahoney "%d errors while writing supers", 44210b246afaSJeff Mahoney total_errors); 442279787eaaSJeff Mahoney return -EIO; 4423a236aed1SChris Mason } 4424f2984462SChris Mason return 0; 4425f2984462SChris Mason } 4426f2984462SChris Mason 4427cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4428cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4429cb517eabSMiao Xie struct btrfs_root *root) 44302619ba1fSChris Mason { 44314785e24fSJosef Bacik bool drop_ref = false; 44324785e24fSJosef Bacik 4433fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4434fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4435fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4436fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 44374785e24fSJosef Bacik drop_ref = true; 4438fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 443976dda93cSYan, Zheng 444084961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4441ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 44421c1ea4f7SLiu Bo if (root->reloc_root) { 444300246528SJosef Bacik btrfs_put_root(root->reloc_root); 44441c1ea4f7SLiu Bo root->reloc_root = NULL; 44451c1ea4f7SLiu Bo } 44461c1ea4f7SLiu Bo } 44473321719eSLiu Bo 44484785e24fSJosef Bacik if (drop_ref) 444900246528SJosef Bacik btrfs_put_root(root); 44502619ba1fSChris Mason } 44512619ba1fSChris Mason 4452c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4453c146afadSYan Zheng { 4454fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4455fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4456fc7cbcd4SDavid Sterba int i = 0; 445765d33fd7SQu Wenruo int err = 0; 4458fc7cbcd4SDavid Sterba unsigned int ret = 0; 4459c146afadSYan Zheng 4460c146afadSYan Zheng while (1) { 4461fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4462fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4463fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4464fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4465fc7cbcd4SDavid Sterba if (!ret) { 4466fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4467c146afadSYan Zheng break; 446865d33fd7SQu Wenruo } 4469fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 447066b4ffd1SJosef Bacik 4471fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4472fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4473fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4474fc7cbcd4SDavid Sterba gang[i] = NULL; 447565d33fd7SQu Wenruo continue; 4476c146afadSYan Zheng } 4477fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4478fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4479fc7cbcd4SDavid Sterba } 4480fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4481fc7cbcd4SDavid Sterba 4482fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4483fc7cbcd4SDavid Sterba if (!gang[i]) 4484fc7cbcd4SDavid Sterba continue; 4485fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4486fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4487fc7cbcd4SDavid Sterba if (err) 4488fc7cbcd4SDavid Sterba break; 4489fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4490fc7cbcd4SDavid Sterba } 4491fc7cbcd4SDavid Sterba root_objectid++; 4492c146afadSYan Zheng } 449365d33fd7SQu Wenruo 4494fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4495fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4496fc7cbcd4SDavid Sterba if (gang[i]) 4497fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 449865d33fd7SQu Wenruo } 449965d33fd7SQu Wenruo return err; 4500c146afadSYan Zheng } 4501c146afadSYan Zheng 45026bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4503c146afadSYan Zheng { 45046bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4505c146afadSYan Zheng struct btrfs_trans_handle *trans; 4506c146afadSYan Zheng 45070b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45082ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45090b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 45100b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4511c71bf099SYan, Zheng 4512c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 45130b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45140b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4515c71bf099SYan, Zheng 45167a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 45173612b495STsutomu Itoh if (IS_ERR(trans)) 45183612b495STsutomu Itoh return PTR_ERR(trans); 45193a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4520c146afadSYan Zheng } 4521c146afadSYan Zheng 452236c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 452336c86a9eSQu Wenruo { 452436c86a9eSQu Wenruo struct btrfs_transaction *trans; 452536c86a9eSQu Wenruo struct btrfs_transaction *tmp; 452636c86a9eSQu Wenruo bool found = false; 452736c86a9eSQu Wenruo 452836c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 452936c86a9eSQu Wenruo return; 453036c86a9eSQu Wenruo 453136c86a9eSQu Wenruo /* 453236c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 453336c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 453436c86a9eSQu Wenruo */ 453536c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 453636c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 453736c86a9eSQu Wenruo struct extent_state *cached = NULL; 453836c86a9eSQu Wenruo u64 dirty_bytes = 0; 453936c86a9eSQu Wenruo u64 cur = 0; 454036c86a9eSQu Wenruo u64 found_start; 454136c86a9eSQu Wenruo u64 found_end; 454236c86a9eSQu Wenruo 454336c86a9eSQu Wenruo found = true; 454436c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 454536c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 454636c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 454736c86a9eSQu Wenruo cur = found_end + 1; 454836c86a9eSQu Wenruo } 454936c86a9eSQu Wenruo btrfs_warn(fs_info, 455036c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 455136c86a9eSQu Wenruo trans->transid, dirty_bytes); 455236c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 455336c86a9eSQu Wenruo 455436c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 455536c86a9eSQu Wenruo fs_info->running_transaction = NULL; 455636c86a9eSQu Wenruo list_del_init(&trans->list); 455736c86a9eSQu Wenruo 455836c86a9eSQu Wenruo btrfs_put_transaction(trans); 455936c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 456036c86a9eSQu Wenruo } 456136c86a9eSQu Wenruo ASSERT(!found); 456236c86a9eSQu Wenruo } 456336c86a9eSQu Wenruo 4564b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4565eb60ceacSChris Mason { 4566c146afadSYan Zheng int ret; 4567e089f05cSChris Mason 4568afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 456931e70e52SFilipe Manana 457031e70e52SFilipe Manana /* 45718a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 45728a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 45738a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 45748a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 45758a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 45768a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 45778a1f1e3dSFilipe Manana * return 1. 45788a1f1e3dSFilipe Manana */ 45798a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 45808a1f1e3dSFilipe Manana 45818a1f1e3dSFilipe Manana /* 458231e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 458331e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 458431e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 458531e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 458631e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 458731e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 458831e70e52SFilipe Manana * trying to stop the async reclaim task. 458931e70e52SFilipe Manana */ 459031e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4591d6fd0ae2SOmar Sandoval /* 4592d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4593d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4594d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4595d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4596d6fd0ae2SOmar Sandoval */ 4597d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4598a2135011SChris Mason 45997343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4600d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 46017343dd61SJustin Maggard 4602803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4603803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4604803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4605803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4606803b2f54SStefan Behrens 4607837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4608aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4609837d5b6eSIlya Dryomov 46108dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 46118dabb742SStefan Behrens 4612aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 46134cb5300bSChris Mason 46144cb5300bSChris Mason /* wait for any defraggers to finish */ 46154cb5300bSChris Mason wait_event(fs_info->transaction_wait, 46164cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 46174cb5300bSChris Mason 46184cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 461926176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 46204cb5300bSChris Mason 4621a362bb86SFilipe Manana /* 4622a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4623a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4624a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4625a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4626a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4627a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4628a362bb86SFilipe Manana * no one else to run iputs. 4629a362bb86SFilipe Manana * 4630a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4631a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4632a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4633a362bb86SFilipe Manana * through some other means. 4634a362bb86SFilipe Manana * 4635a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4636a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4637a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4638a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4639a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4640a362bb86SFilipe Manana */ 4641a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4642a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4643a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4644a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4645a362bb86SFilipe Manana 464621c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 464757056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4648576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 464921c7e756SMiao Xie 4650b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4651b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4652b0643e59SDennis Zhou 4653bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4654e44163e1SJeff Mahoney /* 4655d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4656d6fd0ae2SOmar Sandoval * unused block groups. 4657e44163e1SJeff Mahoney */ 46580b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4659e44163e1SJeff Mahoney 4660f0cc2cd7SFilipe Manana /* 4661f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4662f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4663f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4664f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4665f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4666f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4667f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4668f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4669f0cc2cd7SFilipe Manana * This is a very rare case. 4670f0cc2cd7SFilipe Manana */ 4671f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4672f0cc2cd7SFilipe Manana 46736bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4674d397712bSChris Mason if (ret) 467504892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4676c146afadSYan Zheng } 4677ed2ff2cbSChris Mason 467884961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 46792ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4680acce952bSliubo 4681e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4682e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 46838929ecfaSYan, Zheng 4684e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4685afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4686f25784b3SYan Zheng 46875958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 46885958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 46895958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 46905958253cSQu Wenruo } 46915958253cSQu Wenruo 469204892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4693fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4694bcef60f2SArne Jansen 4695963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 469604892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4697963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4698b0c68f8bSChris Mason } 469931153d81SYan Zheng 47005deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 47014297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 47025deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 47034297ff84SJosef Bacik 47046618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4705b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 47065ac1d209SJeff Mahoney 47071a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 47081a4319ccSLiu Bo 4709de348ee0SWang Shilong /* 4710de348ee0SWang Shilong * we must make sure there is not any read request to 4711de348ee0SWang Shilong * submit after we stopping all workers. 4712de348ee0SWang Shilong */ 4713de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 471496192499SJosef Bacik btrfs_stop_all_workers(fs_info); 471596192499SJosef Bacik 47160a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 471736c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 47180a31daa4SFilipe Manana 4719afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 47204273eaffSAnand Jain free_root_pointers(fs_info, true); 47218c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 47229ad6b7bcSChris Mason 47234e19443dSJosef Bacik /* 47244e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 47254e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 47264e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 47274e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 47284e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 47294e19443dSJosef Bacik */ 47304e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 47314e19443dSJosef Bacik 473213e6c37bSJosef Bacik iput(fs_info->btree_inode); 4733d6bfde87SChris Mason 473421adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 47350b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 47362ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 473721adbd5cSStefan Behrens #endif 473821adbd5cSStefan Behrens 47390b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 474068c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4741eb60ceacSChris Mason } 4742eb60ceacSChris Mason 4743b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4744b9fab919SChris Mason int atomic) 4745ccd467d6SChris Mason { 47461259ab75SChris Mason int ret; 4747727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 47481259ab75SChris Mason 47490b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 47501259ab75SChris Mason if (!ret) 47511259ab75SChris Mason return ret; 47521259ab75SChris Mason 47531259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4754b9fab919SChris Mason parent_transid, atomic); 4755b9fab919SChris Mason if (ret == -EAGAIN) 4756b9fab919SChris Mason return ret; 47571259ab75SChris Mason return !ret; 47585f39d397SChris Mason } 47596702ed49SChris Mason 47605f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 47615f39d397SChris Mason { 47622f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 47635f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4764b9473439SChris Mason int was_dirty; 4765b4ce94deSChris Mason 476606ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 476706ea65a3SJosef Bacik /* 476806ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 476952042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 477006ea65a3SJosef Bacik * outside of the sanity tests. 477106ea65a3SJosef Bacik */ 4772b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 477306ea65a3SJosef Bacik return; 477406ea65a3SJosef Bacik #endif 477549d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 47760b246afaSJeff Mahoney if (transid != fs_info->generation) 47775d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 47780b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 47790b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4780e2d84521SMiao Xie if (!was_dirty) 4781104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4782e2d84521SMiao Xie buf->len, 47830b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 47841f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 478569fc6cbbSQu Wenruo /* 478669fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 478769fc6cbbSQu Wenruo * but item data not updated. 478869fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 478969fc6cbbSQu Wenruo */ 479069fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4791cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4792a4f78750SDavid Sterba btrfs_print_leaf(buf); 47931f21ef0aSFilipe Manana ASSERT(0); 47941f21ef0aSFilipe Manana } 47951f21ef0aSFilipe Manana #endif 4796eb60ceacSChris Mason } 4797eb60ceacSChris Mason 47982ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4799b53d3f5dSLiu Bo int flush_delayed) 480035b7e476SChris Mason { 4801188de649SChris Mason /* 4802188de649SChris Mason * looks as though older kernels can get into trouble with 4803188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4804188de649SChris Mason */ 4805e2d84521SMiao Xie int ret; 4806d6bfde87SChris Mason 48076933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4808d6bfde87SChris Mason return; 4809d6bfde87SChris Mason 4810b53d3f5dSLiu Bo if (flush_delayed) 48112ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 481216cdcec7SMiao Xie 4813d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4814d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4815d814a491SEthan Lien fs_info->dirty_metadata_batch); 4816e2d84521SMiao Xie if (ret > 0) { 48170b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 481816cdcec7SMiao Xie } 481916cdcec7SMiao Xie } 482016cdcec7SMiao Xie 48212ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 482216cdcec7SMiao Xie { 48232ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 482435b7e476SChris Mason } 4825b53d3f5dSLiu Bo 48262ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4827b53d3f5dSLiu Bo { 48282ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4829d6bfde87SChris Mason } 48306b80053dSChris Mason 48312ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4832acce952bSliubo { 4833fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4834fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4835fe816d0fSNikolay Borisov 48360b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 48372ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 48380b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4839acce952bSliubo 48400b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 48410b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4842acce952bSliubo } 4843acce952bSliubo 4844ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4845ef67963dSJosef Bacik { 4846fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4847fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4848fc7cbcd4SDavid Sterba int ret; 4849ef67963dSJosef Bacik 4850fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4851fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4852fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4853fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4854fc7cbcd4SDavid Sterba int i; 4855ef67963dSJosef Bacik 4856fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4857fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4858fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4859fc7cbcd4SDavid Sterba 4860fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4861fc7cbcd4SDavid Sterba if (!gang[i]) 4862ef67963dSJosef Bacik continue; 4863fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4864fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4865fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4866ef67963dSJosef Bacik } 4867fc7cbcd4SDavid Sterba root_objectid++; 4868fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4869ef67963dSJosef Bacik } 4870fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4871ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4872ef67963dSJosef Bacik } 4873ef67963dSJosef Bacik 4874143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4875acce952bSliubo { 4876acce952bSliubo struct btrfs_ordered_extent *ordered; 4877acce952bSliubo 4878199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4879779880efSJosef Bacik /* 4880779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4881779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4882779880efSJosef Bacik */ 4883199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4884779880efSJosef Bacik root_extent_list) 4885779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4886199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4887199c2a9cSMiao Xie } 4888199c2a9cSMiao Xie 4889199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4890199c2a9cSMiao Xie { 4891199c2a9cSMiao Xie struct btrfs_root *root; 4892199c2a9cSMiao Xie struct list_head splice; 4893199c2a9cSMiao Xie 4894199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4895199c2a9cSMiao Xie 4896199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4897199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4898199c2a9cSMiao Xie while (!list_empty(&splice)) { 4899199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4900199c2a9cSMiao Xie ordered_root); 49011de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 49021de2cfdeSJosef Bacik &fs_info->ordered_roots); 4903199c2a9cSMiao Xie 49042a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4905199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4906199c2a9cSMiao Xie 49072a85d9caSLiu Bo cond_resched(); 49082a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4909199c2a9cSMiao Xie } 4910199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 491174d5d229SJosef Bacik 491274d5d229SJosef Bacik /* 491374d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 491474d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 491574d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 491674d5d229SJosef Bacik * actually get run and error out properly. 491774d5d229SJosef Bacik */ 491874d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4919acce952bSliubo } 4920acce952bSliubo 492135a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 49222ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4923acce952bSliubo { 4924acce952bSliubo struct rb_node *node; 4925acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4926acce952bSliubo struct btrfs_delayed_ref_node *ref; 4927acce952bSliubo int ret = 0; 4928acce952bSliubo 4929acce952bSliubo delayed_refs = &trans->delayed_refs; 4930acce952bSliubo 4931acce952bSliubo spin_lock(&delayed_refs->lock); 4932d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4933cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4934b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4935acce952bSliubo return ret; 4936acce952bSliubo } 4937acce952bSliubo 49385c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4939d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 49400e0adbcfSJosef Bacik struct rb_node *n; 4941e78417d1SJosef Bacik bool pin_bytes = false; 4942acce952bSliubo 4943d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4944d7df2c79SJosef Bacik href_node); 49453069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4946b939d1abSJosef Bacik continue; 49473069bd26SJosef Bacik 4948d7df2c79SJosef Bacik spin_lock(&head->lock); 4949e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 49500e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 49510e0adbcfSJosef Bacik ref_node); 4952d7df2c79SJosef Bacik ref->in_tree = 0; 4953e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 49540e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 49551d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 49561d57ee94SWang Xiaoguang list_del(&ref->add_list); 4957d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4958d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4959d7df2c79SJosef Bacik } 496054067ae9SJosef Bacik if (head->must_insert_reserved) 4961e78417d1SJosef Bacik pin_bytes = true; 496278a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4963fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4964d7df2c79SJosef Bacik spin_unlock(&head->lock); 4965acce952bSliubo spin_unlock(&delayed_refs->lock); 4966e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4967acce952bSliubo 4968f603bb94SNikolay Borisov if (pin_bytes) { 4969f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4970f603bb94SNikolay Borisov 4971f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4972f603bb94SNikolay Borisov BUG_ON(!cache); 4973f603bb94SNikolay Borisov 4974f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4975f603bb94SNikolay Borisov spin_lock(&cache->lock); 4976f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4977f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4978f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4979f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4980f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4981f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4982f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4983f603bb94SNikolay Borisov 4984f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4985f603bb94SNikolay Borisov 4986f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4987f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4988f603bb94SNikolay Borisov } 498931890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4990d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4991acce952bSliubo cond_resched(); 4992acce952bSliubo spin_lock(&delayed_refs->lock); 4993acce952bSliubo } 499481f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4995acce952bSliubo 4996acce952bSliubo spin_unlock(&delayed_refs->lock); 4997acce952bSliubo 4998acce952bSliubo return ret; 4999acce952bSliubo } 5000acce952bSliubo 5001143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 5002acce952bSliubo { 5003acce952bSliubo struct btrfs_inode *btrfs_inode; 5004acce952bSliubo struct list_head splice; 5005acce952bSliubo 5006acce952bSliubo INIT_LIST_HEAD(&splice); 5007acce952bSliubo 5008eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5009eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 5010acce952bSliubo 5011acce952bSliubo while (!list_empty(&splice)) { 5012fe816d0fSNikolay Borisov struct inode *inode = NULL; 5013eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 5014acce952bSliubo delalloc_inodes); 5015fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 5016eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5017acce952bSliubo 5018fe816d0fSNikolay Borisov /* 5019fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 5020fe816d0fSNikolay Borisov * meanwhile. 5021fe816d0fSNikolay Borisov */ 5022fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 5023fe816d0fSNikolay Borisov if (inode) { 5024fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 5025fe816d0fSNikolay Borisov iput(inode); 5026fe816d0fSNikolay Borisov } 5027eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5028acce952bSliubo } 5029eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5030eb73c1b7SMiao Xie } 5031eb73c1b7SMiao Xie 5032eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 5033eb73c1b7SMiao Xie { 5034eb73c1b7SMiao Xie struct btrfs_root *root; 5035eb73c1b7SMiao Xie struct list_head splice; 5036eb73c1b7SMiao Xie 5037eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 5038eb73c1b7SMiao Xie 5039eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5040eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 5041eb73c1b7SMiao Xie while (!list_empty(&splice)) { 5042eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 5043eb73c1b7SMiao Xie delalloc_root); 504400246528SJosef Bacik root = btrfs_grab_root(root); 5045eb73c1b7SMiao Xie BUG_ON(!root); 5046eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5047eb73c1b7SMiao Xie 5048eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 504900246528SJosef Bacik btrfs_put_root(root); 5050eb73c1b7SMiao Xie 5051eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5052eb73c1b7SMiao Xie } 5053eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5054acce952bSliubo } 5055acce952bSliubo 50562ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 5057acce952bSliubo struct extent_io_tree *dirty_pages, 5058acce952bSliubo int mark) 5059acce952bSliubo { 5060acce952bSliubo int ret; 5061acce952bSliubo struct extent_buffer *eb; 5062acce952bSliubo u64 start = 0; 5063acce952bSliubo u64 end; 5064acce952bSliubo 5065acce952bSliubo while (1) { 5066acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 5067e6138876SJosef Bacik mark, NULL); 5068acce952bSliubo if (ret) 5069acce952bSliubo break; 5070acce952bSliubo 507191166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 5072acce952bSliubo while (start <= end) { 50730b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 50740b246afaSJeff Mahoney start += fs_info->nodesize; 5075fd8b2b61SJosef Bacik if (!eb) 5076acce952bSliubo continue; 5077fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 5078acce952bSliubo 5079fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 5080fd8b2b61SJosef Bacik &eb->bflags)) 5081fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 5082fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 5083acce952bSliubo } 5084acce952bSliubo } 5085acce952bSliubo 5086acce952bSliubo return ret; 5087acce952bSliubo } 5088acce952bSliubo 50892ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 5090fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 5091acce952bSliubo { 5092acce952bSliubo u64 start; 5093acce952bSliubo u64 end; 5094acce952bSliubo int ret; 5095acce952bSliubo 5096acce952bSliubo while (1) { 50970e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 50980e6ec385SFilipe Manana 5099fcd5e742SLu Fengqi /* 5100fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 5101fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 5102fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 5103fcd5e742SLu Fengqi * the same extent range. 5104fcd5e742SLu Fengqi */ 5105fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 5106acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 51070e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 5108fcd5e742SLu Fengqi if (ret) { 5109fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5110acce952bSliubo break; 5111fcd5e742SLu Fengqi } 5112acce952bSliubo 51130e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 51140e6ec385SFilipe Manana free_extent_state(cached_state); 51152ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 5116fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5117acce952bSliubo cond_resched(); 5118acce952bSliubo } 5119acce952bSliubo 5120acce952bSliubo return 0; 5121acce952bSliubo } 5122acce952bSliubo 512332da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 5124c79a1751SLiu Bo { 5125c79a1751SLiu Bo struct inode *inode; 5126c79a1751SLiu Bo 5127c79a1751SLiu Bo inode = cache->io_ctl.inode; 5128c79a1751SLiu Bo if (inode) { 5129c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 5130c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 5131c79a1751SLiu Bo cache->io_ctl.inode = NULL; 5132c79a1751SLiu Bo iput(inode); 5133c79a1751SLiu Bo } 5134bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 5135c79a1751SLiu Bo btrfs_put_block_group(cache); 5136c79a1751SLiu Bo } 5137c79a1751SLiu Bo 5138c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 51392ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 5140c79a1751SLiu Bo { 514132da5386SDavid Sterba struct btrfs_block_group *cache; 5142c79a1751SLiu Bo 5143c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5144c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 5145c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 514632da5386SDavid Sterba struct btrfs_block_group, 5147c79a1751SLiu Bo dirty_list); 5148c79a1751SLiu Bo 5149c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 5150c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5151c79a1751SLiu Bo list_del_init(&cache->io_list); 5152c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5153c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5154c79a1751SLiu Bo } 5155c79a1751SLiu Bo 5156c79a1751SLiu Bo list_del_init(&cache->dirty_list); 5157c79a1751SLiu Bo spin_lock(&cache->lock); 5158c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5159c79a1751SLiu Bo spin_unlock(&cache->lock); 5160c79a1751SLiu Bo 5161c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5162c79a1751SLiu Bo btrfs_put_block_group(cache); 5163ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 5164c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5165c79a1751SLiu Bo } 5166c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5167c79a1751SLiu Bo 516845ae2c18SNikolay Borisov /* 516945ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 517045ae2c18SNikolay Borisov * to use it without any locking 517145ae2c18SNikolay Borisov */ 5172c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 5173c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 517432da5386SDavid Sterba struct btrfs_block_group, 5175c79a1751SLiu Bo io_list); 5176c79a1751SLiu Bo 5177c79a1751SLiu Bo list_del_init(&cache->io_list); 5178c79a1751SLiu Bo spin_lock(&cache->lock); 5179c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5180c79a1751SLiu Bo spin_unlock(&cache->lock); 5181c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5182c79a1751SLiu Bo } 5183c79a1751SLiu Bo } 5184c79a1751SLiu Bo 518549b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 51862ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 518749b25e05SJeff Mahoney { 5188bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 5189bbbf7243SNikolay Borisov 51902ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 5191c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 5192c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 5193c79a1751SLiu Bo 5194bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 5195bbbf7243SNikolay Borisov post_commit_list) { 5196bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 5197bbbf7243SNikolay Borisov } 5198bbbf7243SNikolay Borisov 51992ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 520049b25e05SJeff Mahoney 52014a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 52020b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 520349b25e05SJeff Mahoney 52044a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 52050b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 520649b25e05SJeff Mahoney 5207ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 520867cde344SMiao Xie 52092ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 521049b25e05SJeff Mahoney EXTENT_DIRTY); 5211fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 521249b25e05SJeff Mahoney 5213d3575156SNaohiro Aota btrfs_free_redirty_list(cur_trans); 5214d3575156SNaohiro Aota 52154a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 52164a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 521749b25e05SJeff Mahoney } 521849b25e05SJeff Mahoney 52192ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 5220acce952bSliubo { 5221acce952bSliubo struct btrfs_transaction *t; 5222acce952bSliubo 52230b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 5224acce952bSliubo 52250b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 52260b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 52270b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 5228724e2315SJosef Bacik struct btrfs_transaction, list); 5229724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 52309b64f57dSElena Reshetova refcount_inc(&t->use_count); 52310b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 52322ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 5233724e2315SJosef Bacik btrfs_put_transaction(t); 52340b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5235724e2315SJosef Bacik continue; 5236724e2315SJosef Bacik } 52370b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 5238724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 52390b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5240724e2315SJosef Bacik /* 5241724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 5242724e2315SJosef Bacik * handle open currently for this transaction. 5243724e2315SJosef Bacik */ 5244724e2315SJosef Bacik wait_event(t->writer_wait, 5245724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 5246724e2315SJosef Bacik } else { 52470b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5248724e2315SJosef Bacik } 52492ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 5250724e2315SJosef Bacik 52510b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 52520b246afaSJeff Mahoney if (t == fs_info->running_transaction) 52530b246afaSJeff Mahoney fs_info->running_transaction = NULL; 5254724e2315SJosef Bacik list_del_init(&t->list); 52550b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5256a4abeea4SJosef Bacik 5257724e2315SJosef Bacik btrfs_put_transaction(t); 52582e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 52590b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5260724e2315SJosef Bacik } 52610b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 52620b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 5263ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 5264ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 52650b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 5266ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 52670b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 5268acce952bSliubo 5269acce952bSliubo return 0; 5270acce952bSliubo } 5271ec7d6dfdSNikolay Borisov 5272453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 5273ec7d6dfdSNikolay Borisov { 5274ec7d6dfdSNikolay Borisov struct btrfs_path *path; 5275ec7d6dfdSNikolay Borisov int ret; 5276ec7d6dfdSNikolay Borisov struct extent_buffer *l; 5277ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 5278ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 5279ec7d6dfdSNikolay Borisov int slot; 5280ec7d6dfdSNikolay Borisov 5281ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 5282ec7d6dfdSNikolay Borisov if (!path) 5283ec7d6dfdSNikolay Borisov return -ENOMEM; 5284ec7d6dfdSNikolay Borisov 5285ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 5286ec7d6dfdSNikolay Borisov search_key.type = -1; 5287ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 5288ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 5289ec7d6dfdSNikolay Borisov if (ret < 0) 5290ec7d6dfdSNikolay Borisov goto error; 5291ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 5292ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 5293ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 5294ec7d6dfdSNikolay Borisov l = path->nodes[0]; 5295ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 529623125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 529723125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 5298ec7d6dfdSNikolay Borisov } else { 529923125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 5300ec7d6dfdSNikolay Borisov } 5301ec7d6dfdSNikolay Borisov ret = 0; 5302ec7d6dfdSNikolay Borisov error: 5303ec7d6dfdSNikolay Borisov btrfs_free_path(path); 5304ec7d6dfdSNikolay Borisov return ret; 5305ec7d6dfdSNikolay Borisov } 5306ec7d6dfdSNikolay Borisov 5307543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 5308ec7d6dfdSNikolay Borisov { 5309ec7d6dfdSNikolay Borisov int ret; 5310ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5311ec7d6dfdSNikolay Borisov 53126b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5313ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5314ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5315ec7d6dfdSNikolay Borisov root->root_key.objectid); 5316ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5317ec7d6dfdSNikolay Borisov goto out; 5318ec7d6dfdSNikolay Borisov } 5319ec7d6dfdSNikolay Borisov 532023125104SNikolay Borisov *objectid = root->free_objectid++; 5321ec7d6dfdSNikolay Borisov ret = 0; 5322ec7d6dfdSNikolay Borisov out: 5323ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5324ec7d6dfdSNikolay Borisov return ret; 5325ec7d6dfdSNikolay Borisov } 5326