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" 532fc6822cSJosef Bacik #include "scrub.h" 54c03b2207SJosef Bacik #include "super.h" 55eb60ceacSChris Mason 56319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 57319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 58319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 59319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 60e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 61e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 62319e4d06SQu Wenruo 63143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 64acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 652ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 66143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 672ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 68acce952bSliubo struct extent_io_tree *dirty_pages, 69acce952bSliubo int mark); 702ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 71acce952bSliubo struct extent_io_tree *pinned_extents); 722ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 732ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 74ce9adaa5SChris Mason 75141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 76141386e1SJosef Bacik { 77141386e1SJosef Bacik if (fs_info->csum_shash) 78141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 79141386e1SJosef Bacik } 80141386e1SJosef Bacik 81d352ac68SChris Mason /* 82d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 83d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 84d352ac68SChris Mason * just before they are sent down the IO stack. 85d352ac68SChris Mason */ 8644b8bd7eSChris Mason struct async_submit_bio { 878896a08dSQu Wenruo struct inode *inode; 8844b8bd7eSChris Mason struct bio *bio; 89a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 9044b8bd7eSChris Mason int mirror_num; 911941b64bSQu Wenruo 921941b64bSQu Wenruo /* Optional parameter for submit_bio_start used by direct io */ 931941b64bSQu Wenruo u64 dio_file_offset; 948b712842SChris Mason struct btrfs_work work; 954e4cbee9SChristoph Hellwig blk_status_t status; 9644b8bd7eSChris Mason }; 9744b8bd7eSChris Mason 9885d4e461SChris Mason /* 992996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 100d352ac68SChris Mason */ 101c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 10219c00ddcSChris Mason { 103d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 1047280305eSDavid Sterba const int num_pages = num_extent_pages(buf); 105a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 106d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 10719c00ddcSChris Mason char *kaddr; 108e9be5a30SDavid Sterba int i; 109d5178578SJohannes Thumshirn 110d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 111d5178578SJohannes Thumshirn crypto_shash_init(shash); 112a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 113e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 114a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 11519c00ddcSChris Mason 116e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 117e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 118e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 11919c00ddcSChris Mason } 12071a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 121d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 12219c00ddcSChris Mason } 12319c00ddcSChris Mason 124d352ac68SChris Mason /* 125d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 126d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 127d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 128d352ac68SChris Mason * in the wrong place. 129d352ac68SChris Mason */ 1301259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 131b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 132b9fab919SChris Mason int atomic) 1331259ab75SChris Mason { 1342ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 1351259ab75SChris Mason int ret; 1361259ab75SChris Mason 1371259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 1381259ab75SChris Mason return 0; 1391259ab75SChris Mason 140b9fab919SChris Mason if (atomic) 141b9fab919SChris Mason return -EAGAIN; 142b9fab919SChris Mason 143570eb97bSJosef Bacik lock_extent(io_tree, eb->start, eb->start + eb->len - 1, &cached_state); 1440b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 1451259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 1461259ab75SChris Mason ret = 0; 1471259ab75SChris Mason goto out; 1481259ab75SChris Mason } 14994647322SDavid Sterba btrfs_err_rl(eb->fs_info, 1508f0ed7d4SQu Wenruo "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", 1518f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 15229549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 1531259ab75SChris Mason ret = 1; 1540b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 15533958dc6SChris Mason out: 156570eb97bSJosef Bacik unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, 157e43bbe5eSDavid Sterba &cached_state); 1581259ab75SChris Mason return ret; 1591259ab75SChris Mason } 1601259ab75SChris Mason 161e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 162e7e16f48SJohannes Thumshirn { 163e7e16f48SJohannes Thumshirn switch (csum_type) { 164e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 1653951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 1663831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 167352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 168e7e16f48SJohannes Thumshirn return true; 169e7e16f48SJohannes Thumshirn default: 170e7e16f48SJohannes Thumshirn return false; 171e7e16f48SJohannes Thumshirn } 172e7e16f48SJohannes Thumshirn } 173e7e16f48SJohannes Thumshirn 174d352ac68SChris Mason /* 1751104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 1761104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 1771104a885SDavid Sterba */ 1783d17adeaSQu Wenruo int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 1793d17adeaSQu Wenruo const struct btrfs_super_block *disk_sb) 1801104a885SDavid Sterba { 18151bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 182d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 183d5178578SJohannes Thumshirn 184d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 1851104a885SDavid Sterba 1861104a885SDavid Sterba /* 1871104a885SDavid Sterba * The super_block structure does not span the whole 18851bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 18951bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 1901104a885SDavid Sterba */ 1913d17adeaSQu Wenruo crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, 192fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 1931104a885SDavid Sterba 19455fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 195e7e16f48SJohannes Thumshirn return 1; 1961104a885SDavid Sterba 197e7e16f48SJohannes Thumshirn return 0; 1981104a885SDavid Sterba } 1991104a885SDavid Sterba 200e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 201ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 202581c1760SQu Wenruo { 203e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 204581c1760SQu Wenruo int found_level; 205581c1760SQu Wenruo struct btrfs_key found_key; 206581c1760SQu Wenruo int ret; 207581c1760SQu Wenruo 208581c1760SQu Wenruo found_level = btrfs_header_level(eb); 209581c1760SQu Wenruo if (found_level != level) { 21063489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 21163489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 212581c1760SQu Wenruo btrfs_err(fs_info, 213581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 214581c1760SQu Wenruo eb->start, level, found_level); 215581c1760SQu Wenruo return -EIO; 216581c1760SQu Wenruo } 217581c1760SQu Wenruo 218581c1760SQu Wenruo if (!first_key) 219581c1760SQu Wenruo return 0; 220581c1760SQu Wenruo 2215d41be6fSQu Wenruo /* 2225d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 2235d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 2245d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 2255d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 2265d41be6fSQu Wenruo */ 2275d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 2285d41be6fSQu Wenruo return 0; 22962fdaa52SQu Wenruo 23062fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 23162fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 23262fdaa52SQu Wenruo btrfs_err(fs_info, 23362fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 23462fdaa52SQu Wenruo eb->start); 23562fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 23662fdaa52SQu Wenruo return -EUCLEAN; 23762fdaa52SQu Wenruo } 23862fdaa52SQu Wenruo 239581c1760SQu Wenruo if (found_level) 240581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 241581c1760SQu Wenruo else 242581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 243581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 244581c1760SQu Wenruo 245581c1760SQu Wenruo if (ret) { 24663489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 24763489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 248581c1760SQu Wenruo btrfs_err(fs_info, 249ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 250ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 251ff76a864SLiu Bo first_key->type, first_key->offset, 252ff76a864SLiu Bo found_key.objectid, found_key.type, 253ff76a864SLiu Bo found_key.offset); 254581c1760SQu Wenruo } 255581c1760SQu Wenruo return ret; 256581c1760SQu Wenruo } 257581c1760SQu Wenruo 2581104a885SDavid Sterba /* 259d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 260d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 261581c1760SQu Wenruo * 262581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 263581c1760SQu Wenruo * @level: expected level, mandatory check 264581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 265d352ac68SChris Mason */ 2666a2e9dc4SFilipe Manana int btrfs_read_extent_buffer(struct extent_buffer *eb, 267581c1760SQu Wenruo u64 parent_transid, int level, 268581c1760SQu Wenruo struct btrfs_key *first_key) 269f188591eSChris Mason { 2705ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 271f188591eSChris Mason struct extent_io_tree *io_tree; 272ea466794SJosef Bacik int failed = 0; 273f188591eSChris Mason int ret; 274f188591eSChris Mason int num_copies = 0; 275f188591eSChris Mason int mirror_num = 0; 276ea466794SJosef Bacik int failed_mirror = 0; 277f188591eSChris Mason 2780b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 279f188591eSChris Mason while (1) { 280f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 281c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); 282256dd1bbSStefan Behrens if (!ret) { 283581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 284b9fab919SChris Mason parent_transid, 0)) 285256dd1bbSStefan Behrens ret = -EIO; 286e064d5e9SDavid Sterba else if (btrfs_verify_level_key(eb, level, 287ff76a864SLiu Bo first_key, parent_transid)) 288581c1760SQu Wenruo ret = -EUCLEAN; 289581c1760SQu Wenruo else 290581c1760SQu Wenruo break; 291256dd1bbSStefan Behrens } 292d397712bSChris Mason 2930b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 294f188591eSChris Mason eb->start, eb->len); 2954235298eSChris Mason if (num_copies == 1) 296ea466794SJosef Bacik break; 2974235298eSChris Mason 2985cf1ab56SJosef Bacik if (!failed_mirror) { 2995cf1ab56SJosef Bacik failed = 1; 3005cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 3015cf1ab56SJosef Bacik } 3025cf1ab56SJosef Bacik 303f188591eSChris Mason mirror_num++; 304ea466794SJosef Bacik if (mirror_num == failed_mirror) 305ea466794SJosef Bacik mirror_num++; 306ea466794SJosef Bacik 3074235298eSChris Mason if (mirror_num > num_copies) 308ea466794SJosef Bacik break; 309f188591eSChris Mason } 310ea466794SJosef Bacik 311c0901581SStefan Behrens if (failed && !ret && failed_mirror) 31220a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 313ea466794SJosef Bacik 314ea466794SJosef Bacik return ret; 315f188591eSChris Mason } 31619c00ddcSChris Mason 317eca0f6f6SQu Wenruo static int csum_one_extent_buffer(struct extent_buffer *eb) 318eca0f6f6SQu Wenruo { 319eca0f6f6SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 320eca0f6f6SQu Wenruo u8 result[BTRFS_CSUM_SIZE]; 321eca0f6f6SQu Wenruo int ret; 322eca0f6f6SQu Wenruo 323eca0f6f6SQu Wenruo ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 324eca0f6f6SQu Wenruo offsetof(struct btrfs_header, fsid), 325eca0f6f6SQu Wenruo BTRFS_FSID_SIZE) == 0); 326eca0f6f6SQu Wenruo csum_tree_block(eb, result); 327eca0f6f6SQu Wenruo 328eca0f6f6SQu Wenruo if (btrfs_header_level(eb)) 329eca0f6f6SQu Wenruo ret = btrfs_check_node(eb); 330eca0f6f6SQu Wenruo else 331eca0f6f6SQu Wenruo ret = btrfs_check_leaf_full(eb); 332eca0f6f6SQu Wenruo 3333777369fSQu Wenruo if (ret < 0) 3343777369fSQu Wenruo goto error; 3353777369fSQu Wenruo 3363777369fSQu Wenruo /* 3373777369fSQu Wenruo * Also check the generation, the eb reached here must be newer than 3383777369fSQu Wenruo * last committed. Or something seriously wrong happened. 3393777369fSQu Wenruo */ 3403777369fSQu Wenruo if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { 3413777369fSQu Wenruo ret = -EUCLEAN; 342eca0f6f6SQu Wenruo btrfs_err(fs_info, 3433777369fSQu Wenruo "block=%llu bad generation, have %llu expect > %llu", 3443777369fSQu Wenruo eb->start, btrfs_header_generation(eb), 3453777369fSQu Wenruo fs_info->last_trans_committed); 3463777369fSQu Wenruo goto error; 347eca0f6f6SQu Wenruo } 348eca0f6f6SQu Wenruo write_extent_buffer(eb, result, 0, fs_info->csum_size); 349eca0f6f6SQu Wenruo 350eca0f6f6SQu Wenruo return 0; 3513777369fSQu Wenruo 3523777369fSQu Wenruo error: 3533777369fSQu Wenruo btrfs_print_tree(eb, 0); 3543777369fSQu Wenruo btrfs_err(fs_info, "block=%llu write time tree block corruption detected", 3553777369fSQu Wenruo eb->start); 3563777369fSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 3573777369fSQu Wenruo return ret; 358eca0f6f6SQu Wenruo } 359eca0f6f6SQu Wenruo 360eca0f6f6SQu Wenruo /* Checksum all dirty extent buffers in one bio_vec */ 361eca0f6f6SQu Wenruo static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, 362eca0f6f6SQu Wenruo struct bio_vec *bvec) 363eca0f6f6SQu Wenruo { 364eca0f6f6SQu Wenruo struct page *page = bvec->bv_page; 365eca0f6f6SQu Wenruo u64 bvec_start = page_offset(page) + bvec->bv_offset; 366eca0f6f6SQu Wenruo u64 cur; 367eca0f6f6SQu Wenruo int ret = 0; 368eca0f6f6SQu Wenruo 369eca0f6f6SQu Wenruo for (cur = bvec_start; cur < bvec_start + bvec->bv_len; 370eca0f6f6SQu Wenruo cur += fs_info->nodesize) { 371eca0f6f6SQu Wenruo struct extent_buffer *eb; 372eca0f6f6SQu Wenruo bool uptodate; 373eca0f6f6SQu Wenruo 374eca0f6f6SQu Wenruo eb = find_extent_buffer(fs_info, cur); 375eca0f6f6SQu Wenruo uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, 376eca0f6f6SQu Wenruo fs_info->nodesize); 377eca0f6f6SQu Wenruo 37801cd3909SDavid Sterba /* A dirty eb shouldn't disappear from buffer_radix */ 379eca0f6f6SQu Wenruo if (WARN_ON(!eb)) 380eca0f6f6SQu Wenruo return -EUCLEAN; 381eca0f6f6SQu Wenruo 382eca0f6f6SQu Wenruo if (WARN_ON(cur != btrfs_header_bytenr(eb))) { 383eca0f6f6SQu Wenruo free_extent_buffer(eb); 384eca0f6f6SQu Wenruo return -EUCLEAN; 385eca0f6f6SQu Wenruo } 386eca0f6f6SQu Wenruo if (WARN_ON(!uptodate)) { 387eca0f6f6SQu Wenruo free_extent_buffer(eb); 388eca0f6f6SQu Wenruo return -EUCLEAN; 389eca0f6f6SQu Wenruo } 390eca0f6f6SQu Wenruo 391eca0f6f6SQu Wenruo ret = csum_one_extent_buffer(eb); 392eca0f6f6SQu Wenruo free_extent_buffer(eb); 393eca0f6f6SQu Wenruo if (ret < 0) 394eca0f6f6SQu Wenruo return ret; 395eca0f6f6SQu Wenruo } 396eca0f6f6SQu Wenruo return ret; 397eca0f6f6SQu Wenruo } 398eca0f6f6SQu Wenruo 399d352ac68SChris Mason /* 400ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 401ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 402ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 403d352ac68SChris Mason */ 404ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 40519c00ddcSChris Mason { 406ac303b69SQu Wenruo struct page *page = bvec->bv_page; 4074eee4fa4SMiao Xie u64 start = page_offset(page); 40819c00ddcSChris Mason u64 found_start; 40919c00ddcSChris Mason struct extent_buffer *eb; 410eca0f6f6SQu Wenruo 411fbca46ebSQu Wenruo if (fs_info->nodesize < PAGE_SIZE) 412eca0f6f6SQu Wenruo return csum_dirty_subpage_buffers(fs_info, bvec); 413f188591eSChris Mason 4144f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4154f2de97aSJosef Bacik if (page != eb->pages[0]) 4164f2de97aSJosef Bacik return 0; 4170f805531SAlex Lyakas 41819c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 419d3575156SNaohiro Aota 420d3575156SNaohiro Aota if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { 421d3575156SNaohiro Aota WARN_ON(found_start != 0); 422d3575156SNaohiro Aota return 0; 423d3575156SNaohiro Aota } 424d3575156SNaohiro Aota 4250f805531SAlex Lyakas /* 4260f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 4270f805531SAlex Lyakas * It is useful to know what went wrong. 4280f805531SAlex Lyakas */ 4290f805531SAlex Lyakas if (WARN_ON(found_start != start)) 4300f805531SAlex Lyakas return -EUCLEAN; 4310f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 4320f805531SAlex Lyakas return -EUCLEAN; 4330f805531SAlex Lyakas 434eca0f6f6SQu Wenruo return csum_one_extent_buffer(eb); 43519c00ddcSChris Mason } 43619c00ddcSChris Mason 437b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 4382b82032cSYan Zheng { 439b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 440944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 44144880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 442944d3f9fSNikolay Borisov u8 *metadata_uuid; 4432b82032cSYan Zheng 4449a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 4459a8658e3SDavid Sterba BTRFS_FSID_SIZE); 4467239ff4bSNikolay Borisov /* 447944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 448944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 449944d3f9fSNikolay Borisov * ->fsid in this case is fine 4507239ff4bSNikolay Borisov */ 451944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 4527239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 4537239ff4bSNikolay Borisov else 4547239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 4557239ff4bSNikolay Borisov 456944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 457944d3f9fSNikolay Borisov return 0; 458944d3f9fSNikolay Borisov 459944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 460944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 461944d3f9fSNikolay Borisov return 0; 462944d3f9fSNikolay Borisov 463944d3f9fSNikolay Borisov return 1; 4642b82032cSYan Zheng } 4652b82032cSYan Zheng 46677bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 46777bf40a2SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb) 468ce9adaa5SChris Mason { 46977bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 470ce9adaa5SChris Mason u64 found_start; 47177bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 47277bf40a2SQu Wenruo u8 found_level; 4732996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 474dfd29eedSDavid Sterba const u8 *header_csum; 47577bf40a2SQu Wenruo int ret = 0; 476ea466794SJosef Bacik 477ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 478727011e0SChris Mason if (found_start != eb->start) { 4798f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, 4808f0ed7d4SQu Wenruo "bad tree block start, mirror %u want %llu have %llu", 4818f0ed7d4SQu Wenruo eb->read_mirror, eb->start, found_start); 482f188591eSChris Mason ret = -EIO; 48377bf40a2SQu Wenruo goto out; 484ce9adaa5SChris Mason } 485b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 4868f0ed7d4SQu Wenruo btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", 4878f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 4881259ab75SChris Mason ret = -EIO; 48977bf40a2SQu Wenruo goto out; 4901259ab75SChris Mason } 491ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 4921c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 4938f0ed7d4SQu Wenruo btrfs_err(fs_info, 4948f0ed7d4SQu Wenruo "bad tree block level, mirror %u level %d on logical %llu", 4958f0ed7d4SQu Wenruo eb->read_mirror, btrfs_header_level(eb), eb->start); 4961c24c3ceSJosef Bacik ret = -EIO; 49777bf40a2SQu Wenruo goto out; 4981c24c3ceSJosef Bacik } 499ce9adaa5SChris Mason 500c67b3892SDavid Sterba csum_tree_block(eb, result); 501dfd29eedSDavid Sterba header_csum = page_address(eb->pages[0]) + 502dfd29eedSDavid Sterba get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); 503a826d6dcSJosef Bacik 504dfd29eedSDavid Sterba if (memcmp(result, header_csum, csum_size) != 0) { 5052996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 5068f0ed7d4SQu Wenruo "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", 5078f0ed7d4SQu Wenruo eb->start, eb->read_mirror, 508dfd29eedSDavid Sterba CSUM_FMT_VALUE(csum_size, header_csum), 50935be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 51035be8851SJohannes Thumshirn btrfs_header_level(eb)); 5112996e1f8SJohannes Thumshirn ret = -EUCLEAN; 51277bf40a2SQu Wenruo goto out; 5132996e1f8SJohannes Thumshirn } 5142996e1f8SJohannes Thumshirn 515a826d6dcSJosef Bacik /* 516a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 517a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 518a826d6dcSJosef Bacik * return -EIO. 519a826d6dcSJosef Bacik */ 5201c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 521a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 522a826d6dcSJosef Bacik ret = -EIO; 523a826d6dcSJosef Bacik } 524ce9adaa5SChris Mason 525813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 526053ab70fSLiu Bo ret = -EIO; 527053ab70fSLiu Bo 5280b32f4bbSJosef Bacik if (!ret) 5290b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 53075391f0dSQu Wenruo else 53175391f0dSQu Wenruo btrfs_err(fs_info, 5328f0ed7d4SQu Wenruo "read time tree block corruption detected on logical %llu mirror %u", 5338f0ed7d4SQu Wenruo eb->start, eb->read_mirror); 53477bf40a2SQu Wenruo out: 53577bf40a2SQu Wenruo return ret; 53677bf40a2SQu Wenruo } 53777bf40a2SQu Wenruo 538371cdc07SQu Wenruo static int validate_subpage_buffer(struct page *page, u64 start, u64 end, 539371cdc07SQu Wenruo int mirror) 540371cdc07SQu Wenruo { 541371cdc07SQu Wenruo struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); 542371cdc07SQu Wenruo struct extent_buffer *eb; 543371cdc07SQu Wenruo bool reads_done; 544371cdc07SQu Wenruo int ret = 0; 545371cdc07SQu Wenruo 546371cdc07SQu Wenruo /* 547371cdc07SQu Wenruo * We don't allow bio merge for subpage metadata read, so we should 548371cdc07SQu Wenruo * only get one eb for each endio hook. 549371cdc07SQu Wenruo */ 550371cdc07SQu Wenruo ASSERT(end == start + fs_info->nodesize - 1); 551371cdc07SQu Wenruo ASSERT(PagePrivate(page)); 552371cdc07SQu Wenruo 553371cdc07SQu Wenruo eb = find_extent_buffer(fs_info, start); 554371cdc07SQu Wenruo /* 555371cdc07SQu Wenruo * When we are reading one tree block, eb must have been inserted into 556371cdc07SQu Wenruo * the radix tree. If not, something is wrong. 557371cdc07SQu Wenruo */ 558371cdc07SQu Wenruo ASSERT(eb); 559371cdc07SQu Wenruo 560371cdc07SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 561371cdc07SQu Wenruo /* Subpage read must finish in page read */ 562371cdc07SQu Wenruo ASSERT(reads_done); 563371cdc07SQu Wenruo 564371cdc07SQu Wenruo eb->read_mirror = mirror; 565371cdc07SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 566371cdc07SQu Wenruo ret = -EIO; 567371cdc07SQu Wenruo goto err; 568371cdc07SQu Wenruo } 569371cdc07SQu Wenruo ret = validate_extent_buffer(eb); 570371cdc07SQu Wenruo if (ret < 0) 571371cdc07SQu Wenruo goto err; 572371cdc07SQu Wenruo 573371cdc07SQu Wenruo set_extent_buffer_uptodate(eb); 574371cdc07SQu Wenruo 575371cdc07SQu Wenruo free_extent_buffer(eb); 576371cdc07SQu Wenruo return ret; 577371cdc07SQu Wenruo err: 578371cdc07SQu Wenruo /* 579371cdc07SQu Wenruo * end_bio_extent_readpage decrements io_pages in case of error, 580371cdc07SQu Wenruo * make sure it has something to decrement. 581371cdc07SQu Wenruo */ 582371cdc07SQu Wenruo atomic_inc(&eb->io_pages); 583371cdc07SQu Wenruo clear_extent_buffer_uptodate(eb); 584371cdc07SQu Wenruo free_extent_buffer(eb); 585371cdc07SQu Wenruo return ret; 586371cdc07SQu Wenruo } 587371cdc07SQu Wenruo 588c3a3b19bSQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, 58977bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 59077bf40a2SQu Wenruo int mirror) 59177bf40a2SQu Wenruo { 59277bf40a2SQu Wenruo struct extent_buffer *eb; 59377bf40a2SQu Wenruo int ret = 0; 59477bf40a2SQu Wenruo int reads_done; 59577bf40a2SQu Wenruo 59677bf40a2SQu Wenruo ASSERT(page->private); 597371cdc07SQu Wenruo 598fbca46ebSQu Wenruo if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) 599371cdc07SQu Wenruo return validate_subpage_buffer(page, start, end, mirror); 600371cdc07SQu Wenruo 60177bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 60277bf40a2SQu Wenruo 60377bf40a2SQu Wenruo /* 60477bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 60577bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 60677bf40a2SQu Wenruo */ 60777bf40a2SQu Wenruo atomic_inc(&eb->refs); 60877bf40a2SQu Wenruo 60977bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 61077bf40a2SQu Wenruo if (!reads_done) 61177bf40a2SQu Wenruo goto err; 61277bf40a2SQu Wenruo 61377bf40a2SQu Wenruo eb->read_mirror = mirror; 61477bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 61577bf40a2SQu Wenruo ret = -EIO; 61677bf40a2SQu Wenruo goto err; 61777bf40a2SQu Wenruo } 61877bf40a2SQu Wenruo ret = validate_extent_buffer(eb); 619ce9adaa5SChris Mason err: 62053b381b3SDavid Woodhouse if (ret) { 62153b381b3SDavid Woodhouse /* 62253b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 62353b381b3SDavid Woodhouse * again, we have to make sure it has something 62453b381b3SDavid Woodhouse * to decrement 62553b381b3SDavid Woodhouse */ 62653b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6270b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 62853b381b3SDavid Woodhouse } 6290b32f4bbSJosef Bacik free_extent_buffer(eb); 63077bf40a2SQu Wenruo 631f188591eSChris Mason return ret; 632ce9adaa5SChris Mason } 633ce9adaa5SChris Mason 6344a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 6354a69a410SChris Mason { 6364a69a410SChris Mason struct async_submit_bio *async; 6374e4cbee9SChristoph Hellwig blk_status_t ret; 6384a69a410SChris Mason 6394a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 6401941b64bSQu Wenruo ret = async->submit_bio_start(async->inode, async->bio, 6411941b64bSQu Wenruo async->dio_file_offset); 64279787eaaSJeff Mahoney if (ret) 6434e4cbee9SChristoph Hellwig async->status = ret; 6444a69a410SChris Mason } 6454a69a410SChris Mason 64606ea01b1SDavid Sterba /* 64706ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 64806ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 64906ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 65006ea01b1SDavid Sterba * 65106ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 65206ea01b1SDavid Sterba * inserted into the tree. 65306ea01b1SDavid Sterba */ 6544a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 6558b712842SChris Mason { 656917f32a2SChristoph Hellwig struct async_submit_bio *async = 657917f32a2SChristoph Hellwig container_of(work, struct async_submit_bio, work); 658917f32a2SChristoph Hellwig struct inode *inode = async->inode; 659917f32a2SChristoph Hellwig struct btrfs_bio *bbio = btrfs_bio(async->bio); 6604854ddd0SChris Mason 661bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 6624e4cbee9SChristoph Hellwig if (async->status) { 663917f32a2SChristoph Hellwig btrfs_bio_end_io(bbio, async->status); 66479787eaaSJeff Mahoney return; 66579787eaaSJeff Mahoney } 66679787eaaSJeff Mahoney 667ec39f769SChris Mason /* 668ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 669ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 670ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 671ec39f769SChris Mason */ 672ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 6731a722d8fSChristoph Hellwig btrfs_submit_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 6744a69a410SChris Mason } 6754a69a410SChris Mason 6764a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 6774a69a410SChris Mason { 6784a69a410SChris Mason struct async_submit_bio *async; 6794a69a410SChris Mason 6804a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 6818b712842SChris Mason kfree(async); 6828b712842SChris Mason } 6838b712842SChris Mason 684ea1f0cedSChristoph Hellwig /* 685ea1f0cedSChristoph Hellwig * Submit bio to an async queue. 686ea1f0cedSChristoph Hellwig * 687ea1f0cedSChristoph Hellwig * Retrun: 688ea1f0cedSChristoph Hellwig * - true if the work has been succesfuly submitted 689ea1f0cedSChristoph Hellwig * - false in case of error 690ea1f0cedSChristoph Hellwig */ 691ea1f0cedSChristoph Hellwig bool btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, int mirror_num, 692ea1f0cedSChristoph Hellwig u64 dio_file_offset, 693e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 69444b8bd7eSChris Mason { 6958896a08dSQu Wenruo struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 69644b8bd7eSChris Mason struct async_submit_bio *async; 69744b8bd7eSChris Mason 69844b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 69944b8bd7eSChris Mason if (!async) 700ea1f0cedSChristoph Hellwig return false; 70144b8bd7eSChris Mason 7028896a08dSQu Wenruo async->inode = inode; 70344b8bd7eSChris Mason async->bio = bio; 70444b8bd7eSChris Mason async->mirror_num = mirror_num; 7054a69a410SChris Mason async->submit_bio_start = submit_bio_start; 7064a69a410SChris Mason 707a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 708a0cac0ecSOmar Sandoval run_one_async_free); 7094a69a410SChris Mason 7101941b64bSQu Wenruo async->dio_file_offset = dio_file_offset; 7118c8bee1dSChris Mason 7124e4cbee9SChristoph Hellwig async->status = 0; 71379787eaaSJeff Mahoney 71467f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 715a31b4a43SChristoph Hellwig btrfs_queue_work(fs_info->hipri_workers, &async->work); 716a31b4a43SChristoph Hellwig else 7175cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 718ea1f0cedSChristoph Hellwig return true; 71944b8bd7eSChris Mason } 72044b8bd7eSChris Mason 7214e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 722ce3ed71aSChris Mason { 7232c30c71bSKent Overstreet struct bio_vec *bvec; 724ce3ed71aSChris Mason struct btrfs_root *root; 7252b070cfeSChristoph Hellwig int ret = 0; 7266dc4f100SMing Lei struct bvec_iter_all iter_all; 727ce3ed71aSChris Mason 728c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 7292b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 730ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 731ac303b69SQu Wenruo ret = csum_dirty_buffer(root->fs_info, bvec); 73279787eaaSJeff Mahoney if (ret) 73379787eaaSJeff Mahoney break; 734ce3ed71aSChris Mason } 7352c30c71bSKent Overstreet 7364e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 737ce3ed71aSChris Mason } 738ce3ed71aSChris Mason 7398896a08dSQu Wenruo static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, 7401941b64bSQu Wenruo u64 dio_file_offset) 74122c59948SChris Mason { 7428b712842SChris Mason /* 7438b712842SChris Mason * when we're called for a write, we're already in the async 7441a722d8fSChristoph Hellwig * submission context. Just jump into btrfs_submit_bio. 7458b712842SChris Mason */ 74679787eaaSJeff Mahoney return btree_csum_one_bio(bio); 74722c59948SChris Mason } 74822c59948SChris Mason 749f4dcfb30SJohannes Thumshirn static bool should_async_write(struct btrfs_fs_info *fs_info, 7509b4e675aSDavid Sterba struct btrfs_inode *bi) 751de0022b9SJosef Bacik { 7524eef29efSNaohiro Aota if (btrfs_is_zoned(fs_info)) 753f4dcfb30SJohannes Thumshirn return false; 7546300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 755f4dcfb30SJohannes Thumshirn return false; 7569b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 757f4dcfb30SJohannes Thumshirn return false; 758f4dcfb30SJohannes Thumshirn return true; 759de0022b9SJosef Bacik } 760de0022b9SJosef Bacik 76194d9e11bSChristoph Hellwig void btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, int mirror_num) 76244b8bd7eSChris Mason { 7630b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 764917f32a2SChristoph Hellwig struct btrfs_bio *bbio = btrfs_bio(bio); 7654e4cbee9SChristoph Hellwig blk_status_t ret; 766cad321adSChris Mason 76708a6f464SChristoph Hellwig bio->bi_opf |= REQ_META; 76808a6f464SChristoph Hellwig 769cfe94440SNaohiro Aota if (btrfs_op(bio) != BTRFS_MAP_WRITE) { 7701a722d8fSChristoph Hellwig btrfs_submit_bio(fs_info, bio, mirror_num); 7711a722d8fSChristoph Hellwig return; 77244b8bd7eSChris Mason } 77344b8bd7eSChris Mason 774ea1f0cedSChristoph Hellwig /* 775ea1f0cedSChristoph Hellwig * Kthread helpers are used to submit writes so that checksumming can 776ea1f0cedSChristoph Hellwig * happen in parallel across all CPUs. 777ea1f0cedSChristoph Hellwig */ 778ea1f0cedSChristoph Hellwig if (should_async_write(fs_info, BTRFS_I(inode)) && 779ea1f0cedSChristoph Hellwig btrfs_wq_submit_bio(inode, bio, mirror_num, 0, btree_submit_bio_start)) 780ea1f0cedSChristoph Hellwig return; 781ea1f0cedSChristoph Hellwig 782ea1f0cedSChristoph Hellwig ret = btree_csum_one_bio(bio); 78394d9e11bSChristoph Hellwig if (ret) { 784917f32a2SChristoph Hellwig btrfs_bio_end_io(bbio, ret); 785ea1f0cedSChristoph Hellwig return; 78694d9e11bSChristoph Hellwig } 787ea1f0cedSChristoph Hellwig 788ea1f0cedSChristoph Hellwig btrfs_submit_bio(fs_info, bio, mirror_num); 78961891923SStefan Behrens } 79061891923SStefan Behrens 7913dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 7928958b551SMatthew Wilcox (Oracle) static int btree_migrate_folio(struct address_space *mapping, 7938958b551SMatthew Wilcox (Oracle) struct folio *dst, struct folio *src, enum migrate_mode mode) 794784b4e29SChris Mason { 795784b4e29SChris Mason /* 796784b4e29SChris Mason * we can't safely write a btree page from here, 797784b4e29SChris Mason * we haven't done the locking hook 798784b4e29SChris Mason */ 7998958b551SMatthew Wilcox (Oracle) if (folio_test_dirty(src)) 800784b4e29SChris Mason return -EAGAIN; 801784b4e29SChris Mason /* 802784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 803784b4e29SChris Mason * We must have no buffers or drop them. 804784b4e29SChris Mason */ 8058958b551SMatthew Wilcox (Oracle) if (folio_get_private(src) && 8068958b551SMatthew Wilcox (Oracle) !filemap_release_folio(src, GFP_KERNEL)) 807784b4e29SChris Mason return -EAGAIN; 80854184650SMatthew Wilcox (Oracle) return migrate_folio(mapping, dst, src, mode); 809784b4e29SChris Mason } 8108958b551SMatthew Wilcox (Oracle) #else 8118958b551SMatthew Wilcox (Oracle) #define btree_migrate_folio NULL 8123dd1462eSJan Beulich #endif 813784b4e29SChris Mason 8140da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 8150da5468fSChris Mason struct writeback_control *wbc) 8160da5468fSChris Mason { 817e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 818e2d84521SMiao Xie int ret; 819e2d84521SMiao Xie 820d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 821448d640bSChris Mason 822448d640bSChris Mason if (wbc->for_kupdate) 823448d640bSChris Mason return 0; 824448d640bSChris Mason 825e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 826b9473439SChris Mason /* this is a bit racy, but that's ok */ 827d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 828d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 829d814a491SEthan Lien fs_info->dirty_metadata_batch); 830e2d84521SMiao Xie if (ret < 0) 831793955bcSChris Mason return 0; 832793955bcSChris Mason } 8330b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 8340da5468fSChris Mason } 8350da5468fSChris Mason 836f913cff3SMatthew Wilcox (Oracle) static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) 8375f39d397SChris Mason { 838f913cff3SMatthew Wilcox (Oracle) if (folio_test_writeback(folio) || folio_test_dirty(folio)) 839f913cff3SMatthew Wilcox (Oracle) return false; 8400c4e538bSDavid Sterba 841f913cff3SMatthew Wilcox (Oracle) return try_release_extent_buffer(&folio->page); 842d98237b3SChris Mason } 843d98237b3SChris Mason 844895586ebSMatthew Wilcox (Oracle) static void btree_invalidate_folio(struct folio *folio, size_t offset, 845895586ebSMatthew Wilcox (Oracle) size_t length) 846d98237b3SChris Mason { 847d1310b2eSChris Mason struct extent_io_tree *tree; 848895586ebSMatthew Wilcox (Oracle) tree = &BTRFS_I(folio->mapping->host)->io_tree; 849895586ebSMatthew Wilcox (Oracle) extent_invalidate_folio(tree, folio, offset); 850f913cff3SMatthew Wilcox (Oracle) btree_release_folio(folio, GFP_NOFS); 851895586ebSMatthew Wilcox (Oracle) if (folio_get_private(folio)) { 852895586ebSMatthew Wilcox (Oracle) btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, 853895586ebSMatthew Wilcox (Oracle) "folio private not zero on folio %llu", 854895586ebSMatthew Wilcox (Oracle) (unsigned long long)folio_pos(folio)); 855895586ebSMatthew Wilcox (Oracle) folio_detach_private(folio); 8569ad6b7bcSChris Mason } 857d98237b3SChris Mason } 858d98237b3SChris Mason 859bb146eb2SJosef Bacik #ifdef DEBUG 8600079c3b1SMatthew Wilcox (Oracle) static bool btree_dirty_folio(struct address_space *mapping, 8610079c3b1SMatthew Wilcox (Oracle) struct folio *folio) 8620079c3b1SMatthew Wilcox (Oracle) { 8630079c3b1SMatthew Wilcox (Oracle) struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); 864139e8cd3SQu Wenruo struct btrfs_subpage *subpage; 8650b32f4bbSJosef Bacik struct extent_buffer *eb; 866139e8cd3SQu Wenruo int cur_bit = 0; 8670079c3b1SMatthew Wilcox (Oracle) u64 page_start = folio_pos(folio); 8680b32f4bbSJosef Bacik 869139e8cd3SQu Wenruo if (fs_info->sectorsize == PAGE_SIZE) { 8700079c3b1SMatthew Wilcox (Oracle) eb = folio_get_private(folio); 8710b32f4bbSJosef Bacik BUG_ON(!eb); 8720b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 8730b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 87449d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 8750079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 876139e8cd3SQu Wenruo } 8770079c3b1SMatthew Wilcox (Oracle) subpage = folio_get_private(folio); 878139e8cd3SQu Wenruo 879139e8cd3SQu Wenruo ASSERT(subpage->dirty_bitmap); 880139e8cd3SQu Wenruo while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { 881139e8cd3SQu Wenruo unsigned long flags; 882139e8cd3SQu Wenruo u64 cur; 883139e8cd3SQu Wenruo u16 tmp = (1 << cur_bit); 884139e8cd3SQu Wenruo 885139e8cd3SQu Wenruo spin_lock_irqsave(&subpage->lock, flags); 886139e8cd3SQu Wenruo if (!(tmp & subpage->dirty_bitmap)) { 887139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 888139e8cd3SQu Wenruo cur_bit++; 889139e8cd3SQu Wenruo continue; 890139e8cd3SQu Wenruo } 891139e8cd3SQu Wenruo spin_unlock_irqrestore(&subpage->lock, flags); 892139e8cd3SQu Wenruo cur = page_start + cur_bit * fs_info->sectorsize; 893139e8cd3SQu Wenruo 894139e8cd3SQu Wenruo eb = find_extent_buffer(fs_info, cur); 895139e8cd3SQu Wenruo ASSERT(eb); 896139e8cd3SQu Wenruo ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 897139e8cd3SQu Wenruo ASSERT(atomic_read(&eb->refs)); 89849d0c642SFilipe Manana btrfs_assert_tree_write_locked(eb); 899139e8cd3SQu Wenruo free_extent_buffer(eb); 900139e8cd3SQu Wenruo 901139e8cd3SQu Wenruo cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); 902139e8cd3SQu Wenruo } 9030079c3b1SMatthew Wilcox (Oracle) return filemap_dirty_folio(mapping, folio); 9040b32f4bbSJosef Bacik } 9050079c3b1SMatthew Wilcox (Oracle) #else 9060079c3b1SMatthew Wilcox (Oracle) #define btree_dirty_folio filemap_dirty_folio 9070079c3b1SMatthew Wilcox (Oracle) #endif 9080b32f4bbSJosef Bacik 9097f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 9100da5468fSChris Mason .writepages = btree_writepages, 911f913cff3SMatthew Wilcox (Oracle) .release_folio = btree_release_folio, 912895586ebSMatthew Wilcox (Oracle) .invalidate_folio = btree_invalidate_folio, 9138958b551SMatthew Wilcox (Oracle) .migrate_folio = btree_migrate_folio, 9140079c3b1SMatthew Wilcox (Oracle) .dirty_folio = btree_dirty_folio, 915d98237b3SChris Mason }; 916123abc88SChris Mason 9172ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 9182ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 9193fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 9203fbaf258SJosef Bacik int level) 9210999df54SChris Mason { 9220b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 9230b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 9243fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 9250999df54SChris Mason } 9260999df54SChris Mason 927581c1760SQu Wenruo /* 928581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 929581c1760SQu Wenruo * verification. 930581c1760SQu Wenruo * 9311b7ec85eSJosef Bacik * @owner_root: the objectid of the root owner for this block. 932581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 933581c1760SQu Wenruo * @level: expected level, mandatory check 934581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 935581c1760SQu Wenruo */ 9362ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 9371b7ec85eSJosef Bacik u64 owner_root, u64 parent_transid, 9381b7ec85eSJosef Bacik int level, struct btrfs_key *first_key) 939e20d96d6SChris Mason { 9405f39d397SChris Mason struct extent_buffer *buf = NULL; 94119c00ddcSChris Mason int ret; 94219c00ddcSChris Mason 9433fbaf258SJosef Bacik buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); 944c871b0f2SLiu Bo if (IS_ERR(buf)) 945c871b0f2SLiu Bo return buf; 946e4204dedSChris Mason 9476a2e9dc4SFilipe Manana ret = btrfs_read_extent_buffer(buf, parent_transid, level, first_key); 9480f0fe8f7SFilipe David Borba Manana if (ret) { 949537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 95064c043deSLiu Bo return ERR_PTR(ret); 9510f0fe8f7SFilipe David Borba Manana } 95288c602abSQu Wenruo if (btrfs_check_eb_owner(buf, owner_root)) { 95388c602abSQu Wenruo free_extent_buffer_stale(buf); 95488c602abSQu Wenruo return ERR_PTR(-EUCLEAN); 95588c602abSQu Wenruo } 9565f39d397SChris Mason return buf; 957ce9adaa5SChris Mason 958eb60ceacSChris Mason } 959eb60ceacSChris Mason 9606a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 961ed2ff2cbSChris Mason { 9626a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 96355c69072SChris Mason if (btrfs_header_generation(buf) == 964e2d84521SMiao Xie fs_info->running_transaction->transid) { 96549d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 966b4ce94deSChris Mason 967b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 968104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 969e2d84521SMiao Xie -buf->len, 970e2d84521SMiao Xie fs_info->dirty_metadata_batch); 9710b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 972925baeddSChris Mason } 9735f39d397SChris Mason } 974ed7b63ebSJosef Bacik } 9755f39d397SChris Mason 976da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 977e20d96d6SChris Mason u64 objectid) 978d97e63b6SChris Mason { 9797c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 9802e608bd1SJosef Bacik 9812e608bd1SJosef Bacik memset(&root->root_key, 0, sizeof(root->root_key)); 9822e608bd1SJosef Bacik memset(&root->root_item, 0, sizeof(root->root_item)); 9832e608bd1SJosef Bacik memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 98496dfcb46SJosef Bacik root->fs_info = fs_info; 9852e608bd1SJosef Bacik root->root_key.objectid = objectid; 986cfaa7295SChris Mason root->node = NULL; 987a28ec197SChris Mason root->commit_root = NULL; 98827cdeb70SMiao Xie root->state = 0; 989abed4aaaSJosef Bacik RB_CLEAR_NODE(&root->rb_node); 9900b86a832SChris Mason 9910f7d52f4SChris Mason root->last_trans = 0; 9926b8fad57SNikolay Borisov root->free_objectid = 0; 993eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 994199c2a9cSMiao Xie root->nr_ordered_extents = 0; 9956bef4d31SEric Paris root->inode_tree = RB_ROOT; 996088aea3bSDavid Sterba INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 9972e608bd1SJosef Bacik 9982e608bd1SJosef Bacik btrfs_init_root_block_rsv(root); 9990b86a832SChris Mason 10000b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 10015d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1002eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1003eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1004199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1005199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1006d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 10072ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 10082ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 10095d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1010eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1011199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1012f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 10132ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 10142ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 10158287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1016a2135011SChris Mason mutex_init(&root->objectid_mutex); 1017e02119d5SChris Mason mutex_init(&root->log_mutex); 101831f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1019573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1020c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 10217237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 10227237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 10237237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 10248b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 10258b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 10267237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 10277237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 10287237f183SYan Zheng atomic_set(&root->log_writers, 0); 10292ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 10300700cea7SElena Reshetova refcount_set(&root->refs, 1); 10318ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1032eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 10337237f183SYan Zheng root->log_transid = 0; 1034d1433debSMiao Xie root->log_transid_committed = -1; 1035257c62e1SChris Mason root->last_log_commit = 0; 10362e608bd1SJosef Bacik root->anon_dev = 0; 1037e289f03eSFilipe Manana if (!dummy) { 103843eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 103943eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1040e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1041e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1042e289f03eSFilipe Manana } 1043017e5369SChris Mason 10445f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1045370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1046bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1047bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1048fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1049bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1050fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1051bd647ce3SJosef Bacik #endif 10523768f368SChris Mason } 10533768f368SChris Mason 105474e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 105596dfcb46SJosef Bacik u64 objectid, gfp_t flags) 10566f07e42eSAl Viro { 105774e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 10586f07e42eSAl Viro if (root) 105996dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 10606f07e42eSAl Viro return root; 10616f07e42eSAl Viro } 10626f07e42eSAl Viro 106306ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 106406ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1065da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 106606ea65a3SJosef Bacik { 106706ea65a3SJosef Bacik struct btrfs_root *root; 106806ea65a3SJosef Bacik 10697c0260eeSJeff Mahoney if (!fs_info) 10707c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 10717c0260eeSJeff Mahoney 107296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 107306ea65a3SJosef Bacik if (!root) 107406ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1075da17066cSJeff Mahoney 1076b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1077faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 107806ea65a3SJosef Bacik 107906ea65a3SJosef Bacik return root; 108006ea65a3SJosef Bacik } 108106ea65a3SJosef Bacik #endif 108206ea65a3SJosef Bacik 1083abed4aaaSJosef Bacik static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) 1084abed4aaaSJosef Bacik { 1085abed4aaaSJosef Bacik const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); 1086abed4aaaSJosef Bacik const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); 1087abed4aaaSJosef Bacik 1088abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); 1089abed4aaaSJosef Bacik } 1090abed4aaaSJosef Bacik 1091abed4aaaSJosef Bacik static int global_root_key_cmp(const void *k, const struct rb_node *node) 1092abed4aaaSJosef Bacik { 1093abed4aaaSJosef Bacik const struct btrfs_key *key = k; 1094abed4aaaSJosef Bacik const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); 1095abed4aaaSJosef Bacik 1096abed4aaaSJosef Bacik return btrfs_comp_cpu_keys(key, &root->root_key); 1097abed4aaaSJosef Bacik } 1098abed4aaaSJosef Bacik 1099abed4aaaSJosef Bacik int btrfs_global_root_insert(struct btrfs_root *root) 1100abed4aaaSJosef Bacik { 1101abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1102abed4aaaSJosef Bacik struct rb_node *tmp; 1103abed4aaaSJosef Bacik 1104abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1105abed4aaaSJosef Bacik tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); 1106abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1107abed4aaaSJosef Bacik ASSERT(!tmp); 1108abed4aaaSJosef Bacik 1109abed4aaaSJosef Bacik return tmp ? -EEXIST : 0; 1110abed4aaaSJosef Bacik } 1111abed4aaaSJosef Bacik 1112abed4aaaSJosef Bacik void btrfs_global_root_delete(struct btrfs_root *root) 1113abed4aaaSJosef Bacik { 1114abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = root->fs_info; 1115abed4aaaSJosef Bacik 1116abed4aaaSJosef Bacik write_lock(&fs_info->global_root_lock); 1117abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1118abed4aaaSJosef Bacik write_unlock(&fs_info->global_root_lock); 1119abed4aaaSJosef Bacik } 1120abed4aaaSJosef Bacik 1121abed4aaaSJosef Bacik struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, 1122abed4aaaSJosef Bacik struct btrfs_key *key) 1123abed4aaaSJosef Bacik { 1124abed4aaaSJosef Bacik struct rb_node *node; 1125abed4aaaSJosef Bacik struct btrfs_root *root = NULL; 1126abed4aaaSJosef Bacik 1127abed4aaaSJosef Bacik read_lock(&fs_info->global_root_lock); 1128abed4aaaSJosef Bacik node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); 1129abed4aaaSJosef Bacik if (node) 1130abed4aaaSJosef Bacik root = container_of(node, struct btrfs_root, rb_node); 1131abed4aaaSJosef Bacik read_unlock(&fs_info->global_root_lock); 1132abed4aaaSJosef Bacik 1133abed4aaaSJosef Bacik return root; 1134abed4aaaSJosef Bacik } 1135abed4aaaSJosef Bacik 1136f7238e50SJosef Bacik static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) 1137f7238e50SJosef Bacik { 1138f7238e50SJosef Bacik struct btrfs_block_group *block_group; 1139f7238e50SJosef Bacik u64 ret; 1140f7238e50SJosef Bacik 1141f7238e50SJosef Bacik if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) 1142f7238e50SJosef Bacik return 0; 1143f7238e50SJosef Bacik 1144f7238e50SJosef Bacik if (bytenr) 1145f7238e50SJosef Bacik block_group = btrfs_lookup_block_group(fs_info, bytenr); 1146f7238e50SJosef Bacik else 1147f7238e50SJosef Bacik block_group = btrfs_lookup_first_block_group(fs_info, bytenr); 1148f7238e50SJosef Bacik ASSERT(block_group); 1149f7238e50SJosef Bacik if (!block_group) 1150f7238e50SJosef Bacik return 0; 1151f7238e50SJosef Bacik ret = block_group->global_root_id; 1152f7238e50SJosef Bacik btrfs_put_block_group(block_group); 1153f7238e50SJosef Bacik 1154f7238e50SJosef Bacik return ret; 1155f7238e50SJosef Bacik } 1156f7238e50SJosef Bacik 1157abed4aaaSJosef Bacik struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1158abed4aaaSJosef Bacik { 1159abed4aaaSJosef Bacik struct btrfs_key key = { 1160abed4aaaSJosef Bacik .objectid = BTRFS_CSUM_TREE_OBJECTID, 1161abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1162f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1163abed4aaaSJosef Bacik }; 1164abed4aaaSJosef Bacik 1165abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1166abed4aaaSJosef Bacik } 1167abed4aaaSJosef Bacik 1168abed4aaaSJosef Bacik struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) 1169abed4aaaSJosef Bacik { 1170abed4aaaSJosef Bacik struct btrfs_key key = { 1171abed4aaaSJosef Bacik .objectid = BTRFS_EXTENT_TREE_OBJECTID, 1172abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1173f7238e50SJosef Bacik .offset = btrfs_global_root_id(fs_info, bytenr), 1174abed4aaaSJosef Bacik }; 1175abed4aaaSJosef Bacik 1176abed4aaaSJosef Bacik return btrfs_global_root(fs_info, &key); 1177abed4aaaSJosef Bacik } 1178abed4aaaSJosef Bacik 117951129b33SJosef Bacik struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) 118051129b33SJosef Bacik { 118151129b33SJosef Bacik if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) 118251129b33SJosef Bacik return fs_info->block_group_root; 118351129b33SJosef Bacik return btrfs_extent_root(fs_info, 0); 118451129b33SJosef Bacik } 118551129b33SJosef Bacik 118620897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 118720897f5cSArne Jansen u64 objectid) 118820897f5cSArne Jansen { 11899b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 119020897f5cSArne Jansen struct extent_buffer *leaf; 119120897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 119220897f5cSArne Jansen struct btrfs_root *root; 119320897f5cSArne Jansen struct btrfs_key key; 1194b89f6d1fSFilipe Manana unsigned int nofs_flag; 119520897f5cSArne Jansen int ret = 0; 119620897f5cSArne Jansen 1197b89f6d1fSFilipe Manana /* 1198b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1199b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1200b89f6d1fSFilipe Manana */ 1201b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 120296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1203b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 120420897f5cSArne Jansen if (!root) 120520897f5cSArne Jansen return ERR_PTR(-ENOMEM); 120620897f5cSArne Jansen 120720897f5cSArne Jansen root->root_key.objectid = objectid; 120820897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 120920897f5cSArne Jansen root->root_key.offset = 0; 121020897f5cSArne Jansen 12119631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 12129631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 121320897f5cSArne Jansen if (IS_ERR(leaf)) { 121420897f5cSArne Jansen ret = PTR_ERR(leaf); 12151dd05682STsutomu Itoh leaf = NULL; 1216c1b07854SPeng Hao goto fail; 121720897f5cSArne Jansen } 121820897f5cSArne Jansen 121920897f5cSArne Jansen root->node = leaf; 122020897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 122120897f5cSArne Jansen 122220897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 122327cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 122420897f5cSArne Jansen 1225f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1226f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 122720897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 122820897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 122920897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 123020897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 123120897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 123220897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 123320897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 123433d85fdaSQu Wenruo if (is_fstree(objectid)) 1235807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1236807fc790SAndy Shevchenko else 1237807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1238c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 123920897f5cSArne Jansen 12408a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 12418a6a87cdSBoris Burkov 124220897f5cSArne Jansen key.objectid = objectid; 124320897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 124420897f5cSArne Jansen key.offset = 0; 124520897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 124620897f5cSArne Jansen if (ret) 124720897f5cSArne Jansen goto fail; 124820897f5cSArne Jansen 124920897f5cSArne Jansen return root; 12501dd05682STsutomu Itoh 12518a6a87cdSBoris Burkov fail: 125200246528SJosef Bacik btrfs_put_root(root); 12531dd05682STsutomu Itoh 12541dd05682STsutomu Itoh return ERR_PTR(ret); 125520897f5cSArne Jansen } 125620897f5cSArne Jansen 12577237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1258e02119d5SChris Mason struct btrfs_fs_info *fs_info) 12590f7d52f4SChris Mason { 12600f7d52f4SChris Mason struct btrfs_root *root; 1261e02119d5SChris Mason 126296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1263e02119d5SChris Mason if (!root) 12647237f183SYan Zheng return ERR_PTR(-ENOMEM); 1265e02119d5SChris Mason 1266e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1267e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1268e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 126927cdeb70SMiao Xie 12706ab6ebb7SNaohiro Aota return root; 12716ab6ebb7SNaohiro Aota } 12726ab6ebb7SNaohiro Aota 12736ab6ebb7SNaohiro Aota int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, 12746ab6ebb7SNaohiro Aota struct btrfs_root *root) 12756ab6ebb7SNaohiro Aota { 12766ab6ebb7SNaohiro Aota struct extent_buffer *leaf; 12776ab6ebb7SNaohiro Aota 12787237f183SYan Zheng /* 127992a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 128027cdeb70SMiao Xie * 128192a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 128292a7cc42SQu Wenruo * and they go away before a real commit is actually done. 128392a7cc42SQu Wenruo * 128492a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 128592a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 12867237f183SYan Zheng */ 1287e02119d5SChris Mason 12884d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 12899631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 12906ab6ebb7SNaohiro Aota if (IS_ERR(leaf)) 12916ab6ebb7SNaohiro Aota return PTR_ERR(leaf); 1292e02119d5SChris Mason 12937237f183SYan Zheng root->node = leaf; 1294e02119d5SChris Mason 1295e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1296e02119d5SChris Mason btrfs_tree_unlock(root->node); 12976ab6ebb7SNaohiro Aota 12986ab6ebb7SNaohiro Aota return 0; 12997237f183SYan Zheng } 13007237f183SYan Zheng 13017237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 13027237f183SYan Zheng struct btrfs_fs_info *fs_info) 13037237f183SYan Zheng { 13047237f183SYan Zheng struct btrfs_root *log_root; 13057237f183SYan Zheng 13067237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 13077237f183SYan Zheng if (IS_ERR(log_root)) 13087237f183SYan Zheng return PTR_ERR(log_root); 13096ab6ebb7SNaohiro Aota 13103ddebf27SNaohiro Aota if (!btrfs_is_zoned(fs_info)) { 13113ddebf27SNaohiro Aota int ret = btrfs_alloc_log_tree_node(trans, log_root); 13123ddebf27SNaohiro Aota 13136ab6ebb7SNaohiro Aota if (ret) { 13146ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13156ab6ebb7SNaohiro Aota return ret; 13166ab6ebb7SNaohiro Aota } 13173ddebf27SNaohiro Aota } 13186ab6ebb7SNaohiro Aota 13197237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 13207237f183SYan Zheng fs_info->log_root_tree = log_root; 13217237f183SYan Zheng return 0; 13227237f183SYan Zheng } 13237237f183SYan Zheng 13247237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 13257237f183SYan Zheng struct btrfs_root *root) 13267237f183SYan Zheng { 13270b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 13287237f183SYan Zheng struct btrfs_root *log_root; 13297237f183SYan Zheng struct btrfs_inode_item *inode_item; 13306ab6ebb7SNaohiro Aota int ret; 13317237f183SYan Zheng 13320b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 13337237f183SYan Zheng if (IS_ERR(log_root)) 13347237f183SYan Zheng return PTR_ERR(log_root); 13357237f183SYan Zheng 13366ab6ebb7SNaohiro Aota ret = btrfs_alloc_log_tree_node(trans, log_root); 13376ab6ebb7SNaohiro Aota if (ret) { 13386ab6ebb7SNaohiro Aota btrfs_put_root(log_root); 13396ab6ebb7SNaohiro Aota return ret; 13406ab6ebb7SNaohiro Aota } 13416ab6ebb7SNaohiro Aota 13427237f183SYan Zheng log_root->last_trans = trans->transid; 13437237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 13447237f183SYan Zheng 13457237f183SYan Zheng inode_item = &log_root->root_item.inode; 13463cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 13473cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 13483cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1349da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 13500b246afaSJeff Mahoney fs_info->nodesize); 13513cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 13527237f183SYan Zheng 13535d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 13547237f183SYan Zheng 13557237f183SYan Zheng WARN_ON(root->log_root); 13567237f183SYan Zheng root->log_root = log_root; 13577237f183SYan Zheng root->log_transid = 0; 1358d1433debSMiao Xie root->log_transid_committed = -1; 1359257c62e1SChris Mason root->last_log_commit = 0; 1360e02119d5SChris Mason return 0; 1361e02119d5SChris Mason } 1362e02119d5SChris Mason 136349d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 136449d11beaSJosef Bacik struct btrfs_path *path, 1365cb517eabSMiao Xie struct btrfs_key *key) 1366e02119d5SChris Mason { 1367e02119d5SChris Mason struct btrfs_root *root; 1368e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 136984234f3aSYan Zheng u64 generation; 1370cb517eabSMiao Xie int ret; 1371581c1760SQu Wenruo int level; 1372cb517eabSMiao Xie 137396dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 137449d11beaSJosef Bacik if (!root) 137549d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 13760f7d52f4SChris Mason 1377cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1378cb517eabSMiao Xie &root->root_item, &root->root_key); 13790f7d52f4SChris Mason if (ret) { 138013a8a7c8SYan, Zheng if (ret > 0) 138113a8a7c8SYan, Zheng ret = -ENOENT; 138249d11beaSJosef Bacik goto fail; 13830f7d52f4SChris Mason } 138413a8a7c8SYan, Zheng 138584234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1386581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 13872ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 13882ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 13891b7ec85eSJosef Bacik key->objectid, generation, level, NULL); 139064c043deSLiu Bo if (IS_ERR(root->node)) { 139164c043deSLiu Bo ret = PTR_ERR(root->node); 13928c38938cSJosef Bacik root->node = NULL; 139349d11beaSJosef Bacik goto fail; 13944eb150d6SQu Wenruo } 13954eb150d6SQu Wenruo if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1396cb517eabSMiao Xie ret = -EIO; 139749d11beaSJosef Bacik goto fail; 1398416bc658SJosef Bacik } 139988c602abSQu Wenruo 140088c602abSQu Wenruo /* 140188c602abSQu Wenruo * For real fs, and not log/reloc trees, root owner must 140288c602abSQu Wenruo * match its root node owner 140388c602abSQu Wenruo */ 140488c602abSQu Wenruo if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && 140588c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 140688c602abSQu Wenruo root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 140788c602abSQu Wenruo root->root_key.objectid != btrfs_header_owner(root->node)) { 140888c602abSQu Wenruo btrfs_crit(fs_info, 140988c602abSQu Wenruo "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", 141088c602abSQu Wenruo root->root_key.objectid, root->node->start, 141188c602abSQu Wenruo btrfs_header_owner(root->node), 141288c602abSQu Wenruo root->root_key.objectid); 141388c602abSQu Wenruo ret = -EUCLEAN; 141488c602abSQu Wenruo goto fail; 141588c602abSQu Wenruo } 14165d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1417cb517eabSMiao Xie return root; 141849d11beaSJosef Bacik fail: 141900246528SJosef Bacik btrfs_put_root(root); 142049d11beaSJosef Bacik return ERR_PTR(ret); 142149d11beaSJosef Bacik } 142249d11beaSJosef Bacik 142349d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 142449d11beaSJosef Bacik struct btrfs_key *key) 142549d11beaSJosef Bacik { 142649d11beaSJosef Bacik struct btrfs_root *root; 142749d11beaSJosef Bacik struct btrfs_path *path; 142849d11beaSJosef Bacik 142949d11beaSJosef Bacik path = btrfs_alloc_path(); 143049d11beaSJosef Bacik if (!path) 143149d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 143249d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 143349d11beaSJosef Bacik btrfs_free_path(path); 143449d11beaSJosef Bacik 143549d11beaSJosef Bacik return root; 1436cb517eabSMiao Xie } 1437cb517eabSMiao Xie 14382dfb1e43SQu Wenruo /* 14392dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 14402dfb1e43SQu Wenruo * 14412dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 14422dfb1e43SQu Wenruo */ 14432dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1444cb517eabSMiao Xie { 1445cb517eabSMiao Xie int ret; 1446dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1447cb517eabSMiao Xie 1448dcc3eb96SNikolay Borisov /* 1449dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1450dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1451dcc3eb96SNikolay Borisov */ 1452dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1453dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1454dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1455dcc3eb96SNikolay Borisov if (ret) 14568257b2dcSMiao Xie goto fail; 14578257b2dcSMiao Xie 1458aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 145937f00a6dSJohannes Thumshirn !btrfs_is_data_reloc_root(root)) { 146092a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1461f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1462f39e4571SJosef Bacik } 1463f39e4571SJosef Bacik 1464851fd730SQu Wenruo /* 1465851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1466851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1467851fd730SQu Wenruo */ 1468851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1469851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 14702dfb1e43SQu Wenruo if (!anon_dev) { 1471cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1472cb517eabSMiao Xie if (ret) 1473876d2cf1SLiu Bo goto fail; 14742dfb1e43SQu Wenruo } else { 14752dfb1e43SQu Wenruo root->anon_dev = anon_dev; 14762dfb1e43SQu Wenruo } 1477851fd730SQu Wenruo } 1478f32e48e9SChandan Rajendra 1479f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1480453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(root); 1481f32e48e9SChandan Rajendra if (ret) { 1482f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1483876d2cf1SLiu Bo goto fail; 1484f32e48e9SChandan Rajendra } 1485f32e48e9SChandan Rajendra 14866b8fad57SNikolay Borisov ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 1487f32e48e9SChandan Rajendra 1488f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1489f32e48e9SChandan Rajendra 1490cb517eabSMiao Xie return 0; 1491cb517eabSMiao Xie fail: 149284db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1493cb517eabSMiao Xie return ret; 1494cb517eabSMiao Xie } 1495cb517eabSMiao Xie 1496a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1497cb517eabSMiao Xie u64 root_id) 1498cb517eabSMiao Xie { 1499cb517eabSMiao Xie struct btrfs_root *root; 1500cb517eabSMiao Xie 1501fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1502fc7cbcd4SDavid Sterba root = radix_tree_lookup(&fs_info->fs_roots_radix, 1503fc7cbcd4SDavid Sterba (unsigned long)root_id); 1504bc44d7c4SJosef Bacik if (root) 150500246528SJosef Bacik root = btrfs_grab_root(root); 1506fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1507cb517eabSMiao Xie return root; 1508cb517eabSMiao Xie } 1509cb517eabSMiao Xie 151049d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 151149d11beaSJosef Bacik u64 objectid) 151249d11beaSJosef Bacik { 1513abed4aaaSJosef Bacik struct btrfs_key key = { 1514abed4aaaSJosef Bacik .objectid = objectid, 1515abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 1516abed4aaaSJosef Bacik .offset = 0, 1517abed4aaaSJosef Bacik }; 1518abed4aaaSJosef Bacik 151949d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 152049d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 152149d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 1522abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 152349d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 152449d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 152549d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 152649d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 152749d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 1528abed4aaaSJosef Bacik return btrfs_grab_root(btrfs_global_root(fs_info, &key)); 152949d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 153049d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 153149d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 153249d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 153349d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 153449d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 153514033b08SQu Wenruo if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) 153614033b08SQu Wenruo return btrfs_grab_root(fs_info->block_group_root) ? 153714033b08SQu Wenruo fs_info->block_group_root : ERR_PTR(-ENOENT); 1538abed4aaaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { 1539abed4aaaSJosef Bacik struct btrfs_root *root = btrfs_global_root(fs_info, &key); 1540abed4aaaSJosef Bacik 1541abed4aaaSJosef Bacik return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); 1542abed4aaaSJosef Bacik } 154349d11beaSJosef Bacik return NULL; 154449d11beaSJosef Bacik } 154549d11beaSJosef Bacik 1546cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1547cb517eabSMiao Xie struct btrfs_root *root) 1548cb517eabSMiao Xie { 1549cb517eabSMiao Xie int ret; 1550cb517eabSMiao Xie 1551fc7cbcd4SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1552fc7cbcd4SDavid Sterba if (ret) 1553fc7cbcd4SDavid Sterba return ret; 1554fc7cbcd4SDavid Sterba 1555fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 1556fc7cbcd4SDavid Sterba ret = radix_tree_insert(&fs_info->fs_roots_radix, 1557fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid, 1558fc7cbcd4SDavid Sterba root); 1559af01d2e5SJosef Bacik if (ret == 0) { 156000246528SJosef Bacik btrfs_grab_root(root); 1561fc7cbcd4SDavid Sterba set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1562af01d2e5SJosef Bacik } 1563fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 1564fc7cbcd4SDavid Sterba radix_tree_preload_end(); 1565cb517eabSMiao Xie 1566cb517eabSMiao Xie return ret; 1567cb517eabSMiao Xie } 1568cb517eabSMiao Xie 1569bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1570bd647ce3SJosef Bacik { 1571bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1572bd647ce3SJosef Bacik struct btrfs_root *root; 1573bd647ce3SJosef Bacik 1574bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1575457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1576457f1864SJosef Bacik 1577bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1578bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1579457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 158071008734SJosef Bacik btrfs_root_name(&root->root_key, buf), 1581bd647ce3SJosef Bacik refcount_read(&root->refs)); 1582bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 158300246528SJosef Bacik btrfs_put_root(root); 158400246528SJosef Bacik btrfs_put_root(root); 1585bd647ce3SJosef Bacik } 1586bd647ce3SJosef Bacik #endif 1587bd647ce3SJosef Bacik } 1588bd647ce3SJosef Bacik 1589abed4aaaSJosef Bacik static void free_global_roots(struct btrfs_fs_info *fs_info) 1590abed4aaaSJosef Bacik { 1591abed4aaaSJosef Bacik struct btrfs_root *root; 1592abed4aaaSJosef Bacik struct rb_node *node; 1593abed4aaaSJosef Bacik 1594abed4aaaSJosef Bacik while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { 1595abed4aaaSJosef Bacik root = rb_entry(node, struct btrfs_root, rb_node); 1596abed4aaaSJosef Bacik rb_erase(&root->rb_node, &fs_info->global_root_tree); 1597abed4aaaSJosef Bacik btrfs_put_root(root); 1598abed4aaaSJosef Bacik } 1599abed4aaaSJosef Bacik } 1600abed4aaaSJosef Bacik 16010d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 16020d4b0463SJosef Bacik { 1603141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1604141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 16055deb17e1SJosef Bacik percpu_counter_destroy(&fs_info->ordered_bytes); 1606141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1607141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1608141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1609141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 16100d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 16110d4b0463SJosef Bacik kfree(fs_info->delayed_root); 1612abed4aaaSJosef Bacik free_global_roots(fs_info); 161300246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 161400246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 161500246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 161600246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 161700246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 161800246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1619aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 16209c54e80dSJosef Bacik btrfs_put_root(fs_info->block_group_root); 1621bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 16223fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 16230d4b0463SJosef Bacik kfree(fs_info->super_copy); 16240d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 16258481dd80SQu Wenruo kfree(fs_info->subpage_info); 16260d4b0463SJosef Bacik kvfree(fs_info); 16270d4b0463SJosef Bacik } 16280d4b0463SJosef Bacik 16290d4b0463SJosef Bacik 16302dfb1e43SQu Wenruo /* 16312dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 16322dfb1e43SQu Wenruo * 16332dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 16342dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 16352dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 16362dfb1e43SQu Wenruo * 16372dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 16382dfb1e43SQu Wenruo * 16392dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 16402dfb1e43SQu Wenruo * same root objectid. 16412dfb1e43SQu Wenruo * 16422dfb1e43SQu Wenruo * @objectid: root id 16432dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 16442dfb1e43SQu Wenruo * pass 0 for new allocation. 16452dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 16462dfb1e43SQu Wenruo * for orphan roots 16472dfb1e43SQu Wenruo */ 16482dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 16492dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 16502dfb1e43SQu Wenruo bool check_ref) 16515eda7b5eSChris Mason { 16525eda7b5eSChris Mason struct btrfs_root *root; 1653381cf658SDavid Sterba struct btrfs_path *path; 16541d4c08e0SDavid Sterba struct btrfs_key key; 16555eda7b5eSChris Mason int ret; 16565eda7b5eSChris Mason 165749d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 165849d11beaSJosef Bacik if (root) 165949d11beaSJosef Bacik return root; 16604df27c4dSYan, Zheng again: 166156e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 166248475471SStefan Behrens if (root) { 16632dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 16642dfb1e43SQu Wenruo ASSERT(!anon_dev); 1665bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 166600246528SJosef Bacik btrfs_put_root(root); 166748475471SStefan Behrens return ERR_PTR(-ENOENT); 1668bc44d7c4SJosef Bacik } 16695eda7b5eSChris Mason return root; 167048475471SStefan Behrens } 16715eda7b5eSChris Mason 167256e9357aSDavid Sterba key.objectid = objectid; 167356e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 167456e9357aSDavid Sterba key.offset = (u64)-1; 167556e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 16765eda7b5eSChris Mason if (IS_ERR(root)) 16775eda7b5eSChris Mason return root; 16783394e160SChris Mason 1679c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1680d68fc57bSYan, Zheng ret = -ENOENT; 1681d68fc57bSYan, Zheng goto fail; 1682d68fc57bSYan, Zheng } 1683d68fc57bSYan, Zheng 16842dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1685cb517eabSMiao Xie if (ret) 1686cb517eabSMiao Xie goto fail; 1687cb517eabSMiao Xie 1688381cf658SDavid Sterba path = btrfs_alloc_path(); 1689381cf658SDavid Sterba if (!path) { 1690381cf658SDavid Sterba ret = -ENOMEM; 1691381cf658SDavid Sterba goto fail; 1692381cf658SDavid Sterba } 16931d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 16941d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 169556e9357aSDavid Sterba key.offset = objectid; 16961d4c08e0SDavid Sterba 16971d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1698381cf658SDavid Sterba btrfs_free_path(path); 1699d68fc57bSYan, Zheng if (ret < 0) 1700d68fc57bSYan, Zheng goto fail; 1701d68fc57bSYan, Zheng if (ret == 0) 170227cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1703d68fc57bSYan, Zheng 1704cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 17050f7d52f4SChris Mason if (ret) { 1706168a2f77SJia-Ju Bai if (ret == -EEXIST) { 170700246528SJosef Bacik btrfs_put_root(root); 17084df27c4dSYan, Zheng goto again; 1709168a2f77SJia-Ju Bai } 17104df27c4dSYan, Zheng goto fail; 17114df27c4dSYan, Zheng } 1712edbd8d4eSChris Mason return root; 17134df27c4dSYan, Zheng fail: 171433fab972SFilipe Manana /* 171533fab972SFilipe Manana * If our caller provided us an anonymous device, then it's his 1716143823cfSDavid Sterba * responsibility to free it in case we fail. So we have to set our 171733fab972SFilipe Manana * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() 171833fab972SFilipe Manana * and once again by our caller. 171933fab972SFilipe Manana */ 172033fab972SFilipe Manana if (anon_dev) 172133fab972SFilipe Manana root->anon_dev = 0; 17228c38938cSJosef Bacik btrfs_put_root(root); 17234df27c4dSYan, Zheng return ERR_PTR(ret); 1724edbd8d4eSChris Mason } 1725edbd8d4eSChris Mason 17262dfb1e43SQu Wenruo /* 17272dfb1e43SQu Wenruo * Get in-memory reference of a root structure 17282dfb1e43SQu Wenruo * 17292dfb1e43SQu Wenruo * @objectid: tree objectid 17302dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 17312dfb1e43SQu Wenruo * one reference 17322dfb1e43SQu Wenruo */ 17332dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 17342dfb1e43SQu Wenruo u64 objectid, bool check_ref) 17352dfb1e43SQu Wenruo { 17362dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 17372dfb1e43SQu Wenruo } 17382dfb1e43SQu Wenruo 17392dfb1e43SQu Wenruo /* 17402dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 17412dfb1e43SQu Wenruo * the anonymous block device id 17422dfb1e43SQu Wenruo * 17432dfb1e43SQu Wenruo * @objectid: tree objectid 17442dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 17452dfb1e43SQu Wenruo * parameter value 17462dfb1e43SQu Wenruo */ 17472dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 17482dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 17492dfb1e43SQu Wenruo { 17502dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 17512dfb1e43SQu Wenruo } 17522dfb1e43SQu Wenruo 17538b712842SChris Mason /* 175449d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 175549d11beaSJosef Bacik * @fs_info: the fs_info 175649d11beaSJosef Bacik * @objectid: the objectid we need to lookup 175749d11beaSJosef Bacik * 175849d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 175949d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 176049d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 176149d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 176249d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 176349d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 176449d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 176549d11beaSJosef Bacik * backref code is finished using the root. 176649d11beaSJosef Bacik */ 176749d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 176849d11beaSJosef Bacik struct btrfs_path *path, 176949d11beaSJosef Bacik u64 objectid) 177049d11beaSJosef Bacik { 177149d11beaSJosef Bacik struct btrfs_root *root; 177249d11beaSJosef Bacik struct btrfs_key key; 177349d11beaSJosef Bacik 177449d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 177549d11beaSJosef Bacik 177649d11beaSJosef Bacik /* 177749d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 177849d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 177949d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 178049d11beaSJosef Bacik * != NULL just return it. 178149d11beaSJosef Bacik */ 178249d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 178349d11beaSJosef Bacik if (root) 178449d11beaSJosef Bacik return root; 178549d11beaSJosef Bacik 178649d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 178749d11beaSJosef Bacik if (root) 178849d11beaSJosef Bacik return root; 178949d11beaSJosef Bacik 179049d11beaSJosef Bacik key.objectid = objectid; 179149d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 179249d11beaSJosef Bacik key.offset = (u64)-1; 179349d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 179449d11beaSJosef Bacik btrfs_release_path(path); 179549d11beaSJosef Bacik 179649d11beaSJosef Bacik return root; 179749d11beaSJosef Bacik } 179849d11beaSJosef Bacik 1799a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1800a74a4b97SChris Mason { 18010d031dc4SYu Zhe struct btrfs_fs_info *fs_info = arg; 1802d0278245SMiao Xie int again; 1803a74a4b97SChris Mason 1804d6fd0ae2SOmar Sandoval while (1) { 1805d0278245SMiao Xie again = 0; 18069d1a2a3aSDavid Sterba 1807fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1808fd340d0fSJosef Bacik 1809d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 18102ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1811d0278245SMiao Xie goto sleep; 1812d0278245SMiao Xie 181390c711abSZygo Blaxell /* 181490c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 181590c711abSZygo Blaxell * before we have finished mounting the filesystem. 181690c711abSZygo Blaxell */ 18170b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 181890c711abSZygo Blaxell goto sleep; 181990c711abSZygo Blaxell 18200b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1821d0278245SMiao Xie goto sleep; 1822d0278245SMiao Xie 1823dc7f370cSMiao Xie /* 1824dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1825dc7f370cSMiao Xie * during the above check and trylock. 1826dc7f370cSMiao Xie */ 18272ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 18280b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1829dc7f370cSMiao Xie goto sleep; 1830dc7f370cSMiao Xie } 1831dc7f370cSMiao Xie 18322ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1833c2d6cb16SFilipe Manana 183433c44184SJosef Bacik again = btrfs_clean_one_deleted_snapshot(fs_info); 18350b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1836a74a4b97SChris Mason 1837d0278245SMiao Xie /* 183805323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 183905323cd1SMiao Xie * needn't do anything special here. 1840d0278245SMiao Xie */ 18410b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 184267c5e7d4SFilipe Manana 184367c5e7d4SFilipe Manana /* 1844f3372065SJohannes Thumshirn * Acquires fs_info->reclaim_bgs_lock to avoid racing 184567c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 184667c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 1847f3372065SJohannes Thumshirn * after acquiring fs_info->reclaim_bgs_lock. So we 184867c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 184967c5e7d4SFilipe Manana * unused block groups. 185067c5e7d4SFilipe Manana */ 18510b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 185218bb8bbfSJohannes Thumshirn 185318bb8bbfSJohannes Thumshirn /* 185418bb8bbfSJohannes Thumshirn * Reclaim block groups in the reclaim_bgs list after we deleted 185518bb8bbfSJohannes Thumshirn * all unused block_groups. This possibly gives us some more free 185618bb8bbfSJohannes Thumshirn * space. 185718bb8bbfSJohannes Thumshirn */ 185818bb8bbfSJohannes Thumshirn btrfs_reclaim_bgs(fs_info); 1859d0278245SMiao Xie sleep: 1860a0a1db70SFilipe Manana clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1861d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1862d6fd0ae2SOmar Sandoval kthread_parkme(); 1863d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1864d6fd0ae2SOmar Sandoval return 0; 1865838fe188SJiri Kosina if (!again) { 1866a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1867a74a4b97SChris Mason schedule(); 1868a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1869a74a4b97SChris Mason } 1870da288d28SFilipe Manana } 1871a74a4b97SChris Mason } 1872a74a4b97SChris Mason 1873a74a4b97SChris Mason static int transaction_kthread(void *arg) 1874a74a4b97SChris Mason { 1875a74a4b97SChris Mason struct btrfs_root *root = arg; 18760b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1877a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1878a74a4b97SChris Mason struct btrfs_transaction *cur; 18798929ecfaSYan, Zheng u64 transid; 1880643900beSNikolay Borisov time64_t delta; 1881a74a4b97SChris Mason unsigned long delay; 1882914b2007SJan Kara bool cannot_commit; 1883a74a4b97SChris Mason 1884a74a4b97SChris Mason do { 1885914b2007SJan Kara cannot_commit = false; 1886ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 18870b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1888a74a4b97SChris Mason 18890b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 18900b246afaSJeff Mahoney cur = fs_info->running_transaction; 1891a74a4b97SChris Mason if (!cur) { 18920b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1893a74a4b97SChris Mason goto sleep; 1894a74a4b97SChris Mason } 189531153d81SYan Zheng 1896643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 1897fdfbf020SJosef Bacik if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && 1898fdfbf020SJosef Bacik cur->state < TRANS_STATE_COMMIT_START && 1899643900beSNikolay Borisov delta < fs_info->commit_interval) { 19000b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1901fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1902fb8a7e94SNikolay Borisov delay = min(delay, 1903fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1904a74a4b97SChris Mason goto sleep; 1905a74a4b97SChris Mason } 19068929ecfaSYan, Zheng transid = cur->transid; 19070b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 190856bec294SChris Mason 190979787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1910354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1911914b2007SJan Kara if (IS_ERR(trans)) { 1912354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1913914b2007SJan Kara cannot_commit = true; 191479787eaaSJeff Mahoney goto sleep; 1915914b2007SJan Kara } 19168929ecfaSYan, Zheng if (transid == trans->transid) { 19173a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 19188929ecfaSYan, Zheng } else { 19193a45bb20SJeff Mahoney btrfs_end_transaction(trans); 19208929ecfaSYan, Zheng } 1921a74a4b97SChris Mason sleep: 19220b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 19230b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1924a74a4b97SChris Mason 192584961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 19262ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 19278929ecfaSYan, Zheng if (!kthread_should_stop() && 19280b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1929914b2007SJan Kara cannot_commit)) 1930bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1931a74a4b97SChris Mason } while (!kthread_should_stop()); 1932a74a4b97SChris Mason return 0; 1933a74a4b97SChris Mason } 1934a74a4b97SChris Mason 1935af31f5e5SChris Mason /* 193601f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 193701f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 193801f0f9daSNikolay Borisov * anything. 1939af31f5e5SChris Mason * 1940af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1941af31f5e5SChris Mason * generation of the latest root in the array with the generation 1942af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1943af31f5e5SChris Mason */ 194401f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1945af31f5e5SChris Mason { 194601f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1947af31f5e5SChris Mason u64 cur; 1948af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1949af31f5e5SChris Mason int i; 1950af31f5e5SChris Mason 1951af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1952af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1953af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1954af31f5e5SChris Mason if (cur == newest_gen) 195501f0f9daSNikolay Borisov return i; 1956af31f5e5SChris Mason } 1957af31f5e5SChris Mason 195801f0f9daSNikolay Borisov return -EINVAL; 1959af31f5e5SChris Mason } 1960af31f5e5SChris Mason 1961af31f5e5SChris Mason /* 1962af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1963af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1964af31f5e5SChris Mason * done 1965af31f5e5SChris Mason */ 1966af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1967af31f5e5SChris Mason { 19686ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1969af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1970af31f5e5SChris Mason 1971af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1972af31f5e5SChris Mason 1973af31f5e5SChris Mason /* 1974af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1975af31f5e5SChris Mason * regardless of which ones we use today 1976af31f5e5SChris Mason */ 1977af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1978af31f5e5SChris Mason 1979af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1980af31f5e5SChris Mason 1981af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1982af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1983af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1984af31f5e5SChris Mason 1985af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1986af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1987af31f5e5SChris Mason 1988af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1989af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1990af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1991af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1992af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1993af31f5e5SChris Mason 19941c56ab99SQu Wenruo if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { 19959c54e80dSJosef Bacik struct btrfs_root *extent_root = btrfs_extent_root(info, 0); 1996f7238e50SJosef Bacik struct btrfs_root *csum_root = btrfs_csum_root(info, 0); 19979c54e80dSJosef Bacik 19989c54e80dSJosef Bacik btrfs_set_backup_extent_root(root_backup, 19999c54e80dSJosef Bacik extent_root->node->start); 2000af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 200129cbcf40SJosef Bacik btrfs_header_generation(extent_root->node)); 2002af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 200329cbcf40SJosef Bacik btrfs_header_level(extent_root->node)); 2004af31f5e5SChris Mason 2005f7238e50SJosef Bacik btrfs_set_backup_csum_root(root_backup, csum_root->node->start); 2006f7238e50SJosef Bacik btrfs_set_backup_csum_root_gen(root_backup, 2007f7238e50SJosef Bacik btrfs_header_generation(csum_root->node)); 2008f7238e50SJosef Bacik btrfs_set_backup_csum_root_level(root_backup, 2009f7238e50SJosef Bacik btrfs_header_level(csum_root->node)); 20109c54e80dSJosef Bacik } 2011af31f5e5SChris Mason 20127c7e82a7SChris Mason /* 20137c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 20147c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 20157c7e82a7SChris Mason */ 20167c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 20177c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 20187c7e82a7SChris Mason info->fs_root->node->start); 2019af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 2020af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 2021af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 2022af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 20237c7e82a7SChris Mason } 2024af31f5e5SChris Mason 2025af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 2026af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 2027af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 2028af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 2029af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 2030af31f5e5SChris Mason 2031af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 2032af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 2033af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 2034af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 2035af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 2036af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 2037af31f5e5SChris Mason 2038af31f5e5SChris Mason /* 2039af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 2040af31f5e5SChris Mason * for the next commit 2041af31f5e5SChris Mason */ 2042af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 2043af31f5e5SChris Mason &info->super_for_commit->super_roots, 2044af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 2045af31f5e5SChris Mason } 2046af31f5e5SChris Mason 2047af31f5e5SChris Mason /* 2048bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 2049bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 2050bd2336b2SNikolay Borisov * 2051bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 2052bd2336b2SNikolay Borisov * priority - priority of backup root required 2053bd2336b2SNikolay Borisov * 2054bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 2055bd2336b2SNikolay Borisov */ 2056bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 2057bd2336b2SNikolay Borisov { 2058bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2059bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 2060bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 2061bd2336b2SNikolay Borisov 2062bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 2063bd2336b2SNikolay Borisov if (priority == 0) 2064bd2336b2SNikolay Borisov return backup_index; 2065bd2336b2SNikolay Borisov 2066bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 2067bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 2068bd2336b2SNikolay Borisov } else { 2069bd2336b2SNikolay Borisov return -EINVAL; 2070bd2336b2SNikolay Borisov } 2071bd2336b2SNikolay Borisov 2072bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 2073bd2336b2SNikolay Borisov 2074bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 2075bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 2076bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2077bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 2078bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 2079bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2080bd2336b2SNikolay Borisov 2081bd2336b2SNikolay Borisov /* 2082bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 2083bd2336b2SNikolay Borisov * need a fsck 2084bd2336b2SNikolay Borisov */ 2085bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2086bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2087bd2336b2SNikolay Borisov 2088bd2336b2SNikolay Borisov return backup_index; 2089bd2336b2SNikolay Borisov } 2090bd2336b2SNikolay Borisov 20917abadb64SLiu Bo /* helper to cleanup workers */ 20927abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20937abadb64SLiu Bo { 2094dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2095afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 2096a31b4a43SChristoph Hellwig btrfs_destroy_workqueue(fs_info->hipri_workers); 20975cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2098d7b9416fSChristoph Hellwig if (fs_info->endio_workers) 2099d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_workers); 2100385de0efSChristoph Hellwig if (fs_info->rmw_workers) 2101385de0efSChristoph Hellwig destroy_workqueue(fs_info->rmw_workers); 2102fed8a72dSChristoph Hellwig if (fs_info->compressed_write_workers) 2103fed8a72dSChristoph Hellwig destroy_workqueue(fs_info->compressed_write_workers); 2104fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2105fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 21065b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2107e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2108a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2109fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2110b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 2111b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 2112a9b9477dSFilipe Manana /* 2113a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 2114a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 2115a9b9477dSFilipe Manana * queues can do metadata I/O operations. 2116a9b9477dSFilipe Manana */ 2117d7b9416fSChristoph Hellwig if (fs_info->endio_meta_workers) 2118d7b9416fSChristoph Hellwig destroy_workqueue(fs_info->endio_meta_workers); 21197abadb64SLiu Bo } 21207abadb64SLiu Bo 21212e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 21222e9f5954SRashika { 21232e9f5954SRashika if (root) { 21242e9f5954SRashika free_extent_buffer(root->node); 21252e9f5954SRashika free_extent_buffer(root->commit_root); 21262e9f5954SRashika root->node = NULL; 21272e9f5954SRashika root->commit_root = NULL; 21282e9f5954SRashika } 21292e9f5954SRashika } 21302e9f5954SRashika 2131abed4aaaSJosef Bacik static void free_global_root_pointers(struct btrfs_fs_info *fs_info) 2132abed4aaaSJosef Bacik { 2133abed4aaaSJosef Bacik struct btrfs_root *root, *tmp; 2134abed4aaaSJosef Bacik 2135abed4aaaSJosef Bacik rbtree_postorder_for_each_entry_safe(root, tmp, 2136abed4aaaSJosef Bacik &fs_info->global_root_tree, 2137abed4aaaSJosef Bacik rb_node) 2138abed4aaaSJosef Bacik free_root_extent_buffers(root); 2139abed4aaaSJosef Bacik } 2140abed4aaaSJosef Bacik 2141af31f5e5SChris Mason /* helper to cleanup tree roots */ 21424273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 2143af31f5e5SChris Mason { 21442e9f5954SRashika free_root_extent_buffers(info->tree_root); 2145655b09feSJosef Bacik 2146abed4aaaSJosef Bacik free_global_root_pointers(info); 21472e9f5954SRashika free_root_extent_buffers(info->dev_root); 21482e9f5954SRashika free_root_extent_buffers(info->quota_root); 21492e9f5954SRashika free_root_extent_buffers(info->uuid_root); 21508c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2151aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 21529c54e80dSJosef Bacik free_root_extent_buffers(info->block_group_root); 21534273eaffSAnand Jain if (free_chunk_root) 21542e9f5954SRashika free_root_extent_buffers(info->chunk_root); 2155af31f5e5SChris Mason } 2156af31f5e5SChris Mason 21578c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 21588c38938cSJosef Bacik { 21598c38938cSJosef Bacik if (!root) 21608c38938cSJosef Bacik return; 21618c38938cSJosef Bacik 21628c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 21638c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 21641dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 21658c38938cSJosef Bacik if (root->anon_dev) 21668c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 21678c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2168923eb523SJohannes Thumshirn free_root_extent_buffers(root); 21698c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2170fc7cbcd4SDavid Sterba spin_lock(&root->fs_info->fs_roots_radix_lock); 21718c38938cSJosef Bacik list_del_init(&root->leak_list); 2172fc7cbcd4SDavid Sterba spin_unlock(&root->fs_info->fs_roots_radix_lock); 21738c38938cSJosef Bacik #endif 21748c38938cSJosef Bacik kfree(root); 21758c38938cSJosef Bacik } 21768c38938cSJosef Bacik } 21778c38938cSJosef Bacik 2178faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2179171f6537SJosef Bacik { 2180fc7cbcd4SDavid Sterba int ret; 2181fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 2182fc7cbcd4SDavid Sterba int i; 2183171f6537SJosef Bacik 2184171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2185fc7cbcd4SDavid Sterba gang[0] = list_entry(fs_info->dead_roots.next, 2186171f6537SJosef Bacik struct btrfs_root, root_list); 2187fc7cbcd4SDavid Sterba list_del(&gang[0]->root_list); 2188171f6537SJosef Bacik 2189fc7cbcd4SDavid Sterba if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2190fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2191fc7cbcd4SDavid Sterba btrfs_put_root(gang[0]); 2192171f6537SJosef Bacik } 2193171f6537SJosef Bacik 2194fc7cbcd4SDavid Sterba while (1) { 2195fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2196fc7cbcd4SDavid Sterba (void **)gang, 0, 2197fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 2198fc7cbcd4SDavid Sterba if (!ret) 2199fc7cbcd4SDavid Sterba break; 2200fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 2201fc7cbcd4SDavid Sterba btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2202171f6537SJosef Bacik } 2203171f6537SJosef Bacik } 2204af31f5e5SChris Mason 2205638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2206638aa7edSEric Sandeen { 2207638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2208638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2209638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2210638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2211638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2212638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2213ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2214638aa7edSEric Sandeen } 2215638aa7edSEric Sandeen 2216779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2217779a65a4SEric Sandeen { 2218779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2219779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2220779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2221779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2222779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2223779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2224907d2710SDavid Sterba atomic_set(&fs_info->reloc_cancel_req, 0); 2225779a65a4SEric Sandeen } 2226779a65a4SEric Sandeen 22276bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2228f37938e0SEric Sandeen { 22292ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 2230e256927bSJosef Bacik unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, 2231e256927bSJosef Bacik fs_info->tree_root); 22322ff7e61eSJeff Mahoney 22332ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 22342ff7e61eSJeff Mahoney set_nlink(inode, 1); 2235f37938e0SEric Sandeen /* 2236f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2237f37938e0SEric Sandeen * the real end of the address space is determined by all of 2238f37938e0SEric Sandeen * the devices in the system 2239f37938e0SEric Sandeen */ 22402ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 22412ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2242f37938e0SEric Sandeen 22432ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 224443eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 2245efb0645bSJosef Bacik IO_TREE_BTREE_INODE_IO, NULL); 22462ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2247f37938e0SEric Sandeen 22485c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 2249adac5584SFilipe Manana BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; 2250adac5584SFilipe Manana BTRFS_I(inode)->location.type = 0; 2251adac5584SFilipe Manana BTRFS_I(inode)->location.offset = 0; 22522ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 2253e256927bSJosef Bacik __insert_inode_hash(inode, hash); 2254f37938e0SEric Sandeen } 2255f37938e0SEric Sandeen 2256ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2257ad618368SEric Sandeen { 2258ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2259129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 22607f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2261ad618368SEric Sandeen } 2262ad618368SEric Sandeen 2263f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2264f9e92e40SEric Sandeen { 2265f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2266f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2267f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2268f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2269f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2270f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2271d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2272011b46c3SQu Wenruo fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; 2273f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2274f9e92e40SEric Sandeen } 2275f9e92e40SEric Sandeen 2276d21deec5SSu Yue static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) 22772a458198SEric Sandeen { 2278f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 22796f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 22802a458198SEric Sandeen 22812a458198SEric Sandeen fs_info->workers = 2282a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); 2283a31b4a43SChristoph Hellwig fs_info->hipri_workers = 2284a31b4a43SChristoph Hellwig btrfs_alloc_workqueue(fs_info, "worker-high", 2285cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 22862a458198SEric Sandeen 22872a458198SEric Sandeen fs_info->delalloc_workers = 2288cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2289cb001095SJeff Mahoney flags, max_active, 2); 22902a458198SEric Sandeen 22912a458198SEric Sandeen fs_info->flush_workers = 2292cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2293cb001095SJeff Mahoney flags, max_active, 0); 22942a458198SEric Sandeen 22952a458198SEric Sandeen fs_info->caching_workers = 2296cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 22972a458198SEric Sandeen 22982a458198SEric Sandeen fs_info->fixup_workers = 2299cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 23002a458198SEric Sandeen 23012a458198SEric Sandeen fs_info->endio_workers = 2302d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio", flags, max_active); 23032a458198SEric Sandeen fs_info->endio_meta_workers = 2304d7b9416fSChristoph Hellwig alloc_workqueue("btrfs-endio-meta", flags, max_active); 2305385de0efSChristoph Hellwig fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); 23062a458198SEric Sandeen fs_info->endio_write_workers = 2307cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2308cb001095SJeff Mahoney max_active, 2); 2309fed8a72dSChristoph Hellwig fs_info->compressed_write_workers = 2310fed8a72dSChristoph Hellwig alloc_workqueue("btrfs-compressed-write", flags, max_active); 23112a458198SEric Sandeen fs_info->endio_freespace_worker = 2312cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2313cb001095SJeff Mahoney max_active, 0); 23142a458198SEric Sandeen fs_info->delayed_workers = 2315cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2316cb001095SJeff Mahoney max_active, 0); 23172a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2318cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2319b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2320b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 23212a458198SEric Sandeen 2322a31b4a43SChristoph Hellwig if (!(fs_info->workers && fs_info->hipri_workers && 2323a31b4a43SChristoph Hellwig fs_info->delalloc_workers && fs_info->flush_workers && 23242a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 2325fed8a72dSChristoph Hellwig fs_info->compressed_write_workers && 2326*1a1a2851SQu Wenruo fs_info->endio_write_workers && 23272a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 2328f26c9238SQu Wenruo fs_info->caching_workers && fs_info->fixup_workers && 2329f26c9238SQu Wenruo fs_info->delayed_workers && fs_info->qgroup_rescan_workers && 2330b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 23312a458198SEric Sandeen return -ENOMEM; 23322a458198SEric Sandeen } 23332a458198SEric Sandeen 23342a458198SEric Sandeen return 0; 23352a458198SEric Sandeen } 23362a458198SEric Sandeen 23376d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 23386d97c6e3SJohannes Thumshirn { 23396d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2340b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 23416d97c6e3SJohannes Thumshirn 2342b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 23436d97c6e3SJohannes Thumshirn 23446d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 23456d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2346b4e967beSDavid Sterba csum_driver); 23476d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 23486d97c6e3SJohannes Thumshirn } 23496d97c6e3SJohannes Thumshirn 23506d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 23516d97c6e3SJohannes Thumshirn 2352c8a5f8caSDavid Sterba btrfs_info(fs_info, "using %s (%s) checksum algorithm", 2353c8a5f8caSDavid Sterba btrfs_super_csum_name(csum_type), 2354c8a5f8caSDavid Sterba crypto_shash_driver_name(csum_shash)); 23556d97c6e3SJohannes Thumshirn return 0; 23566d97c6e3SJohannes Thumshirn } 23576d97c6e3SJohannes Thumshirn 235863443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 235963443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 236063443bf5SEric Sandeen { 236163443bf5SEric Sandeen int ret; 236263443bf5SEric Sandeen struct btrfs_root *log_tree_root; 236363443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 236463443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2365581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 236663443bf5SEric Sandeen 236763443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2368f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 236963443bf5SEric Sandeen return -EIO; 237063443bf5SEric Sandeen } 237163443bf5SEric Sandeen 237296dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 237396dfcb46SJosef Bacik GFP_KERNEL); 237463443bf5SEric Sandeen if (!log_tree_root) 237563443bf5SEric Sandeen return -ENOMEM; 237663443bf5SEric Sandeen 23772ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 23781b7ec85eSJosef Bacik BTRFS_TREE_LOG_OBJECTID, 23791b7ec85eSJosef Bacik fs_info->generation + 1, level, 23801b7ec85eSJosef Bacik NULL); 238164c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2382f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 23830eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 23848c38938cSJosef Bacik log_tree_root->node = NULL; 238500246528SJosef Bacik btrfs_put_root(log_tree_root); 23860eeff236SLiu Bo return ret; 23874eb150d6SQu Wenruo } 23884eb150d6SQu Wenruo if (!extent_buffer_uptodate(log_tree_root->node)) { 2389f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 239000246528SJosef Bacik btrfs_put_root(log_tree_root); 239163443bf5SEric Sandeen return -EIO; 239263443bf5SEric Sandeen } 23934eb150d6SQu Wenruo 239463443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 239563443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 239663443bf5SEric Sandeen if (ret) { 23970b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 239863443bf5SEric Sandeen "Failed to recover log tree"); 239900246528SJosef Bacik btrfs_put_root(log_tree_root); 240063443bf5SEric Sandeen return ret; 240163443bf5SEric Sandeen } 240263443bf5SEric Sandeen 2403bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 24046bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 240563443bf5SEric Sandeen if (ret) 240663443bf5SEric Sandeen return ret; 240763443bf5SEric Sandeen } 240863443bf5SEric Sandeen 240963443bf5SEric Sandeen return 0; 241063443bf5SEric Sandeen } 241163443bf5SEric Sandeen 2412abed4aaaSJosef Bacik static int load_global_roots_objectid(struct btrfs_root *tree_root, 2413abed4aaaSJosef Bacik struct btrfs_path *path, u64 objectid, 2414abed4aaaSJosef Bacik const char *name) 2415abed4aaaSJosef Bacik { 2416abed4aaaSJosef Bacik struct btrfs_fs_info *fs_info = tree_root->fs_info; 2417abed4aaaSJosef Bacik struct btrfs_root *root; 2418f7238e50SJosef Bacik u64 max_global_id = 0; 2419abed4aaaSJosef Bacik int ret; 2420abed4aaaSJosef Bacik struct btrfs_key key = { 2421abed4aaaSJosef Bacik .objectid = objectid, 2422abed4aaaSJosef Bacik .type = BTRFS_ROOT_ITEM_KEY, 2423abed4aaaSJosef Bacik .offset = 0, 2424abed4aaaSJosef Bacik }; 2425abed4aaaSJosef Bacik bool found = false; 2426abed4aaaSJosef Bacik 2427abed4aaaSJosef Bacik /* If we have IGNOREDATACSUMS skip loading these roots. */ 2428abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID && 2429abed4aaaSJosef Bacik btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 2430abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2431abed4aaaSJosef Bacik return 0; 2432abed4aaaSJosef Bacik } 2433abed4aaaSJosef Bacik 2434abed4aaaSJosef Bacik while (1) { 2435abed4aaaSJosef Bacik ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 2436abed4aaaSJosef Bacik if (ret < 0) 2437abed4aaaSJosef Bacik break; 2438abed4aaaSJosef Bacik 2439abed4aaaSJosef Bacik if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { 2440abed4aaaSJosef Bacik ret = btrfs_next_leaf(tree_root, path); 2441abed4aaaSJosef Bacik if (ret) { 2442abed4aaaSJosef Bacik if (ret > 0) 2443abed4aaaSJosef Bacik ret = 0; 2444abed4aaaSJosef Bacik break; 2445abed4aaaSJosef Bacik } 2446abed4aaaSJosef Bacik } 2447abed4aaaSJosef Bacik ret = 0; 2448abed4aaaSJosef Bacik 2449abed4aaaSJosef Bacik btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); 2450abed4aaaSJosef Bacik if (key.objectid != objectid) 2451abed4aaaSJosef Bacik break; 2452abed4aaaSJosef Bacik btrfs_release_path(path); 2453abed4aaaSJosef Bacik 2454f7238e50SJosef Bacik /* 2455f7238e50SJosef Bacik * Just worry about this for extent tree, it'll be the same for 2456f7238e50SJosef Bacik * everybody. 2457f7238e50SJosef Bacik */ 2458f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2459f7238e50SJosef Bacik max_global_id = max(max_global_id, key.offset); 2460f7238e50SJosef Bacik 2461abed4aaaSJosef Bacik found = true; 2462abed4aaaSJosef Bacik root = read_tree_root_path(tree_root, path, &key); 2463abed4aaaSJosef Bacik if (IS_ERR(root)) { 2464abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2465abed4aaaSJosef Bacik ret = PTR_ERR(root); 2466abed4aaaSJosef Bacik break; 2467abed4aaaSJosef Bacik } 2468abed4aaaSJosef Bacik set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2469abed4aaaSJosef Bacik ret = btrfs_global_root_insert(root); 2470abed4aaaSJosef Bacik if (ret) { 2471abed4aaaSJosef Bacik btrfs_put_root(root); 2472abed4aaaSJosef Bacik break; 2473abed4aaaSJosef Bacik } 2474abed4aaaSJosef Bacik key.offset++; 2475abed4aaaSJosef Bacik } 2476abed4aaaSJosef Bacik btrfs_release_path(path); 2477abed4aaaSJosef Bacik 2478f7238e50SJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 2479f7238e50SJosef Bacik fs_info->nr_global_roots = max_global_id + 1; 2480f7238e50SJosef Bacik 2481abed4aaaSJosef Bacik if (!found || ret) { 2482abed4aaaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 2483abed4aaaSJosef Bacik set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); 2484abed4aaaSJosef Bacik 2485abed4aaaSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) 2486abed4aaaSJosef Bacik ret = ret ? ret : -ENOENT; 2487abed4aaaSJosef Bacik else 2488abed4aaaSJosef Bacik ret = 0; 2489abed4aaaSJosef Bacik btrfs_err(fs_info, "failed to load root %s", name); 2490abed4aaaSJosef Bacik } 2491abed4aaaSJosef Bacik return ret; 2492abed4aaaSJosef Bacik } 2493abed4aaaSJosef Bacik 2494abed4aaaSJosef Bacik static int load_global_roots(struct btrfs_root *tree_root) 2495abed4aaaSJosef Bacik { 2496abed4aaaSJosef Bacik struct btrfs_path *path; 2497abed4aaaSJosef Bacik int ret = 0; 2498abed4aaaSJosef Bacik 2499abed4aaaSJosef Bacik path = btrfs_alloc_path(); 2500abed4aaaSJosef Bacik if (!path) 2501abed4aaaSJosef Bacik return -ENOMEM; 2502abed4aaaSJosef Bacik 2503abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2504abed4aaaSJosef Bacik BTRFS_EXTENT_TREE_OBJECTID, "extent"); 2505abed4aaaSJosef Bacik if (ret) 2506abed4aaaSJosef Bacik goto out; 2507abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2508abed4aaaSJosef Bacik BTRFS_CSUM_TREE_OBJECTID, "csum"); 2509abed4aaaSJosef Bacik if (ret) 2510abed4aaaSJosef Bacik goto out; 2511abed4aaaSJosef Bacik if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) 2512abed4aaaSJosef Bacik goto out; 2513abed4aaaSJosef Bacik ret = load_global_roots_objectid(tree_root, path, 2514abed4aaaSJosef Bacik BTRFS_FREE_SPACE_TREE_OBJECTID, 2515abed4aaaSJosef Bacik "free space"); 2516abed4aaaSJosef Bacik out: 2517abed4aaaSJosef Bacik btrfs_free_path(path); 2518abed4aaaSJosef Bacik return ret; 2519abed4aaaSJosef Bacik } 2520abed4aaaSJosef Bacik 25216bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 25224bbcaa64SEric Sandeen { 25236bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2524a4f3d2c4SDavid Sterba struct btrfs_root *root; 25254bbcaa64SEric Sandeen struct btrfs_key location; 25264bbcaa64SEric Sandeen int ret; 25274bbcaa64SEric Sandeen 25286bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 25296bccf3abSJeff Mahoney 2530abed4aaaSJosef Bacik ret = load_global_roots(tree_root); 2531abed4aaaSJosef Bacik if (ret) 2532abed4aaaSJosef Bacik return ret; 2533abed4aaaSJosef Bacik 25344bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 25354bbcaa64SEric Sandeen location.offset = 0; 25364bbcaa64SEric Sandeen 25371c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { 253814033b08SQu Wenruo location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; 253914033b08SQu Wenruo root = btrfs_read_tree_root(tree_root, &location); 254014033b08SQu Wenruo if (IS_ERR(root)) { 254114033b08SQu Wenruo if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 254214033b08SQu Wenruo ret = PTR_ERR(root); 254314033b08SQu Wenruo goto out; 254414033b08SQu Wenruo } 254514033b08SQu Wenruo } else { 254614033b08SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 254714033b08SQu Wenruo fs_info->block_group_root = root; 254814033b08SQu Wenruo } 254914033b08SQu Wenruo } 255014033b08SQu Wenruo 255114033b08SQu Wenruo location.objectid = BTRFS_DEV_TREE_OBJECTID; 2552a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2553f50f4353SLiu Bo if (IS_ERR(root)) { 255442437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2555f50f4353SLiu Bo ret = PTR_ERR(root); 2556f50f4353SLiu Bo goto out; 2557f50f4353SLiu Bo } 255842437a63SJosef Bacik } else { 2559a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2560a4f3d2c4SDavid Sterba fs_info->dev_root = root; 256142437a63SJosef Bacik } 2562820a49daSJosef Bacik /* Initialize fs_info for all devices in any case */ 2563a8d1b164SJohannes Thumshirn ret = btrfs_init_devices_late(fs_info); 2564a8d1b164SJohannes Thumshirn if (ret) 2565a8d1b164SJohannes Thumshirn goto out; 25664bbcaa64SEric Sandeen 2567aeb935a4SQu Wenruo /* 2568aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2569aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2570aeb935a4SQu Wenruo */ 257156e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 257256e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2573aeb935a4SQu Wenruo if (IS_ERR(root)) { 257442437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2575aeb935a4SQu Wenruo ret = PTR_ERR(root); 2576aeb935a4SQu Wenruo goto out; 2577aeb935a4SQu Wenruo } 257842437a63SJosef Bacik } else { 2579aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2580aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 258142437a63SJosef Bacik } 2582aeb935a4SQu Wenruo 25834bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2584a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2585a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2586a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2587afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2588a4f3d2c4SDavid Sterba fs_info->quota_root = root; 25894bbcaa64SEric Sandeen } 25904bbcaa64SEric Sandeen 25914bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2592a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2593a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 259442437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2595a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 25964bbcaa64SEric Sandeen if (ret != -ENOENT) 2597f50f4353SLiu Bo goto out; 259842437a63SJosef Bacik } 25994bbcaa64SEric Sandeen } else { 2600a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2601a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 26024bbcaa64SEric Sandeen } 26034bbcaa64SEric Sandeen 26044bbcaa64SEric Sandeen return 0; 2605f50f4353SLiu Bo out: 2606f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2607f50f4353SLiu Bo location.objectid, ret); 2608f50f4353SLiu Bo return ret; 26094bbcaa64SEric Sandeen } 26104bbcaa64SEric Sandeen 2611069ec957SQu Wenruo /* 2612069ec957SQu Wenruo * Real super block validation 2613069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2614069ec957SQu Wenruo * 2615069ec957SQu Wenruo * @sb: super block to check 2616069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2617069ec957SQu Wenruo * 0 the primary (1st) sb 2618069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2619069ec957SQu Wenruo * -1 skip bytenr check 2620069ec957SQu Wenruo */ 2621a05d3c91SQu Wenruo int btrfs_validate_super(struct btrfs_fs_info *fs_info, 2622069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 262321a852b0SQu Wenruo { 262421a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 262521a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 262621a852b0SQu Wenruo int ret = 0; 262721a852b0SQu Wenruo 262821a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 262921a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 263021a852b0SQu Wenruo ret = -EINVAL; 263121a852b0SQu Wenruo } 263221a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 263321a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 263421a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 263521a852b0SQu Wenruo ret = -EINVAL; 263621a852b0SQu Wenruo } 263721a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 263821a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 263921a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 264021a852b0SQu Wenruo ret = -EINVAL; 264121a852b0SQu Wenruo } 264221a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 264321a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 264421a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 264521a852b0SQu Wenruo ret = -EINVAL; 264621a852b0SQu Wenruo } 264721a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 264821a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 264921a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 265021a852b0SQu Wenruo ret = -EINVAL; 265121a852b0SQu Wenruo } 265221a852b0SQu Wenruo 265321a852b0SQu Wenruo /* 265421a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 265521a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 265621a852b0SQu Wenruo */ 265721a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 265821a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 265921a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 266021a852b0SQu Wenruo ret = -EINVAL; 266121a852b0SQu Wenruo } 26620bb3eb3eSQu Wenruo 26630bb3eb3eSQu Wenruo /* 26641a42daabSQu Wenruo * We only support at most two sectorsizes: 4K and PAGE_SIZE. 26651a42daabSQu Wenruo * 26661a42daabSQu Wenruo * We can support 16K sectorsize with 64K page size without problem, 26671a42daabSQu Wenruo * but such sectorsize/pagesize combination doesn't make much sense. 26681a42daabSQu Wenruo * 4K will be our future standard, PAGE_SIZE is supported from the very 26691a42daabSQu Wenruo * beginning. 26700bb3eb3eSQu Wenruo */ 26711a42daabSQu Wenruo if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { 267221a852b0SQu Wenruo btrfs_err(fs_info, 26730bb3eb3eSQu Wenruo "sectorsize %llu not yet supported for page size %lu", 267421a852b0SQu Wenruo sectorsize, PAGE_SIZE); 267521a852b0SQu Wenruo ret = -EINVAL; 267621a852b0SQu Wenruo } 26770bb3eb3eSQu Wenruo 267821a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 267921a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 268021a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 268121a852b0SQu Wenruo ret = -EINVAL; 268221a852b0SQu Wenruo } 268321a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 268421a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 268521a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 268621a852b0SQu Wenruo ret = -EINVAL; 268721a852b0SQu Wenruo } 268821a852b0SQu Wenruo 268921a852b0SQu Wenruo /* Root alignment check */ 269021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 269121a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 269221a852b0SQu Wenruo btrfs_super_root(sb)); 269321a852b0SQu Wenruo ret = -EINVAL; 269421a852b0SQu Wenruo } 269521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 269621a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 269721a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 269821a852b0SQu Wenruo ret = -EINVAL; 269921a852b0SQu Wenruo } 270021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 270121a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 270221a852b0SQu Wenruo btrfs_super_log_root(sb)); 270321a852b0SQu Wenruo ret = -EINVAL; 270421a852b0SQu Wenruo } 270521a852b0SQu Wenruo 2706aefd7f70SNikolay Borisov if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 2707aefd7f70SNikolay Borisov BTRFS_FSID_SIZE)) { 2708aefd7f70SNikolay Borisov btrfs_err(fs_info, 2709aefd7f70SNikolay Borisov "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", 2710aefd7f70SNikolay Borisov fs_info->super_copy->fsid, fs_info->fs_devices->fsid); 2711aefd7f70SNikolay Borisov ret = -EINVAL; 2712aefd7f70SNikolay Borisov } 2713aefd7f70SNikolay Borisov 2714aefd7f70SNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID) && 2715aefd7f70SNikolay Borisov memcmp(fs_info->fs_devices->metadata_uuid, 2716aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { 2717aefd7f70SNikolay Borisov btrfs_err(fs_info, 2718aefd7f70SNikolay Borisov "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", 2719aefd7f70SNikolay Borisov fs_info->super_copy->metadata_uuid, 2720aefd7f70SNikolay Borisov fs_info->fs_devices->metadata_uuid); 2721aefd7f70SNikolay Borisov ret = -EINVAL; 2722aefd7f70SNikolay Borisov } 2723aefd7f70SNikolay Borisov 27241c56ab99SQu Wenruo /* 27251c56ab99SQu Wenruo * Artificial requirement for block-group-tree to force newer features 27261c56ab99SQu Wenruo * (free-space-tree, no-holes) so the test matrix is smaller. 27271c56ab99SQu Wenruo */ 27281c56ab99SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 27291c56ab99SQu Wenruo (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || 27301c56ab99SQu Wenruo !btrfs_fs_incompat(fs_info, NO_HOLES))) { 27311c56ab99SQu Wenruo btrfs_err(fs_info, 27321c56ab99SQu Wenruo "block-group-tree feature requires fres-space-tree and no-holes"); 27331c56ab99SQu Wenruo ret = -EINVAL; 27341c56ab99SQu Wenruo } 27351c56ab99SQu Wenruo 2736de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 27377239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 273821a852b0SQu Wenruo btrfs_err(fs_info, 27397239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2740de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 274121a852b0SQu Wenruo ret = -EINVAL; 274221a852b0SQu Wenruo } 274321a852b0SQu Wenruo 274421a852b0SQu Wenruo /* 274521a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 274621a852b0SQu Wenruo * done later 274721a852b0SQu Wenruo */ 274821a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 274921a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 275021a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 275121a852b0SQu Wenruo ret = -EINVAL; 275221a852b0SQu Wenruo } 275321a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 275421a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 275521a852b0SQu Wenruo btrfs_super_stripesize(sb)); 275621a852b0SQu Wenruo ret = -EINVAL; 275721a852b0SQu Wenruo } 275821a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 275921a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 276021a852b0SQu Wenruo btrfs_super_num_devices(sb)); 276121a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 276221a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 276321a852b0SQu Wenruo ret = -EINVAL; 276421a852b0SQu Wenruo } 276521a852b0SQu Wenruo 2766069ec957SQu Wenruo if (mirror_num >= 0 && 2767069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 276821a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 276921a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 277021a852b0SQu Wenruo ret = -EINVAL; 277121a852b0SQu Wenruo } 277221a852b0SQu Wenruo 277321a852b0SQu Wenruo /* 277421a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 277521a852b0SQu Wenruo * and one chunk 277621a852b0SQu Wenruo */ 277721a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 277821a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 277921a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 278021a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 278121a852b0SQu Wenruo ret = -EINVAL; 278221a852b0SQu Wenruo } 278321a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 278421a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 278521a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 278621a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 278721a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 278821a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 278921a852b0SQu Wenruo ret = -EINVAL; 279021a852b0SQu Wenruo } 279121a852b0SQu Wenruo 279221a852b0SQu Wenruo /* 279321a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 279421a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 279521a852b0SQu Wenruo */ 279621a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 279721a852b0SQu Wenruo btrfs_warn(fs_info, 279821a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 279921a852b0SQu Wenruo btrfs_super_generation(sb), 280021a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 280121a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 280221a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 280321a852b0SQu Wenruo btrfs_warn(fs_info, 280421a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 280521a852b0SQu Wenruo btrfs_super_generation(sb), 280621a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 280721a852b0SQu Wenruo 280821a852b0SQu Wenruo return ret; 280921a852b0SQu Wenruo } 281021a852b0SQu Wenruo 2811069ec957SQu Wenruo /* 2812069ec957SQu Wenruo * Validation of super block at mount time. 2813069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2814069ec957SQu Wenruo * flags will be skipped. 2815069ec957SQu Wenruo */ 2816069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2817069ec957SQu Wenruo { 2818a05d3c91SQu Wenruo return btrfs_validate_super(fs_info, fs_info->super_copy, 0); 2819069ec957SQu Wenruo } 2820069ec957SQu Wenruo 282175cb857dSQu Wenruo /* 282275cb857dSQu Wenruo * Validation of super block at write time. 282375cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 282475cb857dSQu Wenruo * overwritten soon. 282575cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 282675cb857dSQu Wenruo */ 282775cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 282875cb857dSQu Wenruo struct btrfs_super_block *sb) 282975cb857dSQu Wenruo { 283075cb857dSQu Wenruo int ret; 283175cb857dSQu Wenruo 2832a05d3c91SQu Wenruo ret = btrfs_validate_super(fs_info, sb, -1); 283375cb857dSQu Wenruo if (ret < 0) 283475cb857dSQu Wenruo goto out; 2835e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 283675cb857dSQu Wenruo ret = -EUCLEAN; 283775cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 283875cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 283975cb857dSQu Wenruo goto out; 284075cb857dSQu Wenruo } 284175cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 284275cb857dSQu Wenruo ret = -EUCLEAN; 284375cb857dSQu Wenruo btrfs_err(fs_info, 284475cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 284575cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 284675cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 284775cb857dSQu Wenruo goto out; 284875cb857dSQu Wenruo } 284975cb857dSQu Wenruo out: 285075cb857dSQu Wenruo if (ret < 0) 285175cb857dSQu Wenruo btrfs_err(fs_info, 285275cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 285375cb857dSQu Wenruo return ret; 285475cb857dSQu Wenruo } 285575cb857dSQu Wenruo 2856bd676446SJosef Bacik static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) 2857bd676446SJosef Bacik { 2858bd676446SJosef Bacik int ret = 0; 2859bd676446SJosef Bacik 2860bd676446SJosef Bacik root->node = read_tree_block(root->fs_info, bytenr, 2861bd676446SJosef Bacik root->root_key.objectid, gen, level, NULL); 2862bd676446SJosef Bacik if (IS_ERR(root->node)) { 2863bd676446SJosef Bacik ret = PTR_ERR(root->node); 2864bd676446SJosef Bacik root->node = NULL; 28654eb150d6SQu Wenruo return ret; 28664eb150d6SQu Wenruo } 28674eb150d6SQu Wenruo if (!extent_buffer_uptodate(root->node)) { 2868bd676446SJosef Bacik free_extent_buffer(root->node); 2869bd676446SJosef Bacik root->node = NULL; 28704eb150d6SQu Wenruo return -EIO; 2871bd676446SJosef Bacik } 2872bd676446SJosef Bacik 2873bd676446SJosef Bacik btrfs_set_root_node(&root->root_item, root->node); 2874bd676446SJosef Bacik root->commit_root = btrfs_root_node(root); 2875bd676446SJosef Bacik btrfs_set_root_refs(&root->root_item, 1); 2876bd676446SJosef Bacik return ret; 2877bd676446SJosef Bacik } 2878bd676446SJosef Bacik 2879bd676446SJosef Bacik static int load_important_roots(struct btrfs_fs_info *fs_info) 2880bd676446SJosef Bacik { 2881bd676446SJosef Bacik struct btrfs_super_block *sb = fs_info->super_copy; 2882bd676446SJosef Bacik u64 gen, bytenr; 2883bd676446SJosef Bacik int level, ret; 2884bd676446SJosef Bacik 2885bd676446SJosef Bacik bytenr = btrfs_super_root(sb); 2886bd676446SJosef Bacik gen = btrfs_super_generation(sb); 2887bd676446SJosef Bacik level = btrfs_super_root_level(sb); 2888bd676446SJosef Bacik ret = load_super_root(fs_info->tree_root, bytenr, gen, level); 28899c54e80dSJosef Bacik if (ret) { 2890bd676446SJosef Bacik btrfs_warn(fs_info, "couldn't read tree root"); 2891bd676446SJosef Bacik return ret; 2892bd676446SJosef Bacik } 28939c54e80dSJosef Bacik return 0; 28949c54e80dSJosef Bacik } 28959c54e80dSJosef Bacik 28966ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2897b8522a1eSNikolay Borisov { 28986ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2899b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2900b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2901b8522a1eSNikolay Borisov bool handle_error = false; 2902b8522a1eSNikolay Borisov int ret = 0; 2903b8522a1eSNikolay Borisov int i; 2904b8522a1eSNikolay Borisov 2905b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2906b8522a1eSNikolay Borisov if (handle_error) { 2907b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2908b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2909b8522a1eSNikolay Borisov tree_root->node = NULL; 2910b8522a1eSNikolay Borisov 2911b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2912b8522a1eSNikolay Borisov break; 2913b8522a1eSNikolay Borisov 2914b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2915b8522a1eSNikolay Borisov 2916b8522a1eSNikolay Borisov /* 2917b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2918b8522a1eSNikolay Borisov * valid 2919b8522a1eSNikolay Borisov */ 2920b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2921b8522a1eSNikolay Borisov 2922b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2923b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2924b8522a1eSNikolay Borisov 2925b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 29266ef108ddSNikolay Borisov backup_index = ret; 2927b8522a1eSNikolay Borisov if (ret < 0) 2928b8522a1eSNikolay Borisov return ret; 2929b8522a1eSNikolay Borisov } 2930b8522a1eSNikolay Borisov 2931bd676446SJosef Bacik ret = load_important_roots(fs_info); 2932bd676446SJosef Bacik if (ret) { 2933217f5004SNikolay Borisov handle_error = true; 2934b8522a1eSNikolay Borisov continue; 2935b8522a1eSNikolay Borisov } 2936b8522a1eSNikolay Borisov 2937336a0d8dSNikolay Borisov /* 2938336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2939336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2940336a0d8dSNikolay Borisov */ 2941453e4873SNikolay Borisov ret = btrfs_init_root_free_objectid(tree_root); 2942b8522a1eSNikolay Borisov if (ret < 0) { 2943b8522a1eSNikolay Borisov handle_error = true; 2944b8522a1eSNikolay Borisov continue; 2945b8522a1eSNikolay Borisov } 2946b8522a1eSNikolay Borisov 29476b8fad57SNikolay Borisov ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); 2948b8522a1eSNikolay Borisov 2949b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2950b8522a1eSNikolay Borisov if (ret < 0) { 2951b8522a1eSNikolay Borisov handle_error = true; 2952b8522a1eSNikolay Borisov continue; 2953b8522a1eSNikolay Borisov } 2954b8522a1eSNikolay Borisov 2955b8522a1eSNikolay Borisov /* All successful */ 2956bd676446SJosef Bacik fs_info->generation = btrfs_header_generation(tree_root->node); 2957bd676446SJosef Bacik fs_info->last_trans_committed = fs_info->generation; 2958d96b3424SFilipe Manana fs_info->last_reloc_trans = 0; 29596ef108ddSNikolay Borisov 29606ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 29616ef108ddSNikolay Borisov if (backup_index < 0) { 29626ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 29636ef108ddSNikolay Borisov } else { 29646ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 29656ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 29666ef108ddSNikolay Borisov } 2967b8522a1eSNikolay Borisov break; 2968b8522a1eSNikolay Borisov } 2969b8522a1eSNikolay Borisov 2970b8522a1eSNikolay Borisov return ret; 2971b8522a1eSNikolay Borisov } 2972b8522a1eSNikolay Borisov 29738260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2974eb60ceacSChris Mason { 2975fc7cbcd4SDavid Sterba INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 297601cd3909SDavid Sterba INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 29778fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2978facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 297924bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2980eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 298111833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2982eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2983a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 2984fc7cbcd4SDavid Sterba spin_lock_init(&fs_info->fs_roots_radix_lock); 298524bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 29864cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2987ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2988f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 298947ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 299040ab3be1SNaohiro Aota spin_lock_init(&fs_info->treelog_bg_lock); 2991afba2bc0SNaohiro Aota spin_lock_init(&fs_info->zone_active_bgs_lock); 2992c2707a25SJohannes Thumshirn spin_lock_init(&fs_info->relocation_bg_lock); 2993f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2994abed4aaaSJosef Bacik rwlock_init(&fs_info->global_root_lock); 2995d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 2996f3372065SJohannes Thumshirn mutex_init(&fs_info->reclaim_bgs_lock); 29977585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2998573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 29990bc09ca1SNaohiro Aota mutex_init(&fs_info->zoned_meta_io_lock); 30005f0addf7SNaohiro Aota mutex_init(&fs_info->zoned_data_reloc_io_lock); 3001de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 300219c00ddcSChris Mason 3003e1489b4fSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); 30045a9ba670SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); 30058b53779eSIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); 30065f4403e1SIoannis Angelakopoulos btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); 30073e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, 30083e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMMIT_START); 30093e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, 30103e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_UNBLOCKED); 30113e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, 30123e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); 30133e738c53SIoannis Angelakopoulos btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, 30143e738c53SIoannis Angelakopoulos BTRFS_LOCKDEP_TRANS_COMPLETED); 3015e1489b4fSIoannis Angelakopoulos 30160b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 30176324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 3018f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 301947ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 302018bb8bbfSJohannes Thumshirn INIT_LIST_HEAD(&fs_info->reclaim_bgs); 3021afba2bc0SNaohiro Aota INIT_LIST_HEAD(&fs_info->zone_active_bgs); 3022bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 3023bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 30243fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 30253fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 3026bd647ce3SJosef Bacik #endif 3027c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 302866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 302966d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 303066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 303166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 303266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 303366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 303466d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 3035ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 3036ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 3037ba2c4d4eSJosef Bacik 3038771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 30394cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 3040034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 3041fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 3042abed4aaaSJosef Bacik fs_info->global_root_tree = RB_ROOT; 304395ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 30449ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 30454cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 3046a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 3047f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 30488b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 3049f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 3050fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 3051c8b97818SChris Mason 3052b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 3053b34b086cSChris Mason num_online_cpus() + 2, 8); 30540afbaf8cSChris Mason 3055199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 3056199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 305769fe2d75SJosef Bacik 3058638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 305921adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 306021adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 306121adbd5cSStefan Behrens #endif 3062779a65a4SEric Sandeen btrfs_init_balance(fs_info); 306357056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 3064a2de733cSArne Jansen 306516b0c258SFilipe Manana rwlock_init(&fs_info->block_group_cache_lock); 306608dddb29SFilipe Manana fs_info->block_group_cache_tree = RB_ROOT_CACHED; 30670f9dd46cSJosef Bacik 3068fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 3069fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 307039279cc3SChris Mason 30715a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 3072e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 3073925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 3074a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 3075a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 30761bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 30779e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 3078c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 307976dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 3080803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 3081fa9c0d79SChris Mason 3082ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 3083f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 3084b0643e59SDennis Zhou btrfs_discard_init(fs_info); 3085416ac51dSArne Jansen 3086fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 3087fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 3088fa9c0d79SChris Mason 3089e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 3090f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 3091bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 30924854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 3093034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 30943768f368SChris Mason 3095da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 3096da17066cSJeff Mahoney fs_info->nodesize = 4096; 3097da17066cSJeff Mahoney fs_info->sectorsize = 4096; 3098ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 3099da17066cSJeff Mahoney fs_info->stripesize = 4096; 3100da17066cSJeff Mahoney 3101f7b12a62SNaohiro Aota fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; 3102f7b12a62SNaohiro Aota 3103eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 3104eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 3105eede2bf3SOmar Sandoval 310618bb8bbfSJohannes Thumshirn fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; 310718bb8bbfSJohannes Thumshirn INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); 31088260edbaSJosef Bacik } 31098260edbaSJosef Bacik 31108260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 31118260edbaSJosef Bacik { 31128260edbaSJosef Bacik int ret; 31138260edbaSJosef Bacik 31148260edbaSJosef Bacik fs_info->sb = sb; 31158260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 31168260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 31179e967495SFilipe Manana 31185deb17e1SJosef Bacik ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); 3119ae18c37aSJosef Bacik if (ret) 3120ae18c37aSJosef Bacik return ret; 3121ae18c37aSJosef Bacik 3122ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 3123ae18c37aSJosef Bacik if (ret) 3124c75e8394SJosef Bacik return ret; 3125ae18c37aSJosef Bacik 3126ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 3127ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 3128ae18c37aSJosef Bacik 3129ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 3130ae18c37aSJosef Bacik if (ret) 3131c75e8394SJosef Bacik return ret; 3132ae18c37aSJosef Bacik 3133ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 3134ae18c37aSJosef Bacik GFP_KERNEL); 3135ae18c37aSJosef Bacik if (ret) 3136c75e8394SJosef Bacik return ret; 3137ae18c37aSJosef Bacik 3138ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 3139ae18c37aSJosef Bacik GFP_KERNEL); 3140c75e8394SJosef Bacik if (!fs_info->delayed_root) 3141c75e8394SJosef Bacik return -ENOMEM; 3142ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 3143ae18c37aSJosef Bacik 3144a0a1db70SFilipe Manana if (sb_rdonly(sb)) 3145a0a1db70SFilipe Manana set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 3146a0a1db70SFilipe Manana 3147c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 3148ae18c37aSJosef Bacik } 3149ae18c37aSJosef Bacik 315097f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 315197f4dd09SNikolay Borisov { 31520d031dc4SYu Zhe struct btrfs_fs_info *fs_info = data; 315397f4dd09SNikolay Borisov int ret; 315497f4dd09SNikolay Borisov 315597f4dd09SNikolay Borisov /* 315697f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 315797f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 315897f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 315997f4dd09SNikolay Borisov */ 316097f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 316197f4dd09SNikolay Borisov if (ret < 0) { 3162c94bec2cSJosef Bacik if (ret != -EINTR) 3163c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 3164c94bec2cSJosef Bacik ret); 316597f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 316697f4dd09SNikolay Borisov return ret; 316797f4dd09SNikolay Borisov } 316897f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 316997f4dd09SNikolay Borisov } 317097f4dd09SNikolay Borisov 317197f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 317297f4dd09SNikolay Borisov { 317397f4dd09SNikolay Borisov struct task_struct *task; 317497f4dd09SNikolay Borisov 317597f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 317697f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 317797f4dd09SNikolay Borisov if (IS_ERR(task)) { 317897f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 317997f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 318097f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 318197f4dd09SNikolay Borisov return PTR_ERR(task); 318297f4dd09SNikolay Borisov } 318397f4dd09SNikolay Borisov 318497f4dd09SNikolay Borisov return 0; 318597f4dd09SNikolay Borisov } 318697f4dd09SNikolay Borisov 318744c0ca21SBoris Burkov /* 31888cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 31898cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 31908cd29088SBoris Burkov * code paths. 31918cd29088SBoris Burkov */ 31928cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 31938cd29088SBoris Burkov { 31948cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 31958b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 31968cd29088SBoris Burkov } 31978cd29088SBoris Burkov 31988cd29088SBoris Burkov /* 319944c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 320044c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 320144c0ca21SBoris Burkov */ 320244c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 320344c0ca21SBoris Burkov { 320444c0ca21SBoris Burkov int ret; 320594846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 32068b228324SBoris Burkov bool clear_free_space_tree = false; 32078b228324SBoris Burkov 32088b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 32098b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 32108b228324SBoris Burkov clear_free_space_tree = true; 32118b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 32128b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 32138b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 32148b228324SBoris Burkov clear_free_space_tree = true; 32158b228324SBoris Burkov } 32168b228324SBoris Burkov 32178b228324SBoris Burkov if (clear_free_space_tree) { 32188b228324SBoris Burkov btrfs_info(fs_info, "clearing free space tree"); 32198b228324SBoris Burkov ret = btrfs_clear_free_space_tree(fs_info); 32208b228324SBoris Burkov if (ret) { 32218b228324SBoris Burkov btrfs_warn(fs_info, 32228b228324SBoris Burkov "failed to clear free space tree: %d", ret); 32238b228324SBoris Burkov goto out; 32248b228324SBoris Burkov } 32258b228324SBoris Burkov } 322644c0ca21SBoris Burkov 32278d488a8cSFilipe Manana /* 32288d488a8cSFilipe Manana * btrfs_find_orphan_roots() is responsible for finding all the dead 32298d488a8cSFilipe Manana * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load 3230fc7cbcd4SDavid Sterba * them into the fs_info->fs_roots_radix tree. This must be done before 32318d488a8cSFilipe Manana * calling btrfs_orphan_cleanup() on the tree root. If we don't do it 32328d488a8cSFilipe Manana * first, then btrfs_orphan_cleanup() will delete a dead root's orphan 32338d488a8cSFilipe Manana * item before the root's tree is deleted - this means that if we unmount 32348d488a8cSFilipe Manana * or crash before the deletion completes, on the next mount we will not 32358d488a8cSFilipe Manana * delete what remains of the tree because the orphan item does not 32368d488a8cSFilipe Manana * exists anymore, which is what tells us we have a pending deletion. 32378d488a8cSFilipe Manana */ 32388d488a8cSFilipe Manana ret = btrfs_find_orphan_roots(fs_info); 32398d488a8cSFilipe Manana if (ret) 32408d488a8cSFilipe Manana goto out; 32418d488a8cSFilipe Manana 324244c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 324344c0ca21SBoris Burkov if (ret) 324444c0ca21SBoris Burkov goto out; 324544c0ca21SBoris Burkov 32468f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 32478f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 32488f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 32498f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 32508f1c21d7SBoris Burkov goto out; 32518f1c21d7SBoris Burkov } 32528f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 32538f1c21d7SBoris Burkov 325444c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 32557eefae6bSJosef Bacik ret = btrfs_recover_relocation(fs_info); 325644c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 325744c0ca21SBoris Burkov if (ret < 0) { 325844c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 325944c0ca21SBoris Burkov goto out; 326044c0ca21SBoris Burkov } 326144c0ca21SBoris Burkov 32625011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 32635011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 32645011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 32655011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 32665011139aSBoris Burkov if (ret) { 32675011139aSBoris Burkov btrfs_warn(fs_info, 32685011139aSBoris Burkov "failed to create free space tree: %d", ret); 32695011139aSBoris Burkov goto out; 32705011139aSBoris Burkov } 32715011139aSBoris Burkov } 32725011139aSBoris Burkov 327394846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 327494846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 327594846229SBoris Burkov if (ret) 327694846229SBoris Burkov goto out; 327794846229SBoris Burkov } 327894846229SBoris Burkov 327944c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 328044c0ca21SBoris Burkov if (ret) 328144c0ca21SBoris Burkov goto out; 328244c0ca21SBoris Burkov 328344c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 328444c0ca21SBoris Burkov if (ret) { 328544c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 328644c0ca21SBoris Burkov goto out; 328744c0ca21SBoris Burkov } 328844c0ca21SBoris Burkov 328944c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 329044c0ca21SBoris Burkov 329144c0ca21SBoris Burkov if (!fs_info->uuid_root) { 329244c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 329344c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 329444c0ca21SBoris Burkov if (ret) { 329544c0ca21SBoris Burkov btrfs_warn(fs_info, 329644c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 329744c0ca21SBoris Burkov goto out; 329844c0ca21SBoris Burkov } 329944c0ca21SBoris Burkov } 330044c0ca21SBoris Burkov 330144c0ca21SBoris Burkov out: 330244c0ca21SBoris Burkov return ret; 330344c0ca21SBoris Burkov } 330444c0ca21SBoris Burkov 3305d7f67ac9SQu Wenruo /* 3306d7f67ac9SQu Wenruo * Do various sanity and dependency checks of different features. 3307d7f67ac9SQu Wenruo * 3308d7f67ac9SQu Wenruo * This is the place for less strict checks (like for subpage or artificial 3309d7f67ac9SQu Wenruo * feature dependencies). 3310d7f67ac9SQu Wenruo * 3311d7f67ac9SQu Wenruo * For strict checks or possible corruption detection, see 3312d7f67ac9SQu Wenruo * btrfs_validate_super(). 3313d7f67ac9SQu Wenruo * 3314d7f67ac9SQu Wenruo * This should be called after btrfs_parse_options(), as some mount options 3315d7f67ac9SQu Wenruo * (space cache related) can modify on-disk format like free space tree and 3316d7f67ac9SQu Wenruo * screw up certain feature dependencies. 3317d7f67ac9SQu Wenruo */ 3318d7f67ac9SQu Wenruo int btrfs_check_features(struct btrfs_fs_info *fs_info, struct super_block *sb) 3319d7f67ac9SQu Wenruo { 3320d7f67ac9SQu Wenruo struct btrfs_super_block *disk_super = fs_info->super_copy; 3321d7f67ac9SQu Wenruo u64 incompat = btrfs_super_incompat_flags(disk_super); 3322d7f67ac9SQu Wenruo const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); 3323d7f67ac9SQu Wenruo const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); 3324d7f67ac9SQu Wenruo 3325d7f67ac9SQu Wenruo if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 3326d7f67ac9SQu Wenruo btrfs_err(fs_info, 3327d7f67ac9SQu Wenruo "cannot mount because of unknown incompat features (0x%llx)", 3328d7f67ac9SQu Wenruo incompat); 3329d7f67ac9SQu Wenruo return -EINVAL; 3330d7f67ac9SQu Wenruo } 3331d7f67ac9SQu Wenruo 3332d7f67ac9SQu Wenruo /* Runtime limitation for mixed block groups. */ 3333d7f67ac9SQu Wenruo if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3334d7f67ac9SQu Wenruo (fs_info->sectorsize != fs_info->nodesize)) { 3335d7f67ac9SQu Wenruo btrfs_err(fs_info, 3336d7f67ac9SQu Wenruo "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 3337d7f67ac9SQu Wenruo fs_info->nodesize, fs_info->sectorsize); 3338d7f67ac9SQu Wenruo return -EINVAL; 3339d7f67ac9SQu Wenruo } 3340d7f67ac9SQu Wenruo 3341d7f67ac9SQu Wenruo /* Mixed backref is an always-enabled feature. */ 3342d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 3343d7f67ac9SQu Wenruo 3344d7f67ac9SQu Wenruo /* Set compression related flags just in case. */ 3345d7f67ac9SQu Wenruo if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3346d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 3347d7f67ac9SQu Wenruo else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 3348d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3349d7f67ac9SQu Wenruo 3350d7f67ac9SQu Wenruo /* 3351d7f67ac9SQu Wenruo * An ancient flag, which should really be marked deprecated. 3352d7f67ac9SQu Wenruo * Such runtime limitation doesn't really need a incompat flag. 3353d7f67ac9SQu Wenruo */ 3354d7f67ac9SQu Wenruo if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) 3355d7f67ac9SQu Wenruo incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3356d7f67ac9SQu Wenruo 3357d7f67ac9SQu Wenruo if (compat_ro_unsupp && !sb_rdonly(sb)) { 3358d7f67ac9SQu Wenruo btrfs_err(fs_info, 3359d7f67ac9SQu Wenruo "cannot mount read-write because of unknown compat_ro features (0x%llx)", 3360d7f67ac9SQu Wenruo compat_ro); 3361d7f67ac9SQu Wenruo return -EINVAL; 3362d7f67ac9SQu Wenruo } 3363d7f67ac9SQu Wenruo 3364d7f67ac9SQu Wenruo /* 3365d7f67ac9SQu Wenruo * We have unsupported RO compat features, although RO mounted, we 3366d7f67ac9SQu Wenruo * should not cause any metadata writes, including log replay. 3367d7f67ac9SQu Wenruo * Or we could screw up whatever the new feature requires. 3368d7f67ac9SQu Wenruo */ 3369d7f67ac9SQu Wenruo if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && 3370d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3371d7f67ac9SQu Wenruo btrfs_err(fs_info, 3372d7f67ac9SQu Wenruo "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", 3373d7f67ac9SQu Wenruo compat_ro); 3374d7f67ac9SQu Wenruo return -EINVAL; 3375d7f67ac9SQu Wenruo } 3376d7f67ac9SQu Wenruo 3377d7f67ac9SQu Wenruo /* 3378d7f67ac9SQu Wenruo * Artificial limitations for block group tree, to force 3379d7f67ac9SQu Wenruo * block-group-tree to rely on no-holes and free-space-tree. 3380d7f67ac9SQu Wenruo */ 3381d7f67ac9SQu Wenruo if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && 3382d7f67ac9SQu Wenruo (!btrfs_fs_incompat(fs_info, NO_HOLES) || 3383d7f67ac9SQu Wenruo !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { 3384d7f67ac9SQu Wenruo btrfs_err(fs_info, 3385d7f67ac9SQu Wenruo "block-group-tree feature requires no-holes and free-space-tree features"); 3386d7f67ac9SQu Wenruo return -EINVAL; 3387d7f67ac9SQu Wenruo } 3388d7f67ac9SQu Wenruo 3389d7f67ac9SQu Wenruo /* 3390d7f67ac9SQu Wenruo * Subpage runtime limitation on v1 cache. 3391d7f67ac9SQu Wenruo * 3392d7f67ac9SQu Wenruo * V1 space cache still has some hard codeed PAGE_SIZE usage, while 3393d7f67ac9SQu Wenruo * we're already defaulting to v2 cache, no need to bother v1 as it's 3394d7f67ac9SQu Wenruo * going to be deprecated anyway. 3395d7f67ac9SQu Wenruo */ 3396d7f67ac9SQu Wenruo if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { 3397d7f67ac9SQu Wenruo btrfs_warn(fs_info, 3398d7f67ac9SQu Wenruo "v1 space cache is not supported for page size %lu with sectorsize %u", 3399d7f67ac9SQu Wenruo PAGE_SIZE, fs_info->sectorsize); 3400d7f67ac9SQu Wenruo return -EINVAL; 3401d7f67ac9SQu Wenruo } 3402d7f67ac9SQu Wenruo 3403d7f67ac9SQu Wenruo /* This can be called by remount, we need to protect the super block. */ 3404d7f67ac9SQu Wenruo spin_lock(&fs_info->super_lock); 3405d7f67ac9SQu Wenruo btrfs_set_super_incompat_flags(disk_super, incompat); 3406d7f67ac9SQu Wenruo spin_unlock(&fs_info->super_lock); 3407d7f67ac9SQu Wenruo 3408d7f67ac9SQu Wenruo return 0; 3409d7f67ac9SQu Wenruo } 3410d7f67ac9SQu Wenruo 3411ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 3412ae18c37aSJosef Bacik char *options) 3413ae18c37aSJosef Bacik { 3414ae18c37aSJosef Bacik u32 sectorsize; 3415ae18c37aSJosef Bacik u32 nodesize; 3416ae18c37aSJosef Bacik u32 stripesize; 3417ae18c37aSJosef Bacik u64 generation; 3418ae18c37aSJosef Bacik u64 features; 3419ae18c37aSJosef Bacik u16 csum_type; 3420ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 3421ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 3422ae18c37aSJosef Bacik struct btrfs_root *tree_root; 3423ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 3424ae18c37aSJosef Bacik int ret; 3425ae18c37aSJosef Bacik int err = -EINVAL; 3426ae18c37aSJosef Bacik int level; 3427ae18c37aSJosef Bacik 34288260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 342953b381b3SDavid Woodhouse if (ret) { 343083c8266aSDavid Sterba err = ret; 3431ae18c37aSJosef Bacik goto fail; 343253b381b3SDavid Woodhouse } 343353b381b3SDavid Woodhouse 3434ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3435ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3436ae18c37aSJosef Bacik GFP_KERNEL); 3437ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3438ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3439ae18c37aSJosef Bacik GFP_KERNEL); 3440ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3441ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 3442ae18c37aSJosef Bacik err = -ENOMEM; 3443c75e8394SJosef Bacik goto fail; 3444ae18c37aSJosef Bacik } 3445ae18c37aSJosef Bacik 3446ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 3447ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 3448ae18c37aSJosef Bacik err = -ENOMEM; 3449c75e8394SJosef Bacik goto fail; 3450ae18c37aSJosef Bacik } 3451ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 3452ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 3453ae18c37aSJosef Bacik 3454d24fa5c1SAnand Jain invalidate_bdev(fs_devices->latest_dev->bdev); 34551104a885SDavid Sterba 34561104a885SDavid Sterba /* 34571104a885SDavid Sterba * Read super block and check the signature bytes only 34581104a885SDavid Sterba */ 3459d24fa5c1SAnand Jain disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); 34608f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 34618f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 346216cdcec7SMiao Xie goto fail_alloc; 346320b45077SDave Young } 346439279cc3SChris Mason 34651104a885SDavid Sterba /* 3466260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 34678dc3f22cSJohannes Thumshirn * corrupted, we'll find out 34688dc3f22cSJohannes Thumshirn */ 34698f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 347051bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 34718dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 347251bce6c9SJohannes Thumshirn csum_type); 34738dc3f22cSJohannes Thumshirn err = -EINVAL; 34748f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 34758dc3f22cSJohannes Thumshirn goto fail_alloc; 34768dc3f22cSJohannes Thumshirn } 34778dc3f22cSJohannes Thumshirn 347883c68bbcSSu Yue fs_info->csum_size = btrfs_super_csum_size(disk_super); 347983c68bbcSSu Yue 34806d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 34816d97c6e3SJohannes Thumshirn if (ret) { 34826d97c6e3SJohannes Thumshirn err = ret; 34838f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 34846d97c6e3SJohannes Thumshirn goto fail_alloc; 34856d97c6e3SJohannes Thumshirn } 34866d97c6e3SJohannes Thumshirn 34878dc3f22cSJohannes Thumshirn /* 34881104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 34891104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 34901104a885SDavid Sterba */ 34913d17adeaSQu Wenruo if (btrfs_check_super_csum(fs_info, disk_super)) { 349205135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 34931104a885SDavid Sterba err = -EINVAL; 34948f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3495141386e1SJosef Bacik goto fail_alloc; 34961104a885SDavid Sterba } 34971104a885SDavid Sterba 34981104a885SDavid Sterba /* 34991104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 35001104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 35011104a885SDavid Sterba * the whole block of INFO_SIZE 35021104a885SDavid Sterba */ 35038f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 35048f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 35055f39d397SChris Mason 3506fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3507fbc6feaeSNikolay Borisov 35080b86a832SChris Mason 3509fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3510fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3511fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3512fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3513fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3514fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3515fbc6feaeSNikolay Borisov } 3516fbc6feaeSNikolay Borisov 3517fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3518fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3519de37aa51SNikolay Borisov 3520069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 35211104a885SDavid Sterba if (ret) { 352205135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 35231104a885SDavid Sterba err = -EINVAL; 3524141386e1SJosef Bacik goto fail_alloc; 35251104a885SDavid Sterba } 35261104a885SDavid Sterba 35270f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3528141386e1SJosef Bacik goto fail_alloc; 35290f7d52f4SChris Mason 3530acce952bSliubo /* check FS state, whether FS is broken. */ 353187533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 353287533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3533acce952bSliubo 353475e7cb7fSLiu Bo /* 353575e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 353675e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 353775e7cb7fSLiu Bo */ 353875e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 353975e7cb7fSLiu Bo 35406f93e834SAnand Jain 35416f93e834SAnand Jain /* Set up fs_info before parsing mount options */ 35426f93e834SAnand Jain nodesize = btrfs_super_nodesize(disk_super); 35436f93e834SAnand Jain sectorsize = btrfs_super_sectorsize(disk_super); 35446f93e834SAnand Jain stripesize = sectorsize; 35456f93e834SAnand Jain fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 35466f93e834SAnand Jain fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 35476f93e834SAnand Jain 35486f93e834SAnand Jain fs_info->nodesize = nodesize; 35496f93e834SAnand Jain fs_info->sectorsize = sectorsize; 35506f93e834SAnand Jain fs_info->sectorsize_bits = ilog2(sectorsize); 35516f93e834SAnand Jain fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 35526f93e834SAnand Jain fs_info->stripesize = stripesize; 35536f93e834SAnand Jain 35542ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 35552b82032cSYan Zheng if (ret) { 35562b82032cSYan Zheng err = ret; 3557141386e1SJosef Bacik goto fail_alloc; 35582b82032cSYan Zheng } 3559dfe25020SChris Mason 3560d7f67ac9SQu Wenruo ret = btrfs_check_features(fs_info, sb); 3561d7f67ac9SQu Wenruo if (ret < 0) { 3562d7f67ac9SQu Wenruo err = ret; 3563141386e1SJosef Bacik goto fail_alloc; 3564f2b636e8SJosef Bacik } 3565f2b636e8SJosef Bacik 35668481dd80SQu Wenruo if (sectorsize < PAGE_SIZE) { 35678481dd80SQu Wenruo struct btrfs_subpage_info *subpage_info; 35688481dd80SQu Wenruo 35699f73f1aeSQu Wenruo /* 35709f73f1aeSQu Wenruo * V1 space cache has some hardcoded PAGE_SIZE usage, and is 35719f73f1aeSQu Wenruo * going to be deprecated. 35729f73f1aeSQu Wenruo * 35739f73f1aeSQu Wenruo * Force to use v2 cache for subpage case. 35749f73f1aeSQu Wenruo */ 35759f73f1aeSQu Wenruo btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 35769f73f1aeSQu Wenruo btrfs_set_and_info(fs_info, FREE_SPACE_TREE, 35779f73f1aeSQu Wenruo "forcing free space tree for sector size %u with page size %lu", 35789f73f1aeSQu Wenruo sectorsize, PAGE_SIZE); 35799f73f1aeSQu Wenruo 358095ea0486SQu Wenruo btrfs_warn(fs_info, 358195ea0486SQu Wenruo "read-write for sector size %u with page size %lu is experimental", 35820bb3eb3eSQu Wenruo sectorsize, PAGE_SIZE); 35838481dd80SQu Wenruo subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); 35848481dd80SQu Wenruo if (!subpage_info) 35858481dd80SQu Wenruo goto fail_alloc; 35868481dd80SQu Wenruo btrfs_init_subpage_info(subpage_info, sectorsize); 35878481dd80SQu Wenruo fs_info->subpage_info = subpage_info; 3588c8050b3bSQu Wenruo } 35890bb3eb3eSQu Wenruo 3590d21deec5SSu Yue ret = btrfs_init_workqueues(fs_info); 35912a458198SEric Sandeen if (ret) { 35922a458198SEric Sandeen err = ret; 35930dc3b84aSJosef Bacik goto fail_sb_buffer; 35940dc3b84aSJosef Bacik } 35954543df7eSChris Mason 35969e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 35979e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 35984575c9ccSChris Mason 3599a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3600a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3601de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3602db94535dSChris Mason 3603925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 36046bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3605925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 360684eed90fSChris Mason if (ret) { 360705135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 36085d4f98a2SYan Zheng goto fail_sb_buffer; 360984eed90fSChris Mason } 36100b86a832SChris Mason 361184234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3612581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 3613bd676446SJosef Bacik ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), 3614bd676446SJosef Bacik generation, level); 3615bd676446SJosef Bacik if (ret) { 361605135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3617af31f5e5SChris Mason goto fail_tree_roots; 361883121942SDavid Woodhouse } 36190b86a832SChris Mason 3620e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3621c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3622c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3623e17cade2SChris Mason 36245b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 36252b82032cSYan Zheng if (ret) { 362605135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3627af31f5e5SChris Mason goto fail_tree_roots; 36282b82032cSYan Zheng } 36290b86a832SChris Mason 36308dabb742SStefan Behrens /* 3631bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3632bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3633bacce86aSAnand Jain * target is required. So free devices that are not part of this 36341a9fd417SDavid Sterba * filesystem but skip the replace target device which is checked 3635bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 36368dabb742SStefan Behrens */ 3637bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3638d24fa5c1SAnand Jain if (!fs_devices->latest_dev->bdev) { 363905135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3640a6b0d5c8SChris Mason goto fail_tree_roots; 3641a6b0d5c8SChris Mason } 3642a6b0d5c8SChris Mason 3643b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 36444bbcaa64SEric Sandeen if (ret) 3645b8522a1eSNikolay Borisov goto fail_tree_roots; 36468929ecfaSYan, Zheng 364775ec1db8SJosef Bacik /* 364873651042SNaohiro Aota * Get zone type information of zoned block devices. This will also 364973651042SNaohiro Aota * handle emulation of a zoned filesystem if a regular device has the 365073651042SNaohiro Aota * zoned incompat feature flag set. 365173651042SNaohiro Aota */ 365273651042SNaohiro Aota ret = btrfs_get_dev_zone_info_all_devices(fs_info); 365373651042SNaohiro Aota if (ret) { 365473651042SNaohiro Aota btrfs_err(fs_info, 365573651042SNaohiro Aota "zoned: failed to read device zone info: %d", 365673651042SNaohiro Aota ret); 365773651042SNaohiro Aota goto fail_block_groups; 365873651042SNaohiro Aota } 365973651042SNaohiro Aota 366073651042SNaohiro Aota /* 366175ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 366275ec1db8SJosef Bacik * check the generation here so we can set the 366375ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 366475ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 366575ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 366675ec1db8SJosef Bacik * even though it was perfectly fine. 366775ec1db8SJosef Bacik */ 366875ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 366975ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 367075ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 367175ec1db8SJosef Bacik 3672cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3673cf90d884SQu Wenruo if (ret) { 3674cf90d884SQu Wenruo btrfs_err(fs_info, 3675cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3676cf90d884SQu Wenruo ret); 3677cf90d884SQu Wenruo goto fail_block_groups; 3678cf90d884SQu Wenruo } 367968310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 368068310a5eSIlya Dryomov if (ret) { 368105135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 368268310a5eSIlya Dryomov goto fail_block_groups; 368368310a5eSIlya Dryomov } 368468310a5eSIlya Dryomov 3685733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3686733f4fbbSStefan Behrens if (ret) { 368705135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3688733f4fbbSStefan Behrens goto fail_block_groups; 3689733f4fbbSStefan Behrens } 3690733f4fbbSStefan Behrens 36918dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 36928dabb742SStefan Behrens if (ret) { 369305135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 36948dabb742SStefan Behrens goto fail_block_groups; 36958dabb742SStefan Behrens } 36968dabb742SStefan Behrens 3697b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3698b70f5097SNaohiro Aota if (ret) { 3699b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3700b70f5097SNaohiro Aota ret); 3701b70f5097SNaohiro Aota goto fail_block_groups; 3702b70f5097SNaohiro Aota } 3703b70f5097SNaohiro Aota 3704c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3705b7c35e81SAnand Jain if (ret) { 370605135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 370705135f59SDavid Sterba ret); 3708b7c35e81SAnand Jain goto fail_block_groups; 3709b7c35e81SAnand Jain } 3710b7c35e81SAnand Jain 371196f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 37125ac1d209SJeff Mahoney if (ret) { 371305135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3714b7c35e81SAnand Jain goto fail_fsdev_sysfs; 37155ac1d209SJeff Mahoney } 37165ac1d209SJeff Mahoney 3717c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3718c59021f8Sliubo if (ret) { 371905135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 37202365dd3cSAnand Jain goto fail_sysfs; 3721c59021f8Sliubo } 3722c59021f8Sliubo 37235b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 37241b1d1f66SJosef Bacik if (ret) { 372505135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 37262365dd3cSAnand Jain goto fail_sysfs; 37271b1d1f66SJosef Bacik } 37284330e183SQu Wenruo 372916beac87SNaohiro Aota btrfs_free_zone_cache(fs_info); 373016beac87SNaohiro Aota 37315c78a5e7SAnand Jain if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && 37325c78a5e7SAnand Jain !btrfs_check_rw_degradable(fs_info, NULL)) { 373305135f59SDavid Sterba btrfs_warn(fs_info, 373452042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 37352365dd3cSAnand Jain goto fail_sysfs; 3736292fd7fcSStefan Behrens } 37379078a3e1SChris Mason 373833c44184SJosef Bacik fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, 3739a74a4b97SChris Mason "btrfs-cleaner"); 374057506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 37412365dd3cSAnand Jain goto fail_sysfs; 3742a74a4b97SChris Mason 3743a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3744a74a4b97SChris Mason tree_root, 3745a74a4b97SChris Mason "btrfs-transaction"); 374657506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 37473f157a2fSChris Mason goto fail_cleaner; 3748a74a4b97SChris Mason 3749583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3750c289811cSChris Mason !fs_info->fs_devices->rotating) { 3751583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3752c289811cSChris Mason } 3753c289811cSChris Mason 3754572d9ab7SDavid Sterba /* 375563a7cb13SDavid Sterba * For devices supporting discard turn on discard=async automatically, 375663a7cb13SDavid Sterba * unless it's already set or disabled. This could be turned off by 375763a7cb13SDavid Sterba * nodiscard for the same mount. 375863a7cb13SDavid Sterba */ 375963a7cb13SDavid Sterba if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 376063a7cb13SDavid Sterba btrfs_test_opt(fs_info, DISCARD_ASYNC) || 376163a7cb13SDavid Sterba btrfs_test_opt(fs_info, NODISCARD)) && 376263a7cb13SDavid Sterba fs_info->fs_devices->discardable) { 376363a7cb13SDavid Sterba btrfs_set_and_info(fs_info, DISCARD_ASYNC, 376463a7cb13SDavid Sterba "auto enabling async discard"); 376563a7cb13SDavid Sterba btrfs_clear_opt(fs_info->mount_opt, NODISCARD); 376663a7cb13SDavid Sterba } 376763a7cb13SDavid Sterba 376821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 37690b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 37702ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 37710b246afaSJeff Mahoney btrfs_test_opt(fs_info, 3772cbeaae4fSDavid Sterba CHECK_INTEGRITY_DATA) ? 1 : 0, 377321adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 377421adbd5cSStefan Behrens if (ret) 377505135f59SDavid Sterba btrfs_warn(fs_info, 377605135f59SDavid Sterba "failed to initialize integrity check module: %d", 377705135f59SDavid Sterba ret); 377821adbd5cSStefan Behrens } 377921adbd5cSStefan Behrens #endif 3780bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3781bcef60f2SArne Jansen if (ret) 3782bcef60f2SArne Jansen goto fail_trans_kthread; 378321adbd5cSStefan Behrens 3784fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3785fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3786fd708b81SJosef Bacik 378796da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 378896da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 37890b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3790e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 379163443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 379279787eaaSJeff Mahoney if (ret) { 379363443bf5SEric Sandeen err = ret; 379428c16cbbSWang Shilong goto fail_qgroup; 3795e556ce2cSYan Zheng } 3796e02119d5SChris Mason } 37971a40e23bSZheng Yan 379856e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 37993140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 38003140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3801f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3802315bf8efSJosef Bacik fs_info->fs_root = NULL; 3803bcef60f2SArne Jansen goto fail_qgroup; 38043140c9a3SDan Carpenter } 3805c289811cSChris Mason 3806bc98a42cSDavid Howells if (sb_rdonly(sb)) 38078cd29088SBoris Burkov goto clear_oneshot; 38082b6ba629SIlya Dryomov 380944c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 38102b6ba629SIlya Dryomov if (ret) { 38116bccf3abSJeff Mahoney close_ctree(fs_info); 38122b6ba629SIlya Dryomov return ret; 3813e3acc2a6SJosef Bacik } 3814b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3815b382a324SJan Schmidt 381644c0ca21SBoris Burkov if (fs_info->uuid_root && 381744c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 381844c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 381905135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 382070f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 382170f80175SStefan Behrens if (ret) { 382205135f59SDavid Sterba btrfs_warn(fs_info, 382305135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 38246bccf3abSJeff Mahoney close_ctree(fs_info); 382570f80175SStefan Behrens return ret; 382670f80175SStefan Behrens } 3827f7a81ea4SStefan Behrens } 382894846229SBoris Burkov 3829afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 383047ab2a6cSJosef Bacik 3831b4be6aefSJosef Bacik /* Kick the cleaner thread so it'll start deleting snapshots. */ 3832b4be6aefSJosef Bacik if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) 3833b4be6aefSJosef Bacik wake_up_process(fs_info->cleaner_kthread); 3834b4be6aefSJosef Bacik 38358cd29088SBoris Burkov clear_oneshot: 38368cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3837ad2b2c80SAl Viro return 0; 383839279cc3SChris Mason 3839bcef60f2SArne Jansen fail_qgroup: 3840bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 38417c2ca468SChris Mason fail_trans_kthread: 38427c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 38432ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3844faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 38453f157a2fSChris Mason fail_cleaner: 3846a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 38477c2ca468SChris Mason 38487c2ca468SChris Mason /* 38497c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 38507c2ca468SChris Mason * kthreads 38517c2ca468SChris Mason */ 38527c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 38537c2ca468SChris Mason 38542365dd3cSAnand Jain fail_sysfs: 38556618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 38562365dd3cSAnand Jain 3857b7c35e81SAnand Jain fail_fsdev_sysfs: 3858b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3859b7c35e81SAnand Jain 38601b1d1f66SJosef Bacik fail_block_groups: 386154067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3862af31f5e5SChris Mason 3863af31f5e5SChris Mason fail_tree_roots: 38649e3aa805SJosef Bacik if (fs_info->data_reloc_root) 38659e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 38664273eaffSAnand Jain free_root_pointers(fs_info, true); 38672b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3868af31f5e5SChris Mason 386939279cc3SChris Mason fail_sb_buffer: 38707abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 38715cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 387216cdcec7SMiao Xie fail_alloc: 3873586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3874586e46e2SIlya Dryomov 38754543df7eSChris Mason iput(fs_info->btree_inode); 38767e662854SQinghuang Feng fail: 3877586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3878ad2b2c80SAl Viro return err; 3879eb60ceacSChris Mason } 3880663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3881eb60ceacSChris Mason 3882314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3883f2984462SChris Mason { 3884314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3885314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3886314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3887314b6dd0SJohannes Thumshirn struct page *page; 3888442a4f63SStefan Behrens 3889314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3890314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3891314b6dd0SJohannes Thumshirn 3892314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3893fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3894314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3895314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3896314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3897314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3898314b6dd0SJohannes Thumshirn SetPageError(page); 3899314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3900314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3901314b6dd0SJohannes Thumshirn } else { 3902314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3903f2984462SChris Mason } 3904314b6dd0SJohannes Thumshirn 3905314b6dd0SJohannes Thumshirn put_page(page); 3906314b6dd0SJohannes Thumshirn unlock_page(page); 3907314b6dd0SJohannes Thumshirn } 3908314b6dd0SJohannes Thumshirn 3909314b6dd0SJohannes Thumshirn bio_put(bio); 3910f2984462SChris Mason } 3911f2984462SChris Mason 39128f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 3913a05d3c91SQu Wenruo int copy_num, bool drop_cache) 391429c36d72SAnand Jain { 391529c36d72SAnand Jain struct btrfs_super_block *super; 39168f32380dSJohannes Thumshirn struct page *page; 391712659251SNaohiro Aota u64 bytenr, bytenr_orig; 39188f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 391912659251SNaohiro Aota int ret; 392029c36d72SAnand Jain 392112659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 392212659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 392312659251SNaohiro Aota if (ret == -ENOENT) 392412659251SNaohiro Aota return ERR_PTR(-EINVAL); 392512659251SNaohiro Aota else if (ret) 392612659251SNaohiro Aota return ERR_PTR(ret); 392712659251SNaohiro Aota 3928cda00ebaSChristoph Hellwig if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) 39298f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 393029c36d72SAnand Jain 3931a05d3c91SQu Wenruo if (drop_cache) { 3932a05d3c91SQu Wenruo /* This should only be called with the primary sb. */ 3933a05d3c91SQu Wenruo ASSERT(copy_num == 0); 3934a05d3c91SQu Wenruo 3935a05d3c91SQu Wenruo /* 3936a05d3c91SQu Wenruo * Drop the page of the primary superblock, so later read will 3937a05d3c91SQu Wenruo * always read from the device. 3938a05d3c91SQu Wenruo */ 3939a05d3c91SQu Wenruo invalidate_inode_pages2_range(mapping, 3940a05d3c91SQu Wenruo bytenr >> PAGE_SHIFT, 3941a05d3c91SQu Wenruo (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); 3942a05d3c91SQu Wenruo } 3943a05d3c91SQu Wenruo 39448f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 39458f32380dSJohannes Thumshirn if (IS_ERR(page)) 39468f32380dSJohannes Thumshirn return ERR_CAST(page); 394729c36d72SAnand Jain 39488f32380dSJohannes Thumshirn super = page_address(page); 394996c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 395096c2e067SAnand Jain btrfs_release_disk_super(super); 395196c2e067SAnand Jain return ERR_PTR(-ENODATA); 395296c2e067SAnand Jain } 395396c2e067SAnand Jain 395412659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 39558f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 39568f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 395729c36d72SAnand Jain } 395829c36d72SAnand Jain 39598f32380dSJohannes Thumshirn return super; 396029c36d72SAnand Jain } 396129c36d72SAnand Jain 396229c36d72SAnand Jain 39638f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3964a512bbf8SYan Zheng { 39658f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3966a512bbf8SYan Zheng int i; 3967a512bbf8SYan Zheng u64 transid = 0; 3968a512bbf8SYan Zheng 3969a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3970a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3971a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3972a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3973a512bbf8SYan Zheng */ 3974a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3975a05d3c91SQu Wenruo super = btrfs_read_dev_one_super(bdev, i, false); 39768f32380dSJohannes Thumshirn if (IS_ERR(super)) 3977a512bbf8SYan Zheng continue; 3978a512bbf8SYan Zheng 3979a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 39808f32380dSJohannes Thumshirn if (latest) 39818f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 39828f32380dSJohannes Thumshirn 39838f32380dSJohannes Thumshirn latest = super; 3984a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3985a512bbf8SYan Zheng } 3986a512bbf8SYan Zheng } 398792fc03fbSAnand Jain 39888f32380dSJohannes Thumshirn return super; 3989a512bbf8SYan Zheng } 3990a512bbf8SYan Zheng 39914eedeb75SHisashi Hifumi /* 3992abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3993314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 39944eedeb75SHisashi Hifumi * 3995abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3996abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3997abbb3b8eSDavid Sterba * same for write and wait phases. 39984eedeb75SHisashi Hifumi * 3999314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 40004eedeb75SHisashi Hifumi */ 4001a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 4002abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 4003a512bbf8SYan Zheng { 4004d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 4005314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 4006d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 4007a512bbf8SYan Zheng int i; 4008a512bbf8SYan Zheng int errors = 0; 400912659251SNaohiro Aota int ret; 401012659251SNaohiro Aota u64 bytenr, bytenr_orig; 4011a512bbf8SYan Zheng 4012a512bbf8SYan Zheng if (max_mirrors == 0) 4013a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4014a512bbf8SYan Zheng 4015d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 4016d5178578SJohannes Thumshirn 4017a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 4018314b6dd0SJohannes Thumshirn struct page *page; 4019314b6dd0SJohannes Thumshirn struct bio *bio; 4020314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 4021314b6dd0SJohannes Thumshirn 402212659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 402312659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 402412659251SNaohiro Aota if (ret == -ENOENT) { 402512659251SNaohiro Aota continue; 402612659251SNaohiro Aota } else if (ret < 0) { 402712659251SNaohiro Aota btrfs_err(device->fs_info, 402812659251SNaohiro Aota "couldn't get super block location for mirror %d", 402912659251SNaohiro Aota i); 403012659251SNaohiro Aota errors++; 403112659251SNaohiro Aota continue; 403212659251SNaohiro Aota } 4033935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4034935e5cc9SMiao Xie device->commit_total_bytes) 4035a512bbf8SYan Zheng break; 4036a512bbf8SYan Zheng 403712659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 4038a512bbf8SYan Zheng 4039fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 4040fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 4041fd08001fSEric Biggers sb->csum); 4042a512bbf8SYan Zheng 4043314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 4044314b6dd0SJohannes Thumshirn GFP_NOFS); 4045314b6dd0SJohannes Thumshirn if (!page) { 4046fb456252SJeff Mahoney btrfs_err(device->fs_info, 4047314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 4048f14d104dSDavid Sterba bytenr); 4049634554dcSJosef Bacik errors++; 4050634554dcSJosef Bacik continue; 4051634554dcSJosef Bacik } 4052634554dcSJosef Bacik 4053314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 4054314b6dd0SJohannes Thumshirn get_page(page); 4055a512bbf8SYan Zheng 4056314b6dd0SJohannes Thumshirn disk_super = page_address(page); 4057314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 4058a512bbf8SYan Zheng 4059387125fcSChris Mason /* 4060314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 4061314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 4062314b6dd0SJohannes Thumshirn * checking. 4063387125fcSChris Mason */ 406407888c66SChristoph Hellwig bio = bio_alloc(device->bdev, 1, 406507888c66SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, 406607888c66SChristoph Hellwig GFP_NOFS); 4067314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 4068314b6dd0SJohannes Thumshirn bio->bi_private = device; 4069314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 4070314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 4071314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 4072314b6dd0SJohannes Thumshirn 4073314b6dd0SJohannes Thumshirn /* 4074314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 4075314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 4076314b6dd0SJohannes Thumshirn * copies might still contain the older version. 4077314b6dd0SJohannes Thumshirn */ 40781b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 4079314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 4080314b6dd0SJohannes Thumshirn 408158ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 408258ff51f1SChristoph Hellwig submit_bio(bio); 40838376d9e1SNaohiro Aota 40848376d9e1SNaohiro Aota if (btrfs_advance_sb_log(device, i)) 40858376d9e1SNaohiro Aota errors++; 4086a512bbf8SYan Zheng } 4087a512bbf8SYan Zheng return errors < i ? 0 : -1; 4088a512bbf8SYan Zheng } 4089a512bbf8SYan Zheng 4090387125fcSChris Mason /* 4091abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 4092abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 4093abbb3b8eSDavid Sterba * 4094314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 4095abbb3b8eSDavid Sterba * date. 4096abbb3b8eSDavid Sterba */ 4097abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 4098abbb3b8eSDavid Sterba { 4099abbb3b8eSDavid Sterba int i; 4100abbb3b8eSDavid Sterba int errors = 0; 4101b6a535faSHoward McLauchlan bool primary_failed = false; 410212659251SNaohiro Aota int ret; 4103abbb3b8eSDavid Sterba u64 bytenr; 4104abbb3b8eSDavid Sterba 4105abbb3b8eSDavid Sterba if (max_mirrors == 0) 4106abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 4107abbb3b8eSDavid Sterba 4108abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 4109314b6dd0SJohannes Thumshirn struct page *page; 4110314b6dd0SJohannes Thumshirn 411112659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 411212659251SNaohiro Aota if (ret == -ENOENT) { 411312659251SNaohiro Aota break; 411412659251SNaohiro Aota } else if (ret < 0) { 411512659251SNaohiro Aota errors++; 411612659251SNaohiro Aota if (i == 0) 411712659251SNaohiro Aota primary_failed = true; 411812659251SNaohiro Aota continue; 411912659251SNaohiro Aota } 4120abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 4121abbb3b8eSDavid Sterba device->commit_total_bytes) 4122abbb3b8eSDavid Sterba break; 4123abbb3b8eSDavid Sterba 4124314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 4125314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 4126314b6dd0SJohannes Thumshirn if (!page) { 4127abbb3b8eSDavid Sterba errors++; 4128b6a535faSHoward McLauchlan if (i == 0) 4129b6a535faSHoward McLauchlan primary_failed = true; 4130abbb3b8eSDavid Sterba continue; 4131abbb3b8eSDavid Sterba } 4132314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 4133314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 4134314b6dd0SJohannes Thumshirn if (PageError(page)) { 4135abbb3b8eSDavid Sterba errors++; 4136b6a535faSHoward McLauchlan if (i == 0) 4137b6a535faSHoward McLauchlan primary_failed = true; 4138b6a535faSHoward McLauchlan } 4139abbb3b8eSDavid Sterba 4140314b6dd0SJohannes Thumshirn /* Drop our reference */ 4141314b6dd0SJohannes Thumshirn put_page(page); 4142abbb3b8eSDavid Sterba 4143314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 4144314b6dd0SJohannes Thumshirn put_page(page); 4145abbb3b8eSDavid Sterba } 4146abbb3b8eSDavid Sterba 4147b6a535faSHoward McLauchlan /* log error, force error return */ 4148b6a535faSHoward McLauchlan if (primary_failed) { 4149b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 4150b6a535faSHoward McLauchlan device->devid); 4151b6a535faSHoward McLauchlan return -1; 4152b6a535faSHoward McLauchlan } 4153b6a535faSHoward McLauchlan 4154abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 4155abbb3b8eSDavid Sterba } 4156abbb3b8eSDavid Sterba 4157abbb3b8eSDavid Sterba /* 4158387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 4159387125fcSChris Mason * for the barrier when it is done 4160387125fcSChris Mason */ 41614246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 4162387125fcSChris Mason { 4163f9e69aa9SChristoph Hellwig bio_uninit(bio); 4164387125fcSChris Mason complete(bio->bi_private); 4165387125fcSChris Mason } 4166387125fcSChris Mason 4167387125fcSChris Mason /* 41684fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 41694fc6441aSAnand Jain * done in the waiting counterpart. 4170387125fcSChris Mason */ 41714fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 4172387125fcSChris Mason { 4173f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 4174387125fcSChris Mason 4175a91cf0ffSWang Yugui #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY 4176a91cf0ffSWang Yugui /* 4177a91cf0ffSWang Yugui * When a disk has write caching disabled, we skip submission of a bio 4178a91cf0ffSWang Yugui * with flush and sync requests before writing the superblock, since 4179a91cf0ffSWang Yugui * it's not needed. However when the integrity checker is enabled, this 4180a91cf0ffSWang Yugui * results in reports that there are metadata blocks referred by a 4181a91cf0ffSWang Yugui * superblock that were not properly flushed. So don't skip the bio 4182a91cf0ffSWang Yugui * submission only when the integrity checker is enabled for the sake 4183a91cf0ffSWang Yugui * of simplicity, since this is a debug tool and not meant for use in 4184a91cf0ffSWang Yugui * non-debug builds. 4185a91cf0ffSWang Yugui */ 418608e688fdSChristoph Hellwig if (!bdev_write_cache(device->bdev)) 41874fc6441aSAnand Jain return; 4188a91cf0ffSWang Yugui #endif 4189387125fcSChris Mason 4190f9e69aa9SChristoph Hellwig bio_init(bio, device->bdev, NULL, 0, 4191f9e69aa9SChristoph Hellwig REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); 4192387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 4193387125fcSChris Mason init_completion(&device->flush_wait); 4194387125fcSChris Mason bio->bi_private = &device->flush_wait; 4195387125fcSChris Mason 419658ff51f1SChristoph Hellwig btrfsic_check_bio(bio); 419758ff51f1SChristoph Hellwig submit_bio(bio); 41981c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 41994fc6441aSAnand Jain } 4200387125fcSChris Mason 42014fc6441aSAnand Jain /* 42024fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 42034fc6441aSAnand Jain */ 42048c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 42054fc6441aSAnand Jain { 4206f9e69aa9SChristoph Hellwig struct bio *bio = &device->flush_bio; 42074fc6441aSAnand Jain 42081c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 420958efbc9fSOmar Sandoval return BLK_STS_OK; 42104fc6441aSAnand Jain 42111c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 42122980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 42134fc6441aSAnand Jain 42148c27cb35SLinus Torvalds return bio->bi_status; 4215387125fcSChris Mason } 4216387125fcSChris Mason 4217d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 4218401b41e5SAnand Jain { 42196528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 4220401b41e5SAnand Jain return -EIO; 4221387125fcSChris Mason return 0; 4222387125fcSChris Mason } 4223387125fcSChris Mason 4224387125fcSChris Mason /* 4225387125fcSChris Mason * send an empty flush down to each device in parallel, 4226387125fcSChris Mason * then wait for them 4227387125fcSChris Mason */ 4228387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 4229387125fcSChris Mason { 4230387125fcSChris Mason struct list_head *head; 4231387125fcSChris Mason struct btrfs_device *dev; 42325af3e8ccSStefan Behrens int errors_wait = 0; 42334e4cbee9SChristoph Hellwig blk_status_t ret; 4234387125fcSChris Mason 42351538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 4236387125fcSChris Mason /* send down all the barriers */ 4237387125fcSChris Mason head = &info->fs_devices->devices; 42381538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4239e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4240f88ba6a2SHidetoshi Seto continue; 4241cea7c8bfSAnand Jain if (!dev->bdev) 4242387125fcSChris Mason continue; 4243e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4244ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4245387125fcSChris Mason continue; 4246387125fcSChris Mason 42474fc6441aSAnand Jain write_dev_flush(dev); 424858efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 4249387125fcSChris Mason } 4250387125fcSChris Mason 4251387125fcSChris Mason /* wait for all the barriers */ 42521538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4253e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 4254f88ba6a2SHidetoshi Seto continue; 4255387125fcSChris Mason if (!dev->bdev) { 42565af3e8ccSStefan Behrens errors_wait++; 4257387125fcSChris Mason continue; 4258387125fcSChris Mason } 4259e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4260ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4261387125fcSChris Mason continue; 4262387125fcSChris Mason 42634fc6441aSAnand Jain ret = wait_dev_flush(dev); 4264401b41e5SAnand Jain if (ret) { 4265401b41e5SAnand Jain dev->last_flush_error = ret; 426666b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 426766b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 42685af3e8ccSStefan Behrens errors_wait++; 4269387125fcSChris Mason } 4270401b41e5SAnand Jain } 4271401b41e5SAnand Jain 4272cea7c8bfSAnand Jain if (errors_wait) { 4273401b41e5SAnand Jain /* 4274401b41e5SAnand Jain * At some point we need the status of all disks 4275401b41e5SAnand Jain * to arrive at the volume status. So error checking 4276401b41e5SAnand Jain * is being pushed to a separate loop. 4277401b41e5SAnand Jain */ 4278d10b82feSQu Wenruo return check_barrier_error(info); 4279401b41e5SAnand Jain } 4280387125fcSChris Mason return 0; 4281387125fcSChris Mason } 4282387125fcSChris Mason 4283943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 4284943c6e99SZhao Lei { 42858789f4feSZhao Lei int raid_type; 42868789f4feSZhao Lei int min_tolerated = INT_MAX; 4287943c6e99SZhao Lei 42888789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 42898789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 42908c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 42918789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 42928789f4feSZhao Lei tolerated_failures); 4293943c6e99SZhao Lei 42948789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 42958789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 42968789f4feSZhao Lei continue; 429741a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 42988789f4feSZhao Lei continue; 42998c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 43008789f4feSZhao Lei btrfs_raid_array[raid_type]. 43018789f4feSZhao Lei tolerated_failures); 43028789f4feSZhao Lei } 4303943c6e99SZhao Lei 43048789f4feSZhao Lei if (min_tolerated == INT_MAX) { 4305ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 43068789f4feSZhao Lei min_tolerated = 0; 43078789f4feSZhao Lei } 43088789f4feSZhao Lei 43098789f4feSZhao Lei return min_tolerated; 4310943c6e99SZhao Lei } 4311943c6e99SZhao Lei 4312eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 4313f2984462SChris Mason { 4314e5e9a520SChris Mason struct list_head *head; 4315f2984462SChris Mason struct btrfs_device *dev; 4316a061fc8dSChris Mason struct btrfs_super_block *sb; 4317f2984462SChris Mason struct btrfs_dev_item *dev_item; 4318f2984462SChris Mason int ret; 4319f2984462SChris Mason int do_barriers; 4320a236aed1SChris Mason int max_errors; 4321a236aed1SChris Mason int total_errors = 0; 4322a061fc8dSChris Mason u64 flags; 4323f2984462SChris Mason 43240b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 4325fed3b381SLiu Bo 4326fed3b381SLiu Bo /* 4327fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 4328fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 4329fed3b381SLiu Bo * been consistent on disk. 4330fed3b381SLiu Bo */ 4331fed3b381SLiu Bo if (max_mirrors == 0) 43320b246afaSJeff Mahoney backup_super_roots(fs_info); 4333f2984462SChris Mason 43340b246afaSJeff Mahoney sb = fs_info->super_for_commit; 4335a061fc8dSChris Mason dev_item = &sb->dev_item; 4336e5e9a520SChris Mason 43370b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 43380b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 43390b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 4340387125fcSChris Mason 43415af3e8ccSStefan Behrens if (do_barriers) { 43420b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 43435af3e8ccSStefan Behrens if (ret) { 43445af3e8ccSStefan Behrens mutex_unlock( 43450b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 43460b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 43475af3e8ccSStefan Behrens "errors while submitting device barriers."); 43485af3e8ccSStefan Behrens return ret; 43495af3e8ccSStefan Behrens } 43505af3e8ccSStefan Behrens } 4351387125fcSChris Mason 43521538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4353dfe25020SChris Mason if (!dev->bdev) { 4354dfe25020SChris Mason total_errors++; 4355dfe25020SChris Mason continue; 4356dfe25020SChris Mason } 4357e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4358ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4359dfe25020SChris Mason continue; 4360dfe25020SChris Mason 43612b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 4362a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 4363a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 43647df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 4365935e5cc9SMiao Xie dev->commit_total_bytes); 4366ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 4367ce7213c7SMiao Xie dev->commit_bytes_used); 4368a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 4369a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 4370a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 4371a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 43727239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 43737239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 4374a512bbf8SYan Zheng 4375a061fc8dSChris Mason flags = btrfs_super_flags(sb); 4376a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 4377f2984462SChris Mason 437875cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 437975cb857dSQu Wenruo if (ret < 0) { 438075cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 438175cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 438275cb857dSQu Wenruo "unexpected superblock corruption detected"); 438375cb857dSQu Wenruo return -EUCLEAN; 438475cb857dSQu Wenruo } 438575cb857dSQu Wenruo 4386abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 4387a236aed1SChris Mason if (ret) 4388a236aed1SChris Mason total_errors++; 4389f2984462SChris Mason } 4390a236aed1SChris Mason if (total_errors > max_errors) { 43910b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 4392d397712bSChris Mason total_errors); 43930b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 439479787eaaSJeff Mahoney 43959d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 43960b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 43970b246afaSJeff Mahoney "%d errors while writing supers", 43980b246afaSJeff Mahoney total_errors); 43999d565ba4SStefan Behrens return -EIO; 4400a236aed1SChris Mason } 4401f2984462SChris Mason 4402a512bbf8SYan Zheng total_errors = 0; 44031538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 4404dfe25020SChris Mason if (!dev->bdev) 4405dfe25020SChris Mason continue; 4406e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 4407ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 4408dfe25020SChris Mason continue; 4409dfe25020SChris Mason 4410abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 4411a512bbf8SYan Zheng if (ret) 44121259ab75SChris Mason total_errors++; 4413f2984462SChris Mason } 44140b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 4415a236aed1SChris Mason if (total_errors > max_errors) { 44160b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 44170b246afaSJeff Mahoney "%d errors while writing supers", 44180b246afaSJeff Mahoney total_errors); 441979787eaaSJeff Mahoney return -EIO; 4420a236aed1SChris Mason } 4421f2984462SChris Mason return 0; 4422f2984462SChris Mason } 4423f2984462SChris Mason 4424cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 4425cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 4426cb517eabSMiao Xie struct btrfs_root *root) 44272619ba1fSChris Mason { 44284785e24fSJosef Bacik bool drop_ref = false; 44294785e24fSJosef Bacik 4430fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4431fc7cbcd4SDavid Sterba radix_tree_delete(&fs_info->fs_roots_radix, 4432fc7cbcd4SDavid Sterba (unsigned long)root->root_key.objectid); 4433fc7cbcd4SDavid Sterba if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 44344785e24fSJosef Bacik drop_ref = true; 4435fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 443676dda93cSYan, Zheng 443784961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) { 4438ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 44391c1ea4f7SLiu Bo if (root->reloc_root) { 444000246528SJosef Bacik btrfs_put_root(root->reloc_root); 44411c1ea4f7SLiu Bo root->reloc_root = NULL; 44421c1ea4f7SLiu Bo } 44431c1ea4f7SLiu Bo } 44443321719eSLiu Bo 44454785e24fSJosef Bacik if (drop_ref) 444600246528SJosef Bacik btrfs_put_root(root); 44472619ba1fSChris Mason } 44482619ba1fSChris Mason 4449c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4450c146afadSYan Zheng { 4451fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4452fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4453fc7cbcd4SDavid Sterba int i = 0; 445465d33fd7SQu Wenruo int err = 0; 4455fc7cbcd4SDavid Sterba unsigned int ret = 0; 4456c146afadSYan Zheng 4457c146afadSYan Zheng while (1) { 4458fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4459fc7cbcd4SDavid Sterba ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4460fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4461fc7cbcd4SDavid Sterba ARRAY_SIZE(gang)); 4462fc7cbcd4SDavid Sterba if (!ret) { 4463fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4464c146afadSYan Zheng break; 446565d33fd7SQu Wenruo } 4466fc7cbcd4SDavid Sterba root_objectid = gang[ret - 1]->root_key.objectid + 1; 446766b4ffd1SJosef Bacik 4468fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4469fc7cbcd4SDavid Sterba /* Avoid to grab roots in dead_roots */ 4470fc7cbcd4SDavid Sterba if (btrfs_root_refs(&gang[i]->root_item) == 0) { 4471fc7cbcd4SDavid Sterba gang[i] = NULL; 447265d33fd7SQu Wenruo continue; 4473c146afadSYan Zheng } 4474fc7cbcd4SDavid Sterba /* grab all the search result for later use */ 4475fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4476fc7cbcd4SDavid Sterba } 4477fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4478fc7cbcd4SDavid Sterba 4479fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4480fc7cbcd4SDavid Sterba if (!gang[i]) 4481fc7cbcd4SDavid Sterba continue; 4482fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4483fc7cbcd4SDavid Sterba err = btrfs_orphan_cleanup(gang[i]); 4484fc7cbcd4SDavid Sterba if (err) 4485fc7cbcd4SDavid Sterba break; 4486fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4487fc7cbcd4SDavid Sterba } 4488fc7cbcd4SDavid Sterba root_objectid++; 4489c146afadSYan Zheng } 449065d33fd7SQu Wenruo 4491fc7cbcd4SDavid Sterba /* release the uncleaned roots due to error */ 4492fc7cbcd4SDavid Sterba for (; i < ret; i++) { 4493fc7cbcd4SDavid Sterba if (gang[i]) 4494fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 449565d33fd7SQu Wenruo } 449665d33fd7SQu Wenruo return err; 4497c146afadSYan Zheng } 4498c146afadSYan Zheng 44996bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4500c146afadSYan Zheng { 45016bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4502c146afadSYan Zheng struct btrfs_trans_handle *trans; 4503c146afadSYan Zheng 45040b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 45052ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 45060b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 45070b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4508c71bf099SYan, Zheng 4509c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 45100b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 45110b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4512c71bf099SYan, Zheng 45137a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 45143612b495STsutomu Itoh if (IS_ERR(trans)) 45153612b495STsutomu Itoh return PTR_ERR(trans); 45163a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4517c146afadSYan Zheng } 4518c146afadSYan Zheng 451936c86a9eSQu Wenruo static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) 452036c86a9eSQu Wenruo { 452136c86a9eSQu Wenruo struct btrfs_transaction *trans; 452236c86a9eSQu Wenruo struct btrfs_transaction *tmp; 452336c86a9eSQu Wenruo bool found = false; 452436c86a9eSQu Wenruo 452536c86a9eSQu Wenruo if (list_empty(&fs_info->trans_list)) 452636c86a9eSQu Wenruo return; 452736c86a9eSQu Wenruo 452836c86a9eSQu Wenruo /* 452936c86a9eSQu Wenruo * This function is only called at the very end of close_ctree(), 453036c86a9eSQu Wenruo * thus no other running transaction, no need to take trans_lock. 453136c86a9eSQu Wenruo */ 453236c86a9eSQu Wenruo ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); 453336c86a9eSQu Wenruo list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { 453436c86a9eSQu Wenruo struct extent_state *cached = NULL; 453536c86a9eSQu Wenruo u64 dirty_bytes = 0; 453636c86a9eSQu Wenruo u64 cur = 0; 453736c86a9eSQu Wenruo u64 found_start; 453836c86a9eSQu Wenruo u64 found_end; 453936c86a9eSQu Wenruo 454036c86a9eSQu Wenruo found = true; 454136c86a9eSQu Wenruo while (!find_first_extent_bit(&trans->dirty_pages, cur, 454236c86a9eSQu Wenruo &found_start, &found_end, EXTENT_DIRTY, &cached)) { 454336c86a9eSQu Wenruo dirty_bytes += found_end + 1 - found_start; 454436c86a9eSQu Wenruo cur = found_end + 1; 454536c86a9eSQu Wenruo } 454636c86a9eSQu Wenruo btrfs_warn(fs_info, 454736c86a9eSQu Wenruo "transaction %llu (with %llu dirty metadata bytes) is not committed", 454836c86a9eSQu Wenruo trans->transid, dirty_bytes); 454936c86a9eSQu Wenruo btrfs_cleanup_one_transaction(trans, fs_info); 455036c86a9eSQu Wenruo 455136c86a9eSQu Wenruo if (trans == fs_info->running_transaction) 455236c86a9eSQu Wenruo fs_info->running_transaction = NULL; 455336c86a9eSQu Wenruo list_del_init(&trans->list); 455436c86a9eSQu Wenruo 455536c86a9eSQu Wenruo btrfs_put_transaction(trans); 455636c86a9eSQu Wenruo trace_btrfs_transaction_commit(fs_info); 455736c86a9eSQu Wenruo } 455836c86a9eSQu Wenruo ASSERT(!found); 455936c86a9eSQu Wenruo } 456036c86a9eSQu Wenruo 4561b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4562eb60ceacSChris Mason { 4563c146afadSYan Zheng int ret; 4564e089f05cSChris Mason 4565afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 456631e70e52SFilipe Manana 456731e70e52SFilipe Manana /* 45688a1f1e3dSFilipe Manana * If we had UNFINISHED_DROPS we could still be processing them, so 45698a1f1e3dSFilipe Manana * clear that bit and wake up relocation so it can stop. 45708a1f1e3dSFilipe Manana * We must do this before stopping the block group reclaim task, because 45718a1f1e3dSFilipe Manana * at btrfs_relocate_block_group() we wait for this bit, and after the 45728a1f1e3dSFilipe Manana * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we 45738a1f1e3dSFilipe Manana * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will 45748a1f1e3dSFilipe Manana * return 1. 45758a1f1e3dSFilipe Manana */ 45768a1f1e3dSFilipe Manana btrfs_wake_unfinished_drop(fs_info); 45778a1f1e3dSFilipe Manana 45788a1f1e3dSFilipe Manana /* 457931e70e52SFilipe Manana * We may have the reclaim task running and relocating a data block group, 458031e70e52SFilipe Manana * in which case it may create delayed iputs. So stop it before we park 458131e70e52SFilipe Manana * the cleaner kthread otherwise we can get new delayed iputs after 458231e70e52SFilipe Manana * parking the cleaner, and that can make the async reclaim task to hang 458331e70e52SFilipe Manana * if it's waiting for delayed iputs to complete, since the cleaner is 458431e70e52SFilipe Manana * parked and can not run delayed iputs - this will make us hang when 458531e70e52SFilipe Manana * trying to stop the async reclaim task. 458631e70e52SFilipe Manana */ 458731e70e52SFilipe Manana cancel_work_sync(&fs_info->reclaim_bgs_work); 4588d6fd0ae2SOmar Sandoval /* 4589d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4590d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4591d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4592d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4593d6fd0ae2SOmar Sandoval */ 4594d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4595a2135011SChris Mason 45967343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4597d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 45987343dd61SJustin Maggard 4599803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4600803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4601803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4602803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4603803b2f54SStefan Behrens 4604837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4605aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4606837d5b6eSIlya Dryomov 46078dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 46088dabb742SStefan Behrens 4609aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 46104cb5300bSChris Mason 46114cb5300bSChris Mason /* wait for any defraggers to finish */ 46124cb5300bSChris Mason wait_event(fs_info->transaction_wait, 46134cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 46144cb5300bSChris Mason 46154cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 461626176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 46174cb5300bSChris Mason 4618a362bb86SFilipe Manana /* 4619a362bb86SFilipe Manana * After we parked the cleaner kthread, ordered extents may have 4620a362bb86SFilipe Manana * completed and created new delayed iputs. If one of the async reclaim 4621a362bb86SFilipe Manana * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we 4622a362bb86SFilipe Manana * can hang forever trying to stop it, because if a delayed iput is 4623a362bb86SFilipe Manana * added after it ran btrfs_run_delayed_iputs() and before it called 4624a362bb86SFilipe Manana * btrfs_wait_on_delayed_iputs(), it will hang forever since there is 4625a362bb86SFilipe Manana * no one else to run iputs. 4626a362bb86SFilipe Manana * 4627a362bb86SFilipe Manana * So wait for all ongoing ordered extents to complete and then run 4628a362bb86SFilipe Manana * delayed iputs. This works because once we reach this point no one 4629a362bb86SFilipe Manana * can either create new ordered extents nor create delayed iputs 4630a362bb86SFilipe Manana * through some other means. 4631a362bb86SFilipe Manana * 4632a362bb86SFilipe Manana * Also note that btrfs_wait_ordered_roots() is not safe here, because 4633a362bb86SFilipe Manana * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, 4634a362bb86SFilipe Manana * but the delayed iput for the respective inode is made only when doing 4635a362bb86SFilipe Manana * the final btrfs_put_ordered_extent() (which must happen at 4636a362bb86SFilipe Manana * btrfs_finish_ordered_io() when we are unmounting). 4637a362bb86SFilipe Manana */ 4638a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_write_workers); 4639a362bb86SFilipe Manana /* Ordered extents for free space inodes. */ 4640a362bb86SFilipe Manana btrfs_flush_workqueue(fs_info->endio_freespace_worker); 4641a362bb86SFilipe Manana btrfs_run_delayed_iputs(fs_info); 4642a362bb86SFilipe Manana 464321c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 464457056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 4645576fa348SJosef Bacik cancel_work_sync(&fs_info->preempt_reclaim_work); 464621c7e756SMiao Xie 4647b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4648b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4649b0643e59SDennis Zhou 4650bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4651e44163e1SJeff Mahoney /* 4652d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4653d6fd0ae2SOmar Sandoval * unused block groups. 4654e44163e1SJeff Mahoney */ 46550b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4656e44163e1SJeff Mahoney 4657f0cc2cd7SFilipe Manana /* 4658f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4659f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4660f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4661f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4662f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4663f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4664f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4665f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4666f0cc2cd7SFilipe Manana * This is a very rare case. 4667f0cc2cd7SFilipe Manana */ 4668f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4669f0cc2cd7SFilipe Manana 46706bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4671d397712bSChris Mason if (ret) 467204892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4673c146afadSYan Zheng } 4674ed2ff2cbSChris Mason 467584961539SJosef Bacik if (BTRFS_FS_ERROR(fs_info)) 46762ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4677acce952bSliubo 4678e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4679e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 46808929ecfaSYan, Zheng 4681e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4682afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4683f25784b3SYan Zheng 46845958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 46855958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 46865958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 46875958253cSQu Wenruo } 46885958253cSQu Wenruo 468904892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4690fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4691bcef60f2SArne Jansen 4692963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 469304892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4694963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4695b0c68f8bSChris Mason } 469631153d81SYan Zheng 46975deb17e1SJosef Bacik if (percpu_counter_sum(&fs_info->ordered_bytes)) 46984297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 46995deb17e1SJosef Bacik percpu_counter_sum(&fs_info->ordered_bytes)); 47004297ff84SJosef Bacik 47016618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4702b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 47035ac1d209SJeff Mahoney 47041a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 47051a4319ccSLiu Bo 4706de348ee0SWang Shilong /* 4707de348ee0SWang Shilong * we must make sure there is not any read request to 4708de348ee0SWang Shilong * submit after we stopping all workers. 4709de348ee0SWang Shilong */ 4710de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 471196192499SJosef Bacik btrfs_stop_all_workers(fs_info); 471296192499SJosef Bacik 47130a31daa4SFilipe Manana /* We shouldn't have any transaction open at this point */ 471436c86a9eSQu Wenruo warn_about_uncommitted_trans(fs_info); 47150a31daa4SFilipe Manana 4716afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 47174273eaffSAnand Jain free_root_pointers(fs_info, true); 47188c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 47199ad6b7bcSChris Mason 47204e19443dSJosef Bacik /* 47214e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 47224e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 47234e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 47244e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 47254e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 47264e19443dSJosef Bacik */ 47274e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 47284e19443dSJosef Bacik 472913e6c37bSJosef Bacik iput(fs_info->btree_inode); 4730d6bfde87SChris Mason 473121adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 47320b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 47332ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 473421adbd5cSStefan Behrens #endif 473521adbd5cSStefan Behrens 47360b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 473768c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4738eb60ceacSChris Mason } 4739eb60ceacSChris Mason 4740b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4741b9fab919SChris Mason int atomic) 4742ccd467d6SChris Mason { 47431259ab75SChris Mason int ret; 4744727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 47451259ab75SChris Mason 47460b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 47471259ab75SChris Mason if (!ret) 47481259ab75SChris Mason return ret; 47491259ab75SChris Mason 47501259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4751b9fab919SChris Mason parent_transid, atomic); 4752b9fab919SChris Mason if (ret == -EAGAIN) 4753b9fab919SChris Mason return ret; 47541259ab75SChris Mason return !ret; 47555f39d397SChris Mason } 47566702ed49SChris Mason 47575f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 47585f39d397SChris Mason { 47592f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 47605f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4761b9473439SChris Mason int was_dirty; 4762b4ce94deSChris Mason 476306ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 476406ea65a3SJosef Bacik /* 476506ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 476652042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 476706ea65a3SJosef Bacik * outside of the sanity tests. 476806ea65a3SJosef Bacik */ 4769b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 477006ea65a3SJosef Bacik return; 477106ea65a3SJosef Bacik #endif 477249d0c642SFilipe Manana btrfs_assert_tree_write_locked(buf); 47730b246afaSJeff Mahoney if (transid != fs_info->generation) 47745d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 47750b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 47760b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4777e2d84521SMiao Xie if (!was_dirty) 4778104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4779e2d84521SMiao Xie buf->len, 47800b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 47811f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 478269fc6cbbSQu Wenruo /* 478369fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 478469fc6cbbSQu Wenruo * but item data not updated. 478569fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 478669fc6cbbSQu Wenruo */ 478769fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4788cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4789a4f78750SDavid Sterba btrfs_print_leaf(buf); 47901f21ef0aSFilipe Manana ASSERT(0); 47911f21ef0aSFilipe Manana } 47921f21ef0aSFilipe Manana #endif 4793eb60ceacSChris Mason } 4794eb60ceacSChris Mason 47952ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4796b53d3f5dSLiu Bo int flush_delayed) 479735b7e476SChris Mason { 4798188de649SChris Mason /* 4799188de649SChris Mason * looks as though older kernels can get into trouble with 4800188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4801188de649SChris Mason */ 4802e2d84521SMiao Xie int ret; 4803d6bfde87SChris Mason 48046933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4805d6bfde87SChris Mason return; 4806d6bfde87SChris Mason 4807b53d3f5dSLiu Bo if (flush_delayed) 48082ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 480916cdcec7SMiao Xie 4810d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4811d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4812d814a491SEthan Lien fs_info->dirty_metadata_batch); 4813e2d84521SMiao Xie if (ret > 0) { 48140b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 481516cdcec7SMiao Xie } 481616cdcec7SMiao Xie } 481716cdcec7SMiao Xie 48182ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 481916cdcec7SMiao Xie { 48202ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 482135b7e476SChris Mason } 4822b53d3f5dSLiu Bo 48232ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4824b53d3f5dSLiu Bo { 48252ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4826d6bfde87SChris Mason } 48276b80053dSChris Mason 48282ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4829acce952bSliubo { 4830fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4831fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4832fe816d0fSNikolay Borisov 48330b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 48342ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 48350b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4836acce952bSliubo 48370b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 48380b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4839acce952bSliubo } 4840acce952bSliubo 4841ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4842ef67963dSJosef Bacik { 4843fc7cbcd4SDavid Sterba struct btrfs_root *gang[8]; 4844fc7cbcd4SDavid Sterba u64 root_objectid = 0; 4845fc7cbcd4SDavid Sterba int ret; 4846ef67963dSJosef Bacik 4847fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4848fc7cbcd4SDavid Sterba while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4849fc7cbcd4SDavid Sterba (void **)gang, root_objectid, 4850fc7cbcd4SDavid Sterba ARRAY_SIZE(gang))) != 0) { 4851fc7cbcd4SDavid Sterba int i; 4852ef67963dSJosef Bacik 4853fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) 4854fc7cbcd4SDavid Sterba gang[i] = btrfs_grab_root(gang[i]); 4855fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4856fc7cbcd4SDavid Sterba 4857fc7cbcd4SDavid Sterba for (i = 0; i < ret; i++) { 4858fc7cbcd4SDavid Sterba if (!gang[i]) 4859ef67963dSJosef Bacik continue; 4860fc7cbcd4SDavid Sterba root_objectid = gang[i]->root_key.objectid; 4861fc7cbcd4SDavid Sterba btrfs_free_log(NULL, gang[i]); 4862fc7cbcd4SDavid Sterba btrfs_put_root(gang[i]); 4863ef67963dSJosef Bacik } 4864fc7cbcd4SDavid Sterba root_objectid++; 4865fc7cbcd4SDavid Sterba spin_lock(&fs_info->fs_roots_radix_lock); 4866ef67963dSJosef Bacik } 4867fc7cbcd4SDavid Sterba spin_unlock(&fs_info->fs_roots_radix_lock); 4868ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4869ef67963dSJosef Bacik } 4870ef67963dSJosef Bacik 4871143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4872acce952bSliubo { 4873acce952bSliubo struct btrfs_ordered_extent *ordered; 4874acce952bSliubo 4875199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4876779880efSJosef Bacik /* 4877779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4878779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4879779880efSJosef Bacik */ 4880199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4881779880efSJosef Bacik root_extent_list) 4882779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4883199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4884199c2a9cSMiao Xie } 4885199c2a9cSMiao Xie 4886199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4887199c2a9cSMiao Xie { 4888199c2a9cSMiao Xie struct btrfs_root *root; 4889199c2a9cSMiao Xie struct list_head splice; 4890199c2a9cSMiao Xie 4891199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4892199c2a9cSMiao Xie 4893199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4894199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4895199c2a9cSMiao Xie while (!list_empty(&splice)) { 4896199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4897199c2a9cSMiao Xie ordered_root); 48981de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 48991de2cfdeSJosef Bacik &fs_info->ordered_roots); 4900199c2a9cSMiao Xie 49012a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4902199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4903199c2a9cSMiao Xie 49042a85d9caSLiu Bo cond_resched(); 49052a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4906199c2a9cSMiao Xie } 4907199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 490874d5d229SJosef Bacik 490974d5d229SJosef Bacik /* 491074d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 491174d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 491274d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 491374d5d229SJosef Bacik * actually get run and error out properly. 491474d5d229SJosef Bacik */ 491574d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4916acce952bSliubo } 4917acce952bSliubo 491835a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 49192ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4920acce952bSliubo { 4921acce952bSliubo struct rb_node *node; 4922acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4923acce952bSliubo struct btrfs_delayed_ref_node *ref; 4924acce952bSliubo int ret = 0; 4925acce952bSliubo 4926acce952bSliubo delayed_refs = &trans->delayed_refs; 4927acce952bSliubo 4928acce952bSliubo spin_lock(&delayed_refs->lock); 4929d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4930cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4931b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4932acce952bSliubo return ret; 4933acce952bSliubo } 4934acce952bSliubo 49355c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4936d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 49370e0adbcfSJosef Bacik struct rb_node *n; 4938e78417d1SJosef Bacik bool pin_bytes = false; 4939acce952bSliubo 4940d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4941d7df2c79SJosef Bacik href_node); 49423069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4943b939d1abSJosef Bacik continue; 49443069bd26SJosef Bacik 4945d7df2c79SJosef Bacik spin_lock(&head->lock); 4946e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 49470e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 49480e0adbcfSJosef Bacik ref_node); 4949d7df2c79SJosef Bacik ref->in_tree = 0; 4950e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 49510e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 49521d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 49531d57ee94SWang Xiaoguang list_del(&ref->add_list); 4954d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4955d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4956d7df2c79SJosef Bacik } 495754067ae9SJosef Bacik if (head->must_insert_reserved) 4958e78417d1SJosef Bacik pin_bytes = true; 495978a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4960fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4961d7df2c79SJosef Bacik spin_unlock(&head->lock); 4962acce952bSliubo spin_unlock(&delayed_refs->lock); 4963e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4964acce952bSliubo 4965f603bb94SNikolay Borisov if (pin_bytes) { 4966f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4967f603bb94SNikolay Borisov 4968f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4969f603bb94SNikolay Borisov BUG_ON(!cache); 4970f603bb94SNikolay Borisov 4971f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4972f603bb94SNikolay Borisov spin_lock(&cache->lock); 4973f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4974f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4975f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4976f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4977f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4978f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4979f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4980f603bb94SNikolay Borisov 4981f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4982f603bb94SNikolay Borisov 4983f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4984f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4985f603bb94SNikolay Borisov } 498631890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4987d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4988acce952bSliubo cond_resched(); 4989acce952bSliubo spin_lock(&delayed_refs->lock); 4990acce952bSliubo } 499181f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4992acce952bSliubo 4993acce952bSliubo spin_unlock(&delayed_refs->lock); 4994acce952bSliubo 4995acce952bSliubo return ret; 4996acce952bSliubo } 4997acce952bSliubo 4998143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4999acce952bSliubo { 5000acce952bSliubo struct btrfs_inode *btrfs_inode; 5001acce952bSliubo struct list_head splice; 5002acce952bSliubo 5003acce952bSliubo INIT_LIST_HEAD(&splice); 5004acce952bSliubo 5005eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5006eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 5007acce952bSliubo 5008acce952bSliubo while (!list_empty(&splice)) { 5009fe816d0fSNikolay Borisov struct inode *inode = NULL; 5010eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 5011acce952bSliubo delalloc_inodes); 5012fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 5013eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5014acce952bSliubo 5015fe816d0fSNikolay Borisov /* 5016fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 5017fe816d0fSNikolay Borisov * meanwhile. 5018fe816d0fSNikolay Borisov */ 5019fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 5020fe816d0fSNikolay Borisov if (inode) { 5021fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 5022fe816d0fSNikolay Borisov iput(inode); 5023fe816d0fSNikolay Borisov } 5024eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 5025acce952bSliubo } 5026eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 5027eb73c1b7SMiao Xie } 5028eb73c1b7SMiao Xie 5029eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 5030eb73c1b7SMiao Xie { 5031eb73c1b7SMiao Xie struct btrfs_root *root; 5032eb73c1b7SMiao Xie struct list_head splice; 5033eb73c1b7SMiao Xie 5034eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 5035eb73c1b7SMiao Xie 5036eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5037eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 5038eb73c1b7SMiao Xie while (!list_empty(&splice)) { 5039eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 5040eb73c1b7SMiao Xie delalloc_root); 504100246528SJosef Bacik root = btrfs_grab_root(root); 5042eb73c1b7SMiao Xie BUG_ON(!root); 5043eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5044eb73c1b7SMiao Xie 5045eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 504600246528SJosef Bacik btrfs_put_root(root); 5047eb73c1b7SMiao Xie 5048eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 5049eb73c1b7SMiao Xie } 5050eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 5051acce952bSliubo } 5052acce952bSliubo 50532ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 5054acce952bSliubo struct extent_io_tree *dirty_pages, 5055acce952bSliubo int mark) 5056acce952bSliubo { 5057acce952bSliubo int ret; 5058acce952bSliubo struct extent_buffer *eb; 5059acce952bSliubo u64 start = 0; 5060acce952bSliubo u64 end; 5061acce952bSliubo 5062acce952bSliubo while (1) { 5063acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 5064e6138876SJosef Bacik mark, NULL); 5065acce952bSliubo if (ret) 5066acce952bSliubo break; 5067acce952bSliubo 506891166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 5069acce952bSliubo while (start <= end) { 50700b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 50710b246afaSJeff Mahoney start += fs_info->nodesize; 5072fd8b2b61SJosef Bacik if (!eb) 5073acce952bSliubo continue; 5074fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 5075acce952bSliubo 5076fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 5077fd8b2b61SJosef Bacik &eb->bflags)) 5078fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 5079fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 5080acce952bSliubo } 5081acce952bSliubo } 5082acce952bSliubo 5083acce952bSliubo return ret; 5084acce952bSliubo } 5085acce952bSliubo 50862ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 5087fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 5088acce952bSliubo { 5089acce952bSliubo u64 start; 5090acce952bSliubo u64 end; 5091acce952bSliubo int ret; 5092acce952bSliubo 5093acce952bSliubo while (1) { 50940e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 50950e6ec385SFilipe Manana 5096fcd5e742SLu Fengqi /* 5097fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 5098fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 5099fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 5100fcd5e742SLu Fengqi * the same extent range. 5101fcd5e742SLu Fengqi */ 5102fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 5103acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 51040e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 5105fcd5e742SLu Fengqi if (ret) { 5106fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5107acce952bSliubo break; 5108fcd5e742SLu Fengqi } 5109acce952bSliubo 51100e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 51110e6ec385SFilipe Manana free_extent_state(cached_state); 51122ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 5113fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 5114acce952bSliubo cond_resched(); 5115acce952bSliubo } 5116acce952bSliubo 5117acce952bSliubo return 0; 5118acce952bSliubo } 5119acce952bSliubo 512032da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 5121c79a1751SLiu Bo { 5122c79a1751SLiu Bo struct inode *inode; 5123c79a1751SLiu Bo 5124c79a1751SLiu Bo inode = cache->io_ctl.inode; 5125c79a1751SLiu Bo if (inode) { 5126c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 5127c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 5128c79a1751SLiu Bo cache->io_ctl.inode = NULL; 5129c79a1751SLiu Bo iput(inode); 5130c79a1751SLiu Bo } 5131bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 5132c79a1751SLiu Bo btrfs_put_block_group(cache); 5133c79a1751SLiu Bo } 5134c79a1751SLiu Bo 5135c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 51362ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 5137c79a1751SLiu Bo { 513832da5386SDavid Sterba struct btrfs_block_group *cache; 5139c79a1751SLiu Bo 5140c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5141c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 5142c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 514332da5386SDavid Sterba struct btrfs_block_group, 5144c79a1751SLiu Bo dirty_list); 5145c79a1751SLiu Bo 5146c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 5147c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5148c79a1751SLiu Bo list_del_init(&cache->io_list); 5149c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5150c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5151c79a1751SLiu Bo } 5152c79a1751SLiu Bo 5153c79a1751SLiu Bo list_del_init(&cache->dirty_list); 5154c79a1751SLiu Bo spin_lock(&cache->lock); 5155c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5156c79a1751SLiu Bo spin_unlock(&cache->lock); 5157c79a1751SLiu Bo 5158c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5159c79a1751SLiu Bo btrfs_put_block_group(cache); 5160ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 5161c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 5162c79a1751SLiu Bo } 5163c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 5164c79a1751SLiu Bo 516545ae2c18SNikolay Borisov /* 516645ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 516745ae2c18SNikolay Borisov * to use it without any locking 516845ae2c18SNikolay Borisov */ 5169c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 5170c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 517132da5386SDavid Sterba struct btrfs_block_group, 5172c79a1751SLiu Bo io_list); 5173c79a1751SLiu Bo 5174c79a1751SLiu Bo list_del_init(&cache->io_list); 5175c79a1751SLiu Bo spin_lock(&cache->lock); 5176c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 5177c79a1751SLiu Bo spin_unlock(&cache->lock); 5178c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 5179c79a1751SLiu Bo } 5180c79a1751SLiu Bo } 5181c79a1751SLiu Bo 518249b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 51832ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 518449b25e05SJeff Mahoney { 5185bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 5186bbbf7243SNikolay Borisov 51872ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 5188c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 5189c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 5190c79a1751SLiu Bo 5191bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 5192bbbf7243SNikolay Borisov post_commit_list) { 5193bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 5194bbbf7243SNikolay Borisov } 5195bbbf7243SNikolay Borisov 51962ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 519749b25e05SJeff Mahoney 51984a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 51990b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 520049b25e05SJeff Mahoney 52014a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 52020b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 520349b25e05SJeff Mahoney 5204ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 520567cde344SMiao Xie 52062ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 520749b25e05SJeff Mahoney EXTENT_DIRTY); 5208fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 520949b25e05SJeff Mahoney 5210d3575156SNaohiro Aota btrfs_free_redirty_list(cur_trans); 5211d3575156SNaohiro Aota 52124a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 52134a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 521449b25e05SJeff Mahoney } 521549b25e05SJeff Mahoney 52162ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 5217acce952bSliubo { 5218acce952bSliubo struct btrfs_transaction *t; 5219acce952bSliubo 52200b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 5221acce952bSliubo 52220b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 52230b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 52240b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 5225724e2315SJosef Bacik struct btrfs_transaction, list); 5226724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 52279b64f57dSElena Reshetova refcount_inc(&t->use_count); 52280b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 52292ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 5230724e2315SJosef Bacik btrfs_put_transaction(t); 52310b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5232724e2315SJosef Bacik continue; 5233724e2315SJosef Bacik } 52340b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 5235724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 52360b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5237724e2315SJosef Bacik /* 5238724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 5239724e2315SJosef Bacik * handle open currently for this transaction. 5240724e2315SJosef Bacik */ 5241724e2315SJosef Bacik wait_event(t->writer_wait, 5242724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 5243724e2315SJosef Bacik } else { 52440b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5245724e2315SJosef Bacik } 52462ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 5247724e2315SJosef Bacik 52480b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 52490b246afaSJeff Mahoney if (t == fs_info->running_transaction) 52500b246afaSJeff Mahoney fs_info->running_transaction = NULL; 5251724e2315SJosef Bacik list_del_init(&t->list); 52520b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 5253a4abeea4SJosef Bacik 5254724e2315SJosef Bacik btrfs_put_transaction(t); 52552e4e97abSJosef Bacik trace_btrfs_transaction_commit(fs_info); 52560b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 5257724e2315SJosef Bacik } 52580b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 52590b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 5260ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 5261ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 52620b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 5263ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 52640b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 5265acce952bSliubo 5266acce952bSliubo return 0; 5267acce952bSliubo } 5268ec7d6dfdSNikolay Borisov 5269453e4873SNikolay Borisov int btrfs_init_root_free_objectid(struct btrfs_root *root) 5270ec7d6dfdSNikolay Borisov { 5271ec7d6dfdSNikolay Borisov struct btrfs_path *path; 5272ec7d6dfdSNikolay Borisov int ret; 5273ec7d6dfdSNikolay Borisov struct extent_buffer *l; 5274ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 5275ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 5276ec7d6dfdSNikolay Borisov int slot; 5277ec7d6dfdSNikolay Borisov 5278ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 5279ec7d6dfdSNikolay Borisov if (!path) 5280ec7d6dfdSNikolay Borisov return -ENOMEM; 5281ec7d6dfdSNikolay Borisov 5282ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 5283ec7d6dfdSNikolay Borisov search_key.type = -1; 5284ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 5285ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 5286ec7d6dfdSNikolay Borisov if (ret < 0) 5287ec7d6dfdSNikolay Borisov goto error; 5288ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 5289ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 5290ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 5291ec7d6dfdSNikolay Borisov l = path->nodes[0]; 5292ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 529323125104SNikolay Borisov root->free_objectid = max_t(u64, found_key.objectid + 1, 529423125104SNikolay Borisov BTRFS_FIRST_FREE_OBJECTID); 5295ec7d6dfdSNikolay Borisov } else { 529623125104SNikolay Borisov root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; 5297ec7d6dfdSNikolay Borisov } 5298ec7d6dfdSNikolay Borisov ret = 0; 5299ec7d6dfdSNikolay Borisov error: 5300ec7d6dfdSNikolay Borisov btrfs_free_path(path); 5301ec7d6dfdSNikolay Borisov return ret; 5302ec7d6dfdSNikolay Borisov } 5303ec7d6dfdSNikolay Borisov 5304543068a2SNikolay Borisov int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) 5305ec7d6dfdSNikolay Borisov { 5306ec7d6dfdSNikolay Borisov int ret; 5307ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 5308ec7d6dfdSNikolay Borisov 53096b8fad57SNikolay Borisov if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 5310ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 5311ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 5312ec7d6dfdSNikolay Borisov root->root_key.objectid); 5313ec7d6dfdSNikolay Borisov ret = -ENOSPC; 5314ec7d6dfdSNikolay Borisov goto out; 5315ec7d6dfdSNikolay Borisov } 5316ec7d6dfdSNikolay Borisov 531723125104SNikolay Borisov *objectid = root->free_objectid++; 5318ec7d6dfdSNikolay Borisov ret = 0; 5319ec7d6dfdSNikolay Borisov out: 5320ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 5321ec7d6dfdSNikolay Borisov return ret; 5322ec7d6dfdSNikolay Borisov } 5323