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> 80f7d52f4SChris Mason #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" 45eb60ceacSChris Mason 46319e4d06SQu Wenruo #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ 47319e4d06SQu Wenruo BTRFS_HEADER_FLAG_RELOC |\ 48319e4d06SQu Wenruo BTRFS_SUPER_FLAG_ERROR |\ 49319e4d06SQu Wenruo BTRFS_SUPER_FLAG_SEEDING |\ 50e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP |\ 51e2731e55SAnand Jain BTRFS_SUPER_FLAG_METADUMP_V2) 52319e4d06SQu Wenruo 538b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 54143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 55acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 562ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info); 57143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 582ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 59acce952bSliubo struct extent_io_tree *dirty_pages, 60acce952bSliubo int mark); 612ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 62acce952bSliubo struct extent_io_tree *pinned_extents); 632ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); 642ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); 65ce9adaa5SChris Mason 66d352ac68SChris Mason /* 6797eb6b69SDavid Sterba * btrfs_end_io_wq structs are used to do processing in task context when an IO 6897eb6b69SDavid Sterba * is complete. This is used during reads to verify checksums, and it is used 69d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 70d352ac68SChris Mason */ 7197eb6b69SDavid Sterba struct btrfs_end_io_wq { 72ce9adaa5SChris Mason struct bio *bio; 73ce9adaa5SChris Mason bio_end_io_t *end_io; 74ce9adaa5SChris Mason void *private; 75ce9adaa5SChris Mason struct btrfs_fs_info *info; 764e4cbee9SChristoph Hellwig blk_status_t status; 77bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata; 788b712842SChris Mason struct btrfs_work work; 79ce9adaa5SChris Mason }; 800da5468fSChris Mason 8197eb6b69SDavid Sterba static struct kmem_cache *btrfs_end_io_wq_cache; 8297eb6b69SDavid Sterba 8397eb6b69SDavid Sterba int __init btrfs_end_io_wq_init(void) 8497eb6b69SDavid Sterba { 8597eb6b69SDavid Sterba btrfs_end_io_wq_cache = kmem_cache_create("btrfs_end_io_wq", 8697eb6b69SDavid Sterba sizeof(struct btrfs_end_io_wq), 8797eb6b69SDavid Sterba 0, 88fba4b697SNikolay Borisov SLAB_MEM_SPREAD, 8997eb6b69SDavid Sterba NULL); 9097eb6b69SDavid Sterba if (!btrfs_end_io_wq_cache) 9197eb6b69SDavid Sterba return -ENOMEM; 9297eb6b69SDavid Sterba return 0; 9397eb6b69SDavid Sterba } 9497eb6b69SDavid Sterba 95e67c718bSDavid Sterba void __cold btrfs_end_io_wq_exit(void) 9697eb6b69SDavid Sterba { 9797eb6b69SDavid Sterba kmem_cache_destroy(btrfs_end_io_wq_cache); 9897eb6b69SDavid Sterba } 9997eb6b69SDavid Sterba 100141386e1SJosef Bacik static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) 101141386e1SJosef Bacik { 102141386e1SJosef Bacik if (fs_info->csum_shash) 103141386e1SJosef Bacik crypto_free_shash(fs_info->csum_shash); 104141386e1SJosef Bacik } 105141386e1SJosef Bacik 106d352ac68SChris Mason /* 107d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 108d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 109d352ac68SChris Mason * just before they are sent down the IO stack. 110d352ac68SChris Mason */ 11144b8bd7eSChris Mason struct async_submit_bio { 1128896a08dSQu Wenruo struct inode *inode; 11344b8bd7eSChris Mason struct bio *bio; 114a758781dSDavid Sterba extent_submit_bio_start_t *submit_bio_start; 11544b8bd7eSChris Mason int mirror_num; 116eaf25d93SChris Mason /* 117eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 118eaf25d93SChris Mason * can't tell us where in the file the bio should go 119eaf25d93SChris Mason */ 120eaf25d93SChris Mason u64 bio_offset; 1218b712842SChris Mason struct btrfs_work work; 1224e4cbee9SChristoph Hellwig blk_status_t status; 12344b8bd7eSChris Mason }; 12444b8bd7eSChris Mason 12585d4e461SChris Mason /* 12685d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 12785d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 12885d4e461SChris Mason * the level the eb occupies in the tree. 1294008c04aSChris Mason * 13085d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 13185d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 13285d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 1334fd786e6SMisono Tomohiro * by btrfs_root->root_key.objectid. This ensures that all special purpose 1344fd786e6SMisono Tomohiro * roots have separate keysets. 1354008c04aSChris Mason * 13685d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 13785d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 13885d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1394008c04aSChris Mason * 14085d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 14185d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 14285d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 14385d4e461SChris Mason * 14485d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 14585d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 14685d4e461SChris Mason * needs update as well. 1474008c04aSChris Mason */ 1484008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1494008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1504008c04aSChris Mason # error 1514008c04aSChris Mason # endif 15285d4e461SChris Mason 153ab1405aaSDavid Sterba #define DEFINE_LEVEL(stem, level) \ 154ab1405aaSDavid Sterba .names[level] = "btrfs-" stem "-0" #level, 155ab1405aaSDavid Sterba 156ab1405aaSDavid Sterba #define DEFINE_NAME(stem) \ 157ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 0) \ 158ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 1) \ 159ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 2) \ 160ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 3) \ 161ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 4) \ 162ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 5) \ 163ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 6) \ 164ab1405aaSDavid Sterba DEFINE_LEVEL(stem, 7) 165ab1405aaSDavid Sterba 16685d4e461SChris Mason static struct btrfs_lockdep_keyset { 16785d4e461SChris Mason u64 id; /* root objectid */ 168ab1405aaSDavid Sterba /* Longest entry: btrfs-free-space-00 */ 169387824afSDavid Sterba char names[BTRFS_MAX_LEVEL][20]; 170387824afSDavid Sterba struct lock_class_key keys[BTRFS_MAX_LEVEL]; 17185d4e461SChris Mason } btrfs_lockdep_keysets[] = { 172ab1405aaSDavid Sterba { .id = BTRFS_ROOT_TREE_OBJECTID, DEFINE_NAME("root") }, 173ab1405aaSDavid Sterba { .id = BTRFS_EXTENT_TREE_OBJECTID, DEFINE_NAME("extent") }, 174ab1405aaSDavid Sterba { .id = BTRFS_CHUNK_TREE_OBJECTID, DEFINE_NAME("chunk") }, 175ab1405aaSDavid Sterba { .id = BTRFS_DEV_TREE_OBJECTID, DEFINE_NAME("dev") }, 176ab1405aaSDavid Sterba { .id = BTRFS_CSUM_TREE_OBJECTID, DEFINE_NAME("csum") }, 177ab1405aaSDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, DEFINE_NAME("quota") }, 178ab1405aaSDavid Sterba { .id = BTRFS_TREE_LOG_OBJECTID, DEFINE_NAME("log") }, 179ab1405aaSDavid Sterba { .id = BTRFS_TREE_RELOC_OBJECTID, DEFINE_NAME("treloc") }, 180ab1405aaSDavid Sterba { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, DEFINE_NAME("dreloc") }, 181ab1405aaSDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, DEFINE_NAME("uuid") }, 182ab1405aaSDavid Sterba { .id = BTRFS_FREE_SPACE_TREE_OBJECTID, DEFINE_NAME("free-space") }, 183ab1405aaSDavid Sterba { .id = 0, DEFINE_NAME("tree") }, 1844008c04aSChris Mason }; 18585d4e461SChris Mason 186ab1405aaSDavid Sterba #undef DEFINE_LEVEL 187ab1405aaSDavid Sterba #undef DEFINE_NAME 18885d4e461SChris Mason 18985d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 19085d4e461SChris Mason int level) 19185d4e461SChris Mason { 19285d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 19385d4e461SChris Mason 19485d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 19585d4e461SChris Mason 19685d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 19785d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 19885d4e461SChris Mason if (ks->id == objectid) 19985d4e461SChris Mason break; 20085d4e461SChris Mason 20185d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 20285d4e461SChris Mason &ks->keys[level], ks->names[level]); 20385d4e461SChris Mason } 20485d4e461SChris Mason 2054008c04aSChris Mason #endif 2064008c04aSChris Mason 207d352ac68SChris Mason /* 2082996e1f8SJohannes Thumshirn * Compute the csum of a btree block and store the result to provided buffer. 209d352ac68SChris Mason */ 210c67b3892SDavid Sterba static void csum_tree_block(struct extent_buffer *buf, u8 *result) 21119c00ddcSChris Mason { 212d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = buf->fs_info; 213e9be5a30SDavid Sterba const int num_pages = fs_info->nodesize >> PAGE_SHIFT; 214a26663e7SQu Wenruo const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); 215d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 21619c00ddcSChris Mason char *kaddr; 217e9be5a30SDavid Sterba int i; 218d5178578SJohannes Thumshirn 219d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 220d5178578SJohannes Thumshirn crypto_shash_init(shash); 221a26663e7SQu Wenruo kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); 222e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, 223a26663e7SQu Wenruo first_page_part - BTRFS_CSUM_SIZE); 22419c00ddcSChris Mason 225e9be5a30SDavid Sterba for (i = 1; i < num_pages; i++) { 226e9be5a30SDavid Sterba kaddr = page_address(buf->pages[i]); 227e9be5a30SDavid Sterba crypto_shash_update(shash, kaddr, PAGE_SIZE); 22819c00ddcSChris Mason } 22971a63551SDavid Sterba memset(result, 0, BTRFS_CSUM_SIZE); 230d5178578SJohannes Thumshirn crypto_shash_final(shash, result); 23119c00ddcSChris Mason } 23219c00ddcSChris Mason 233d352ac68SChris Mason /* 234d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 235d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 236d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 237d352ac68SChris Mason * in the wrong place. 238d352ac68SChris Mason */ 2391259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 240b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 241b9fab919SChris Mason int atomic) 2421259ab75SChris Mason { 2432ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 2441259ab75SChris Mason int ret; 2452755a0deSDavid Sterba bool need_lock = (current->journal_info == BTRFS_SEND_TRANS_STUB); 2461259ab75SChris Mason 2471259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 2481259ab75SChris Mason return 0; 2491259ab75SChris Mason 250b9fab919SChris Mason if (atomic) 251b9fab919SChris Mason return -EAGAIN; 252b9fab919SChris Mason 253ac5887c8SJosef Bacik if (need_lock) 254a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 255a26e8c9fSJosef Bacik 2562ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 257ff13db41SDavid Sterba &cached_state); 2580b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 2591259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 2601259ab75SChris Mason ret = 0; 2611259ab75SChris Mason goto out; 2621259ab75SChris Mason } 26394647322SDavid Sterba btrfs_err_rl(eb->fs_info, 26494647322SDavid Sterba "parent transid verify failed on %llu wanted %llu found %llu", 26594647322SDavid Sterba eb->start, 26629549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 2671259ab75SChris Mason ret = 1; 268a26e8c9fSJosef Bacik 269a26e8c9fSJosef Bacik /* 270a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 271a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 27201327610SNicholas D Steeves * block that has been freed and re-allocated. So don't clear uptodate 273a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 274a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 275a26e8c9fSJosef Bacik * making everybody very sad. 276a26e8c9fSJosef Bacik */ 277a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 2780b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 27933958dc6SChris Mason out: 2802ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 281e43bbe5eSDavid Sterba &cached_state); 282472b909fSJosef Bacik if (need_lock) 283ac5887c8SJosef Bacik btrfs_tree_read_unlock(eb); 2841259ab75SChris Mason return ret; 2851259ab75SChris Mason } 2861259ab75SChris Mason 287e7e16f48SJohannes Thumshirn static bool btrfs_supported_super_csum(u16 csum_type) 288e7e16f48SJohannes Thumshirn { 289e7e16f48SJohannes Thumshirn switch (csum_type) { 290e7e16f48SJohannes Thumshirn case BTRFS_CSUM_TYPE_CRC32: 2913951e7f0SJohannes Thumshirn case BTRFS_CSUM_TYPE_XXHASH: 2923831bf00SJohannes Thumshirn case BTRFS_CSUM_TYPE_SHA256: 293352ae07bSDavid Sterba case BTRFS_CSUM_TYPE_BLAKE2: 294e7e16f48SJohannes Thumshirn return true; 295e7e16f48SJohannes Thumshirn default: 296e7e16f48SJohannes Thumshirn return false; 297e7e16f48SJohannes Thumshirn } 298e7e16f48SJohannes Thumshirn } 299e7e16f48SJohannes Thumshirn 300d352ac68SChris Mason /* 3011104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 3021104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 3031104a885SDavid Sterba */ 304ab8d0fc4SJeff Mahoney static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, 305ab8d0fc4SJeff Mahoney char *raw_disk_sb) 3061104a885SDavid Sterba { 3071104a885SDavid Sterba struct btrfs_super_block *disk_sb = 3081104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 30951bce6c9SJohannes Thumshirn char result[BTRFS_CSUM_SIZE]; 310d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 311d5178578SJohannes Thumshirn 312d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3131104a885SDavid Sterba 3141104a885SDavid Sterba /* 3151104a885SDavid Sterba * The super_block structure does not span the whole 31651bce6c9SJohannes Thumshirn * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is 31751bce6c9SJohannes Thumshirn * filled with zeros and is included in the checksum. 3181104a885SDavid Sterba */ 319fd08001fSEric Biggers crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, 320fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); 3211104a885SDavid Sterba 32255fc29beSDavid Sterba if (memcmp(disk_sb->csum, result, fs_info->csum_size)) 323e7e16f48SJohannes Thumshirn return 1; 3241104a885SDavid Sterba 325e7e16f48SJohannes Thumshirn return 0; 3261104a885SDavid Sterba } 3271104a885SDavid Sterba 328e064d5e9SDavid Sterba int btrfs_verify_level_key(struct extent_buffer *eb, int level, 329ff76a864SLiu Bo struct btrfs_key *first_key, u64 parent_transid) 330581c1760SQu Wenruo { 331e064d5e9SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 332581c1760SQu Wenruo int found_level; 333581c1760SQu Wenruo struct btrfs_key found_key; 334581c1760SQu Wenruo int ret; 335581c1760SQu Wenruo 336581c1760SQu Wenruo found_level = btrfs_header_level(eb); 337581c1760SQu Wenruo if (found_level != level) { 33863489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 33963489055SQu Wenruo KERN_ERR "BTRFS: tree level check failed\n"); 340581c1760SQu Wenruo btrfs_err(fs_info, 341581c1760SQu Wenruo "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", 342581c1760SQu Wenruo eb->start, level, found_level); 343581c1760SQu Wenruo return -EIO; 344581c1760SQu Wenruo } 345581c1760SQu Wenruo 346581c1760SQu Wenruo if (!first_key) 347581c1760SQu Wenruo return 0; 348581c1760SQu Wenruo 3495d41be6fSQu Wenruo /* 3505d41be6fSQu Wenruo * For live tree block (new tree blocks in current transaction), 3515d41be6fSQu Wenruo * we need proper lock context to avoid race, which is impossible here. 3525d41be6fSQu Wenruo * So we only checks tree blocks which is read from disk, whose 3535d41be6fSQu Wenruo * generation <= fs_info->last_trans_committed. 3545d41be6fSQu Wenruo */ 3555d41be6fSQu Wenruo if (btrfs_header_generation(eb) > fs_info->last_trans_committed) 3565d41be6fSQu Wenruo return 0; 35762fdaa52SQu Wenruo 35862fdaa52SQu Wenruo /* We have @first_key, so this @eb must have at least one item */ 35962fdaa52SQu Wenruo if (btrfs_header_nritems(eb) == 0) { 36062fdaa52SQu Wenruo btrfs_err(fs_info, 36162fdaa52SQu Wenruo "invalid tree nritems, bytenr=%llu nritems=0 expect >0", 36262fdaa52SQu Wenruo eb->start); 36362fdaa52SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 36462fdaa52SQu Wenruo return -EUCLEAN; 36562fdaa52SQu Wenruo } 36662fdaa52SQu Wenruo 367581c1760SQu Wenruo if (found_level) 368581c1760SQu Wenruo btrfs_node_key_to_cpu(eb, &found_key, 0); 369581c1760SQu Wenruo else 370581c1760SQu Wenruo btrfs_item_key_to_cpu(eb, &found_key, 0); 371581c1760SQu Wenruo ret = btrfs_comp_cpu_keys(first_key, &found_key); 372581c1760SQu Wenruo 373581c1760SQu Wenruo if (ret) { 37463489055SQu Wenruo WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), 37563489055SQu Wenruo KERN_ERR "BTRFS: tree first key check failed\n"); 376581c1760SQu Wenruo btrfs_err(fs_info, 377ff76a864SLiu Bo "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", 378ff76a864SLiu Bo eb->start, parent_transid, first_key->objectid, 379ff76a864SLiu Bo first_key->type, first_key->offset, 380ff76a864SLiu Bo found_key.objectid, found_key.type, 381ff76a864SLiu Bo found_key.offset); 382581c1760SQu Wenruo } 383581c1760SQu Wenruo return ret; 384581c1760SQu Wenruo } 385581c1760SQu Wenruo 3861104a885SDavid Sterba /* 387d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 388d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 389581c1760SQu Wenruo * 390581c1760SQu Wenruo * @parent_transid: expected transid, skip check if 0 391581c1760SQu Wenruo * @level: expected level, mandatory check 392581c1760SQu Wenruo * @first_key: expected key of first slot, skip check if NULL 393d352ac68SChris Mason */ 3945ab12d1fSDavid Sterba static int btree_read_extent_buffer_pages(struct extent_buffer *eb, 395581c1760SQu Wenruo u64 parent_transid, int level, 396581c1760SQu Wenruo struct btrfs_key *first_key) 397f188591eSChris Mason { 3985ab12d1fSDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 399f188591eSChris Mason struct extent_io_tree *io_tree; 400ea466794SJosef Bacik int failed = 0; 401f188591eSChris Mason int ret; 402f188591eSChris Mason int num_copies = 0; 403f188591eSChris Mason int mirror_num = 0; 404ea466794SJosef Bacik int failed_mirror = 0; 405f188591eSChris Mason 4060b246afaSJeff Mahoney io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; 407f188591eSChris Mason while (1) { 408f8397d69SNikolay Borisov clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 409c2ccfbc6SNikolay Borisov ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); 410256dd1bbSStefan Behrens if (!ret) { 411581c1760SQu Wenruo if (verify_parent_transid(io_tree, eb, 412b9fab919SChris Mason parent_transid, 0)) 413256dd1bbSStefan Behrens ret = -EIO; 414e064d5e9SDavid Sterba else if (btrfs_verify_level_key(eb, level, 415ff76a864SLiu Bo first_key, parent_transid)) 416581c1760SQu Wenruo ret = -EUCLEAN; 417581c1760SQu Wenruo else 418581c1760SQu Wenruo break; 419256dd1bbSStefan Behrens } 420d397712bSChris Mason 4210b246afaSJeff Mahoney num_copies = btrfs_num_copies(fs_info, 422f188591eSChris Mason eb->start, eb->len); 4234235298eSChris Mason if (num_copies == 1) 424ea466794SJosef Bacik break; 4254235298eSChris Mason 4265cf1ab56SJosef Bacik if (!failed_mirror) { 4275cf1ab56SJosef Bacik failed = 1; 4285cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4295cf1ab56SJosef Bacik } 4305cf1ab56SJosef Bacik 431f188591eSChris Mason mirror_num++; 432ea466794SJosef Bacik if (mirror_num == failed_mirror) 433ea466794SJosef Bacik mirror_num++; 434ea466794SJosef Bacik 4354235298eSChris Mason if (mirror_num > num_copies) 436ea466794SJosef Bacik break; 437f188591eSChris Mason } 438ea466794SJosef Bacik 439c0901581SStefan Behrens if (failed && !ret && failed_mirror) 44020a1fbf9SDavid Sterba btrfs_repair_eb_io_failure(eb, failed_mirror); 441ea466794SJosef Bacik 442ea466794SJosef Bacik return ret; 443f188591eSChris Mason } 44419c00ddcSChris Mason 445d352ac68SChris Mason /* 446ac303b69SQu Wenruo * Checksum a dirty tree block before IO. This has extra checks to make sure 447ac303b69SQu Wenruo * we only fill in the checksum field in the first page of a multi-page block. 448ac303b69SQu Wenruo * For subpage extent buffers we need bvec to also read the offset in the page. 449d352ac68SChris Mason */ 450ac303b69SQu Wenruo static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) 45119c00ddcSChris Mason { 452ac303b69SQu Wenruo struct page *page = bvec->bv_page; 4534eee4fa4SMiao Xie u64 start = page_offset(page); 45419c00ddcSChris Mason u64 found_start; 4552996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 45619c00ddcSChris Mason struct extent_buffer *eb; 4578d47a0d8SQu Wenruo int ret; 458f188591eSChris Mason 4594f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4604f2de97aSJosef Bacik if (page != eb->pages[0]) 4614f2de97aSJosef Bacik return 0; 4620f805531SAlex Lyakas 46319c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 4640f805531SAlex Lyakas /* 4650f805531SAlex Lyakas * Please do not consolidate these warnings into a single if. 4660f805531SAlex Lyakas * It is useful to know what went wrong. 4670f805531SAlex Lyakas */ 4680f805531SAlex Lyakas if (WARN_ON(found_start != start)) 4690f805531SAlex Lyakas return -EUCLEAN; 4700f805531SAlex Lyakas if (WARN_ON(!PageUptodate(page))) 4710f805531SAlex Lyakas return -EUCLEAN; 4720f805531SAlex Lyakas 473de37aa51SNikolay Borisov ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, 4749a8658e3SDavid Sterba offsetof(struct btrfs_header, fsid), 4759a8658e3SDavid Sterba BTRFS_FSID_SIZE) == 0); 4760f805531SAlex Lyakas 477c67b3892SDavid Sterba csum_tree_block(eb, result); 4782996e1f8SJohannes Thumshirn 4798d47a0d8SQu Wenruo if (btrfs_header_level(eb)) 4808d47a0d8SQu Wenruo ret = btrfs_check_node(eb); 4818d47a0d8SQu Wenruo else 4828d47a0d8SQu Wenruo ret = btrfs_check_leaf_full(eb); 4838d47a0d8SQu Wenruo 4848d47a0d8SQu Wenruo if (ret < 0) { 485c06631b0SQu Wenruo btrfs_print_tree(eb, 0); 4868d47a0d8SQu Wenruo btrfs_err(fs_info, 4878d47a0d8SQu Wenruo "block=%llu write time tree block corruption detected", 4888d47a0d8SQu Wenruo eb->start); 489c06631b0SQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 4908d47a0d8SQu Wenruo return ret; 4918d47a0d8SQu Wenruo } 492713cebfbSDavid Sterba write_extent_buffer(eb, result, 0, fs_info->csum_size); 4938d47a0d8SQu Wenruo 4942996e1f8SJohannes Thumshirn return 0; 49519c00ddcSChris Mason } 49619c00ddcSChris Mason 497b0c9b3b0SDavid Sterba static int check_tree_block_fsid(struct extent_buffer *eb) 4982b82032cSYan Zheng { 499b0c9b3b0SDavid Sterba struct btrfs_fs_info *fs_info = eb->fs_info; 500944d3f9fSNikolay Borisov struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; 50144880fdcSAnand Jain u8 fsid[BTRFS_FSID_SIZE]; 502944d3f9fSNikolay Borisov u8 *metadata_uuid; 5032b82032cSYan Zheng 5049a8658e3SDavid Sterba read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), 5059a8658e3SDavid Sterba BTRFS_FSID_SIZE); 5067239ff4bSNikolay Borisov /* 507944d3f9fSNikolay Borisov * Checking the incompat flag is only valid for the current fs. For 508944d3f9fSNikolay Borisov * seed devices it's forbidden to have their uuid changed so reading 509944d3f9fSNikolay Borisov * ->fsid in this case is fine 5107239ff4bSNikolay Borisov */ 511944d3f9fSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) 5127239ff4bSNikolay Borisov metadata_uuid = fs_devices->metadata_uuid; 5137239ff4bSNikolay Borisov else 5147239ff4bSNikolay Borisov metadata_uuid = fs_devices->fsid; 5157239ff4bSNikolay Borisov 516944d3f9fSNikolay Borisov if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) 517944d3f9fSNikolay Borisov return 0; 518944d3f9fSNikolay Borisov 519944d3f9fSNikolay Borisov list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) 520944d3f9fSNikolay Borisov if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) 521944d3f9fSNikolay Borisov return 0; 522944d3f9fSNikolay Borisov 523944d3f9fSNikolay Borisov return 1; 5242b82032cSYan Zheng } 5252b82032cSYan Zheng 52677bf40a2SQu Wenruo /* Do basic extent buffer checks at read time */ 52777bf40a2SQu Wenruo static int validate_extent_buffer(struct extent_buffer *eb) 528ce9adaa5SChris Mason { 52977bf40a2SQu Wenruo struct btrfs_fs_info *fs_info = eb->fs_info; 530ce9adaa5SChris Mason u64 found_start; 53177bf40a2SQu Wenruo const u32 csum_size = fs_info->csum_size; 53277bf40a2SQu Wenruo u8 found_level; 5332996e1f8SJohannes Thumshirn u8 result[BTRFS_CSUM_SIZE]; 53477bf40a2SQu Wenruo int ret = 0; 535ea466794SJosef Bacik 536ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 537727011e0SChris Mason if (found_start != eb->start) { 538893bf4b1SSu Yue btrfs_err_rl(fs_info, "bad tree block start, want %llu have %llu", 539893bf4b1SSu Yue eb->start, found_start); 540f188591eSChris Mason ret = -EIO; 54177bf40a2SQu Wenruo goto out; 542ce9adaa5SChris Mason } 543b0c9b3b0SDavid Sterba if (check_tree_block_fsid(eb)) { 54402873e43SZhao Lei btrfs_err_rl(fs_info, "bad fsid on block %llu", 54594647322SDavid Sterba eb->start); 5461259ab75SChris Mason ret = -EIO; 54777bf40a2SQu Wenruo goto out; 5481259ab75SChris Mason } 549ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 5501c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 551893bf4b1SSu Yue btrfs_err(fs_info, "bad tree block level %d on %llu", 552893bf4b1SSu Yue (int)btrfs_header_level(eb), eb->start); 5531c24c3ceSJosef Bacik ret = -EIO; 55477bf40a2SQu Wenruo goto out; 5551c24c3ceSJosef Bacik } 556ce9adaa5SChris Mason 557c67b3892SDavid Sterba csum_tree_block(eb, result); 558a826d6dcSJosef Bacik 5592996e1f8SJohannes Thumshirn if (memcmp_extent_buffer(eb, result, 0, csum_size)) { 56035be8851SJohannes Thumshirn u8 val[BTRFS_CSUM_SIZE] = { 0 }; 5612996e1f8SJohannes Thumshirn 5622996e1f8SJohannes Thumshirn read_extent_buffer(eb, &val, 0, csum_size); 5632996e1f8SJohannes Thumshirn btrfs_warn_rl(fs_info, 56435be8851SJohannes Thumshirn "%s checksum verify failed on %llu wanted " CSUM_FMT " found " CSUM_FMT " level %d", 5652996e1f8SJohannes Thumshirn fs_info->sb->s_id, eb->start, 56635be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, val), 56735be8851SJohannes Thumshirn CSUM_FMT_VALUE(csum_size, result), 56835be8851SJohannes Thumshirn btrfs_header_level(eb)); 5692996e1f8SJohannes Thumshirn ret = -EUCLEAN; 57077bf40a2SQu Wenruo goto out; 5712996e1f8SJohannes Thumshirn } 5722996e1f8SJohannes Thumshirn 573a826d6dcSJosef Bacik /* 574a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 575a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 576a826d6dcSJosef Bacik * return -EIO. 577a826d6dcSJosef Bacik */ 5781c4360eeSDavid Sterba if (found_level == 0 && btrfs_check_leaf_full(eb)) { 579a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 580a826d6dcSJosef Bacik ret = -EIO; 581a826d6dcSJosef Bacik } 582ce9adaa5SChris Mason 583813fd1dcSDavid Sterba if (found_level > 0 && btrfs_check_node(eb)) 584053ab70fSLiu Bo ret = -EIO; 585053ab70fSLiu Bo 5860b32f4bbSJosef Bacik if (!ret) 5870b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 58875391f0dSQu Wenruo else 58975391f0dSQu Wenruo btrfs_err(fs_info, 59075391f0dSQu Wenruo "block=%llu read time tree block corruption detected", 59175391f0dSQu Wenruo eb->start); 59277bf40a2SQu Wenruo out: 59377bf40a2SQu Wenruo return ret; 59477bf40a2SQu Wenruo } 59577bf40a2SQu Wenruo 5968e1dc982SQu Wenruo int btrfs_validate_metadata_buffer(struct btrfs_io_bio *io_bio, 59777bf40a2SQu Wenruo struct page *page, u64 start, u64 end, 59877bf40a2SQu Wenruo int mirror) 59977bf40a2SQu Wenruo { 60077bf40a2SQu Wenruo struct extent_buffer *eb; 60177bf40a2SQu Wenruo int ret = 0; 60277bf40a2SQu Wenruo int reads_done; 60377bf40a2SQu Wenruo 60477bf40a2SQu Wenruo ASSERT(page->private); 60577bf40a2SQu Wenruo eb = (struct extent_buffer *)page->private; 60677bf40a2SQu Wenruo 60777bf40a2SQu Wenruo /* 60877bf40a2SQu Wenruo * The pending IO might have been the only thing that kept this buffer 60977bf40a2SQu Wenruo * in memory. Make sure we have a ref for all this other checks 61077bf40a2SQu Wenruo */ 61177bf40a2SQu Wenruo atomic_inc(&eb->refs); 61277bf40a2SQu Wenruo 61377bf40a2SQu Wenruo reads_done = atomic_dec_and_test(&eb->io_pages); 61477bf40a2SQu Wenruo if (!reads_done) 61577bf40a2SQu Wenruo goto err; 61677bf40a2SQu Wenruo 61777bf40a2SQu Wenruo eb->read_mirror = mirror; 61877bf40a2SQu Wenruo if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { 61977bf40a2SQu Wenruo ret = -EIO; 62077bf40a2SQu Wenruo goto err; 62177bf40a2SQu Wenruo } 62277bf40a2SQu Wenruo ret = validate_extent_buffer(eb); 623ce9adaa5SChris Mason err: 62479fb65a1SJosef Bacik if (reads_done && 62579fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 626d48d71aaSDavid Sterba btree_readahead_hook(eb, ret); 6274bb31e92SArne Jansen 62853b381b3SDavid Woodhouse if (ret) { 62953b381b3SDavid Woodhouse /* 63053b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 63153b381b3SDavid Woodhouse * again, we have to make sure it has something 63253b381b3SDavid Woodhouse * to decrement 63353b381b3SDavid Woodhouse */ 63453b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6350b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 63653b381b3SDavid Woodhouse } 6370b32f4bbSJosef Bacik free_extent_buffer(eb); 63877bf40a2SQu Wenruo 639f188591eSChris Mason return ret; 640ce9adaa5SChris Mason } 641ce9adaa5SChris Mason 6424246a0b6SChristoph Hellwig static void end_workqueue_bio(struct bio *bio) 643ce9adaa5SChris Mason { 64497eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq = bio->bi_private; 645ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 6469e0af237SLiu Bo struct btrfs_workqueue *wq; 647ce9adaa5SChris Mason 648ce9adaa5SChris Mason fs_info = end_io_wq->info; 6494e4cbee9SChristoph Hellwig end_io_wq->status = bio->bi_status; 650d20f7043SChris Mason 65137226b21SMike Christie if (bio_op(bio) == REQ_OP_WRITE) { 652a0cac0ecSOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) 6539e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 654a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) 6559e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 656a0cac0ecSOmar Sandoval else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 6579e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 658a0cac0ecSOmar Sandoval else 6599e0af237SLiu Bo wq = fs_info->endio_write_workers; 6609e0af237SLiu Bo } else { 6615c047a69SOmar Sandoval if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 6629e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 663a0cac0ecSOmar Sandoval else if (end_io_wq->metadata) 6649e0af237SLiu Bo wq = fs_info->endio_meta_workers; 665a0cac0ecSOmar Sandoval else 6669e0af237SLiu Bo wq = fs_info->endio_workers; 6679e0af237SLiu Bo } 6689e0af237SLiu Bo 669a0cac0ecSOmar Sandoval btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); 6709e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 671ce9adaa5SChris Mason } 672ce9adaa5SChris Mason 6734e4cbee9SChristoph Hellwig blk_status_t btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 674bfebd8b5SDavid Sterba enum btrfs_wq_endio_type metadata) 6750b86a832SChris Mason { 67697eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 6778b110e39SMiao Xie 67897eb6b69SDavid Sterba end_io_wq = kmem_cache_alloc(btrfs_end_io_wq_cache, GFP_NOFS); 679ce9adaa5SChris Mason if (!end_io_wq) 6804e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 681ce9adaa5SChris Mason 682ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 683ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 68422c59948SChris Mason end_io_wq->info = info; 6854e4cbee9SChristoph Hellwig end_io_wq->status = 0; 686ce9adaa5SChris Mason end_io_wq->bio = bio; 68722c59948SChris Mason end_io_wq->metadata = metadata; 688ce9adaa5SChris Mason 689ce9adaa5SChris Mason bio->bi_private = end_io_wq; 690ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 69122c59948SChris Mason return 0; 69222c59948SChris Mason } 69322c59948SChris Mason 6944a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 6954a69a410SChris Mason { 6964a69a410SChris Mason struct async_submit_bio *async; 6974e4cbee9SChristoph Hellwig blk_status_t ret; 6984a69a410SChris Mason 6994a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7008896a08dSQu Wenruo ret = async->submit_bio_start(async->inode, async->bio, async->bio_offset); 70179787eaaSJeff Mahoney if (ret) 7024e4cbee9SChristoph Hellwig async->status = ret; 7034a69a410SChris Mason } 7044a69a410SChris Mason 70506ea01b1SDavid Sterba /* 70606ea01b1SDavid Sterba * In order to insert checksums into the metadata in large chunks, we wait 70706ea01b1SDavid Sterba * until bio submission time. All the pages in the bio are checksummed and 70806ea01b1SDavid Sterba * sums are attached onto the ordered extent record. 70906ea01b1SDavid Sterba * 71006ea01b1SDavid Sterba * At IO completion time the csums attached on the ordered extent record are 71106ea01b1SDavid Sterba * inserted into the tree. 71206ea01b1SDavid Sterba */ 7134a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7148b712842SChris Mason { 7158b712842SChris Mason struct async_submit_bio *async; 71606ea01b1SDavid Sterba struct inode *inode; 71706ea01b1SDavid Sterba blk_status_t ret; 7188b712842SChris Mason 7198b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 7208896a08dSQu Wenruo inode = async->inode; 7214854ddd0SChris Mason 722bb7ab3b9SAdam Buchbinder /* If an error occurred we just want to clean up the bio and move on */ 7234e4cbee9SChristoph Hellwig if (async->status) { 7244e4cbee9SChristoph Hellwig async->bio->bi_status = async->status; 7254246a0b6SChristoph Hellwig bio_endio(async->bio); 72679787eaaSJeff Mahoney return; 72779787eaaSJeff Mahoney } 72879787eaaSJeff Mahoney 729ec39f769SChris Mason /* 730ec39f769SChris Mason * All of the bios that pass through here are from async helpers. 731ec39f769SChris Mason * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. 732ec39f769SChris Mason * This changes nothing when cgroups aren't in use. 733ec39f769SChris Mason */ 734ec39f769SChris Mason async->bio->bi_opf |= REQ_CGROUP_PUNT; 73508635baeSChris Mason ret = btrfs_map_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); 73606ea01b1SDavid Sterba if (ret) { 73706ea01b1SDavid Sterba async->bio->bi_status = ret; 73806ea01b1SDavid Sterba bio_endio(async->bio); 73906ea01b1SDavid Sterba } 7404a69a410SChris Mason } 7414a69a410SChris Mason 7424a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 7434a69a410SChris Mason { 7444a69a410SChris Mason struct async_submit_bio *async; 7454a69a410SChris Mason 7464a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 7478b712842SChris Mason kfree(async); 7488b712842SChris Mason } 7498b712842SChris Mason 7508896a08dSQu Wenruo blk_status_t btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, 751c6100a4bSJosef Bacik int mirror_num, unsigned long bio_flags, 7528896a08dSQu Wenruo u64 bio_offset, 753e288c080SDavid Sterba extent_submit_bio_start_t *submit_bio_start) 75444b8bd7eSChris Mason { 7558896a08dSQu Wenruo struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; 75644b8bd7eSChris Mason struct async_submit_bio *async; 75744b8bd7eSChris Mason 75844b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 75944b8bd7eSChris Mason if (!async) 7604e4cbee9SChristoph Hellwig return BLK_STS_RESOURCE; 76144b8bd7eSChris Mason 7628896a08dSQu Wenruo async->inode = inode; 76344b8bd7eSChris Mason async->bio = bio; 76444b8bd7eSChris Mason async->mirror_num = mirror_num; 7654a69a410SChris Mason async->submit_bio_start = submit_bio_start; 7664a69a410SChris Mason 767a0cac0ecSOmar Sandoval btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, 768a0cac0ecSOmar Sandoval run_one_async_free); 7694a69a410SChris Mason 770eaf25d93SChris Mason async->bio_offset = bio_offset; 7718c8bee1dSChris Mason 7724e4cbee9SChristoph Hellwig async->status = 0; 77379787eaaSJeff Mahoney 77467f055c7SChristoph Hellwig if (op_is_sync(bio->bi_opf)) 7755cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 776d313d7a3SChris Mason 7775cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 77844b8bd7eSChris Mason return 0; 77944b8bd7eSChris Mason } 78044b8bd7eSChris Mason 7814e4cbee9SChristoph Hellwig static blk_status_t btree_csum_one_bio(struct bio *bio) 782ce3ed71aSChris Mason { 7832c30c71bSKent Overstreet struct bio_vec *bvec; 784ce3ed71aSChris Mason struct btrfs_root *root; 7852b070cfeSChristoph Hellwig int ret = 0; 7866dc4f100SMing Lei struct bvec_iter_all iter_all; 787ce3ed71aSChris Mason 788c09abff8SDavid Sterba ASSERT(!bio_flagged(bio, BIO_CLONED)); 7892b070cfeSChristoph Hellwig bio_for_each_segment_all(bvec, bio, iter_all) { 790ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 791ac303b69SQu Wenruo ret = csum_dirty_buffer(root->fs_info, bvec); 79279787eaaSJeff Mahoney if (ret) 79379787eaaSJeff Mahoney break; 794ce3ed71aSChris Mason } 7952c30c71bSKent Overstreet 7964e4cbee9SChristoph Hellwig return errno_to_blk_status(ret); 797ce3ed71aSChris Mason } 798ce3ed71aSChris Mason 7998896a08dSQu Wenruo static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, 800eaf25d93SChris Mason u64 bio_offset) 80122c59948SChris Mason { 8028b712842SChris Mason /* 8038b712842SChris Mason * when we're called for a write, we're already in the async 8045443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8058b712842SChris Mason */ 80679787eaaSJeff Mahoney return btree_csum_one_bio(bio); 80722c59948SChris Mason } 80822c59948SChris Mason 8099b4e675aSDavid Sterba static int check_async_write(struct btrfs_fs_info *fs_info, 8109b4e675aSDavid Sterba struct btrfs_inode *bi) 811de0022b9SJosef Bacik { 8126300463bSLiu Bo if (atomic_read(&bi->sync_writers)) 8136300463bSLiu Bo return 0; 8149b4e675aSDavid Sterba if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) 815de0022b9SJosef Bacik return 0; 816de0022b9SJosef Bacik return 1; 817de0022b9SJosef Bacik } 818de0022b9SJosef Bacik 8191b36294aSNikolay Borisov blk_status_t btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, 8201b36294aSNikolay Borisov int mirror_num, unsigned long bio_flags) 82144b8bd7eSChris Mason { 8220b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); 8239b4e675aSDavid Sterba int async = check_async_write(fs_info, BTRFS_I(inode)); 8244e4cbee9SChristoph Hellwig blk_status_t ret; 825cad321adSChris Mason 82637226b21SMike Christie if (bio_op(bio) != REQ_OP_WRITE) { 827cad321adSChris Mason /* 828cad321adSChris Mason * called for a read, do the setup so that checksum validation 829cad321adSChris Mason * can happen in the async kernel threads 830cad321adSChris Mason */ 8310b246afaSJeff Mahoney ret = btrfs_bio_wq_end_io(fs_info, bio, 8320b246afaSJeff Mahoney BTRFS_WQ_ENDIO_METADATA); 8331d4284bdSChris Mason if (ret) 83461891923SStefan Behrens goto out_w_error; 83508635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 836de0022b9SJosef Bacik } else if (!async) { 837de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 838de0022b9SJosef Bacik if (ret) 83961891923SStefan Behrens goto out_w_error; 84008635baeSChris Mason ret = btrfs_map_bio(fs_info, bio, mirror_num); 84161891923SStefan Behrens } else { 842cad321adSChris Mason /* 84361891923SStefan Behrens * kthread helpers are used to submit writes so that 84461891923SStefan Behrens * checksumming can happen in parallel across all CPUs 845cad321adSChris Mason */ 8468896a08dSQu Wenruo ret = btrfs_wq_submit_bio(inode, bio, mirror_num, 0, 8478896a08dSQu Wenruo 0, btree_submit_bio_start); 84844b8bd7eSChris Mason } 84944b8bd7eSChris Mason 8504246a0b6SChristoph Hellwig if (ret) 8514246a0b6SChristoph Hellwig goto out_w_error; 8524246a0b6SChristoph Hellwig return 0; 8534246a0b6SChristoph Hellwig 85461891923SStefan Behrens out_w_error: 8554e4cbee9SChristoph Hellwig bio->bi_status = ret; 8564246a0b6SChristoph Hellwig bio_endio(bio); 85761891923SStefan Behrens return ret; 85861891923SStefan Behrens } 85961891923SStefan Behrens 8603dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 861784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 862a6bc32b8SMel Gorman struct page *newpage, struct page *page, 863a6bc32b8SMel Gorman enum migrate_mode mode) 864784b4e29SChris Mason { 865784b4e29SChris Mason /* 866784b4e29SChris Mason * we can't safely write a btree page from here, 867784b4e29SChris Mason * we haven't done the locking hook 868784b4e29SChris Mason */ 869784b4e29SChris Mason if (PageDirty(page)) 870784b4e29SChris Mason return -EAGAIN; 871784b4e29SChris Mason /* 872784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 873784b4e29SChris Mason * We must have no buffers or drop them. 874784b4e29SChris Mason */ 875784b4e29SChris Mason if (page_has_private(page) && 876784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 877784b4e29SChris Mason return -EAGAIN; 878a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 879784b4e29SChris Mason } 8803dd1462eSJan Beulich #endif 881784b4e29SChris Mason 8820da5468fSChris Mason 8830da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 8840da5468fSChris Mason struct writeback_control *wbc) 8850da5468fSChris Mason { 886e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 887e2d84521SMiao Xie int ret; 888e2d84521SMiao Xie 889d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 890448d640bSChris Mason 891448d640bSChris Mason if (wbc->for_kupdate) 892448d640bSChris Mason return 0; 893448d640bSChris Mason 894e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 895b9473439SChris Mason /* this is a bit racy, but that's ok */ 896d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 897d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 898d814a491SEthan Lien fs_info->dirty_metadata_batch); 899e2d84521SMiao Xie if (ret < 0) 900793955bcSChris Mason return 0; 901793955bcSChris Mason } 9020b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9030da5468fSChris Mason } 9040da5468fSChris Mason 90570dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 9065f39d397SChris Mason { 90798509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 90898509cfcSChris Mason return 0; 9090c4e538bSDavid Sterba 910f7a52a40SDavid Sterba return try_release_extent_buffer(page); 911d98237b3SChris Mason } 912d98237b3SChris Mason 913d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 914d47992f8SLukas Czerner unsigned int length) 915d98237b3SChris Mason { 916d1310b2eSChris Mason struct extent_io_tree *tree; 917d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 9185f39d397SChris Mason extent_invalidatepage(tree, page, offset); 9195f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 9209ad6b7bcSChris Mason if (PagePrivate(page)) { 921efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 922efe120a0SFrank Holton "page private not zero on page %llu", 923efe120a0SFrank Holton (unsigned long long)page_offset(page)); 924d1b89bc0SGuoqing Jiang detach_page_private(page); 9259ad6b7bcSChris Mason } 926d98237b3SChris Mason } 927d98237b3SChris Mason 9280b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 9290b32f4bbSJosef Bacik { 930bb146eb2SJosef Bacik #ifdef DEBUG 9310b32f4bbSJosef Bacik struct extent_buffer *eb; 9320b32f4bbSJosef Bacik 9330b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 9340b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 9350b32f4bbSJosef Bacik BUG_ON(!eb); 9360b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 9370b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 9380b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 939bb146eb2SJosef Bacik #endif 9400b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 9410b32f4bbSJosef Bacik } 9420b32f4bbSJosef Bacik 9437f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 9440da5468fSChris Mason .writepages = btree_writepages, 9455f39d397SChris Mason .releasepage = btree_releasepage, 9465f39d397SChris Mason .invalidatepage = btree_invalidatepage, 9475a92bc88SChris Mason #ifdef CONFIG_MIGRATION 948784b4e29SChris Mason .migratepage = btree_migratepage, 9495a92bc88SChris Mason #endif 9500b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 951d98237b3SChris Mason }; 952123abc88SChris Mason 9532ff7e61eSJeff Mahoney struct extent_buffer *btrfs_find_create_tree_block( 9542ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info, 9553fbaf258SJosef Bacik u64 bytenr, u64 owner_root, 9563fbaf258SJosef Bacik int level) 9570999df54SChris Mason { 9580b246afaSJeff Mahoney if (btrfs_is_testing(fs_info)) 9590b246afaSJeff Mahoney return alloc_test_extent_buffer(fs_info, bytenr); 9603fbaf258SJosef Bacik return alloc_extent_buffer(fs_info, bytenr, owner_root, level); 9610999df54SChris Mason } 9620999df54SChris Mason 963581c1760SQu Wenruo /* 964581c1760SQu Wenruo * Read tree block at logical address @bytenr and do variant basic but critical 965581c1760SQu Wenruo * verification. 966581c1760SQu Wenruo * 9671b7ec85eSJosef Bacik * @owner_root: the objectid of the root owner for this block. 968581c1760SQu Wenruo * @parent_transid: expected transid of this tree block, skip check if 0 969581c1760SQu Wenruo * @level: expected level, mandatory check 970581c1760SQu Wenruo * @first_key: expected key in slot 0, skip check if NULL 971581c1760SQu Wenruo */ 9722ff7e61eSJeff Mahoney struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, 9731b7ec85eSJosef Bacik u64 owner_root, u64 parent_transid, 9741b7ec85eSJosef Bacik int level, struct btrfs_key *first_key) 975e20d96d6SChris Mason { 9765f39d397SChris Mason struct extent_buffer *buf = NULL; 97719c00ddcSChris Mason int ret; 97819c00ddcSChris Mason 9793fbaf258SJosef Bacik buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); 980c871b0f2SLiu Bo if (IS_ERR(buf)) 981c871b0f2SLiu Bo return buf; 982e4204dedSChris Mason 9835ab12d1fSDavid Sterba ret = btree_read_extent_buffer_pages(buf, parent_transid, 984581c1760SQu Wenruo level, first_key); 9850f0fe8f7SFilipe David Borba Manana if (ret) { 986537f38f0SNikolay Borisov free_extent_buffer_stale(buf); 98764c043deSLiu Bo return ERR_PTR(ret); 9880f0fe8f7SFilipe David Borba Manana } 9895f39d397SChris Mason return buf; 990ce9adaa5SChris Mason 991eb60ceacSChris Mason } 992eb60ceacSChris Mason 9936a884d7dSDavid Sterba void btrfs_clean_tree_block(struct extent_buffer *buf) 994ed2ff2cbSChris Mason { 9956a884d7dSDavid Sterba struct btrfs_fs_info *fs_info = buf->fs_info; 99655c69072SChris Mason if (btrfs_header_generation(buf) == 997e2d84521SMiao Xie fs_info->running_transaction->transid) { 998b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 999b4ce94deSChris Mason 1000b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1001104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 1002e2d84521SMiao Xie -buf->len, 1003e2d84521SMiao Xie fs_info->dirty_metadata_batch); 10040b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1005925baeddSChris Mason } 10065f39d397SChris Mason } 1007ed7b63ebSJosef Bacik } 10085f39d397SChris Mason 1009da17066cSJeff Mahoney static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1010e20d96d6SChris Mason u64 objectid) 1011d97e63b6SChris Mason { 10127c0260eeSJeff Mahoney bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 101396dfcb46SJosef Bacik root->fs_info = fs_info; 1014cfaa7295SChris Mason root->node = NULL; 1015a28ec197SChris Mason root->commit_root = NULL; 101627cdeb70SMiao Xie root->state = 0; 1017d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 10180b86a832SChris Mason 10190f7d52f4SChris Mason root->last_trans = 0; 102013a8a7c8SYan, Zheng root->highest_objectid = 0; 1021eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1022199c2a9cSMiao Xie root->nr_ordered_extents = 0; 10236bef4d31SEric Paris root->inode_tree = RB_ROOT; 102416cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1025f0486c68SYan, Zheng root->block_rsv = NULL; 10260b86a832SChris Mason 10270b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 10285d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1029eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1030eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1031199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1032199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 1033d2311e69SQu Wenruo INIT_LIST_HEAD(&root->reloc_dirty_list); 10342ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 10352ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 10365d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1037eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1038199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1039f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 10402ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 10412ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 10428287475aSQu Wenruo spin_lock_init(&root->qgroup_meta_rsv_lock); 1043a2135011SChris Mason mutex_init(&root->objectid_mutex); 1044e02119d5SChris Mason mutex_init(&root->log_mutex); 104531f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1046573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 1047c53e9653SQu Wenruo init_waitqueue_head(&root->qgroup_flush_wait); 10487237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 10497237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 10507237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 10518b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 10528b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 10537237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 10547237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 10557237f183SYan Zheng atomic_set(&root->log_writers, 0); 10562ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 10570700cea7SElena Reshetova refcount_set(&root->refs, 1); 10588ecebf4dSRobbie Ko atomic_set(&root->snapshot_force_cow, 0); 1059eede2bf3SOmar Sandoval atomic_set(&root->nr_swapfiles, 0); 10607237f183SYan Zheng root->log_transid = 0; 1061d1433debSMiao Xie root->log_transid_committed = -1; 1062257c62e1SChris Mason root->last_log_commit = 0; 1063e289f03eSFilipe Manana if (!dummy) { 106443eb5f29SQu Wenruo extent_io_tree_init(fs_info, &root->dirty_log_pages, 106543eb5f29SQu Wenruo IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); 1066e289f03eSFilipe Manana extent_io_tree_init(fs_info, &root->log_csum_range, 1067e289f03eSFilipe Manana IO_TREE_LOG_CSUM_RANGE, NULL); 1068e289f03eSFilipe Manana } 1069017e5369SChris Mason 10703768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 10713768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 10726702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 10734d775673SChris Mason root->root_key.objectid = objectid; 10740ee5dc67SAl Viro root->anon_dev = 0; 10758ea05e3aSAlexander Block 10765f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 1077370a11b8SQu Wenruo btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); 1078bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1079bd647ce3SJosef Bacik INIT_LIST_HEAD(&root->leak_list); 1080bd647ce3SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 1081bd647ce3SJosef Bacik list_add_tail(&root->leak_list, &fs_info->allocated_roots); 1082bd647ce3SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 1083bd647ce3SJosef Bacik #endif 10843768f368SChris Mason } 10853768f368SChris Mason 108674e4d827SDavid Sterba static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, 108796dfcb46SJosef Bacik u64 objectid, gfp_t flags) 10886f07e42eSAl Viro { 108974e4d827SDavid Sterba struct btrfs_root *root = kzalloc(sizeof(*root), flags); 10906f07e42eSAl Viro if (root) 109196dfcb46SJosef Bacik __setup_root(root, fs_info, objectid); 10926f07e42eSAl Viro return root; 10936f07e42eSAl Viro } 10946f07e42eSAl Viro 109506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 109606ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 1097da17066cSJeff Mahoney struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) 109806ea65a3SJosef Bacik { 109906ea65a3SJosef Bacik struct btrfs_root *root; 110006ea65a3SJosef Bacik 11017c0260eeSJeff Mahoney if (!fs_info) 11027c0260eeSJeff Mahoney return ERR_PTR(-EINVAL); 11037c0260eeSJeff Mahoney 110496dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); 110506ea65a3SJosef Bacik if (!root) 110606ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1107da17066cSJeff Mahoney 1108b9ef22deSFeifei Xu /* We don't use the stripesize in selftest, set it as sectorsize */ 1109faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 111006ea65a3SJosef Bacik 111106ea65a3SJosef Bacik return root; 111206ea65a3SJosef Bacik } 111306ea65a3SJosef Bacik #endif 111406ea65a3SJosef Bacik 111520897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 111620897f5cSArne Jansen u64 objectid) 111720897f5cSArne Jansen { 11189b7a2440SDavid Sterba struct btrfs_fs_info *fs_info = trans->fs_info; 111920897f5cSArne Jansen struct extent_buffer *leaf; 112020897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 112120897f5cSArne Jansen struct btrfs_root *root; 112220897f5cSArne Jansen struct btrfs_key key; 1123b89f6d1fSFilipe Manana unsigned int nofs_flag; 112420897f5cSArne Jansen int ret = 0; 112520897f5cSArne Jansen 1126b89f6d1fSFilipe Manana /* 1127b89f6d1fSFilipe Manana * We're holding a transaction handle, so use a NOFS memory allocation 1128b89f6d1fSFilipe Manana * context to avoid deadlock if reclaim happens. 1129b89f6d1fSFilipe Manana */ 1130b89f6d1fSFilipe Manana nofs_flag = memalloc_nofs_save(); 113196dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); 1132b89f6d1fSFilipe Manana memalloc_nofs_restore(nofs_flag); 113320897f5cSArne Jansen if (!root) 113420897f5cSArne Jansen return ERR_PTR(-ENOMEM); 113520897f5cSArne Jansen 113620897f5cSArne Jansen root->root_key.objectid = objectid; 113720897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 113820897f5cSArne Jansen root->root_key.offset = 0; 113920897f5cSArne Jansen 11409631e4ccSJosef Bacik leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, 11419631e4ccSJosef Bacik BTRFS_NESTING_NORMAL); 114220897f5cSArne Jansen if (IS_ERR(leaf)) { 114320897f5cSArne Jansen ret = PTR_ERR(leaf); 11441dd05682STsutomu Itoh leaf = NULL; 1145*8a6a87cdSBoris Burkov goto fail_unlock; 114620897f5cSArne Jansen } 114720897f5cSArne Jansen 114820897f5cSArne Jansen root->node = leaf; 114920897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 115020897f5cSArne Jansen 115120897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 115227cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 115320897f5cSArne Jansen 1154f944d2cbSDavid Sterba btrfs_set_root_flags(&root->root_item, 0); 1155f944d2cbSDavid Sterba btrfs_set_root_limit(&root->root_item, 0); 115620897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 115720897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 115820897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 115920897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 116020897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 116120897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 116220897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 116333d85fdaSQu Wenruo if (is_fstree(objectid)) 1164807fc790SAndy Shevchenko generate_random_guid(root->root_item.uuid); 1165807fc790SAndy Shevchenko else 1166807fc790SAndy Shevchenko export_guid(root->root_item.uuid, &guid_null); 1167c8422684SDavid Sterba btrfs_set_root_drop_level(&root->root_item, 0); 116820897f5cSArne Jansen 1169*8a6a87cdSBoris Burkov btrfs_tree_unlock(leaf); 1170*8a6a87cdSBoris Burkov 117120897f5cSArne Jansen key.objectid = objectid; 117220897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 117320897f5cSArne Jansen key.offset = 0; 117420897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 117520897f5cSArne Jansen if (ret) 117620897f5cSArne Jansen goto fail; 117720897f5cSArne Jansen 117820897f5cSArne Jansen return root; 11791dd05682STsutomu Itoh 1180*8a6a87cdSBoris Burkov fail_unlock: 11818c38938cSJosef Bacik if (leaf) 11821dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 1183*8a6a87cdSBoris Burkov fail: 118400246528SJosef Bacik btrfs_put_root(root); 11851dd05682STsutomu Itoh 11861dd05682STsutomu Itoh return ERR_PTR(ret); 118720897f5cSArne Jansen } 118820897f5cSArne Jansen 11897237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1190e02119d5SChris Mason struct btrfs_fs_info *fs_info) 11910f7d52f4SChris Mason { 11920f7d52f4SChris Mason struct btrfs_root *root; 11937237f183SYan Zheng struct extent_buffer *leaf; 1194e02119d5SChris Mason 119596dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); 1196e02119d5SChris Mason if (!root) 11977237f183SYan Zheng return ERR_PTR(-ENOMEM); 1198e02119d5SChris Mason 1199e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1200e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1201e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 120227cdeb70SMiao Xie 12037237f183SYan Zheng /* 120492a7cc42SQu Wenruo * DON'T set SHAREABLE bit for log trees. 120527cdeb70SMiao Xie * 120692a7cc42SQu Wenruo * Log trees are not exposed to user space thus can't be snapshotted, 120792a7cc42SQu Wenruo * and they go away before a real commit is actually done. 120892a7cc42SQu Wenruo * 120992a7cc42SQu Wenruo * They do store pointers to file data extents, and those reference 121092a7cc42SQu Wenruo * counts still get updated (along with back refs to the log tree). 12117237f183SYan Zheng */ 1212e02119d5SChris Mason 12134d75f8a9SDavid Sterba leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, 12149631e4ccSJosef Bacik NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); 12157237f183SYan Zheng if (IS_ERR(leaf)) { 121600246528SJosef Bacik btrfs_put_root(root); 12177237f183SYan Zheng return ERR_CAST(leaf); 12187237f183SYan Zheng } 1219e02119d5SChris Mason 12207237f183SYan Zheng root->node = leaf; 1221e02119d5SChris Mason 1222e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1223e02119d5SChris Mason btrfs_tree_unlock(root->node); 12247237f183SYan Zheng return root; 12257237f183SYan Zheng } 12267237f183SYan Zheng 12277237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 12287237f183SYan Zheng struct btrfs_fs_info *fs_info) 12297237f183SYan Zheng { 12307237f183SYan Zheng struct btrfs_root *log_root; 12317237f183SYan Zheng 12327237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 12337237f183SYan Zheng if (IS_ERR(log_root)) 12347237f183SYan Zheng return PTR_ERR(log_root); 12357237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 12367237f183SYan Zheng fs_info->log_root_tree = log_root; 12377237f183SYan Zheng return 0; 12387237f183SYan Zheng } 12397237f183SYan Zheng 12407237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 12417237f183SYan Zheng struct btrfs_root *root) 12427237f183SYan Zheng { 12430b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 12447237f183SYan Zheng struct btrfs_root *log_root; 12457237f183SYan Zheng struct btrfs_inode_item *inode_item; 12467237f183SYan Zheng 12470b246afaSJeff Mahoney log_root = alloc_log_tree(trans, fs_info); 12487237f183SYan Zheng if (IS_ERR(log_root)) 12497237f183SYan Zheng return PTR_ERR(log_root); 12507237f183SYan Zheng 12517237f183SYan Zheng log_root->last_trans = trans->transid; 12527237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 12537237f183SYan Zheng 12547237f183SYan Zheng inode_item = &log_root->root_item.inode; 12553cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 12563cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 12573cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1258da17066cSJeff Mahoney btrfs_set_stack_inode_nbytes(inode_item, 12590b246afaSJeff Mahoney fs_info->nodesize); 12603cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 12617237f183SYan Zheng 12625d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 12637237f183SYan Zheng 12647237f183SYan Zheng WARN_ON(root->log_root); 12657237f183SYan Zheng root->log_root = log_root; 12667237f183SYan Zheng root->log_transid = 0; 1267d1433debSMiao Xie root->log_transid_committed = -1; 1268257c62e1SChris Mason root->last_log_commit = 0; 1269e02119d5SChris Mason return 0; 1270e02119d5SChris Mason } 1271e02119d5SChris Mason 127249d11beaSJosef Bacik static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, 127349d11beaSJosef Bacik struct btrfs_path *path, 1274cb517eabSMiao Xie struct btrfs_key *key) 1275e02119d5SChris Mason { 1276e02119d5SChris Mason struct btrfs_root *root; 1277e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 127884234f3aSYan Zheng u64 generation; 1279cb517eabSMiao Xie int ret; 1280581c1760SQu Wenruo int level; 1281cb517eabSMiao Xie 128296dfcb46SJosef Bacik root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); 128349d11beaSJosef Bacik if (!root) 128449d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 12850f7d52f4SChris Mason 1286cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1287cb517eabSMiao Xie &root->root_item, &root->root_key); 12880f7d52f4SChris Mason if (ret) { 128913a8a7c8SYan, Zheng if (ret > 0) 129013a8a7c8SYan, Zheng ret = -ENOENT; 129149d11beaSJosef Bacik goto fail; 12920f7d52f4SChris Mason } 129313a8a7c8SYan, Zheng 129484234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1295581c1760SQu Wenruo level = btrfs_root_level(&root->root_item); 12962ff7e61eSJeff Mahoney root->node = read_tree_block(fs_info, 12972ff7e61eSJeff Mahoney btrfs_root_bytenr(&root->root_item), 12981b7ec85eSJosef Bacik key->objectid, generation, level, NULL); 129964c043deSLiu Bo if (IS_ERR(root->node)) { 130064c043deSLiu Bo ret = PTR_ERR(root->node); 13018c38938cSJosef Bacik root->node = NULL; 130249d11beaSJosef Bacik goto fail; 1303cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1304cb517eabSMiao Xie ret = -EIO; 130549d11beaSJosef Bacik goto fail; 1306416bc658SJosef Bacik } 13075d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 1308cb517eabSMiao Xie return root; 130949d11beaSJosef Bacik fail: 131000246528SJosef Bacik btrfs_put_root(root); 131149d11beaSJosef Bacik return ERR_PTR(ret); 131249d11beaSJosef Bacik } 131349d11beaSJosef Bacik 131449d11beaSJosef Bacik struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 131549d11beaSJosef Bacik struct btrfs_key *key) 131649d11beaSJosef Bacik { 131749d11beaSJosef Bacik struct btrfs_root *root; 131849d11beaSJosef Bacik struct btrfs_path *path; 131949d11beaSJosef Bacik 132049d11beaSJosef Bacik path = btrfs_alloc_path(); 132149d11beaSJosef Bacik if (!path) 132249d11beaSJosef Bacik return ERR_PTR(-ENOMEM); 132349d11beaSJosef Bacik root = read_tree_root_path(tree_root, path, key); 132449d11beaSJosef Bacik btrfs_free_path(path); 132549d11beaSJosef Bacik 132649d11beaSJosef Bacik return root; 1327cb517eabSMiao Xie } 1328cb517eabSMiao Xie 13292dfb1e43SQu Wenruo /* 13302dfb1e43SQu Wenruo * Initialize subvolume root in-memory structure 13312dfb1e43SQu Wenruo * 13322dfb1e43SQu Wenruo * @anon_dev: anonymous device to attach to the root, if zero, allocate new 13332dfb1e43SQu Wenruo */ 13342dfb1e43SQu Wenruo static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) 1335cb517eabSMiao Xie { 1336cb517eabSMiao Xie int ret; 1337dcc3eb96SNikolay Borisov unsigned int nofs_flag; 1338cb517eabSMiao Xie 1339dcc3eb96SNikolay Borisov /* 1340dcc3eb96SNikolay Borisov * We might be called under a transaction (e.g. indirect backref 1341dcc3eb96SNikolay Borisov * resolution) which could deadlock if it triggers memory reclaim 1342dcc3eb96SNikolay Borisov */ 1343dcc3eb96SNikolay Borisov nofs_flag = memalloc_nofs_save(); 1344dcc3eb96SNikolay Borisov ret = btrfs_drew_lock_init(&root->snapshot_lock); 1345dcc3eb96SNikolay Borisov memalloc_nofs_restore(nofs_flag); 1346dcc3eb96SNikolay Borisov if (ret) 13478257b2dcSMiao Xie goto fail; 13488257b2dcSMiao Xie 1349aeb935a4SQu Wenruo if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && 1350aeb935a4SQu Wenruo root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) { 135192a7cc42SQu Wenruo set_bit(BTRFS_ROOT_SHAREABLE, &root->state); 1352f39e4571SJosef Bacik btrfs_check_and_init_root_item(&root->root_item); 1353f39e4571SJosef Bacik } 1354f39e4571SJosef Bacik 1355851fd730SQu Wenruo /* 1356851fd730SQu Wenruo * Don't assign anonymous block device to roots that are not exposed to 1357851fd730SQu Wenruo * userspace, the id pool is limited to 1M 1358851fd730SQu Wenruo */ 1359851fd730SQu Wenruo if (is_fstree(root->root_key.objectid) && 1360851fd730SQu Wenruo btrfs_root_refs(&root->root_item) > 0) { 13612dfb1e43SQu Wenruo if (!anon_dev) { 1362cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1363cb517eabSMiao Xie if (ret) 1364876d2cf1SLiu Bo goto fail; 13652dfb1e43SQu Wenruo } else { 13662dfb1e43SQu Wenruo root->anon_dev = anon_dev; 13672dfb1e43SQu Wenruo } 1368851fd730SQu Wenruo } 1369f32e48e9SChandan Rajendra 1370f32e48e9SChandan Rajendra mutex_lock(&root->objectid_mutex); 1371f32e48e9SChandan Rajendra ret = btrfs_find_highest_objectid(root, 1372f32e48e9SChandan Rajendra &root->highest_objectid); 1373f32e48e9SChandan Rajendra if (ret) { 1374f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1375876d2cf1SLiu Bo goto fail; 1376f32e48e9SChandan Rajendra } 1377f32e48e9SChandan Rajendra 1378f32e48e9SChandan Rajendra ASSERT(root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 1379f32e48e9SChandan Rajendra 1380f32e48e9SChandan Rajendra mutex_unlock(&root->objectid_mutex); 1381f32e48e9SChandan Rajendra 1382cb517eabSMiao Xie return 0; 1383cb517eabSMiao Xie fail: 138484db5ccfSDavid Sterba /* The caller is responsible to call btrfs_free_fs_root */ 1385cb517eabSMiao Xie return ret; 1386cb517eabSMiao Xie } 1387cb517eabSMiao Xie 1388a98db0f3SJosef Bacik static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1389cb517eabSMiao Xie u64 root_id) 1390cb517eabSMiao Xie { 1391cb517eabSMiao Xie struct btrfs_root *root; 1392cb517eabSMiao Xie 1393cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1394cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1395cb517eabSMiao Xie (unsigned long)root_id); 1396bc44d7c4SJosef Bacik if (root) 139700246528SJosef Bacik root = btrfs_grab_root(root); 1398cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1399cb517eabSMiao Xie return root; 1400cb517eabSMiao Xie } 1401cb517eabSMiao Xie 140249d11beaSJosef Bacik static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, 140349d11beaSJosef Bacik u64 objectid) 140449d11beaSJosef Bacik { 140549d11beaSJosef Bacik if (objectid == BTRFS_ROOT_TREE_OBJECTID) 140649d11beaSJosef Bacik return btrfs_grab_root(fs_info->tree_root); 140749d11beaSJosef Bacik if (objectid == BTRFS_EXTENT_TREE_OBJECTID) 140849d11beaSJosef Bacik return btrfs_grab_root(fs_info->extent_root); 140949d11beaSJosef Bacik if (objectid == BTRFS_CHUNK_TREE_OBJECTID) 141049d11beaSJosef Bacik return btrfs_grab_root(fs_info->chunk_root); 141149d11beaSJosef Bacik if (objectid == BTRFS_DEV_TREE_OBJECTID) 141249d11beaSJosef Bacik return btrfs_grab_root(fs_info->dev_root); 141349d11beaSJosef Bacik if (objectid == BTRFS_CSUM_TREE_OBJECTID) 141449d11beaSJosef Bacik return btrfs_grab_root(fs_info->csum_root); 141549d11beaSJosef Bacik if (objectid == BTRFS_QUOTA_TREE_OBJECTID) 141649d11beaSJosef Bacik return btrfs_grab_root(fs_info->quota_root) ? 141749d11beaSJosef Bacik fs_info->quota_root : ERR_PTR(-ENOENT); 141849d11beaSJosef Bacik if (objectid == BTRFS_UUID_TREE_OBJECTID) 141949d11beaSJosef Bacik return btrfs_grab_root(fs_info->uuid_root) ? 142049d11beaSJosef Bacik fs_info->uuid_root : ERR_PTR(-ENOENT); 142149d11beaSJosef Bacik if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) 142249d11beaSJosef Bacik return btrfs_grab_root(fs_info->free_space_root) ? 142349d11beaSJosef Bacik fs_info->free_space_root : ERR_PTR(-ENOENT); 142449d11beaSJosef Bacik return NULL; 142549d11beaSJosef Bacik } 142649d11beaSJosef Bacik 1427cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1428cb517eabSMiao Xie struct btrfs_root *root) 1429cb517eabSMiao Xie { 1430cb517eabSMiao Xie int ret; 1431cb517eabSMiao Xie 1432e1860a77SDavid Sterba ret = radix_tree_preload(GFP_NOFS); 1433cb517eabSMiao Xie if (ret) 1434cb517eabSMiao Xie return ret; 1435cb517eabSMiao Xie 1436cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1437cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1438cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1439cb517eabSMiao Xie root); 1440af01d2e5SJosef Bacik if (ret == 0) { 144100246528SJosef Bacik btrfs_grab_root(root); 144227cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1443af01d2e5SJosef Bacik } 1444cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1445cb517eabSMiao Xie radix_tree_preload_end(); 1446cb517eabSMiao Xie 1447cb517eabSMiao Xie return ret; 1448cb517eabSMiao Xie } 1449cb517eabSMiao Xie 1450bd647ce3SJosef Bacik void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) 1451bd647ce3SJosef Bacik { 1452bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 1453bd647ce3SJosef Bacik struct btrfs_root *root; 1454bd647ce3SJosef Bacik 1455bd647ce3SJosef Bacik while (!list_empty(&fs_info->allocated_roots)) { 1456457f1864SJosef Bacik char buf[BTRFS_ROOT_NAME_BUF_LEN]; 1457457f1864SJosef Bacik 1458bd647ce3SJosef Bacik root = list_first_entry(&fs_info->allocated_roots, 1459bd647ce3SJosef Bacik struct btrfs_root, leak_list); 1460457f1864SJosef Bacik btrfs_err(fs_info, "leaked root %s refcount %d", 1461457f1864SJosef Bacik btrfs_root_name(root->root_key.objectid, buf), 1462bd647ce3SJosef Bacik refcount_read(&root->refs)); 1463bd647ce3SJosef Bacik while (refcount_read(&root->refs) > 1) 146400246528SJosef Bacik btrfs_put_root(root); 146500246528SJosef Bacik btrfs_put_root(root); 1466bd647ce3SJosef Bacik } 1467bd647ce3SJosef Bacik #endif 1468bd647ce3SJosef Bacik } 1469bd647ce3SJosef Bacik 14700d4b0463SJosef Bacik void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) 14710d4b0463SJosef Bacik { 1472141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 1473141386e1SJosef Bacik percpu_counter_destroy(&fs_info->delalloc_bytes); 1474141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dio_bytes); 1475141386e1SJosef Bacik percpu_counter_destroy(&fs_info->dev_replace.bio_counter); 1476141386e1SJosef Bacik btrfs_free_csum_hash(fs_info); 1477141386e1SJosef Bacik btrfs_free_stripe_hash_table(fs_info); 1478141386e1SJosef Bacik btrfs_free_ref_cache(fs_info); 14790d4b0463SJosef Bacik kfree(fs_info->balance_ctl); 14800d4b0463SJosef Bacik kfree(fs_info->delayed_root); 148100246528SJosef Bacik btrfs_put_root(fs_info->extent_root); 148200246528SJosef Bacik btrfs_put_root(fs_info->tree_root); 148300246528SJosef Bacik btrfs_put_root(fs_info->chunk_root); 148400246528SJosef Bacik btrfs_put_root(fs_info->dev_root); 148500246528SJosef Bacik btrfs_put_root(fs_info->csum_root); 148600246528SJosef Bacik btrfs_put_root(fs_info->quota_root); 148700246528SJosef Bacik btrfs_put_root(fs_info->uuid_root); 148800246528SJosef Bacik btrfs_put_root(fs_info->free_space_root); 148900246528SJosef Bacik btrfs_put_root(fs_info->fs_root); 1490aeb935a4SQu Wenruo btrfs_put_root(fs_info->data_reloc_root); 1491bd647ce3SJosef Bacik btrfs_check_leaked_roots(fs_info); 14923fd63727SJosef Bacik btrfs_extent_buffer_leak_debug_check(fs_info); 14930d4b0463SJosef Bacik kfree(fs_info->super_copy); 14940d4b0463SJosef Bacik kfree(fs_info->super_for_commit); 14950d4b0463SJosef Bacik kvfree(fs_info); 14960d4b0463SJosef Bacik } 14970d4b0463SJosef Bacik 14980d4b0463SJosef Bacik 14992dfb1e43SQu Wenruo /* 15002dfb1e43SQu Wenruo * Get an in-memory reference of a root structure. 15012dfb1e43SQu Wenruo * 15022dfb1e43SQu Wenruo * For essential trees like root/extent tree, we grab it from fs_info directly. 15032dfb1e43SQu Wenruo * For subvolume trees, we check the cached filesystem roots first. If not 15042dfb1e43SQu Wenruo * found, then read it from disk and add it to cached fs roots. 15052dfb1e43SQu Wenruo * 15062dfb1e43SQu Wenruo * Caller should release the root by calling btrfs_put_root() after the usage. 15072dfb1e43SQu Wenruo * 15082dfb1e43SQu Wenruo * NOTE: Reloc and log trees can't be read by this function as they share the 15092dfb1e43SQu Wenruo * same root objectid. 15102dfb1e43SQu Wenruo * 15112dfb1e43SQu Wenruo * @objectid: root id 15122dfb1e43SQu Wenruo * @anon_dev: preallocated anonymous block device number for new roots, 15132dfb1e43SQu Wenruo * pass 0 for new allocation. 15142dfb1e43SQu Wenruo * @check_ref: whether to check root item references, If true, return -ENOENT 15152dfb1e43SQu Wenruo * for orphan roots 15162dfb1e43SQu Wenruo */ 15172dfb1e43SQu Wenruo static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, 15182dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev, 15192dfb1e43SQu Wenruo bool check_ref) 15205eda7b5eSChris Mason { 15215eda7b5eSChris Mason struct btrfs_root *root; 1522381cf658SDavid Sterba struct btrfs_path *path; 15231d4c08e0SDavid Sterba struct btrfs_key key; 15245eda7b5eSChris Mason int ret; 15255eda7b5eSChris Mason 152649d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 152749d11beaSJosef Bacik if (root) 152849d11beaSJosef Bacik return root; 15294df27c4dSYan, Zheng again: 153056e9357aSDavid Sterba root = btrfs_lookup_fs_root(fs_info, objectid); 153148475471SStefan Behrens if (root) { 15322dfb1e43SQu Wenruo /* Shouldn't get preallocated anon_dev for cached roots */ 15332dfb1e43SQu Wenruo ASSERT(!anon_dev); 1534bc44d7c4SJosef Bacik if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 153500246528SJosef Bacik btrfs_put_root(root); 153648475471SStefan Behrens return ERR_PTR(-ENOENT); 1537bc44d7c4SJosef Bacik } 15385eda7b5eSChris Mason return root; 153948475471SStefan Behrens } 15405eda7b5eSChris Mason 154156e9357aSDavid Sterba key.objectid = objectid; 154256e9357aSDavid Sterba key.type = BTRFS_ROOT_ITEM_KEY; 154356e9357aSDavid Sterba key.offset = (u64)-1; 154456e9357aSDavid Sterba root = btrfs_read_tree_root(fs_info->tree_root, &key); 15455eda7b5eSChris Mason if (IS_ERR(root)) 15465eda7b5eSChris Mason return root; 15473394e160SChris Mason 1548c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1549d68fc57bSYan, Zheng ret = -ENOENT; 1550d68fc57bSYan, Zheng goto fail; 1551d68fc57bSYan, Zheng } 1552d68fc57bSYan, Zheng 15532dfb1e43SQu Wenruo ret = btrfs_init_fs_root(root, anon_dev); 1554cb517eabSMiao Xie if (ret) 1555cb517eabSMiao Xie goto fail; 1556cb517eabSMiao Xie 1557381cf658SDavid Sterba path = btrfs_alloc_path(); 1558381cf658SDavid Sterba if (!path) { 1559381cf658SDavid Sterba ret = -ENOMEM; 1560381cf658SDavid Sterba goto fail; 1561381cf658SDavid Sterba } 15621d4c08e0SDavid Sterba key.objectid = BTRFS_ORPHAN_OBJECTID; 15631d4c08e0SDavid Sterba key.type = BTRFS_ORPHAN_ITEM_KEY; 156456e9357aSDavid Sterba key.offset = objectid; 15651d4c08e0SDavid Sterba 15661d4c08e0SDavid Sterba ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); 1567381cf658SDavid Sterba btrfs_free_path(path); 1568d68fc57bSYan, Zheng if (ret < 0) 1569d68fc57bSYan, Zheng goto fail; 1570d68fc57bSYan, Zheng if (ret == 0) 157127cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1572d68fc57bSYan, Zheng 1573cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 15740f7d52f4SChris Mason if (ret) { 157500246528SJosef Bacik btrfs_put_root(root); 15764785e24fSJosef Bacik if (ret == -EEXIST) 15774df27c4dSYan, Zheng goto again; 15784df27c4dSYan, Zheng goto fail; 15794df27c4dSYan, Zheng } 1580edbd8d4eSChris Mason return root; 15814df27c4dSYan, Zheng fail: 15828c38938cSJosef Bacik btrfs_put_root(root); 15834df27c4dSYan, Zheng return ERR_PTR(ret); 1584edbd8d4eSChris Mason } 1585edbd8d4eSChris Mason 15862dfb1e43SQu Wenruo /* 15872dfb1e43SQu Wenruo * Get in-memory reference of a root structure 15882dfb1e43SQu Wenruo * 15892dfb1e43SQu Wenruo * @objectid: tree objectid 15902dfb1e43SQu Wenruo * @check_ref: if set, verify that the tree exists and the item has at least 15912dfb1e43SQu Wenruo * one reference 15922dfb1e43SQu Wenruo */ 15932dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 15942dfb1e43SQu Wenruo u64 objectid, bool check_ref) 15952dfb1e43SQu Wenruo { 15962dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); 15972dfb1e43SQu Wenruo } 15982dfb1e43SQu Wenruo 15992dfb1e43SQu Wenruo /* 16002dfb1e43SQu Wenruo * Get in-memory reference of a root structure, created as new, optionally pass 16012dfb1e43SQu Wenruo * the anonymous block device id 16022dfb1e43SQu Wenruo * 16032dfb1e43SQu Wenruo * @objectid: tree objectid 16042dfb1e43SQu Wenruo * @anon_dev: if zero, allocate a new anonymous block device or use the 16052dfb1e43SQu Wenruo * parameter value 16062dfb1e43SQu Wenruo */ 16072dfb1e43SQu Wenruo struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, 16082dfb1e43SQu Wenruo u64 objectid, dev_t anon_dev) 16092dfb1e43SQu Wenruo { 16102dfb1e43SQu Wenruo return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); 16112dfb1e43SQu Wenruo } 16122dfb1e43SQu Wenruo 16138b712842SChris Mason /* 161449d11beaSJosef Bacik * btrfs_get_fs_root_commit_root - return a root for the given objectid 161549d11beaSJosef Bacik * @fs_info: the fs_info 161649d11beaSJosef Bacik * @objectid: the objectid we need to lookup 161749d11beaSJosef Bacik * 161849d11beaSJosef Bacik * This is exclusively used for backref walking, and exists specifically because 161949d11beaSJosef Bacik * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref 162049d11beaSJosef Bacik * creation time, which means we may have to read the tree_root in order to look 162149d11beaSJosef Bacik * up a fs root that is not in memory. If the root is not in memory we will 162249d11beaSJosef Bacik * read the tree root commit root and look up the fs root from there. This is a 162349d11beaSJosef Bacik * temporary root, it will not be inserted into the radix tree as it doesn't 162449d11beaSJosef Bacik * have the most uptodate information, it'll simply be discarded once the 162549d11beaSJosef Bacik * backref code is finished using the root. 162649d11beaSJosef Bacik */ 162749d11beaSJosef Bacik struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, 162849d11beaSJosef Bacik struct btrfs_path *path, 162949d11beaSJosef Bacik u64 objectid) 163049d11beaSJosef Bacik { 163149d11beaSJosef Bacik struct btrfs_root *root; 163249d11beaSJosef Bacik struct btrfs_key key; 163349d11beaSJosef Bacik 163449d11beaSJosef Bacik ASSERT(path->search_commit_root && path->skip_locking); 163549d11beaSJosef Bacik 163649d11beaSJosef Bacik /* 163749d11beaSJosef Bacik * This can return -ENOENT if we ask for a root that doesn't exist, but 163849d11beaSJosef Bacik * since this is called via the backref walking code we won't be looking 163949d11beaSJosef Bacik * up a root that doesn't exist, unless there's corruption. So if root 164049d11beaSJosef Bacik * != NULL just return it. 164149d11beaSJosef Bacik */ 164249d11beaSJosef Bacik root = btrfs_get_global_root(fs_info, objectid); 164349d11beaSJosef Bacik if (root) 164449d11beaSJosef Bacik return root; 164549d11beaSJosef Bacik 164649d11beaSJosef Bacik root = btrfs_lookup_fs_root(fs_info, objectid); 164749d11beaSJosef Bacik if (root) 164849d11beaSJosef Bacik return root; 164949d11beaSJosef Bacik 165049d11beaSJosef Bacik key.objectid = objectid; 165149d11beaSJosef Bacik key.type = BTRFS_ROOT_ITEM_KEY; 165249d11beaSJosef Bacik key.offset = (u64)-1; 165349d11beaSJosef Bacik root = read_tree_root_path(fs_info->tree_root, path, &key); 165449d11beaSJosef Bacik btrfs_release_path(path); 165549d11beaSJosef Bacik 165649d11beaSJosef Bacik return root; 165749d11beaSJosef Bacik } 165849d11beaSJosef Bacik 165949d11beaSJosef Bacik /* 16608b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 16618b712842SChris Mason * functions. This is where read checksum verification actually happens 16628b712842SChris Mason */ 16638b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1664ce9adaa5SChris Mason { 1665ce9adaa5SChris Mason struct bio *bio; 166697eb6b69SDavid Sterba struct btrfs_end_io_wq *end_io_wq; 1667ce9adaa5SChris Mason 166897eb6b69SDavid Sterba end_io_wq = container_of(work, struct btrfs_end_io_wq, work); 1669ce9adaa5SChris Mason bio = end_io_wq->bio; 16708b712842SChris Mason 16714e4cbee9SChristoph Hellwig bio->bi_status = end_io_wq->status; 1672ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1673ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 16744246a0b6SChristoph Hellwig bio_endio(bio); 16759be490f1SOmar Sandoval kmem_cache_free(btrfs_end_io_wq_cache, end_io_wq); 1676ce9adaa5SChris Mason } 167744b8bd7eSChris Mason 1678a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1679a74a4b97SChris Mason { 1680a74a4b97SChris Mason struct btrfs_root *root = arg; 16810b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1682d0278245SMiao Xie int again; 1683a74a4b97SChris Mason 1684d6fd0ae2SOmar Sandoval while (1) { 1685d0278245SMiao Xie again = 0; 16869d1a2a3aSDavid Sterba 1687fd340d0fSJosef Bacik set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1688fd340d0fSJosef Bacik 1689d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 16902ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) 1691d0278245SMiao Xie goto sleep; 1692d0278245SMiao Xie 169390c711abSZygo Blaxell /* 169490c711abSZygo Blaxell * Do not do anything if we might cause open_ctree() to block 169590c711abSZygo Blaxell * before we have finished mounting the filesystem. 169690c711abSZygo Blaxell */ 16970b246afaSJeff Mahoney if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) 169890c711abSZygo Blaxell goto sleep; 169990c711abSZygo Blaxell 17000b246afaSJeff Mahoney if (!mutex_trylock(&fs_info->cleaner_mutex)) 1701d0278245SMiao Xie goto sleep; 1702d0278245SMiao Xie 1703dc7f370cSMiao Xie /* 1704dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1705dc7f370cSMiao Xie * during the above check and trylock. 1706dc7f370cSMiao Xie */ 17072ff7e61eSJeff Mahoney if (btrfs_need_cleaner_sleep(fs_info)) { 17080b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1709dc7f370cSMiao Xie goto sleep; 1710dc7f370cSMiao Xie } 1711dc7f370cSMiao Xie 17122ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 1713c2d6cb16SFilipe Manana 17149d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 17150b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 1716a74a4b97SChris Mason 1717d0278245SMiao Xie /* 171805323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 171905323cd1SMiao Xie * needn't do anything special here. 1720d0278245SMiao Xie */ 17210b246afaSJeff Mahoney btrfs_run_defrag_inodes(fs_info); 172267c5e7d4SFilipe Manana 172367c5e7d4SFilipe Manana /* 172467c5e7d4SFilipe Manana * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 172567c5e7d4SFilipe Manana * with relocation (btrfs_relocate_chunk) and relocation 172667c5e7d4SFilipe Manana * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) 172767c5e7d4SFilipe Manana * after acquiring fs_info->delete_unused_bgs_mutex. So we 172867c5e7d4SFilipe Manana * can't hold, nor need to, fs_info->cleaner_mutex when deleting 172967c5e7d4SFilipe Manana * unused block groups. 173067c5e7d4SFilipe Manana */ 17310b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 1732d0278245SMiao Xie sleep: 1733fd340d0fSJosef Bacik clear_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); 1734d6fd0ae2SOmar Sandoval if (kthread_should_park()) 1735d6fd0ae2SOmar Sandoval kthread_parkme(); 1736d6fd0ae2SOmar Sandoval if (kthread_should_stop()) 1737d6fd0ae2SOmar Sandoval return 0; 1738838fe188SJiri Kosina if (!again) { 1739a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 1740a74a4b97SChris Mason schedule(); 1741a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1742a74a4b97SChris Mason } 1743da288d28SFilipe Manana } 1744a74a4b97SChris Mason } 1745a74a4b97SChris Mason 1746a74a4b97SChris Mason static int transaction_kthread(void *arg) 1747a74a4b97SChris Mason { 1748a74a4b97SChris Mason struct btrfs_root *root = arg; 17490b246afaSJeff Mahoney struct btrfs_fs_info *fs_info = root->fs_info; 1750a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1751a74a4b97SChris Mason struct btrfs_transaction *cur; 17528929ecfaSYan, Zheng u64 transid; 1753643900beSNikolay Borisov time64_t delta; 1754a74a4b97SChris Mason unsigned long delay; 1755914b2007SJan Kara bool cannot_commit; 1756a74a4b97SChris Mason 1757a74a4b97SChris Mason do { 1758914b2007SJan Kara cannot_commit = false; 1759ba1bc00fSNikolay Borisov delay = msecs_to_jiffies(fs_info->commit_interval * 1000); 17600b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 1761a74a4b97SChris Mason 17620b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 17630b246afaSJeff Mahoney cur = fs_info->running_transaction; 1764a74a4b97SChris Mason if (!cur) { 17650b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1766a74a4b97SChris Mason goto sleep; 1767a74a4b97SChris Mason } 176831153d81SYan Zheng 1769643900beSNikolay Borisov delta = ktime_get_seconds() - cur->start_time; 17703296bf56SQu Wenruo if (cur->state < TRANS_STATE_COMMIT_START && 1771643900beSNikolay Borisov delta < fs_info->commit_interval) { 17720b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 1773fb8a7e94SNikolay Borisov delay -= msecs_to_jiffies((delta - 1) * 1000); 1774fb8a7e94SNikolay Borisov delay = min(delay, 1775fb8a7e94SNikolay Borisov msecs_to_jiffies(fs_info->commit_interval * 1000)); 1776a74a4b97SChris Mason goto sleep; 1777a74a4b97SChris Mason } 17788929ecfaSYan, Zheng transid = cur->transid; 17790b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 178056bec294SChris Mason 178179787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1782354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1783914b2007SJan Kara if (IS_ERR(trans)) { 1784354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1785914b2007SJan Kara cannot_commit = true; 178679787eaaSJeff Mahoney goto sleep; 1787914b2007SJan Kara } 17888929ecfaSYan, Zheng if (transid == trans->transid) { 17893a45bb20SJeff Mahoney btrfs_commit_transaction(trans); 17908929ecfaSYan, Zheng } else { 17913a45bb20SJeff Mahoney btrfs_end_transaction(trans); 17928929ecfaSYan, Zheng } 1793a74a4b97SChris Mason sleep: 17940b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 17950b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 1796a74a4b97SChris Mason 17974e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 17980b246afaSJeff Mahoney &fs_info->fs_state))) 17992ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 18008929ecfaSYan, Zheng if (!kthread_should_stop() && 18010b246afaSJeff Mahoney (!btrfs_transaction_blocked(fs_info) || 1802914b2007SJan Kara cannot_commit)) 1803bc5511d0SNikolay Borisov schedule_timeout_interruptible(delay); 1804a74a4b97SChris Mason } while (!kthread_should_stop()); 1805a74a4b97SChris Mason return 0; 1806a74a4b97SChris Mason } 1807a74a4b97SChris Mason 1808af31f5e5SChris Mason /* 180901f0f9daSNikolay Borisov * This will find the highest generation in the array of root backups. The 181001f0f9daSNikolay Borisov * index of the highest array is returned, or -EINVAL if we can't find 181101f0f9daSNikolay Borisov * anything. 1812af31f5e5SChris Mason * 1813af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1814af31f5e5SChris Mason * generation of the latest root in the array with the generation 1815af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1816af31f5e5SChris Mason */ 181701f0f9daSNikolay Borisov static int find_newest_super_backup(struct btrfs_fs_info *info) 1818af31f5e5SChris Mason { 181901f0f9daSNikolay Borisov const u64 newest_gen = btrfs_super_generation(info->super_copy); 1820af31f5e5SChris Mason u64 cur; 1821af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1822af31f5e5SChris Mason int i; 1823af31f5e5SChris Mason 1824af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1825af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1826af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1827af31f5e5SChris Mason if (cur == newest_gen) 182801f0f9daSNikolay Borisov return i; 1829af31f5e5SChris Mason } 1830af31f5e5SChris Mason 183101f0f9daSNikolay Borisov return -EINVAL; 1832af31f5e5SChris Mason } 1833af31f5e5SChris Mason 1834af31f5e5SChris Mason /* 1835af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1836af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1837af31f5e5SChris Mason * done 1838af31f5e5SChris Mason */ 1839af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1840af31f5e5SChris Mason { 18416ef108ddSNikolay Borisov const int next_backup = info->backup_root_index; 1842af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1843af31f5e5SChris Mason 1844af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1845af31f5e5SChris Mason 1846af31f5e5SChris Mason /* 1847af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1848af31f5e5SChris Mason * regardless of which ones we use today 1849af31f5e5SChris Mason */ 1850af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1851af31f5e5SChris Mason 1852af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1853af31f5e5SChris Mason 1854af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1855af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1856af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1857af31f5e5SChris Mason 1858af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1859af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1860af31f5e5SChris Mason 1861af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1862af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1863af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1864af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1865af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1866af31f5e5SChris Mason 1867af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1868af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1869af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1870af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1871af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1872af31f5e5SChris Mason 18737c7e82a7SChris Mason /* 18747c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 18757c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 18767c7e82a7SChris Mason */ 18777c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 18787c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 18797c7e82a7SChris Mason info->fs_root->node->start); 1880af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1881af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1882af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1883af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 18847c7e82a7SChris Mason } 1885af31f5e5SChris Mason 1886af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1887af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1888af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1889af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1890af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1891af31f5e5SChris Mason 1892af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1893af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1894af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1895af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1896af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1897af31f5e5SChris Mason 1898af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1899af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1900af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1901af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1902af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1903af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1904af31f5e5SChris Mason 1905af31f5e5SChris Mason /* 1906af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1907af31f5e5SChris Mason * for the next commit 1908af31f5e5SChris Mason */ 1909af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1910af31f5e5SChris Mason &info->super_for_commit->super_roots, 1911af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1912af31f5e5SChris Mason } 1913af31f5e5SChris Mason 1914af31f5e5SChris Mason /* 1915bd2336b2SNikolay Borisov * read_backup_root - Reads a backup root based on the passed priority. Prio 0 1916bd2336b2SNikolay Borisov * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots 1917bd2336b2SNikolay Borisov * 1918bd2336b2SNikolay Borisov * fs_info - filesystem whose backup roots need to be read 1919bd2336b2SNikolay Borisov * priority - priority of backup root required 1920bd2336b2SNikolay Borisov * 1921bd2336b2SNikolay Borisov * Returns backup root index on success and -EINVAL otherwise. 1922bd2336b2SNikolay Borisov */ 1923bd2336b2SNikolay Borisov static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) 1924bd2336b2SNikolay Borisov { 1925bd2336b2SNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 1926bd2336b2SNikolay Borisov struct btrfs_super_block *super = fs_info->super_copy; 1927bd2336b2SNikolay Borisov struct btrfs_root_backup *root_backup; 1928bd2336b2SNikolay Borisov 1929bd2336b2SNikolay Borisov if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { 1930bd2336b2SNikolay Borisov if (priority == 0) 1931bd2336b2SNikolay Borisov return backup_index; 1932bd2336b2SNikolay Borisov 1933bd2336b2SNikolay Borisov backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; 1934bd2336b2SNikolay Borisov backup_index %= BTRFS_NUM_BACKUP_ROOTS; 1935bd2336b2SNikolay Borisov } else { 1936bd2336b2SNikolay Borisov return -EINVAL; 1937bd2336b2SNikolay Borisov } 1938bd2336b2SNikolay Borisov 1939bd2336b2SNikolay Borisov root_backup = super->super_roots + backup_index; 1940bd2336b2SNikolay Borisov 1941bd2336b2SNikolay Borisov btrfs_set_super_generation(super, 1942bd2336b2SNikolay Borisov btrfs_backup_tree_root_gen(root_backup)); 1943bd2336b2SNikolay Borisov btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 1944bd2336b2SNikolay Borisov btrfs_set_super_root_level(super, 1945bd2336b2SNikolay Borisov btrfs_backup_tree_root_level(root_backup)); 1946bd2336b2SNikolay Borisov btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 1947bd2336b2SNikolay Borisov 1948bd2336b2SNikolay Borisov /* 1949bd2336b2SNikolay Borisov * Fixme: the total bytes and num_devices need to match or we should 1950bd2336b2SNikolay Borisov * need a fsck 1951bd2336b2SNikolay Borisov */ 1952bd2336b2SNikolay Borisov btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 1953bd2336b2SNikolay Borisov btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 1954bd2336b2SNikolay Borisov 1955bd2336b2SNikolay Borisov return backup_index; 1956bd2336b2SNikolay Borisov } 1957bd2336b2SNikolay Borisov 19587abadb64SLiu Bo /* helper to cleanup workers */ 19597abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 19607abadb64SLiu Bo { 1961dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 1962afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 19635cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 1964fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 1965fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 1966d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 1967fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 1968fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 19695b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 1970e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 1971736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 1972a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 1973fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 1974b0643e59SDennis Zhou if (fs_info->discard_ctl.discard_workers) 1975b0643e59SDennis Zhou destroy_workqueue(fs_info->discard_ctl.discard_workers); 1976a9b9477dSFilipe Manana /* 1977a9b9477dSFilipe Manana * Now that all other work queues are destroyed, we can safely destroy 1978a9b9477dSFilipe Manana * the queues used for metadata I/O, since tasks from those other work 1979a9b9477dSFilipe Manana * queues can do metadata I/O operations. 1980a9b9477dSFilipe Manana */ 1981a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_workers); 1982a9b9477dSFilipe Manana btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 19837abadb64SLiu Bo } 19847abadb64SLiu Bo 19852e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 19862e9f5954SRashika { 19872e9f5954SRashika if (root) { 19882e9f5954SRashika free_extent_buffer(root->node); 19892e9f5954SRashika free_extent_buffer(root->commit_root); 19902e9f5954SRashika root->node = NULL; 19912e9f5954SRashika root->commit_root = NULL; 19922e9f5954SRashika } 19932e9f5954SRashika } 19942e9f5954SRashika 1995af31f5e5SChris Mason /* helper to cleanup tree roots */ 19964273eaffSAnand Jain static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) 1997af31f5e5SChris Mason { 19982e9f5954SRashika free_root_extent_buffers(info->tree_root); 1999655b09feSJosef Bacik 20002e9f5954SRashika free_root_extent_buffers(info->dev_root); 20012e9f5954SRashika free_root_extent_buffers(info->extent_root); 20022e9f5954SRashika free_root_extent_buffers(info->csum_root); 20032e9f5954SRashika free_root_extent_buffers(info->quota_root); 20042e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20058c38938cSJosef Bacik free_root_extent_buffers(info->fs_root); 2006aeb935a4SQu Wenruo free_root_extent_buffers(info->data_reloc_root); 20074273eaffSAnand Jain if (free_chunk_root) 20082e9f5954SRashika free_root_extent_buffers(info->chunk_root); 200970f6d82eSOmar Sandoval free_root_extent_buffers(info->free_space_root); 2010af31f5e5SChris Mason } 2011af31f5e5SChris Mason 20128c38938cSJosef Bacik void btrfs_put_root(struct btrfs_root *root) 20138c38938cSJosef Bacik { 20148c38938cSJosef Bacik if (!root) 20158c38938cSJosef Bacik return; 20168c38938cSJosef Bacik 20178c38938cSJosef Bacik if (refcount_dec_and_test(&root->refs)) { 20188c38938cSJosef Bacik WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 20191dae7e0eSQu Wenruo WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); 20208c38938cSJosef Bacik if (root->anon_dev) 20218c38938cSJosef Bacik free_anon_bdev(root->anon_dev); 20228c38938cSJosef Bacik btrfs_drew_lock_destroy(&root->snapshot_lock); 2023923eb523SJohannes Thumshirn free_root_extent_buffers(root); 20248c38938cSJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 20258c38938cSJosef Bacik spin_lock(&root->fs_info->fs_roots_radix_lock); 20268c38938cSJosef Bacik list_del_init(&root->leak_list); 20278c38938cSJosef Bacik spin_unlock(&root->fs_info->fs_roots_radix_lock); 20288c38938cSJosef Bacik #endif 20298c38938cSJosef Bacik kfree(root); 20308c38938cSJosef Bacik } 20318c38938cSJosef Bacik } 20328c38938cSJosef Bacik 2033faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2034171f6537SJosef Bacik { 2035171f6537SJosef Bacik int ret; 2036171f6537SJosef Bacik struct btrfs_root *gang[8]; 2037171f6537SJosef Bacik int i; 2038171f6537SJosef Bacik 2039171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2040171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2041171f6537SJosef Bacik struct btrfs_root, root_list); 2042171f6537SJosef Bacik list_del(&gang[0]->root_list); 2043171f6537SJosef Bacik 20448c38938cSJosef Bacik if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) 2045cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 204600246528SJosef Bacik btrfs_put_root(gang[0]); 2047171f6537SJosef Bacik } 2048171f6537SJosef Bacik 2049171f6537SJosef Bacik while (1) { 2050171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2051171f6537SJosef Bacik (void **)gang, 0, 2052171f6537SJosef Bacik ARRAY_SIZE(gang)); 2053171f6537SJosef Bacik if (!ret) 2054171f6537SJosef Bacik break; 2055171f6537SJosef Bacik for (i = 0; i < ret; i++) 2056cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2057171f6537SJosef Bacik } 2058171f6537SJosef Bacik } 2059af31f5e5SChris Mason 2060638aa7edSEric Sandeen static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) 2061638aa7edSEric Sandeen { 2062638aa7edSEric Sandeen mutex_init(&fs_info->scrub_lock); 2063638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_running, 0); 2064638aa7edSEric Sandeen atomic_set(&fs_info->scrub_pause_req, 0); 2065638aa7edSEric Sandeen atomic_set(&fs_info->scrubs_paused, 0); 2066638aa7edSEric Sandeen atomic_set(&fs_info->scrub_cancel_req, 0); 2067638aa7edSEric Sandeen init_waitqueue_head(&fs_info->scrub_pause_wait); 2068ff09c4caSAnand Jain refcount_set(&fs_info->scrub_workers_refcnt, 0); 2069638aa7edSEric Sandeen } 2070638aa7edSEric Sandeen 2071779a65a4SEric Sandeen static void btrfs_init_balance(struct btrfs_fs_info *fs_info) 2072779a65a4SEric Sandeen { 2073779a65a4SEric Sandeen spin_lock_init(&fs_info->balance_lock); 2074779a65a4SEric Sandeen mutex_init(&fs_info->balance_mutex); 2075779a65a4SEric Sandeen atomic_set(&fs_info->balance_pause_req, 0); 2076779a65a4SEric Sandeen atomic_set(&fs_info->balance_cancel_req, 0); 2077779a65a4SEric Sandeen fs_info->balance_ctl = NULL; 2078779a65a4SEric Sandeen init_waitqueue_head(&fs_info->balance_wait_q); 2079779a65a4SEric Sandeen } 2080779a65a4SEric Sandeen 20816bccf3abSJeff Mahoney static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) 2082f37938e0SEric Sandeen { 20832ff7e61eSJeff Mahoney struct inode *inode = fs_info->btree_inode; 20842ff7e61eSJeff Mahoney 20852ff7e61eSJeff Mahoney inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 20862ff7e61eSJeff Mahoney set_nlink(inode, 1); 2087f37938e0SEric Sandeen /* 2088f37938e0SEric Sandeen * we set the i_size on the btree inode to the max possible int. 2089f37938e0SEric Sandeen * the real end of the address space is determined by all of 2090f37938e0SEric Sandeen * the devices in the system 2091f37938e0SEric Sandeen */ 20922ff7e61eSJeff Mahoney inode->i_size = OFFSET_MAX; 20932ff7e61eSJeff Mahoney inode->i_mapping->a_ops = &btree_aops; 2094f37938e0SEric Sandeen 20952ff7e61eSJeff Mahoney RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); 209643eb5f29SQu Wenruo extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, 20972c53a14dSQu Wenruo IO_TREE_BTREE_INODE_IO, inode); 20987b439738SDavid Sterba BTRFS_I(inode)->io_tree.track_uptodate = false; 20992ff7e61eSJeff Mahoney extent_map_tree_init(&BTRFS_I(inode)->extent_tree); 2100f37938e0SEric Sandeen 21015c8fd99fSJosef Bacik BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); 21022ff7e61eSJeff Mahoney memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key)); 21032ff7e61eSJeff Mahoney set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); 21042ff7e61eSJeff Mahoney btrfs_insert_inode_hash(inode); 2105f37938e0SEric Sandeen } 2106f37938e0SEric Sandeen 2107ad618368SEric Sandeen static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2108ad618368SEric Sandeen { 2109ad618368SEric Sandeen mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2110129827e3SDavid Sterba init_rwsem(&fs_info->dev_replace.rwsem); 21117f8d236aSDavid Sterba init_waitqueue_head(&fs_info->dev_replace.replace_wait); 2112ad618368SEric Sandeen } 2113ad618368SEric Sandeen 2114f9e92e40SEric Sandeen static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) 2115f9e92e40SEric Sandeen { 2116f9e92e40SEric Sandeen spin_lock_init(&fs_info->qgroup_lock); 2117f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_ioctl_lock); 2118f9e92e40SEric Sandeen fs_info->qgroup_tree = RB_ROOT; 2119f9e92e40SEric Sandeen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2120f9e92e40SEric Sandeen fs_info->qgroup_seq = 1; 2121f9e92e40SEric Sandeen fs_info->qgroup_ulist = NULL; 2122d2c609b8SJeff Mahoney fs_info->qgroup_rescan_running = false; 2123f9e92e40SEric Sandeen mutex_init(&fs_info->qgroup_rescan_lock); 2124f9e92e40SEric Sandeen } 2125f9e92e40SEric Sandeen 21262a458198SEric Sandeen static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, 21272a458198SEric Sandeen struct btrfs_fs_devices *fs_devices) 21282a458198SEric Sandeen { 2129f7b885beSAnand Jain u32 max_active = fs_info->thread_pool_size; 21306f011058SDavid Sterba unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 21312a458198SEric Sandeen 21322a458198SEric Sandeen fs_info->workers = 2133cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "worker", 2134cb001095SJeff Mahoney flags | WQ_HIGHPRI, max_active, 16); 21352a458198SEric Sandeen 21362a458198SEric Sandeen fs_info->delalloc_workers = 2137cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delalloc", 2138cb001095SJeff Mahoney flags, max_active, 2); 21392a458198SEric Sandeen 21402a458198SEric Sandeen fs_info->flush_workers = 2141cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "flush_delalloc", 2142cb001095SJeff Mahoney flags, max_active, 0); 21432a458198SEric Sandeen 21442a458198SEric Sandeen fs_info->caching_workers = 2145cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); 21462a458198SEric Sandeen 21472a458198SEric Sandeen fs_info->fixup_workers = 2148cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); 21492a458198SEric Sandeen 21502a458198SEric Sandeen /* 21512a458198SEric Sandeen * endios are largely parallel and should have a very 21522a458198SEric Sandeen * low idle thresh 21532a458198SEric Sandeen */ 21542a458198SEric Sandeen fs_info->endio_workers = 2155cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio", flags, max_active, 4); 21562a458198SEric Sandeen fs_info->endio_meta_workers = 2157cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta", flags, 2158cb001095SJeff Mahoney max_active, 4); 21592a458198SEric Sandeen fs_info->endio_meta_write_workers = 2160cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-meta-write", flags, 2161cb001095SJeff Mahoney max_active, 2); 21622a458198SEric Sandeen fs_info->endio_raid56_workers = 2163cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-raid56", flags, 2164cb001095SJeff Mahoney max_active, 4); 21652a458198SEric Sandeen fs_info->rmw_workers = 2166cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "rmw", flags, max_active, 2); 21672a458198SEric Sandeen fs_info->endio_write_workers = 2168cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "endio-write", flags, 2169cb001095SJeff Mahoney max_active, 2); 21702a458198SEric Sandeen fs_info->endio_freespace_worker = 2171cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "freespace-write", flags, 2172cb001095SJeff Mahoney max_active, 0); 21732a458198SEric Sandeen fs_info->delayed_workers = 2174cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, 2175cb001095SJeff Mahoney max_active, 0); 21762a458198SEric Sandeen fs_info->readahead_workers = 2177cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "readahead", flags, 2178cb001095SJeff Mahoney max_active, 2); 21792a458198SEric Sandeen fs_info->qgroup_rescan_workers = 2180cb001095SJeff Mahoney btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); 2181b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers = 2182b0643e59SDennis Zhou alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); 21832a458198SEric Sandeen 21842a458198SEric Sandeen if (!(fs_info->workers && fs_info->delalloc_workers && 2185ba8a9d07SChris Mason fs_info->flush_workers && 21862a458198SEric Sandeen fs_info->endio_workers && fs_info->endio_meta_workers && 21872a458198SEric Sandeen fs_info->endio_meta_write_workers && 21882a458198SEric Sandeen fs_info->endio_write_workers && fs_info->endio_raid56_workers && 21892a458198SEric Sandeen fs_info->endio_freespace_worker && fs_info->rmw_workers && 21902a458198SEric Sandeen fs_info->caching_workers && fs_info->readahead_workers && 21912a458198SEric Sandeen fs_info->fixup_workers && fs_info->delayed_workers && 2192b0643e59SDennis Zhou fs_info->qgroup_rescan_workers && 2193b0643e59SDennis Zhou fs_info->discard_ctl.discard_workers)) { 21942a458198SEric Sandeen return -ENOMEM; 21952a458198SEric Sandeen } 21962a458198SEric Sandeen 21972a458198SEric Sandeen return 0; 21982a458198SEric Sandeen } 21992a458198SEric Sandeen 22006d97c6e3SJohannes Thumshirn static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) 22016d97c6e3SJohannes Thumshirn { 22026d97c6e3SJohannes Thumshirn struct crypto_shash *csum_shash; 2203b4e967beSDavid Sterba const char *csum_driver = btrfs_super_csum_driver(csum_type); 22046d97c6e3SJohannes Thumshirn 2205b4e967beSDavid Sterba csum_shash = crypto_alloc_shash(csum_driver, 0, 0); 22066d97c6e3SJohannes Thumshirn 22076d97c6e3SJohannes Thumshirn if (IS_ERR(csum_shash)) { 22086d97c6e3SJohannes Thumshirn btrfs_err(fs_info, "error allocating %s hash for checksum", 2209b4e967beSDavid Sterba csum_driver); 22106d97c6e3SJohannes Thumshirn return PTR_ERR(csum_shash); 22116d97c6e3SJohannes Thumshirn } 22126d97c6e3SJohannes Thumshirn 22136d97c6e3SJohannes Thumshirn fs_info->csum_shash = csum_shash; 22146d97c6e3SJohannes Thumshirn 22156d97c6e3SJohannes Thumshirn return 0; 22166d97c6e3SJohannes Thumshirn } 22176d97c6e3SJohannes Thumshirn 221863443bf5SEric Sandeen static int btrfs_replay_log(struct btrfs_fs_info *fs_info, 221963443bf5SEric Sandeen struct btrfs_fs_devices *fs_devices) 222063443bf5SEric Sandeen { 222163443bf5SEric Sandeen int ret; 222263443bf5SEric Sandeen struct btrfs_root *log_tree_root; 222363443bf5SEric Sandeen struct btrfs_super_block *disk_super = fs_info->super_copy; 222463443bf5SEric Sandeen u64 bytenr = btrfs_super_log_root(disk_super); 2225581c1760SQu Wenruo int level = btrfs_super_log_root_level(disk_super); 222663443bf5SEric Sandeen 222763443bf5SEric Sandeen if (fs_devices->rw_devices == 0) { 2228f14d104dSDavid Sterba btrfs_warn(fs_info, "log replay required on RO media"); 222963443bf5SEric Sandeen return -EIO; 223063443bf5SEric Sandeen } 223163443bf5SEric Sandeen 223296dfcb46SJosef Bacik log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, 223396dfcb46SJosef Bacik GFP_KERNEL); 223463443bf5SEric Sandeen if (!log_tree_root) 223563443bf5SEric Sandeen return -ENOMEM; 223663443bf5SEric Sandeen 22372ff7e61eSJeff Mahoney log_tree_root->node = read_tree_block(fs_info, bytenr, 22381b7ec85eSJosef Bacik BTRFS_TREE_LOG_OBJECTID, 22391b7ec85eSJosef Bacik fs_info->generation + 1, level, 22401b7ec85eSJosef Bacik NULL); 224164c043deSLiu Bo if (IS_ERR(log_tree_root->node)) { 2242f14d104dSDavid Sterba btrfs_warn(fs_info, "failed to read log tree"); 22430eeff236SLiu Bo ret = PTR_ERR(log_tree_root->node); 22448c38938cSJosef Bacik log_tree_root->node = NULL; 224500246528SJosef Bacik btrfs_put_root(log_tree_root); 22460eeff236SLiu Bo return ret; 224764c043deSLiu Bo } else if (!extent_buffer_uptodate(log_tree_root->node)) { 2248f14d104dSDavid Sterba btrfs_err(fs_info, "failed to read log tree"); 224900246528SJosef Bacik btrfs_put_root(log_tree_root); 225063443bf5SEric Sandeen return -EIO; 225163443bf5SEric Sandeen } 225263443bf5SEric Sandeen /* returns with log_tree_root freed on success */ 225363443bf5SEric Sandeen ret = btrfs_recover_log_trees(log_tree_root); 225463443bf5SEric Sandeen if (ret) { 22550b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 225663443bf5SEric Sandeen "Failed to recover log tree"); 225700246528SJosef Bacik btrfs_put_root(log_tree_root); 225863443bf5SEric Sandeen return ret; 225963443bf5SEric Sandeen } 226063443bf5SEric Sandeen 2261bc98a42cSDavid Howells if (sb_rdonly(fs_info->sb)) { 22626bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 226363443bf5SEric Sandeen if (ret) 226463443bf5SEric Sandeen return ret; 226563443bf5SEric Sandeen } 226663443bf5SEric Sandeen 226763443bf5SEric Sandeen return 0; 226863443bf5SEric Sandeen } 226963443bf5SEric Sandeen 22706bccf3abSJeff Mahoney static int btrfs_read_roots(struct btrfs_fs_info *fs_info) 22714bbcaa64SEric Sandeen { 22726bccf3abSJeff Mahoney struct btrfs_root *tree_root = fs_info->tree_root; 2273a4f3d2c4SDavid Sterba struct btrfs_root *root; 22744bbcaa64SEric Sandeen struct btrfs_key location; 22754bbcaa64SEric Sandeen int ret; 22764bbcaa64SEric Sandeen 22776bccf3abSJeff Mahoney BUG_ON(!fs_info->tree_root); 22786bccf3abSJeff Mahoney 22794bbcaa64SEric Sandeen location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 22804bbcaa64SEric Sandeen location.type = BTRFS_ROOT_ITEM_KEY; 22814bbcaa64SEric Sandeen location.offset = 0; 22824bbcaa64SEric Sandeen 2283a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2284f50f4353SLiu Bo if (IS_ERR(root)) { 228542437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2286f50f4353SLiu Bo ret = PTR_ERR(root); 2287f50f4353SLiu Bo goto out; 2288f50f4353SLiu Bo } 228942437a63SJosef Bacik } else { 2290a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2291a4f3d2c4SDavid Sterba fs_info->extent_root = root; 229242437a63SJosef Bacik } 22934bbcaa64SEric Sandeen 22944bbcaa64SEric Sandeen location.objectid = BTRFS_DEV_TREE_OBJECTID; 2295a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2296f50f4353SLiu Bo if (IS_ERR(root)) { 229742437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2298f50f4353SLiu Bo ret = PTR_ERR(root); 2299f50f4353SLiu Bo goto out; 2300f50f4353SLiu Bo } 230142437a63SJosef Bacik } else { 2302a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2303a4f3d2c4SDavid Sterba fs_info->dev_root = root; 23044bbcaa64SEric Sandeen btrfs_init_devices_late(fs_info); 230542437a63SJosef Bacik } 23064bbcaa64SEric Sandeen 2307882dbe0cSJosef Bacik /* If IGNOREDATACSUMS is set don't bother reading the csum root. */ 2308882dbe0cSJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { 23094bbcaa64SEric Sandeen location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2310a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2311f50f4353SLiu Bo if (IS_ERR(root)) { 231242437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2313f50f4353SLiu Bo ret = PTR_ERR(root); 2314f50f4353SLiu Bo goto out; 2315f50f4353SLiu Bo } 231642437a63SJosef Bacik } else { 2317a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2318a4f3d2c4SDavid Sterba fs_info->csum_root = root; 231942437a63SJosef Bacik } 2320882dbe0cSJosef Bacik } 23214bbcaa64SEric Sandeen 2322aeb935a4SQu Wenruo /* 2323aeb935a4SQu Wenruo * This tree can share blocks with some other fs tree during relocation 2324aeb935a4SQu Wenruo * and we need a proper setup by btrfs_get_fs_root 2325aeb935a4SQu Wenruo */ 232656e9357aSDavid Sterba root = btrfs_get_fs_root(tree_root->fs_info, 232756e9357aSDavid Sterba BTRFS_DATA_RELOC_TREE_OBJECTID, true); 2328aeb935a4SQu Wenruo if (IS_ERR(root)) { 232942437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2330aeb935a4SQu Wenruo ret = PTR_ERR(root); 2331aeb935a4SQu Wenruo goto out; 2332aeb935a4SQu Wenruo } 233342437a63SJosef Bacik } else { 2334aeb935a4SQu Wenruo set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2335aeb935a4SQu Wenruo fs_info->data_reloc_root = root; 233642437a63SJosef Bacik } 2337aeb935a4SQu Wenruo 23384bbcaa64SEric Sandeen location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2339a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2340a4f3d2c4SDavid Sterba if (!IS_ERR(root)) { 2341a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2342afcdd129SJosef Bacik set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); 2343a4f3d2c4SDavid Sterba fs_info->quota_root = root; 23444bbcaa64SEric Sandeen } 23454bbcaa64SEric Sandeen 23464bbcaa64SEric Sandeen location.objectid = BTRFS_UUID_TREE_OBJECTID; 2347a4f3d2c4SDavid Sterba root = btrfs_read_tree_root(tree_root, &location); 2348a4f3d2c4SDavid Sterba if (IS_ERR(root)) { 234942437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2350a4f3d2c4SDavid Sterba ret = PTR_ERR(root); 23514bbcaa64SEric Sandeen if (ret != -ENOENT) 2352f50f4353SLiu Bo goto out; 235342437a63SJosef Bacik } 23544bbcaa64SEric Sandeen } else { 2355a4f3d2c4SDavid Sterba set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 2356a4f3d2c4SDavid Sterba fs_info->uuid_root = root; 23574bbcaa64SEric Sandeen } 23584bbcaa64SEric Sandeen 235970f6d82eSOmar Sandoval if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 236070f6d82eSOmar Sandoval location.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID; 236170f6d82eSOmar Sandoval root = btrfs_read_tree_root(tree_root, &location); 2362f50f4353SLiu Bo if (IS_ERR(root)) { 236342437a63SJosef Bacik if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { 2364f50f4353SLiu Bo ret = PTR_ERR(root); 2365f50f4353SLiu Bo goto out; 2366f50f4353SLiu Bo } 236742437a63SJosef Bacik } else { 236870f6d82eSOmar Sandoval set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 236970f6d82eSOmar Sandoval fs_info->free_space_root = root; 237070f6d82eSOmar Sandoval } 237142437a63SJosef Bacik } 237270f6d82eSOmar Sandoval 23734bbcaa64SEric Sandeen return 0; 2374f50f4353SLiu Bo out: 2375f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", 2376f50f4353SLiu Bo location.objectid, ret); 2377f50f4353SLiu Bo return ret; 23784bbcaa64SEric Sandeen } 23794bbcaa64SEric Sandeen 2380069ec957SQu Wenruo /* 2381069ec957SQu Wenruo * Real super block validation 2382069ec957SQu Wenruo * NOTE: super csum type and incompat features will not be checked here. 2383069ec957SQu Wenruo * 2384069ec957SQu Wenruo * @sb: super block to check 2385069ec957SQu Wenruo * @mirror_num: the super block number to check its bytenr: 2386069ec957SQu Wenruo * 0 the primary (1st) sb 2387069ec957SQu Wenruo * 1, 2 2nd and 3rd backup copy 2388069ec957SQu Wenruo * -1 skip bytenr check 2389069ec957SQu Wenruo */ 2390069ec957SQu Wenruo static int validate_super(struct btrfs_fs_info *fs_info, 2391069ec957SQu Wenruo struct btrfs_super_block *sb, int mirror_num) 239221a852b0SQu Wenruo { 239321a852b0SQu Wenruo u64 nodesize = btrfs_super_nodesize(sb); 239421a852b0SQu Wenruo u64 sectorsize = btrfs_super_sectorsize(sb); 239521a852b0SQu Wenruo int ret = 0; 239621a852b0SQu Wenruo 239721a852b0SQu Wenruo if (btrfs_super_magic(sb) != BTRFS_MAGIC) { 239821a852b0SQu Wenruo btrfs_err(fs_info, "no valid FS found"); 239921a852b0SQu Wenruo ret = -EINVAL; 240021a852b0SQu Wenruo } 240121a852b0SQu Wenruo if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { 240221a852b0SQu Wenruo btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", 240321a852b0SQu Wenruo btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); 240421a852b0SQu Wenruo ret = -EINVAL; 240521a852b0SQu Wenruo } 240621a852b0SQu Wenruo if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { 240721a852b0SQu Wenruo btrfs_err(fs_info, "tree_root level too big: %d >= %d", 240821a852b0SQu Wenruo btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); 240921a852b0SQu Wenruo ret = -EINVAL; 241021a852b0SQu Wenruo } 241121a852b0SQu Wenruo if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { 241221a852b0SQu Wenruo btrfs_err(fs_info, "chunk_root level too big: %d >= %d", 241321a852b0SQu Wenruo btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); 241421a852b0SQu Wenruo ret = -EINVAL; 241521a852b0SQu Wenruo } 241621a852b0SQu Wenruo if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { 241721a852b0SQu Wenruo btrfs_err(fs_info, "log_root level too big: %d >= %d", 241821a852b0SQu Wenruo btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); 241921a852b0SQu Wenruo ret = -EINVAL; 242021a852b0SQu Wenruo } 242121a852b0SQu Wenruo 242221a852b0SQu Wenruo /* 242321a852b0SQu Wenruo * Check sectorsize and nodesize first, other check will need it. 242421a852b0SQu Wenruo * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. 242521a852b0SQu Wenruo */ 242621a852b0SQu Wenruo if (!is_power_of_2(sectorsize) || sectorsize < 4096 || 242721a852b0SQu Wenruo sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { 242821a852b0SQu Wenruo btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); 242921a852b0SQu Wenruo ret = -EINVAL; 243021a852b0SQu Wenruo } 243121a852b0SQu Wenruo /* Only PAGE SIZE is supported yet */ 243221a852b0SQu Wenruo if (sectorsize != PAGE_SIZE) { 243321a852b0SQu Wenruo btrfs_err(fs_info, 243421a852b0SQu Wenruo "sectorsize %llu not supported yet, only support %lu", 243521a852b0SQu Wenruo sectorsize, PAGE_SIZE); 243621a852b0SQu Wenruo ret = -EINVAL; 243721a852b0SQu Wenruo } 243821a852b0SQu Wenruo if (!is_power_of_2(nodesize) || nodesize < sectorsize || 243921a852b0SQu Wenruo nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { 244021a852b0SQu Wenruo btrfs_err(fs_info, "invalid nodesize %llu", nodesize); 244121a852b0SQu Wenruo ret = -EINVAL; 244221a852b0SQu Wenruo } 244321a852b0SQu Wenruo if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { 244421a852b0SQu Wenruo btrfs_err(fs_info, "invalid leafsize %u, should be %llu", 244521a852b0SQu Wenruo le32_to_cpu(sb->__unused_leafsize), nodesize); 244621a852b0SQu Wenruo ret = -EINVAL; 244721a852b0SQu Wenruo } 244821a852b0SQu Wenruo 244921a852b0SQu Wenruo /* Root alignment check */ 245021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { 245121a852b0SQu Wenruo btrfs_warn(fs_info, "tree_root block unaligned: %llu", 245221a852b0SQu Wenruo btrfs_super_root(sb)); 245321a852b0SQu Wenruo ret = -EINVAL; 245421a852b0SQu Wenruo } 245521a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { 245621a852b0SQu Wenruo btrfs_warn(fs_info, "chunk_root block unaligned: %llu", 245721a852b0SQu Wenruo btrfs_super_chunk_root(sb)); 245821a852b0SQu Wenruo ret = -EINVAL; 245921a852b0SQu Wenruo } 246021a852b0SQu Wenruo if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { 246121a852b0SQu Wenruo btrfs_warn(fs_info, "log_root block unaligned: %llu", 246221a852b0SQu Wenruo btrfs_super_log_root(sb)); 246321a852b0SQu Wenruo ret = -EINVAL; 246421a852b0SQu Wenruo } 246521a852b0SQu Wenruo 2466de37aa51SNikolay Borisov if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, 24677239ff4bSNikolay Borisov BTRFS_FSID_SIZE) != 0) { 246821a852b0SQu Wenruo btrfs_err(fs_info, 24697239ff4bSNikolay Borisov "dev_item UUID does not match metadata fsid: %pU != %pU", 2470de37aa51SNikolay Borisov fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); 247121a852b0SQu Wenruo ret = -EINVAL; 247221a852b0SQu Wenruo } 247321a852b0SQu Wenruo 247421a852b0SQu Wenruo /* 247521a852b0SQu Wenruo * Hint to catch really bogus numbers, bitflips or so, more exact checks are 247621a852b0SQu Wenruo * done later 247721a852b0SQu Wenruo */ 247821a852b0SQu Wenruo if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { 247921a852b0SQu Wenruo btrfs_err(fs_info, "bytes_used is too small %llu", 248021a852b0SQu Wenruo btrfs_super_bytes_used(sb)); 248121a852b0SQu Wenruo ret = -EINVAL; 248221a852b0SQu Wenruo } 248321a852b0SQu Wenruo if (!is_power_of_2(btrfs_super_stripesize(sb))) { 248421a852b0SQu Wenruo btrfs_err(fs_info, "invalid stripesize %u", 248521a852b0SQu Wenruo btrfs_super_stripesize(sb)); 248621a852b0SQu Wenruo ret = -EINVAL; 248721a852b0SQu Wenruo } 248821a852b0SQu Wenruo if (btrfs_super_num_devices(sb) > (1UL << 31)) 248921a852b0SQu Wenruo btrfs_warn(fs_info, "suspicious number of devices: %llu", 249021a852b0SQu Wenruo btrfs_super_num_devices(sb)); 249121a852b0SQu Wenruo if (btrfs_super_num_devices(sb) == 0) { 249221a852b0SQu Wenruo btrfs_err(fs_info, "number of devices is 0"); 249321a852b0SQu Wenruo ret = -EINVAL; 249421a852b0SQu Wenruo } 249521a852b0SQu Wenruo 2496069ec957SQu Wenruo if (mirror_num >= 0 && 2497069ec957SQu Wenruo btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { 249821a852b0SQu Wenruo btrfs_err(fs_info, "super offset mismatch %llu != %u", 249921a852b0SQu Wenruo btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); 250021a852b0SQu Wenruo ret = -EINVAL; 250121a852b0SQu Wenruo } 250221a852b0SQu Wenruo 250321a852b0SQu Wenruo /* 250421a852b0SQu Wenruo * Obvious sys_chunk_array corruptions, it must hold at least one key 250521a852b0SQu Wenruo * and one chunk 250621a852b0SQu Wenruo */ 250721a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { 250821a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too big %u > %u", 250921a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 251021a852b0SQu Wenruo BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); 251121a852b0SQu Wenruo ret = -EINVAL; 251221a852b0SQu Wenruo } 251321a852b0SQu Wenruo if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) 251421a852b0SQu Wenruo + sizeof(struct btrfs_chunk)) { 251521a852b0SQu Wenruo btrfs_err(fs_info, "system chunk array too small %u < %zu", 251621a852b0SQu Wenruo btrfs_super_sys_array_size(sb), 251721a852b0SQu Wenruo sizeof(struct btrfs_disk_key) 251821a852b0SQu Wenruo + sizeof(struct btrfs_chunk)); 251921a852b0SQu Wenruo ret = -EINVAL; 252021a852b0SQu Wenruo } 252121a852b0SQu Wenruo 252221a852b0SQu Wenruo /* 252321a852b0SQu Wenruo * The generation is a global counter, we'll trust it more than the others 252421a852b0SQu Wenruo * but it's still possible that it's the one that's wrong. 252521a852b0SQu Wenruo */ 252621a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) 252721a852b0SQu Wenruo btrfs_warn(fs_info, 252821a852b0SQu Wenruo "suspicious: generation < chunk_root_generation: %llu < %llu", 252921a852b0SQu Wenruo btrfs_super_generation(sb), 253021a852b0SQu Wenruo btrfs_super_chunk_root_generation(sb)); 253121a852b0SQu Wenruo if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) 253221a852b0SQu Wenruo && btrfs_super_cache_generation(sb) != (u64)-1) 253321a852b0SQu Wenruo btrfs_warn(fs_info, 253421a852b0SQu Wenruo "suspicious: generation < cache_generation: %llu < %llu", 253521a852b0SQu Wenruo btrfs_super_generation(sb), 253621a852b0SQu Wenruo btrfs_super_cache_generation(sb)); 253721a852b0SQu Wenruo 253821a852b0SQu Wenruo return ret; 253921a852b0SQu Wenruo } 254021a852b0SQu Wenruo 2541069ec957SQu Wenruo /* 2542069ec957SQu Wenruo * Validation of super block at mount time. 2543069ec957SQu Wenruo * Some checks already done early at mount time, like csum type and incompat 2544069ec957SQu Wenruo * flags will be skipped. 2545069ec957SQu Wenruo */ 2546069ec957SQu Wenruo static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) 2547069ec957SQu Wenruo { 2548069ec957SQu Wenruo return validate_super(fs_info, fs_info->super_copy, 0); 2549069ec957SQu Wenruo } 2550069ec957SQu Wenruo 255175cb857dSQu Wenruo /* 255275cb857dSQu Wenruo * Validation of super block at write time. 255375cb857dSQu Wenruo * Some checks like bytenr check will be skipped as their values will be 255475cb857dSQu Wenruo * overwritten soon. 255575cb857dSQu Wenruo * Extra checks like csum type and incompat flags will be done here. 255675cb857dSQu Wenruo */ 255775cb857dSQu Wenruo static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, 255875cb857dSQu Wenruo struct btrfs_super_block *sb) 255975cb857dSQu Wenruo { 256075cb857dSQu Wenruo int ret; 256175cb857dSQu Wenruo 256275cb857dSQu Wenruo ret = validate_super(fs_info, sb, -1); 256375cb857dSQu Wenruo if (ret < 0) 256475cb857dSQu Wenruo goto out; 2565e7e16f48SJohannes Thumshirn if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { 256675cb857dSQu Wenruo ret = -EUCLEAN; 256775cb857dSQu Wenruo btrfs_err(fs_info, "invalid csum type, has %u want %u", 256875cb857dSQu Wenruo btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); 256975cb857dSQu Wenruo goto out; 257075cb857dSQu Wenruo } 257175cb857dSQu Wenruo if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { 257275cb857dSQu Wenruo ret = -EUCLEAN; 257375cb857dSQu Wenruo btrfs_err(fs_info, 257475cb857dSQu Wenruo "invalid incompat flags, has 0x%llx valid mask 0x%llx", 257575cb857dSQu Wenruo btrfs_super_incompat_flags(sb), 257675cb857dSQu Wenruo (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); 257775cb857dSQu Wenruo goto out; 257875cb857dSQu Wenruo } 257975cb857dSQu Wenruo out: 258075cb857dSQu Wenruo if (ret < 0) 258175cb857dSQu Wenruo btrfs_err(fs_info, 258275cb857dSQu Wenruo "super block corruption detected before writing it to disk"); 258375cb857dSQu Wenruo return ret; 258475cb857dSQu Wenruo } 258575cb857dSQu Wenruo 25866ef108ddSNikolay Borisov static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) 2587b8522a1eSNikolay Borisov { 25886ef108ddSNikolay Borisov int backup_index = find_newest_super_backup(fs_info); 2589b8522a1eSNikolay Borisov struct btrfs_super_block *sb = fs_info->super_copy; 2590b8522a1eSNikolay Borisov struct btrfs_root *tree_root = fs_info->tree_root; 2591b8522a1eSNikolay Borisov bool handle_error = false; 2592b8522a1eSNikolay Borisov int ret = 0; 2593b8522a1eSNikolay Borisov int i; 2594b8522a1eSNikolay Borisov 2595b8522a1eSNikolay Borisov for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 2596b8522a1eSNikolay Borisov u64 generation; 2597b8522a1eSNikolay Borisov int level; 2598b8522a1eSNikolay Borisov 2599b8522a1eSNikolay Borisov if (handle_error) { 2600b8522a1eSNikolay Borisov if (!IS_ERR(tree_root->node)) 2601b8522a1eSNikolay Borisov free_extent_buffer(tree_root->node); 2602b8522a1eSNikolay Borisov tree_root->node = NULL; 2603b8522a1eSNikolay Borisov 2604b8522a1eSNikolay Borisov if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) 2605b8522a1eSNikolay Borisov break; 2606b8522a1eSNikolay Borisov 2607b8522a1eSNikolay Borisov free_root_pointers(fs_info, 0); 2608b8522a1eSNikolay Borisov 2609b8522a1eSNikolay Borisov /* 2610b8522a1eSNikolay Borisov * Don't use the log in recovery mode, it won't be 2611b8522a1eSNikolay Borisov * valid 2612b8522a1eSNikolay Borisov */ 2613b8522a1eSNikolay Borisov btrfs_set_super_log_root(sb, 0); 2614b8522a1eSNikolay Borisov 2615b8522a1eSNikolay Borisov /* We can't trust the free space cache either */ 2616b8522a1eSNikolay Borisov btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 2617b8522a1eSNikolay Borisov 2618b8522a1eSNikolay Borisov ret = read_backup_root(fs_info, i); 26196ef108ddSNikolay Borisov backup_index = ret; 2620b8522a1eSNikolay Borisov if (ret < 0) 2621b8522a1eSNikolay Borisov return ret; 2622b8522a1eSNikolay Borisov } 2623b8522a1eSNikolay Borisov generation = btrfs_super_generation(sb); 2624b8522a1eSNikolay Borisov level = btrfs_super_root_level(sb); 2625b8522a1eSNikolay Borisov tree_root->node = read_tree_block(fs_info, btrfs_super_root(sb), 26261b7ec85eSJosef Bacik BTRFS_ROOT_TREE_OBJECTID, 2627b8522a1eSNikolay Borisov generation, level, NULL); 26280465337cSJosef Bacik if (IS_ERR(tree_root->node)) { 2629217f5004SNikolay Borisov handle_error = true; 2630b8522a1eSNikolay Borisov ret = PTR_ERR(tree_root->node); 26310465337cSJosef Bacik tree_root->node = NULL; 2632217f5004SNikolay Borisov btrfs_warn(fs_info, "couldn't read tree root"); 2633217f5004SNikolay Borisov continue; 2634b8522a1eSNikolay Borisov 2635217f5004SNikolay Borisov } else if (!extent_buffer_uptodate(tree_root->node)) { 2636217f5004SNikolay Borisov handle_error = true; 2637217f5004SNikolay Borisov ret = -EIO; 2638217f5004SNikolay Borisov btrfs_warn(fs_info, "error while reading tree root"); 2639b8522a1eSNikolay Borisov continue; 2640b8522a1eSNikolay Borisov } 2641b8522a1eSNikolay Borisov 2642b8522a1eSNikolay Borisov btrfs_set_root_node(&tree_root->root_item, tree_root->node); 2643b8522a1eSNikolay Borisov tree_root->commit_root = btrfs_root_node(tree_root); 2644b8522a1eSNikolay Borisov btrfs_set_root_refs(&tree_root->root_item, 1); 2645b8522a1eSNikolay Borisov 2646336a0d8dSNikolay Borisov /* 2647336a0d8dSNikolay Borisov * No need to hold btrfs_root::objectid_mutex since the fs 2648336a0d8dSNikolay Borisov * hasn't been fully initialised and we are the only user 2649336a0d8dSNikolay Borisov */ 2650b8522a1eSNikolay Borisov ret = btrfs_find_highest_objectid(tree_root, 2651b8522a1eSNikolay Borisov &tree_root->highest_objectid); 2652b8522a1eSNikolay Borisov if (ret < 0) { 2653b8522a1eSNikolay Borisov handle_error = true; 2654b8522a1eSNikolay Borisov continue; 2655b8522a1eSNikolay Borisov } 2656b8522a1eSNikolay Borisov 2657b8522a1eSNikolay Borisov ASSERT(tree_root->highest_objectid <= BTRFS_LAST_FREE_OBJECTID); 2658b8522a1eSNikolay Borisov 2659b8522a1eSNikolay Borisov ret = btrfs_read_roots(fs_info); 2660b8522a1eSNikolay Borisov if (ret < 0) { 2661b8522a1eSNikolay Borisov handle_error = true; 2662b8522a1eSNikolay Borisov continue; 2663b8522a1eSNikolay Borisov } 2664b8522a1eSNikolay Borisov 2665b8522a1eSNikolay Borisov /* All successful */ 2666b8522a1eSNikolay Borisov fs_info->generation = generation; 2667b8522a1eSNikolay Borisov fs_info->last_trans_committed = generation; 26686ef108ddSNikolay Borisov 26696ef108ddSNikolay Borisov /* Always begin writing backup roots after the one being used */ 26706ef108ddSNikolay Borisov if (backup_index < 0) { 26716ef108ddSNikolay Borisov fs_info->backup_root_index = 0; 26726ef108ddSNikolay Borisov } else { 26736ef108ddSNikolay Borisov fs_info->backup_root_index = backup_index + 1; 26746ef108ddSNikolay Borisov fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; 26756ef108ddSNikolay Borisov } 2676b8522a1eSNikolay Borisov break; 2677b8522a1eSNikolay Borisov } 2678b8522a1eSNikolay Borisov 2679b8522a1eSNikolay Borisov return ret; 2680b8522a1eSNikolay Borisov } 2681b8522a1eSNikolay Borisov 26828260edbaSJosef Bacik void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) 2683eb60ceacSChris Mason { 268476dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2685f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 26868fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2687facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 268824bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2689eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 269011833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2691eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2692a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 269376dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 269424bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 26954cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 2696ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2697f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 269847ab2a6cSJosef Bacik spin_lock_init(&fs_info->unused_bgs_lock); 2699f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 2700d7c15171SZhao Lei mutex_init(&fs_info->unused_bg_unpin_mutex); 270167c5e7d4SFilipe Manana mutex_init(&fs_info->delete_unused_bgs_mutex); 27027585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2703573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2704de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 270519c00ddcSChris Mason 27060b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 27076324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2708f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 270947ab2a6cSJosef Bacik INIT_LIST_HEAD(&fs_info->unused_bgs); 2710bd647ce3SJosef Bacik #ifdef CONFIG_BTRFS_DEBUG 2711bd647ce3SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_roots); 27123fd63727SJosef Bacik INIT_LIST_HEAD(&fs_info->allocated_ebs); 27133fd63727SJosef Bacik spin_lock_init(&fs_info->eb_leak_lock); 2714bd647ce3SJosef Bacik #endif 2715c8bf1b67SDavid Sterba extent_map_tree_init(&fs_info->mapping_tree); 271666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 271766d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 271866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 271966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 272066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 272166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 272266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2723ba2c4d4eSJosef Bacik btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, 2724ba2c4d4eSJosef Bacik BTRFS_BLOCK_RSV_DELREFS); 2725ba2c4d4eSJosef Bacik 2726771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 27274cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 27282fefd558SZhao Lei atomic_set(&fs_info->reada_works_cnt, 0); 2729034f784dSJosef Bacik atomic_set(&fs_info->nr_delayed_iputs, 0); 2730fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 273195ac567aSFilipe David Borba Manana fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; 27329ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 27334cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 2734a5ed45f8SNikolay Borisov atomic64_set(&fs_info->free_chunk_space, 0); 2735f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 27368b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2737f8c269d7SDavid Sterba fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ 273890519d66SArne Jansen /* readahead state */ 2739d0164adcSMel Gorman INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_DIRECT_RECLAIM); 274090519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2741fd708b81SJosef Bacik btrfs_init_ref_verify(fs_info); 2742c8b97818SChris Mason 2743b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2744b34b086cSChris Mason num_online_cpus() + 2, 8); 27450afbaf8cSChris Mason 2746199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2747199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 274869fe2d75SJosef Bacik 2749638aa7edSEric Sandeen btrfs_init_scrub(fs_info); 275021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 275121adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 275221adbd5cSStefan Behrens #endif 2753779a65a4SEric Sandeen btrfs_init_balance(fs_info); 275457056740SJosef Bacik btrfs_init_async_reclaim_work(fs_info); 2755a2de733cSArne Jansen 27560f9dd46cSJosef Bacik spin_lock_init(&fs_info->block_group_cache_lock); 27576bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2758a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 27590f9dd46cSJosef Bacik 2760fe119a6eSNikolay Borisov extent_io_tree_init(fs_info, &fs_info->excluded_extents, 2761fe119a6eSNikolay Borisov IO_TREE_FS_EXCLUDED_EXTENTS, NULL); 2762afcdd129SJosef Bacik set_bit(BTRFS_FS_BARRIER, &fs_info->flags); 276339279cc3SChris Mason 27645a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 2765e02119d5SChris Mason mutex_init(&fs_info->tree_log_mutex); 2766925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2767a74a4b97SChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2768a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 27691bbc621eSChris Mason mutex_init(&fs_info->ro_block_group_mutex); 27709e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2771c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 277276dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2773803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2774fa9c0d79SChris Mason 2775ad618368SEric Sandeen btrfs_init_dev_replace_locks(fs_info); 2776f9e92e40SEric Sandeen btrfs_init_qgroup(fs_info); 2777b0643e59SDennis Zhou btrfs_discard_init(fs_info); 2778416ac51dSArne Jansen 2779fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2780fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2781fa9c0d79SChris Mason 2782e6dcd2dcSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 2783f9295749SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2784bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 27854854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 2786034f784dSJosef Bacik init_waitqueue_head(&fs_info->delayed_iputs_wait); 27873768f368SChris Mason 2788da17066cSJeff Mahoney /* Usable values until the real ones are cached from the superblock */ 2789da17066cSJeff Mahoney fs_info->nodesize = 4096; 2790da17066cSJeff Mahoney fs_info->sectorsize = 4096; 2791ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(4096); 2792da17066cSJeff Mahoney fs_info->stripesize = 4096; 2793da17066cSJeff Mahoney 2794eede2bf3SOmar Sandoval spin_lock_init(&fs_info->swapfile_pins_lock); 2795eede2bf3SOmar Sandoval fs_info->swapfile_pins = RB_ROOT; 2796eede2bf3SOmar Sandoval 27979e967495SFilipe Manana fs_info->send_in_progress = 0; 27988260edbaSJosef Bacik } 27998260edbaSJosef Bacik 28008260edbaSJosef Bacik static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) 28018260edbaSJosef Bacik { 28028260edbaSJosef Bacik int ret; 28038260edbaSJosef Bacik 28048260edbaSJosef Bacik fs_info->sb = sb; 28058260edbaSJosef Bacik sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; 28068260edbaSJosef Bacik sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); 28079e967495SFilipe Manana 2808c75e8394SJosef Bacik ret = percpu_counter_init(&fs_info->dio_bytes, 0, GFP_KERNEL); 2809ae18c37aSJosef Bacik if (ret) 2810ae18c37aSJosef Bacik return ret; 2811ae18c37aSJosef Bacik 2812ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); 2813ae18c37aSJosef Bacik if (ret) 2814c75e8394SJosef Bacik return ret; 2815ae18c37aSJosef Bacik 2816ae18c37aSJosef Bacik fs_info->dirty_metadata_batch = PAGE_SIZE * 2817ae18c37aSJosef Bacik (1 + ilog2(nr_cpu_ids)); 2818ae18c37aSJosef Bacik 2819ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); 2820ae18c37aSJosef Bacik if (ret) 2821c75e8394SJosef Bacik return ret; 2822ae18c37aSJosef Bacik 2823ae18c37aSJosef Bacik ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, 2824ae18c37aSJosef Bacik GFP_KERNEL); 2825ae18c37aSJosef Bacik if (ret) 2826c75e8394SJosef Bacik return ret; 2827ae18c37aSJosef Bacik 2828ae18c37aSJosef Bacik fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 2829ae18c37aSJosef Bacik GFP_KERNEL); 2830c75e8394SJosef Bacik if (!fs_info->delayed_root) 2831c75e8394SJosef Bacik return -ENOMEM; 2832ae18c37aSJosef Bacik btrfs_init_delayed_root(fs_info->delayed_root); 2833ae18c37aSJosef Bacik 2834c75e8394SJosef Bacik return btrfs_alloc_stripe_hash_table(fs_info); 2835ae18c37aSJosef Bacik } 2836ae18c37aSJosef Bacik 283797f4dd09SNikolay Borisov static int btrfs_uuid_rescan_kthread(void *data) 283897f4dd09SNikolay Borisov { 283997f4dd09SNikolay Borisov struct btrfs_fs_info *fs_info = (struct btrfs_fs_info *)data; 284097f4dd09SNikolay Borisov int ret; 284197f4dd09SNikolay Borisov 284297f4dd09SNikolay Borisov /* 284397f4dd09SNikolay Borisov * 1st step is to iterate through the existing UUID tree and 284497f4dd09SNikolay Borisov * to delete all entries that contain outdated data. 284597f4dd09SNikolay Borisov * 2nd step is to add all missing entries to the UUID tree. 284697f4dd09SNikolay Borisov */ 284797f4dd09SNikolay Borisov ret = btrfs_uuid_tree_iterate(fs_info); 284897f4dd09SNikolay Borisov if (ret < 0) { 2849c94bec2cSJosef Bacik if (ret != -EINTR) 2850c94bec2cSJosef Bacik btrfs_warn(fs_info, "iterating uuid_tree failed %d", 2851c94bec2cSJosef Bacik ret); 285297f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 285397f4dd09SNikolay Borisov return ret; 285497f4dd09SNikolay Borisov } 285597f4dd09SNikolay Borisov return btrfs_uuid_scan_kthread(data); 285697f4dd09SNikolay Borisov } 285797f4dd09SNikolay Borisov 285897f4dd09SNikolay Borisov static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) 285997f4dd09SNikolay Borisov { 286097f4dd09SNikolay Borisov struct task_struct *task; 286197f4dd09SNikolay Borisov 286297f4dd09SNikolay Borisov down(&fs_info->uuid_tree_rescan_sem); 286397f4dd09SNikolay Borisov task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); 286497f4dd09SNikolay Borisov if (IS_ERR(task)) { 286597f4dd09SNikolay Borisov /* fs_info->update_uuid_tree_gen remains 0 in all error case */ 286697f4dd09SNikolay Borisov btrfs_warn(fs_info, "failed to start uuid_rescan task"); 286797f4dd09SNikolay Borisov up(&fs_info->uuid_tree_rescan_sem); 286897f4dd09SNikolay Borisov return PTR_ERR(task); 286997f4dd09SNikolay Borisov } 287097f4dd09SNikolay Borisov 287197f4dd09SNikolay Borisov return 0; 287297f4dd09SNikolay Borisov } 287397f4dd09SNikolay Borisov 287444c0ca21SBoris Burkov /* 28758cd29088SBoris Burkov * Some options only have meaning at mount time and shouldn't persist across 28768cd29088SBoris Burkov * remounts, or be displayed. Clear these at the end of mount and remount 28778cd29088SBoris Burkov * code paths. 28788cd29088SBoris Burkov */ 28798cd29088SBoris Burkov void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 28808cd29088SBoris Burkov { 28818cd29088SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 28828b228324SBoris Burkov btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 28838cd29088SBoris Burkov } 28848cd29088SBoris Burkov 28858cd29088SBoris Burkov /* 288644c0ca21SBoris Burkov * Mounting logic specific to read-write file systems. Shared by open_ctree 288744c0ca21SBoris Burkov * and btrfs_remount when remounting from read-only to read-write. 288844c0ca21SBoris Burkov */ 288944c0ca21SBoris Burkov int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) 289044c0ca21SBoris Burkov { 289144c0ca21SBoris Burkov int ret; 289294846229SBoris Burkov const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 28938b228324SBoris Burkov bool clear_free_space_tree = false; 28948b228324SBoris Burkov 28958b228324SBoris Burkov if (btrfs_test_opt(fs_info, CLEAR_CACHE) && 28968b228324SBoris Burkov btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 28978b228324SBoris Burkov clear_free_space_tree = true; 28988b228324SBoris Burkov } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 28998b228324SBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { 29008b228324SBoris Burkov btrfs_warn(fs_info, "free space tree is invalid"); 29018b228324SBoris Burkov clear_free_space_tree = true; 29028b228324SBoris Burkov } 29038b228324SBoris Burkov 29048b228324SBoris Burkov if (clear_free_space_tree) { 29058b228324SBoris Burkov btrfs_info(fs_info, "clearing free space tree"); 29068b228324SBoris Burkov ret = btrfs_clear_free_space_tree(fs_info); 29078b228324SBoris Burkov if (ret) { 29088b228324SBoris Burkov btrfs_warn(fs_info, 29098b228324SBoris Burkov "failed to clear free space tree: %d", ret); 29108b228324SBoris Burkov goto out; 29118b228324SBoris Burkov } 29128b228324SBoris Burkov } 291344c0ca21SBoris Burkov 291444c0ca21SBoris Burkov ret = btrfs_cleanup_fs_roots(fs_info); 291544c0ca21SBoris Burkov if (ret) 291644c0ca21SBoris Burkov goto out; 291744c0ca21SBoris Burkov 29188f1c21d7SBoris Burkov down_read(&fs_info->cleanup_work_sem); 29198f1c21d7SBoris Burkov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29208f1c21d7SBoris Burkov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29218f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29228f1c21d7SBoris Burkov goto out; 29238f1c21d7SBoris Burkov } 29248f1c21d7SBoris Burkov up_read(&fs_info->cleanup_work_sem); 29258f1c21d7SBoris Burkov 292644c0ca21SBoris Burkov mutex_lock(&fs_info->cleaner_mutex); 292744c0ca21SBoris Burkov ret = btrfs_recover_relocation(fs_info->tree_root); 292844c0ca21SBoris Burkov mutex_unlock(&fs_info->cleaner_mutex); 292944c0ca21SBoris Burkov if (ret < 0) { 293044c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to recover relocation: %d", ret); 293144c0ca21SBoris Burkov goto out; 293244c0ca21SBoris Burkov } 293344c0ca21SBoris Burkov 29345011139aSBoris Burkov if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && 29355011139aSBoris Burkov !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 29365011139aSBoris Burkov btrfs_info(fs_info, "creating free space tree"); 29375011139aSBoris Burkov ret = btrfs_create_free_space_tree(fs_info); 29385011139aSBoris Burkov if (ret) { 29395011139aSBoris Burkov btrfs_warn(fs_info, 29405011139aSBoris Burkov "failed to create free space tree: %d", ret); 29415011139aSBoris Burkov goto out; 29425011139aSBoris Burkov } 29435011139aSBoris Burkov } 29445011139aSBoris Burkov 294594846229SBoris Burkov if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { 294694846229SBoris Burkov ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 294794846229SBoris Burkov if (ret) 294894846229SBoris Burkov goto out; 294994846229SBoris Burkov } 295094846229SBoris Burkov 295144c0ca21SBoris Burkov ret = btrfs_resume_balance_async(fs_info); 295244c0ca21SBoris Burkov if (ret) 295344c0ca21SBoris Burkov goto out; 295444c0ca21SBoris Burkov 295544c0ca21SBoris Burkov ret = btrfs_resume_dev_replace_async(fs_info); 295644c0ca21SBoris Burkov if (ret) { 295744c0ca21SBoris Burkov btrfs_warn(fs_info, "failed to resume dev_replace"); 295844c0ca21SBoris Burkov goto out; 295944c0ca21SBoris Burkov } 296044c0ca21SBoris Burkov 296144c0ca21SBoris Burkov btrfs_qgroup_rescan_resume(fs_info); 296244c0ca21SBoris Burkov 296344c0ca21SBoris Burkov if (!fs_info->uuid_root) { 296444c0ca21SBoris Burkov btrfs_info(fs_info, "creating UUID tree"); 296544c0ca21SBoris Burkov ret = btrfs_create_uuid_tree(fs_info); 296644c0ca21SBoris Burkov if (ret) { 296744c0ca21SBoris Burkov btrfs_warn(fs_info, 296844c0ca21SBoris Burkov "failed to create the UUID tree %d", ret); 296944c0ca21SBoris Burkov goto out; 297044c0ca21SBoris Burkov } 297144c0ca21SBoris Burkov } 297244c0ca21SBoris Burkov 297344c0ca21SBoris Burkov out: 297444c0ca21SBoris Burkov return ret; 297544c0ca21SBoris Burkov } 297644c0ca21SBoris Burkov 2977ae18c37aSJosef Bacik int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, 2978ae18c37aSJosef Bacik char *options) 2979ae18c37aSJosef Bacik { 2980ae18c37aSJosef Bacik u32 sectorsize; 2981ae18c37aSJosef Bacik u32 nodesize; 2982ae18c37aSJosef Bacik u32 stripesize; 2983ae18c37aSJosef Bacik u64 generation; 2984ae18c37aSJosef Bacik u64 features; 2985ae18c37aSJosef Bacik u16 csum_type; 2986ae18c37aSJosef Bacik struct btrfs_super_block *disk_super; 2987ae18c37aSJosef Bacik struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2988ae18c37aSJosef Bacik struct btrfs_root *tree_root; 2989ae18c37aSJosef Bacik struct btrfs_root *chunk_root; 2990ae18c37aSJosef Bacik int ret; 2991ae18c37aSJosef Bacik int err = -EINVAL; 2992ae18c37aSJosef Bacik int level; 2993ae18c37aSJosef Bacik 29948260edbaSJosef Bacik ret = init_mount_fs_info(fs_info, sb); 299553b381b3SDavid Woodhouse if (ret) { 299683c8266aSDavid Sterba err = ret; 2997ae18c37aSJosef Bacik goto fail; 299853b381b3SDavid Woodhouse } 299953b381b3SDavid Woodhouse 3000ae18c37aSJosef Bacik /* These need to be init'ed before we start creating inodes and such. */ 3001ae18c37aSJosef Bacik tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, 3002ae18c37aSJosef Bacik GFP_KERNEL); 3003ae18c37aSJosef Bacik fs_info->tree_root = tree_root; 3004ae18c37aSJosef Bacik chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, 3005ae18c37aSJosef Bacik GFP_KERNEL); 3006ae18c37aSJosef Bacik fs_info->chunk_root = chunk_root; 3007ae18c37aSJosef Bacik if (!tree_root || !chunk_root) { 3008ae18c37aSJosef Bacik err = -ENOMEM; 3009c75e8394SJosef Bacik goto fail; 3010ae18c37aSJosef Bacik } 3011ae18c37aSJosef Bacik 3012ae18c37aSJosef Bacik fs_info->btree_inode = new_inode(sb); 3013ae18c37aSJosef Bacik if (!fs_info->btree_inode) { 3014ae18c37aSJosef Bacik err = -ENOMEM; 3015c75e8394SJosef Bacik goto fail; 3016ae18c37aSJosef Bacik } 3017ae18c37aSJosef Bacik mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 3018ae18c37aSJosef Bacik btrfs_init_btree_inode(fs_info); 3019ae18c37aSJosef Bacik 30203c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 30211104a885SDavid Sterba 30221104a885SDavid Sterba /* 30231104a885SDavid Sterba * Read super block and check the signature bytes only 30241104a885SDavid Sterba */ 30258f32380dSJohannes Thumshirn disk_super = btrfs_read_dev_super(fs_devices->latest_bdev); 30268f32380dSJohannes Thumshirn if (IS_ERR(disk_super)) { 30278f32380dSJohannes Thumshirn err = PTR_ERR(disk_super); 302816cdcec7SMiao Xie goto fail_alloc; 302920b45077SDave Young } 303039279cc3SChris Mason 30311104a885SDavid Sterba /* 3032260db43cSRandy Dunlap * Verify the type first, if that or the checksum value are 30338dc3f22cSJohannes Thumshirn * corrupted, we'll find out 30348dc3f22cSJohannes Thumshirn */ 30358f32380dSJohannes Thumshirn csum_type = btrfs_super_csum_type(disk_super); 303651bce6c9SJohannes Thumshirn if (!btrfs_supported_super_csum(csum_type)) { 30378dc3f22cSJohannes Thumshirn btrfs_err(fs_info, "unsupported checksum algorithm: %u", 303851bce6c9SJohannes Thumshirn csum_type); 30398dc3f22cSJohannes Thumshirn err = -EINVAL; 30408f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 30418dc3f22cSJohannes Thumshirn goto fail_alloc; 30428dc3f22cSJohannes Thumshirn } 30438dc3f22cSJohannes Thumshirn 30446d97c6e3SJohannes Thumshirn ret = btrfs_init_csum_hash(fs_info, csum_type); 30456d97c6e3SJohannes Thumshirn if (ret) { 30466d97c6e3SJohannes Thumshirn err = ret; 30478f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 30486d97c6e3SJohannes Thumshirn goto fail_alloc; 30496d97c6e3SJohannes Thumshirn } 30506d97c6e3SJohannes Thumshirn 30518dc3f22cSJohannes Thumshirn /* 30521104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 30531104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 30541104a885SDavid Sterba */ 30558f32380dSJohannes Thumshirn if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { 305605135f59SDavid Sterba btrfs_err(fs_info, "superblock checksum mismatch"); 30571104a885SDavid Sterba err = -EINVAL; 30588f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 3059141386e1SJosef Bacik goto fail_alloc; 30601104a885SDavid Sterba } 30611104a885SDavid Sterba 30621104a885SDavid Sterba /* 30631104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 30641104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 30651104a885SDavid Sterba * the whole block of INFO_SIZE 30661104a885SDavid Sterba */ 30678f32380dSJohannes Thumshirn memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); 30688f32380dSJohannes Thumshirn btrfs_release_disk_super(disk_super); 30695f39d397SChris Mason 3070fbc6feaeSNikolay Borisov disk_super = fs_info->super_copy; 3071fbc6feaeSNikolay Borisov 3072de37aa51SNikolay Borisov ASSERT(!memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, 3073de37aa51SNikolay Borisov BTRFS_FSID_SIZE)); 3074de37aa51SNikolay Borisov 30757239ff4bSNikolay Borisov if (btrfs_fs_incompat(fs_info, METADATA_UUID)) { 3076de37aa51SNikolay Borisov ASSERT(!memcmp(fs_info->fs_devices->metadata_uuid, 3077de37aa51SNikolay Borisov fs_info->super_copy->metadata_uuid, 3078de37aa51SNikolay Borisov BTRFS_FSID_SIZE)); 30797239ff4bSNikolay Borisov } 30800b86a832SChris Mason 3081fbc6feaeSNikolay Borisov features = btrfs_super_flags(disk_super); 3082fbc6feaeSNikolay Borisov if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { 3083fbc6feaeSNikolay Borisov features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; 3084fbc6feaeSNikolay Borisov btrfs_set_super_flags(disk_super, features); 3085fbc6feaeSNikolay Borisov btrfs_info(fs_info, 3086fbc6feaeSNikolay Borisov "found metadata UUID change in progress flag, clearing"); 3087fbc6feaeSNikolay Borisov } 3088fbc6feaeSNikolay Borisov 3089fbc6feaeSNikolay Borisov memcpy(fs_info->super_for_commit, fs_info->super_copy, 3090fbc6feaeSNikolay Borisov sizeof(*fs_info->super_for_commit)); 3091de37aa51SNikolay Borisov 3092069ec957SQu Wenruo ret = btrfs_validate_mount_super(fs_info); 30931104a885SDavid Sterba if (ret) { 309405135f59SDavid Sterba btrfs_err(fs_info, "superblock contains fatal errors"); 30951104a885SDavid Sterba err = -EINVAL; 3096141386e1SJosef Bacik goto fail_alloc; 30971104a885SDavid Sterba } 30981104a885SDavid Sterba 30990f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 3100141386e1SJosef Bacik goto fail_alloc; 31010f7d52f4SChris Mason 3102acce952bSliubo /* check FS state, whether FS is broken. */ 310387533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 310487533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 3105acce952bSliubo 310675e7cb7fSLiu Bo /* 310775e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 310875e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 310975e7cb7fSLiu Bo */ 311075e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 311175e7cb7fSLiu Bo 31122ff7e61eSJeff Mahoney ret = btrfs_parse_options(fs_info, options, sb->s_flags); 31132b82032cSYan Zheng if (ret) { 31142b82032cSYan Zheng err = ret; 3115141386e1SJosef Bacik goto fail_alloc; 31162b82032cSYan Zheng } 3117dfe25020SChris Mason 3118f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 3119f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 3120f2b636e8SJosef Bacik if (features) { 312105135f59SDavid Sterba btrfs_err(fs_info, 312205135f59SDavid Sterba "cannot mount because of unsupported optional features (%llx)", 3123c1c9ff7cSGeert Uytterhoeven features); 3124f2b636e8SJosef Bacik err = -EINVAL; 3125141386e1SJosef Bacik goto fail_alloc; 3126f2b636e8SJosef Bacik } 3127f2b636e8SJosef Bacik 31285d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 31295d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 31300b246afaSJeff Mahoney if (fs_info->compress_type == BTRFS_COMPRESS_LZO) 3131a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 31325c1aab1dSNick Terrell else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) 31335c1aab1dSNick Terrell features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; 3134727011e0SChris Mason 31353173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 313605135f59SDavid Sterba btrfs_info(fs_info, "has skinny extents"); 31373173a18fSJosef Bacik 3138b70f5097SNaohiro Aota fs_info->zoned = (features & BTRFS_FEATURE_INCOMPAT_ZONED); 3139b70f5097SNaohiro Aota 3140727011e0SChris Mason /* 3141727011e0SChris Mason * flag our filesystem as having big metadata blocks if 3142727011e0SChris Mason * they are bigger than the page size 3143727011e0SChris Mason */ 314409cbfeafSKirill A. Shutemov if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) { 3145727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 314605135f59SDavid Sterba btrfs_info(fs_info, 314705135f59SDavid Sterba "flagging fs with big metadata feature"); 3148727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 3149727011e0SChris Mason } 3150727011e0SChris Mason 3151bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 3152bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 3153b7f67055SChandan Rajendra stripesize = sectorsize; 3154707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 3155963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 3156bc3f116fSChris Mason 3157da17066cSJeff Mahoney /* Cache block sizes */ 3158da17066cSJeff Mahoney fs_info->nodesize = nodesize; 3159da17066cSJeff Mahoney fs_info->sectorsize = sectorsize; 3160ab108d99SDavid Sterba fs_info->sectorsize_bits = ilog2(sectorsize); 316122b6331dSDavid Sterba fs_info->csum_size = btrfs_super_csum_size(disk_super); 3162fe5ecbe8SDavid Sterba fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; 3163da17066cSJeff Mahoney fs_info->stripesize = stripesize; 3164da17066cSJeff Mahoney 3165bc3f116fSChris Mason /* 3166bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 3167bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 3168bc3f116fSChris Mason */ 3169bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 3170707e8a07SDavid Sterba (sectorsize != nodesize)) { 317105135f59SDavid Sterba btrfs_err(fs_info, 317205135f59SDavid Sterba "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", 317305135f59SDavid Sterba nodesize, sectorsize); 3174141386e1SJosef Bacik goto fail_alloc; 3175bc3f116fSChris Mason } 3176bc3f116fSChris Mason 3177ceda0864SMiao Xie /* 3178ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 3179ceda0864SMiao Xie * update the flag. 3180ceda0864SMiao Xie */ 31815d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 31825d4f98a2SYan Zheng 3183f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 3184f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 3185bc98a42cSDavid Howells if (!sb_rdonly(sb) && features) { 318605135f59SDavid Sterba btrfs_err(fs_info, 318705135f59SDavid Sterba "cannot mount read-write because of unsupported optional features (%llx)", 3188c1c9ff7cSGeert Uytterhoeven features); 3189f2b636e8SJosef Bacik err = -EINVAL; 3190141386e1SJosef Bacik goto fail_alloc; 3191f2b636e8SJosef Bacik } 319261d92c32SChris Mason 31932a458198SEric Sandeen ret = btrfs_init_workqueues(fs_info, fs_devices); 31942a458198SEric Sandeen if (ret) { 31952a458198SEric Sandeen err = ret; 31960dc3b84aSJosef Bacik goto fail_sb_buffer; 31970dc3b84aSJosef Bacik } 31984543df7eSChris Mason 31999e11ceeeSJan Kara sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); 32009e11ceeeSJan Kara sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); 32014575c9ccSChris Mason 3202a061fc8dSChris Mason sb->s_blocksize = sectorsize; 3203a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 3204de37aa51SNikolay Borisov memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); 3205db94535dSChris Mason 3206925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 32076bccf3abSJeff Mahoney ret = btrfs_read_sys_array(fs_info); 3208925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 320984eed90fSChris Mason if (ret) { 321005135f59SDavid Sterba btrfs_err(fs_info, "failed to read the system array: %d", ret); 32115d4f98a2SYan Zheng goto fail_sb_buffer; 321284eed90fSChris Mason } 32130b86a832SChris Mason 321484234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 3215581c1760SQu Wenruo level = btrfs_super_chunk_root_level(disk_super); 32160b86a832SChris Mason 32172ff7e61eSJeff Mahoney chunk_root->node = read_tree_block(fs_info, 32180b86a832SChris Mason btrfs_super_chunk_root(disk_super), 32191b7ec85eSJosef Bacik BTRFS_CHUNK_TREE_OBJECTID, 3220581c1760SQu Wenruo generation, level, NULL); 322164c043deSLiu Bo if (IS_ERR(chunk_root->node) || 322264c043deSLiu Bo !extent_buffer_uptodate(chunk_root->node)) { 322305135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk root"); 3224e5fffbacSchandan if (!IS_ERR(chunk_root->node)) 3225e5fffbacSchandan free_extent_buffer(chunk_root->node); 322695ab1f64SZhao Lei chunk_root->node = NULL; 3227af31f5e5SChris Mason goto fail_tree_roots; 322883121942SDavid Woodhouse } 32295d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 32305d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 32310b86a832SChris Mason 3232e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 3233c4ac7541SDavid Sterba offsetof(struct btrfs_header, chunk_tree_uuid), 3234c4ac7541SDavid Sterba BTRFS_UUID_SIZE); 3235e17cade2SChris Mason 32365b4aacefSJeff Mahoney ret = btrfs_read_chunk_tree(fs_info); 32372b82032cSYan Zheng if (ret) { 323805135f59SDavid Sterba btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 3239af31f5e5SChris Mason goto fail_tree_roots; 32402b82032cSYan Zheng } 32410b86a832SChris Mason 32428dabb742SStefan Behrens /* 3243bacce86aSAnand Jain * At this point we know all the devices that make this filesystem, 3244bacce86aSAnand Jain * including the seed devices but we don't know yet if the replace 3245bacce86aSAnand Jain * target is required. So free devices that are not part of this 3246bacce86aSAnand Jain * filesystem but skip the replace traget device which is checked 3247bacce86aSAnand Jain * below in btrfs_init_dev_replace(). 32488dabb742SStefan Behrens */ 3249bacce86aSAnand Jain btrfs_free_extra_devids(fs_devices); 3250a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 325105135f59SDavid Sterba btrfs_err(fs_info, "failed to read devices"); 3252a6b0d5c8SChris Mason goto fail_tree_roots; 3253a6b0d5c8SChris Mason } 3254a6b0d5c8SChris Mason 3255b8522a1eSNikolay Borisov ret = init_tree_roots(fs_info); 32564bbcaa64SEric Sandeen if (ret) 3257b8522a1eSNikolay Borisov goto fail_tree_roots; 32588929ecfaSYan, Zheng 325975ec1db8SJosef Bacik /* 326075ec1db8SJosef Bacik * If we have a uuid root and we're not being told to rescan we need to 326175ec1db8SJosef Bacik * check the generation here so we can set the 326275ec1db8SJosef Bacik * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the 326375ec1db8SJosef Bacik * transaction during a balance or the log replay without updating the 326475ec1db8SJosef Bacik * uuid generation, and then if we crash we would rescan the uuid tree, 326575ec1db8SJosef Bacik * even though it was perfectly fine. 326675ec1db8SJosef Bacik */ 326775ec1db8SJosef Bacik if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && 326875ec1db8SJosef Bacik fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) 326975ec1db8SJosef Bacik set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); 327075ec1db8SJosef Bacik 3271cf90d884SQu Wenruo ret = btrfs_verify_dev_extents(fs_info); 3272cf90d884SQu Wenruo if (ret) { 3273cf90d884SQu Wenruo btrfs_err(fs_info, 3274cf90d884SQu Wenruo "failed to verify dev extents against chunks: %d", 3275cf90d884SQu Wenruo ret); 3276cf90d884SQu Wenruo goto fail_block_groups; 3277cf90d884SQu Wenruo } 327868310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 327968310a5eSIlya Dryomov if (ret) { 328005135f59SDavid Sterba btrfs_err(fs_info, "failed to recover balance: %d", ret); 328168310a5eSIlya Dryomov goto fail_block_groups; 328268310a5eSIlya Dryomov } 328368310a5eSIlya Dryomov 3284733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 3285733f4fbbSStefan Behrens if (ret) { 328605135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_stats: %d", ret); 3287733f4fbbSStefan Behrens goto fail_block_groups; 3288733f4fbbSStefan Behrens } 3289733f4fbbSStefan Behrens 32908dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 32918dabb742SStefan Behrens if (ret) { 329205135f59SDavid Sterba btrfs_err(fs_info, "failed to init dev_replace: %d", ret); 32938dabb742SStefan Behrens goto fail_block_groups; 32948dabb742SStefan Behrens } 32958dabb742SStefan Behrens 3296b70f5097SNaohiro Aota ret = btrfs_check_zoned_mode(fs_info); 3297b70f5097SNaohiro Aota if (ret) { 3298b70f5097SNaohiro Aota btrfs_err(fs_info, "failed to initialize zoned mode: %d", 3299b70f5097SNaohiro Aota ret); 3300b70f5097SNaohiro Aota goto fail_block_groups; 3301b70f5097SNaohiro Aota } 3302b70f5097SNaohiro Aota 3303c6761a9eSAnand Jain ret = btrfs_sysfs_add_fsid(fs_devices); 3304b7c35e81SAnand Jain if (ret) { 330505135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", 330605135f59SDavid Sterba ret); 3307b7c35e81SAnand Jain goto fail_block_groups; 3308b7c35e81SAnand Jain } 3309b7c35e81SAnand Jain 331096f3136eSAnand Jain ret = btrfs_sysfs_add_mounted(fs_info); 33115ac1d209SJeff Mahoney if (ret) { 331205135f59SDavid Sterba btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); 3313b7c35e81SAnand Jain goto fail_fsdev_sysfs; 33145ac1d209SJeff Mahoney } 33155ac1d209SJeff Mahoney 3316c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 3317c59021f8Sliubo if (ret) { 331805135f59SDavid Sterba btrfs_err(fs_info, "failed to initialize space info: %d", ret); 33192365dd3cSAnand Jain goto fail_sysfs; 3320c59021f8Sliubo } 3321c59021f8Sliubo 33225b4aacefSJeff Mahoney ret = btrfs_read_block_groups(fs_info); 33231b1d1f66SJosef Bacik if (ret) { 332405135f59SDavid Sterba btrfs_err(fs_info, "failed to read block groups: %d", ret); 33252365dd3cSAnand Jain goto fail_sysfs; 33261b1d1f66SJosef Bacik } 33274330e183SQu Wenruo 33286528b99dSAnand Jain if (!sb_rdonly(sb) && !btrfs_check_rw_degradable(fs_info, NULL)) { 332905135f59SDavid Sterba btrfs_warn(fs_info, 333052042d8eSAndrea Gelmini "writable mount is not allowed due to too many missing devices"); 33312365dd3cSAnand Jain goto fail_sysfs; 3332292fd7fcSStefan Behrens } 33339078a3e1SChris Mason 3334a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 3335a74a4b97SChris Mason "btrfs-cleaner"); 333657506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 33372365dd3cSAnand Jain goto fail_sysfs; 3338a74a4b97SChris Mason 3339a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 3340a74a4b97SChris Mason tree_root, 3341a74a4b97SChris Mason "btrfs-transaction"); 334257506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 33433f157a2fSChris Mason goto fail_cleaner; 3344a74a4b97SChris Mason 3345583b7231SHans van Kranenburg if (!btrfs_test_opt(fs_info, NOSSD) && 3346c289811cSChris Mason !fs_info->fs_devices->rotating) { 3347583b7231SHans van Kranenburg btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); 3348c289811cSChris Mason } 3349c289811cSChris Mason 3350572d9ab7SDavid Sterba /* 335101327610SNicholas D Steeves * Mount does not set all options immediately, we can do it now and do 3352572d9ab7SDavid Sterba * not have to wait for transaction commit 3353572d9ab7SDavid Sterba */ 3354572d9ab7SDavid Sterba btrfs_apply_pending_changes(fs_info); 33553818aea2SQu Wenruo 335621adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 33570b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { 33582ff7e61eSJeff Mahoney ret = btrfsic_mount(fs_info, fs_devices, 33590b246afaSJeff Mahoney btrfs_test_opt(fs_info, 336021adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 336121adbd5cSStefan Behrens 1 : 0, 336221adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 336321adbd5cSStefan Behrens if (ret) 336405135f59SDavid Sterba btrfs_warn(fs_info, 336505135f59SDavid Sterba "failed to initialize integrity check module: %d", 336605135f59SDavid Sterba ret); 336721adbd5cSStefan Behrens } 336821adbd5cSStefan Behrens #endif 3369bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 3370bcef60f2SArne Jansen if (ret) 3371bcef60f2SArne Jansen goto fail_trans_kthread; 337221adbd5cSStefan Behrens 3373fd708b81SJosef Bacik if (btrfs_build_ref_tree(fs_info)) 3374fd708b81SJosef Bacik btrfs_err(fs_info, "couldn't build ref tree"); 3375fd708b81SJosef Bacik 337696da0919SQu Wenruo /* do not make disk changes in broken FS or nologreplay is given */ 337796da0919SQu Wenruo if (btrfs_super_log_root(disk_super) != 0 && 33780b246afaSJeff Mahoney !btrfs_test_opt(fs_info, NOLOGREPLAY)) { 3379e8294f2fSDavid Sterba btrfs_info(fs_info, "start tree-log replay"); 338063443bf5SEric Sandeen ret = btrfs_replay_log(fs_info, fs_devices); 338179787eaaSJeff Mahoney if (ret) { 338263443bf5SEric Sandeen err = ret; 338328c16cbbSWang Shilong goto fail_qgroup; 3384e556ce2cSYan Zheng } 3385e02119d5SChris Mason } 33861a40e23bSZheng Yan 33876bccf3abSJeff Mahoney ret = btrfs_find_orphan_roots(fs_info); 338879787eaaSJeff Mahoney if (ret) 338928c16cbbSWang Shilong goto fail_qgroup; 339076dda93cSYan, Zheng 339156e9357aSDavid Sterba fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); 33923140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 33933140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 3394f50f4353SLiu Bo btrfs_warn(fs_info, "failed to read fs tree: %d", err); 3395315bf8efSJosef Bacik fs_info->fs_root = NULL; 3396bcef60f2SArne Jansen goto fail_qgroup; 33973140c9a3SDan Carpenter } 3398c289811cSChris Mason 3399bc98a42cSDavid Howells if (sb_rdonly(sb)) 34008cd29088SBoris Burkov goto clear_oneshot; 34012b6ba629SIlya Dryomov 340244c0ca21SBoris Burkov ret = btrfs_start_pre_rw_mount(fs_info); 34032b6ba629SIlya Dryomov if (ret) { 34046bccf3abSJeff Mahoney close_ctree(fs_info); 34052b6ba629SIlya Dryomov return ret; 3406e3acc2a6SJosef Bacik } 3407b0643e59SDennis Zhou btrfs_discard_resume(fs_info); 3408b382a324SJan Schmidt 340944c0ca21SBoris Burkov if (fs_info->uuid_root && 341044c0ca21SBoris Burkov (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || 341144c0ca21SBoris Burkov fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { 341205135f59SDavid Sterba btrfs_info(fs_info, "checking UUID tree"); 341370f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 341470f80175SStefan Behrens if (ret) { 341505135f59SDavid Sterba btrfs_warn(fs_info, 341605135f59SDavid Sterba "failed to check the UUID tree: %d", ret); 34176bccf3abSJeff Mahoney close_ctree(fs_info); 341870f80175SStefan Behrens return ret; 341970f80175SStefan Behrens } 3420f7a81ea4SStefan Behrens } 342194846229SBoris Burkov 3422afcdd129SJosef Bacik set_bit(BTRFS_FS_OPEN, &fs_info->flags); 342347ab2a6cSJosef Bacik 34248cd29088SBoris Burkov clear_oneshot: 34258cd29088SBoris Burkov btrfs_clear_oneshot_options(fs_info); 3426ad2b2c80SAl Viro return 0; 342739279cc3SChris Mason 3428bcef60f2SArne Jansen fail_qgroup: 3429bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 34307c2ca468SChris Mason fail_trans_kthread: 34317c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 34322ff7e61eSJeff Mahoney btrfs_cleanup_transaction(fs_info); 3433faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 34343f157a2fSChris Mason fail_cleaner: 3435a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 34367c2ca468SChris Mason 34377c2ca468SChris Mason /* 34387c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 34397c2ca468SChris Mason * kthreads 34407c2ca468SChris Mason */ 34417c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 34427c2ca468SChris Mason 34432365dd3cSAnand Jain fail_sysfs: 34446618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 34452365dd3cSAnand Jain 3446b7c35e81SAnand Jain fail_fsdev_sysfs: 3447b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 3448b7c35e81SAnand Jain 34491b1d1f66SJosef Bacik fail_block_groups: 345054067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 3451af31f5e5SChris Mason 3452af31f5e5SChris Mason fail_tree_roots: 34539e3aa805SJosef Bacik if (fs_info->data_reloc_root) 34549e3aa805SJosef Bacik btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); 34554273eaffSAnand Jain free_root_pointers(fs_info, true); 34562b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3457af31f5e5SChris Mason 345839279cc3SChris Mason fail_sb_buffer: 34597abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 34605cdd7db6SFilipe Manana btrfs_free_block_groups(fs_info); 346116cdcec7SMiao Xie fail_alloc: 3462586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3463586e46e2SIlya Dryomov 34644543df7eSChris Mason iput(fs_info->btree_inode); 34657e662854SQinghuang Feng fail: 3466586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3467ad2b2c80SAl Viro return err; 3468eb60ceacSChris Mason } 3469663faf9fSMasami Hiramatsu ALLOW_ERROR_INJECTION(open_ctree, ERRNO); 3470eb60ceacSChris Mason 3471314b6dd0SJohannes Thumshirn static void btrfs_end_super_write(struct bio *bio) 3472f2984462SChris Mason { 3473314b6dd0SJohannes Thumshirn struct btrfs_device *device = bio->bi_private; 3474314b6dd0SJohannes Thumshirn struct bio_vec *bvec; 3475314b6dd0SJohannes Thumshirn struct bvec_iter_all iter_all; 3476314b6dd0SJohannes Thumshirn struct page *page; 3477442a4f63SStefan Behrens 3478314b6dd0SJohannes Thumshirn bio_for_each_segment_all(bvec, bio, iter_all) { 3479314b6dd0SJohannes Thumshirn page = bvec->bv_page; 3480314b6dd0SJohannes Thumshirn 3481314b6dd0SJohannes Thumshirn if (bio->bi_status) { 3482fb456252SJeff Mahoney btrfs_warn_rl_in_rcu(device->fs_info, 3483314b6dd0SJohannes Thumshirn "lost page write due to IO error on %s (%d)", 3484314b6dd0SJohannes Thumshirn rcu_str_deref(device->name), 3485314b6dd0SJohannes Thumshirn blk_status_to_errno(bio->bi_status)); 3486314b6dd0SJohannes Thumshirn ClearPageUptodate(page); 3487314b6dd0SJohannes Thumshirn SetPageError(page); 3488314b6dd0SJohannes Thumshirn btrfs_dev_stat_inc_and_print(device, 3489314b6dd0SJohannes Thumshirn BTRFS_DEV_STAT_WRITE_ERRS); 3490314b6dd0SJohannes Thumshirn } else { 3491314b6dd0SJohannes Thumshirn SetPageUptodate(page); 3492f2984462SChris Mason } 3493314b6dd0SJohannes Thumshirn 3494314b6dd0SJohannes Thumshirn put_page(page); 3495314b6dd0SJohannes Thumshirn unlock_page(page); 3496314b6dd0SJohannes Thumshirn } 3497314b6dd0SJohannes Thumshirn 3498314b6dd0SJohannes Thumshirn bio_put(bio); 3499f2984462SChris Mason } 3500f2984462SChris Mason 35018f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, 35028f32380dSJohannes Thumshirn int copy_num) 350329c36d72SAnand Jain { 350429c36d72SAnand Jain struct btrfs_super_block *super; 35058f32380dSJohannes Thumshirn struct page *page; 350612659251SNaohiro Aota u64 bytenr, bytenr_orig; 35078f32380dSJohannes Thumshirn struct address_space *mapping = bdev->bd_inode->i_mapping; 350812659251SNaohiro Aota int ret; 350929c36d72SAnand Jain 351012659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(copy_num); 351112659251SNaohiro Aota ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); 351212659251SNaohiro Aota if (ret == -ENOENT) 351312659251SNaohiro Aota return ERR_PTR(-EINVAL); 351412659251SNaohiro Aota else if (ret) 351512659251SNaohiro Aota return ERR_PTR(ret); 351612659251SNaohiro Aota 351729c36d72SAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= i_size_read(bdev->bd_inode)) 35188f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 351929c36d72SAnand Jain 35208f32380dSJohannes Thumshirn page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); 35218f32380dSJohannes Thumshirn if (IS_ERR(page)) 35228f32380dSJohannes Thumshirn return ERR_CAST(page); 352329c36d72SAnand Jain 35248f32380dSJohannes Thumshirn super = page_address(page); 352596c2e067SAnand Jain if (btrfs_super_magic(super) != BTRFS_MAGIC) { 352696c2e067SAnand Jain btrfs_release_disk_super(super); 352796c2e067SAnand Jain return ERR_PTR(-ENODATA); 352896c2e067SAnand Jain } 352996c2e067SAnand Jain 353012659251SNaohiro Aota if (btrfs_super_bytenr(super) != bytenr_orig) { 35318f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 35328f32380dSJohannes Thumshirn return ERR_PTR(-EINVAL); 353329c36d72SAnand Jain } 353429c36d72SAnand Jain 35358f32380dSJohannes Thumshirn return super; 353629c36d72SAnand Jain } 353729c36d72SAnand Jain 353829c36d72SAnand Jain 35398f32380dSJohannes Thumshirn struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) 3540a512bbf8SYan Zheng { 35418f32380dSJohannes Thumshirn struct btrfs_super_block *super, *latest = NULL; 3542a512bbf8SYan Zheng int i; 3543a512bbf8SYan Zheng u64 transid = 0; 3544a512bbf8SYan Zheng 3545a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3546a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3547a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3548a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3549a512bbf8SYan Zheng */ 3550a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 35518f32380dSJohannes Thumshirn super = btrfs_read_dev_one_super(bdev, i); 35528f32380dSJohannes Thumshirn if (IS_ERR(super)) 3553a512bbf8SYan Zheng continue; 3554a512bbf8SYan Zheng 3555a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 35568f32380dSJohannes Thumshirn if (latest) 35578f32380dSJohannes Thumshirn btrfs_release_disk_super(super); 35588f32380dSJohannes Thumshirn 35598f32380dSJohannes Thumshirn latest = super; 3560a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3561a512bbf8SYan Zheng } 3562a512bbf8SYan Zheng } 356392fc03fbSAnand Jain 35648f32380dSJohannes Thumshirn return super; 3565a512bbf8SYan Zheng } 3566a512bbf8SYan Zheng 35674eedeb75SHisashi Hifumi /* 3568abbb3b8eSDavid Sterba * Write superblock @sb to the @device. Do not wait for completion, all the 3569314b6dd0SJohannes Thumshirn * pages we use for writing are locked. 35704eedeb75SHisashi Hifumi * 3571abbb3b8eSDavid Sterba * Write @max_mirrors copies of the superblock, where 0 means default that fit 3572abbb3b8eSDavid Sterba * the expected device size at commit time. Note that max_mirrors must be 3573abbb3b8eSDavid Sterba * same for write and wait phases. 35744eedeb75SHisashi Hifumi * 3575314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or submission fails. 35764eedeb75SHisashi Hifumi */ 3577a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3578abbb3b8eSDavid Sterba struct btrfs_super_block *sb, int max_mirrors) 3579a512bbf8SYan Zheng { 3580d5178578SJohannes Thumshirn struct btrfs_fs_info *fs_info = device->fs_info; 3581314b6dd0SJohannes Thumshirn struct address_space *mapping = device->bdev->bd_inode->i_mapping; 3582d5178578SJohannes Thumshirn SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); 3583a512bbf8SYan Zheng int i; 3584a512bbf8SYan Zheng int errors = 0; 358512659251SNaohiro Aota int ret; 358612659251SNaohiro Aota u64 bytenr, bytenr_orig; 3587a512bbf8SYan Zheng 3588a512bbf8SYan Zheng if (max_mirrors == 0) 3589a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3590a512bbf8SYan Zheng 3591d5178578SJohannes Thumshirn shash->tfm = fs_info->csum_shash; 3592d5178578SJohannes Thumshirn 3593a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3594314b6dd0SJohannes Thumshirn struct page *page; 3595314b6dd0SJohannes Thumshirn struct bio *bio; 3596314b6dd0SJohannes Thumshirn struct btrfs_super_block *disk_super; 3597314b6dd0SJohannes Thumshirn 359812659251SNaohiro Aota bytenr_orig = btrfs_sb_offset(i); 359912659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); 360012659251SNaohiro Aota if (ret == -ENOENT) { 360112659251SNaohiro Aota continue; 360212659251SNaohiro Aota } else if (ret < 0) { 360312659251SNaohiro Aota btrfs_err(device->fs_info, 360412659251SNaohiro Aota "couldn't get super block location for mirror %d", 360512659251SNaohiro Aota i); 360612659251SNaohiro Aota errors++; 360712659251SNaohiro Aota continue; 360812659251SNaohiro Aota } 3609935e5cc9SMiao Xie if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3610935e5cc9SMiao Xie device->commit_total_bytes) 3611a512bbf8SYan Zheng break; 3612a512bbf8SYan Zheng 361312659251SNaohiro Aota btrfs_set_super_bytenr(sb, bytenr_orig); 3614a512bbf8SYan Zheng 3615fd08001fSEric Biggers crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, 3616fd08001fSEric Biggers BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, 3617fd08001fSEric Biggers sb->csum); 3618a512bbf8SYan Zheng 3619314b6dd0SJohannes Thumshirn page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, 3620314b6dd0SJohannes Thumshirn GFP_NOFS); 3621314b6dd0SJohannes Thumshirn if (!page) { 3622fb456252SJeff Mahoney btrfs_err(device->fs_info, 3623314b6dd0SJohannes Thumshirn "couldn't get super block page for bytenr %llu", 3624f14d104dSDavid Sterba bytenr); 3625634554dcSJosef Bacik errors++; 3626634554dcSJosef Bacik continue; 3627634554dcSJosef Bacik } 3628634554dcSJosef Bacik 3629314b6dd0SJohannes Thumshirn /* Bump the refcount for wait_dev_supers() */ 3630314b6dd0SJohannes Thumshirn get_page(page); 3631a512bbf8SYan Zheng 3632314b6dd0SJohannes Thumshirn disk_super = page_address(page); 3633314b6dd0SJohannes Thumshirn memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); 3634a512bbf8SYan Zheng 3635387125fcSChris Mason /* 3636314b6dd0SJohannes Thumshirn * Directly use bios here instead of relying on the page cache 3637314b6dd0SJohannes Thumshirn * to do I/O, so we don't lose the ability to do integrity 3638314b6dd0SJohannes Thumshirn * checking. 3639387125fcSChris Mason */ 3640314b6dd0SJohannes Thumshirn bio = bio_alloc(GFP_NOFS, 1); 3641314b6dd0SJohannes Thumshirn bio_set_dev(bio, device->bdev); 3642314b6dd0SJohannes Thumshirn bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; 3643314b6dd0SJohannes Thumshirn bio->bi_private = device; 3644314b6dd0SJohannes Thumshirn bio->bi_end_io = btrfs_end_super_write; 3645314b6dd0SJohannes Thumshirn __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, 3646314b6dd0SJohannes Thumshirn offset_in_page(bytenr)); 3647314b6dd0SJohannes Thumshirn 3648314b6dd0SJohannes Thumshirn /* 3649314b6dd0SJohannes Thumshirn * We FUA only the first super block. The others we allow to 3650314b6dd0SJohannes Thumshirn * go down lazy and there's a short window where the on-disk 3651314b6dd0SJohannes Thumshirn * copies might still contain the older version. 3652314b6dd0SJohannes Thumshirn */ 3653314b6dd0SJohannes Thumshirn bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO; 36541b9e619cSOmar Sandoval if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) 3655314b6dd0SJohannes Thumshirn bio->bi_opf |= REQ_FUA; 3656314b6dd0SJohannes Thumshirn 3657314b6dd0SJohannes Thumshirn btrfsic_submit_bio(bio); 365812659251SNaohiro Aota btrfs_advance_sb_log(device, i); 3659a512bbf8SYan Zheng } 3660a512bbf8SYan Zheng return errors < i ? 0 : -1; 3661a512bbf8SYan Zheng } 3662a512bbf8SYan Zheng 3663387125fcSChris Mason /* 3664abbb3b8eSDavid Sterba * Wait for write completion of superblocks done by write_dev_supers, 3665abbb3b8eSDavid Sterba * @max_mirrors same for write and wait phases. 3666abbb3b8eSDavid Sterba * 3667314b6dd0SJohannes Thumshirn * Return number of errors when page is not found or not marked up to 3668abbb3b8eSDavid Sterba * date. 3669abbb3b8eSDavid Sterba */ 3670abbb3b8eSDavid Sterba static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) 3671abbb3b8eSDavid Sterba { 3672abbb3b8eSDavid Sterba int i; 3673abbb3b8eSDavid Sterba int errors = 0; 3674b6a535faSHoward McLauchlan bool primary_failed = false; 367512659251SNaohiro Aota int ret; 3676abbb3b8eSDavid Sterba u64 bytenr; 3677abbb3b8eSDavid Sterba 3678abbb3b8eSDavid Sterba if (max_mirrors == 0) 3679abbb3b8eSDavid Sterba max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3680abbb3b8eSDavid Sterba 3681abbb3b8eSDavid Sterba for (i = 0; i < max_mirrors; i++) { 3682314b6dd0SJohannes Thumshirn struct page *page; 3683314b6dd0SJohannes Thumshirn 368412659251SNaohiro Aota ret = btrfs_sb_log_location(device, i, READ, &bytenr); 368512659251SNaohiro Aota if (ret == -ENOENT) { 368612659251SNaohiro Aota break; 368712659251SNaohiro Aota } else if (ret < 0) { 368812659251SNaohiro Aota errors++; 368912659251SNaohiro Aota if (i == 0) 369012659251SNaohiro Aota primary_failed = true; 369112659251SNaohiro Aota continue; 369212659251SNaohiro Aota } 3693abbb3b8eSDavid Sterba if (bytenr + BTRFS_SUPER_INFO_SIZE >= 3694abbb3b8eSDavid Sterba device->commit_total_bytes) 3695abbb3b8eSDavid Sterba break; 3696abbb3b8eSDavid Sterba 3697314b6dd0SJohannes Thumshirn page = find_get_page(device->bdev->bd_inode->i_mapping, 3698314b6dd0SJohannes Thumshirn bytenr >> PAGE_SHIFT); 3699314b6dd0SJohannes Thumshirn if (!page) { 3700abbb3b8eSDavid Sterba errors++; 3701b6a535faSHoward McLauchlan if (i == 0) 3702b6a535faSHoward McLauchlan primary_failed = true; 3703abbb3b8eSDavid Sterba continue; 3704abbb3b8eSDavid Sterba } 3705314b6dd0SJohannes Thumshirn /* Page is submitted locked and unlocked once the IO completes */ 3706314b6dd0SJohannes Thumshirn wait_on_page_locked(page); 3707314b6dd0SJohannes Thumshirn if (PageError(page)) { 3708abbb3b8eSDavid Sterba errors++; 3709b6a535faSHoward McLauchlan if (i == 0) 3710b6a535faSHoward McLauchlan primary_failed = true; 3711b6a535faSHoward McLauchlan } 3712abbb3b8eSDavid Sterba 3713314b6dd0SJohannes Thumshirn /* Drop our reference */ 3714314b6dd0SJohannes Thumshirn put_page(page); 3715abbb3b8eSDavid Sterba 3716314b6dd0SJohannes Thumshirn /* Drop the reference from the writing run */ 3717314b6dd0SJohannes Thumshirn put_page(page); 3718abbb3b8eSDavid Sterba } 3719abbb3b8eSDavid Sterba 3720b6a535faSHoward McLauchlan /* log error, force error return */ 3721b6a535faSHoward McLauchlan if (primary_failed) { 3722b6a535faSHoward McLauchlan btrfs_err(device->fs_info, "error writing primary super block to device %llu", 3723b6a535faSHoward McLauchlan device->devid); 3724b6a535faSHoward McLauchlan return -1; 3725b6a535faSHoward McLauchlan } 3726b6a535faSHoward McLauchlan 3727abbb3b8eSDavid Sterba return errors < i ? 0 : -1; 3728abbb3b8eSDavid Sterba } 3729abbb3b8eSDavid Sterba 3730abbb3b8eSDavid Sterba /* 3731387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3732387125fcSChris Mason * for the barrier when it is done 3733387125fcSChris Mason */ 37344246a0b6SChristoph Hellwig static void btrfs_end_empty_barrier(struct bio *bio) 3735387125fcSChris Mason { 3736387125fcSChris Mason complete(bio->bi_private); 3737387125fcSChris Mason } 3738387125fcSChris Mason 3739387125fcSChris Mason /* 37404fc6441aSAnand Jain * Submit a flush request to the device if it supports it. Error handling is 37414fc6441aSAnand Jain * done in the waiting counterpart. 3742387125fcSChris Mason */ 37434fc6441aSAnand Jain static void write_dev_flush(struct btrfs_device *device) 3744387125fcSChris Mason { 3745c2a9c7abSAnand Jain struct request_queue *q = bdev_get_queue(device->bdev); 3746e0ae9994SDavid Sterba struct bio *bio = device->flush_bio; 3747387125fcSChris Mason 3748c2a9c7abSAnand Jain if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) 37494fc6441aSAnand Jain return; 3750387125fcSChris Mason 3751e0ae9994SDavid Sterba bio_reset(bio); 3752387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 375374d46992SChristoph Hellwig bio_set_dev(bio, device->bdev); 37548d910125SJan Kara bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH; 3755387125fcSChris Mason init_completion(&device->flush_wait); 3756387125fcSChris Mason bio->bi_private = &device->flush_wait; 3757387125fcSChris Mason 375843a01111SLu Fengqi btrfsic_submit_bio(bio); 37591c3063b6SAnand Jain set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 37604fc6441aSAnand Jain } 3761387125fcSChris Mason 37624fc6441aSAnand Jain /* 37634fc6441aSAnand Jain * If the flush bio has been submitted by write_dev_flush, wait for it. 37644fc6441aSAnand Jain */ 37658c27cb35SLinus Torvalds static blk_status_t wait_dev_flush(struct btrfs_device *device) 37664fc6441aSAnand Jain { 37674fc6441aSAnand Jain struct bio *bio = device->flush_bio; 37684fc6441aSAnand Jain 37691c3063b6SAnand Jain if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) 377058efbc9fSOmar Sandoval return BLK_STS_OK; 37714fc6441aSAnand Jain 37721c3063b6SAnand Jain clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); 37732980d574SDavid Sterba wait_for_completion_io(&device->flush_wait); 37744fc6441aSAnand Jain 37758c27cb35SLinus Torvalds return bio->bi_status; 3776387125fcSChris Mason } 3777387125fcSChris Mason 3778d10b82feSQu Wenruo static int check_barrier_error(struct btrfs_fs_info *fs_info) 3779401b41e5SAnand Jain { 37806528b99dSAnand Jain if (!btrfs_check_rw_degradable(fs_info, NULL)) 3781401b41e5SAnand Jain return -EIO; 3782387125fcSChris Mason return 0; 3783387125fcSChris Mason } 3784387125fcSChris Mason 3785387125fcSChris Mason /* 3786387125fcSChris Mason * send an empty flush down to each device in parallel, 3787387125fcSChris Mason * then wait for them 3788387125fcSChris Mason */ 3789387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3790387125fcSChris Mason { 3791387125fcSChris Mason struct list_head *head; 3792387125fcSChris Mason struct btrfs_device *dev; 37935af3e8ccSStefan Behrens int errors_wait = 0; 37944e4cbee9SChristoph Hellwig blk_status_t ret; 3795387125fcSChris Mason 37961538e6c5SDavid Sterba lockdep_assert_held(&info->fs_devices->device_list_mutex); 3797387125fcSChris Mason /* send down all the barriers */ 3798387125fcSChris Mason head = &info->fs_devices->devices; 37991538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3800e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3801f88ba6a2SHidetoshi Seto continue; 3802cea7c8bfSAnand Jain if (!dev->bdev) 3803387125fcSChris Mason continue; 3804e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3805ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3806387125fcSChris Mason continue; 3807387125fcSChris Mason 38084fc6441aSAnand Jain write_dev_flush(dev); 380958efbc9fSOmar Sandoval dev->last_flush_error = BLK_STS_OK; 3810387125fcSChris Mason } 3811387125fcSChris Mason 3812387125fcSChris Mason /* wait for all the barriers */ 38131538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3814e6e674bdSAnand Jain if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) 3815f88ba6a2SHidetoshi Seto continue; 3816387125fcSChris Mason if (!dev->bdev) { 38175af3e8ccSStefan Behrens errors_wait++; 3818387125fcSChris Mason continue; 3819387125fcSChris Mason } 3820e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3821ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3822387125fcSChris Mason continue; 3823387125fcSChris Mason 38244fc6441aSAnand Jain ret = wait_dev_flush(dev); 3825401b41e5SAnand Jain if (ret) { 3826401b41e5SAnand Jain dev->last_flush_error = ret; 382766b4993eSDavid Sterba btrfs_dev_stat_inc_and_print(dev, 382866b4993eSDavid Sterba BTRFS_DEV_STAT_FLUSH_ERRS); 38295af3e8ccSStefan Behrens errors_wait++; 3830387125fcSChris Mason } 3831401b41e5SAnand Jain } 3832401b41e5SAnand Jain 3833cea7c8bfSAnand Jain if (errors_wait) { 3834401b41e5SAnand Jain /* 3835401b41e5SAnand Jain * At some point we need the status of all disks 3836401b41e5SAnand Jain * to arrive at the volume status. So error checking 3837401b41e5SAnand Jain * is being pushed to a separate loop. 3838401b41e5SAnand Jain */ 3839d10b82feSQu Wenruo return check_barrier_error(info); 3840401b41e5SAnand Jain } 3841387125fcSChris Mason return 0; 3842387125fcSChris Mason } 3843387125fcSChris Mason 3844943c6e99SZhao Lei int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) 3845943c6e99SZhao Lei { 38468789f4feSZhao Lei int raid_type; 38478789f4feSZhao Lei int min_tolerated = INT_MAX; 3848943c6e99SZhao Lei 38498789f4feSZhao Lei if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || 38508789f4feSZhao Lei (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) 38518c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 38528789f4feSZhao Lei btrfs_raid_array[BTRFS_RAID_SINGLE]. 38538789f4feSZhao Lei tolerated_failures); 3854943c6e99SZhao Lei 38558789f4feSZhao Lei for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { 38568789f4feSZhao Lei if (raid_type == BTRFS_RAID_SINGLE) 38578789f4feSZhao Lei continue; 385841a6e891SAnand Jain if (!(flags & btrfs_raid_array[raid_type].bg_flag)) 38598789f4feSZhao Lei continue; 38608c3e3582SDavid Sterba min_tolerated = min_t(int, min_tolerated, 38618789f4feSZhao Lei btrfs_raid_array[raid_type]. 38628789f4feSZhao Lei tolerated_failures); 38638789f4feSZhao Lei } 3864943c6e99SZhao Lei 38658789f4feSZhao Lei if (min_tolerated == INT_MAX) { 3866ab8d0fc4SJeff Mahoney pr_warn("BTRFS: unknown raid flag: %llu", flags); 38678789f4feSZhao Lei min_tolerated = 0; 38688789f4feSZhao Lei } 38698789f4feSZhao Lei 38708789f4feSZhao Lei return min_tolerated; 3871943c6e99SZhao Lei } 3872943c6e99SZhao Lei 3873eece6a9cSDavid Sterba int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) 3874f2984462SChris Mason { 3875e5e9a520SChris Mason struct list_head *head; 3876f2984462SChris Mason struct btrfs_device *dev; 3877a061fc8dSChris Mason struct btrfs_super_block *sb; 3878f2984462SChris Mason struct btrfs_dev_item *dev_item; 3879f2984462SChris Mason int ret; 3880f2984462SChris Mason int do_barriers; 3881a236aed1SChris Mason int max_errors; 3882a236aed1SChris Mason int total_errors = 0; 3883a061fc8dSChris Mason u64 flags; 3884f2984462SChris Mason 38850b246afaSJeff Mahoney do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); 3886fed3b381SLiu Bo 3887fed3b381SLiu Bo /* 3888fed3b381SLiu Bo * max_mirrors == 0 indicates we're from commit_transaction, 3889fed3b381SLiu Bo * not from fsync where the tree roots in fs_info have not 3890fed3b381SLiu Bo * been consistent on disk. 3891fed3b381SLiu Bo */ 3892fed3b381SLiu Bo if (max_mirrors == 0) 38930b246afaSJeff Mahoney backup_super_roots(fs_info); 3894f2984462SChris Mason 38950b246afaSJeff Mahoney sb = fs_info->super_for_commit; 3896a061fc8dSChris Mason dev_item = &sb->dev_item; 3897e5e9a520SChris Mason 38980b246afaSJeff Mahoney mutex_lock(&fs_info->fs_devices->device_list_mutex); 38990b246afaSJeff Mahoney head = &fs_info->fs_devices->devices; 39000b246afaSJeff Mahoney max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; 3901387125fcSChris Mason 39025af3e8ccSStefan Behrens if (do_barriers) { 39030b246afaSJeff Mahoney ret = barrier_all_devices(fs_info); 39045af3e8ccSStefan Behrens if (ret) { 39055af3e8ccSStefan Behrens mutex_unlock( 39060b246afaSJeff Mahoney &fs_info->fs_devices->device_list_mutex); 39070b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, ret, 39085af3e8ccSStefan Behrens "errors while submitting device barriers."); 39095af3e8ccSStefan Behrens return ret; 39105af3e8ccSStefan Behrens } 39115af3e8ccSStefan Behrens } 3912387125fcSChris Mason 39131538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3914dfe25020SChris Mason if (!dev->bdev) { 3915dfe25020SChris Mason total_errors++; 3916dfe25020SChris Mason continue; 3917dfe25020SChris Mason } 3918e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3919ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3920dfe25020SChris Mason continue; 3921dfe25020SChris Mason 39222b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3923a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3924a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 39257df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 3926935e5cc9SMiao Xie dev->commit_total_bytes); 3927ce7213c7SMiao Xie btrfs_set_stack_device_bytes_used(dev_item, 3928ce7213c7SMiao Xie dev->commit_bytes_used); 3929a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3930a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3931a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3932a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 39337239ff4bSNikolay Borisov memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, 39347239ff4bSNikolay Borisov BTRFS_FSID_SIZE); 3935a512bbf8SYan Zheng 3936a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3937a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3938f2984462SChris Mason 393975cb857dSQu Wenruo ret = btrfs_validate_write_super(fs_info, sb); 394075cb857dSQu Wenruo if (ret < 0) { 394175cb857dSQu Wenruo mutex_unlock(&fs_info->fs_devices->device_list_mutex); 394275cb857dSQu Wenruo btrfs_handle_fs_error(fs_info, -EUCLEAN, 394375cb857dSQu Wenruo "unexpected superblock corruption detected"); 394475cb857dSQu Wenruo return -EUCLEAN; 394575cb857dSQu Wenruo } 394675cb857dSQu Wenruo 3947abbb3b8eSDavid Sterba ret = write_dev_supers(dev, sb, max_mirrors); 3948a236aed1SChris Mason if (ret) 3949a236aed1SChris Mason total_errors++; 3950f2984462SChris Mason } 3951a236aed1SChris Mason if (total_errors > max_errors) { 39520b246afaSJeff Mahoney btrfs_err(fs_info, "%d errors while writing supers", 3953d397712bSChris Mason total_errors); 39540b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 395579787eaaSJeff Mahoney 39569d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 39570b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 39580b246afaSJeff Mahoney "%d errors while writing supers", 39590b246afaSJeff Mahoney total_errors); 39609d565ba4SStefan Behrens return -EIO; 3961a236aed1SChris Mason } 3962f2984462SChris Mason 3963a512bbf8SYan Zheng total_errors = 0; 39641538e6c5SDavid Sterba list_for_each_entry(dev, head, dev_list) { 3965dfe25020SChris Mason if (!dev->bdev) 3966dfe25020SChris Mason continue; 3967e12c9621SAnand Jain if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || 3968ebbede42SAnand Jain !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) 3969dfe25020SChris Mason continue; 3970dfe25020SChris Mason 3971abbb3b8eSDavid Sterba ret = wait_dev_supers(dev, max_mirrors); 3972a512bbf8SYan Zheng if (ret) 39731259ab75SChris Mason total_errors++; 3974f2984462SChris Mason } 39750b246afaSJeff Mahoney mutex_unlock(&fs_info->fs_devices->device_list_mutex); 3976a236aed1SChris Mason if (total_errors > max_errors) { 39770b246afaSJeff Mahoney btrfs_handle_fs_error(fs_info, -EIO, 39780b246afaSJeff Mahoney "%d errors while writing supers", 39790b246afaSJeff Mahoney total_errors); 398079787eaaSJeff Mahoney return -EIO; 3981a236aed1SChris Mason } 3982f2984462SChris Mason return 0; 3983f2984462SChris Mason } 3984f2984462SChris Mason 3985cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3986cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3987cb517eabSMiao Xie struct btrfs_root *root) 39882619ba1fSChris Mason { 39894785e24fSJosef Bacik bool drop_ref = false; 39904785e24fSJosef Bacik 39914df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 39922619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 39932619ba1fSChris Mason (unsigned long)root->root_key.objectid); 3994af01d2e5SJosef Bacik if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 39954785e24fSJosef Bacik drop_ref = true; 39964df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 399776dda93cSYan, Zheng 39981c1ea4f7SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 3999ef67963dSJosef Bacik ASSERT(root->log_root == NULL); 40001c1ea4f7SLiu Bo if (root->reloc_root) { 400100246528SJosef Bacik btrfs_put_root(root->reloc_root); 40021c1ea4f7SLiu Bo root->reloc_root = NULL; 40031c1ea4f7SLiu Bo } 40041c1ea4f7SLiu Bo } 40053321719eSLiu Bo 40064785e24fSJosef Bacik if (drop_ref) 400700246528SJosef Bacik btrfs_put_root(root); 40082619ba1fSChris Mason } 40092619ba1fSChris Mason 4010c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 4011c146afadSYan Zheng { 4012c146afadSYan Zheng u64 root_objectid = 0; 4013c146afadSYan Zheng struct btrfs_root *gang[8]; 401465d33fd7SQu Wenruo int i = 0; 401565d33fd7SQu Wenruo int err = 0; 401665d33fd7SQu Wenruo unsigned int ret = 0; 4017c146afadSYan Zheng 4018c146afadSYan Zheng while (1) { 4019efc34534SJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4020c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4021c146afadSYan Zheng (void **)gang, root_objectid, 4022c146afadSYan Zheng ARRAY_SIZE(gang)); 402365d33fd7SQu Wenruo if (!ret) { 4024efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4025c146afadSYan Zheng break; 402665d33fd7SQu Wenruo } 40275d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 402866b4ffd1SJosef Bacik 402965d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 403065d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 403165d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 403265d33fd7SQu Wenruo gang[i] = NULL; 403365d33fd7SQu Wenruo continue; 403465d33fd7SQu Wenruo } 403565d33fd7SQu Wenruo /* grab all the search result for later use */ 403600246528SJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 403765d33fd7SQu Wenruo } 4038efc34534SJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 403965d33fd7SQu Wenruo 404065d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 404165d33fd7SQu Wenruo if (!gang[i]) 404265d33fd7SQu Wenruo continue; 4043c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 404466b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 404566b4ffd1SJosef Bacik if (err) 404665d33fd7SQu Wenruo break; 404700246528SJosef Bacik btrfs_put_root(gang[i]); 4048c146afadSYan Zheng } 4049c146afadSYan Zheng root_objectid++; 4050c146afadSYan Zheng } 405165d33fd7SQu Wenruo 405265d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 405365d33fd7SQu Wenruo for (; i < ret; i++) { 405465d33fd7SQu Wenruo if (gang[i]) 405500246528SJosef Bacik btrfs_put_root(gang[i]); 405665d33fd7SQu Wenruo } 405765d33fd7SQu Wenruo return err; 4058c146afadSYan Zheng } 4059c146afadSYan Zheng 40606bccf3abSJeff Mahoney int btrfs_commit_super(struct btrfs_fs_info *fs_info) 4061c146afadSYan Zheng { 40626bccf3abSJeff Mahoney struct btrfs_root *root = fs_info->tree_root; 4063c146afadSYan Zheng struct btrfs_trans_handle *trans; 4064c146afadSYan Zheng 40650b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 40662ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 40670b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 40680b246afaSJeff Mahoney wake_up_process(fs_info->cleaner_kthread); 4069c71bf099SYan, Zheng 4070c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 40710b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 40720b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4073c71bf099SYan, Zheng 40747a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 40753612b495STsutomu Itoh if (IS_ERR(trans)) 40763612b495STsutomu Itoh return PTR_ERR(trans); 40773a45bb20SJeff Mahoney return btrfs_commit_transaction(trans); 4078c146afadSYan Zheng } 4079c146afadSYan Zheng 4080b105e927SDavid Sterba void __cold close_ctree(struct btrfs_fs_info *fs_info) 4081eb60ceacSChris Mason { 4082c146afadSYan Zheng int ret; 4083e089f05cSChris Mason 4084afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 4085d6fd0ae2SOmar Sandoval /* 4086d6fd0ae2SOmar Sandoval * We don't want the cleaner to start new transactions, add more delayed 4087d6fd0ae2SOmar Sandoval * iputs, etc. while we're closing. We can't use kthread_stop() yet 4088d6fd0ae2SOmar Sandoval * because that frees the task_struct, and the transaction kthread might 4089d6fd0ae2SOmar Sandoval * still try to wake up the cleaner. 4090d6fd0ae2SOmar Sandoval */ 4091d6fd0ae2SOmar Sandoval kthread_park(fs_info->cleaner_kthread); 4092a2135011SChris Mason 40937343dd61SJustin Maggard /* wait for the qgroup rescan worker to stop */ 4094d06f23d6SJeff Mahoney btrfs_qgroup_wait_for_completion(fs_info, false); 40957343dd61SJustin Maggard 4096803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 4097803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 4098803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 4099803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 4100803b2f54SStefan Behrens 4101837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 4102aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 4103837d5b6eSIlya Dryomov 41048dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 41058dabb742SStefan Behrens 4106aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 41074cb5300bSChris Mason 41084cb5300bSChris Mason /* wait for any defraggers to finish */ 41094cb5300bSChris Mason wait_event(fs_info->transaction_wait, 41104cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 41114cb5300bSChris Mason 41124cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 411326176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 41144cb5300bSChris Mason 411521c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 411657056740SJosef Bacik cancel_work_sync(&fs_info->async_data_reclaim_work); 411721c7e756SMiao Xie 4118b0643e59SDennis Zhou /* Cancel or finish ongoing discard work */ 4119b0643e59SDennis Zhou btrfs_discard_cleanup(fs_info); 4120b0643e59SDennis Zhou 4121bc98a42cSDavid Howells if (!sb_rdonly(fs_info->sb)) { 4122e44163e1SJeff Mahoney /* 4123d6fd0ae2SOmar Sandoval * The cleaner kthread is stopped, so do one final pass over 4124d6fd0ae2SOmar Sandoval * unused block groups. 4125e44163e1SJeff Mahoney */ 41260b246afaSJeff Mahoney btrfs_delete_unused_bgs(fs_info); 4127e44163e1SJeff Mahoney 4128f0cc2cd7SFilipe Manana /* 4129f0cc2cd7SFilipe Manana * There might be existing delayed inode workers still running 4130f0cc2cd7SFilipe Manana * and holding an empty delayed inode item. We must wait for 4131f0cc2cd7SFilipe Manana * them to complete first because they can create a transaction. 4132f0cc2cd7SFilipe Manana * This happens when someone calls btrfs_balance_delayed_items() 4133f0cc2cd7SFilipe Manana * and then a transaction commit runs the same delayed nodes 4134f0cc2cd7SFilipe Manana * before any delayed worker has done something with the nodes. 4135f0cc2cd7SFilipe Manana * We must wait for any worker here and not at transaction 4136f0cc2cd7SFilipe Manana * commit time since that could cause a deadlock. 4137f0cc2cd7SFilipe Manana * This is a very rare case. 4138f0cc2cd7SFilipe Manana */ 4139f0cc2cd7SFilipe Manana btrfs_flush_workqueue(fs_info->delayed_workers); 4140f0cc2cd7SFilipe Manana 41416bccf3abSJeff Mahoney ret = btrfs_commit_super(fs_info); 4142d397712bSChris Mason if (ret) 414304892340SEric Sandeen btrfs_err(fs_info, "commit super ret %d", ret); 4144c146afadSYan Zheng } 4145ed2ff2cbSChris Mason 4146af722733SLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state) || 4147af722733SLiu Bo test_bit(BTRFS_FS_STATE_TRANS_ABORTED, &fs_info->fs_state)) 41482ff7e61eSJeff Mahoney btrfs_error_commit_super(fs_info); 4149acce952bSliubo 4150e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 4151e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 41528929ecfaSYan, Zheng 4153e187831eSJosef Bacik ASSERT(list_empty(&fs_info->delayed_iputs)); 4154afcdd129SJosef Bacik set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); 4155f25784b3SYan Zheng 41565958253cSQu Wenruo if (btrfs_check_quota_leak(fs_info)) { 41575958253cSQu Wenruo WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 41585958253cSQu Wenruo btrfs_err(fs_info, "qgroup reserved space leaked"); 41595958253cSQu Wenruo } 41605958253cSQu Wenruo 416104892340SEric Sandeen btrfs_free_qgroup_config(fs_info); 4162fe816d0fSNikolay Borisov ASSERT(list_empty(&fs_info->delalloc_roots)); 4163bcef60f2SArne Jansen 4164963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 416504892340SEric Sandeen btrfs_info(fs_info, "at unmount delalloc count %lld", 4166963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 4167b0c68f8bSChris Mason } 416831153d81SYan Zheng 41694297ff84SJosef Bacik if (percpu_counter_sum(&fs_info->dio_bytes)) 41704297ff84SJosef Bacik btrfs_info(fs_info, "at unmount dio bytes count %lld", 41714297ff84SJosef Bacik percpu_counter_sum(&fs_info->dio_bytes)); 41724297ff84SJosef Bacik 41736618a59bSAnand Jain btrfs_sysfs_remove_mounted(fs_info); 4174b7c35e81SAnand Jain btrfs_sysfs_remove_fsid(fs_info->fs_devices); 41755ac1d209SJeff Mahoney 41761a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 41771a4319ccSLiu Bo 4178de348ee0SWang Shilong /* 4179de348ee0SWang Shilong * we must make sure there is not any read request to 4180de348ee0SWang Shilong * submit after we stopping all workers. 4181de348ee0SWang Shilong */ 4182de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 418396192499SJosef Bacik btrfs_stop_all_workers(fs_info); 418496192499SJosef Bacik 4185afcdd129SJosef Bacik clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 41864273eaffSAnand Jain free_root_pointers(fs_info, true); 41878c38938cSJosef Bacik btrfs_free_fs_roots(fs_info); 41889ad6b7bcSChris Mason 41894e19443dSJosef Bacik /* 41904e19443dSJosef Bacik * We must free the block groups after dropping the fs_roots as we could 41914e19443dSJosef Bacik * have had an IO error and have left over tree log blocks that aren't 41924e19443dSJosef Bacik * cleaned up until the fs roots are freed. This makes the block group 41934e19443dSJosef Bacik * accounting appear to be wrong because there's pending reserved bytes, 41944e19443dSJosef Bacik * so make sure we do the block group cleanup afterwards. 41954e19443dSJosef Bacik */ 41964e19443dSJosef Bacik btrfs_free_block_groups(fs_info); 41974e19443dSJosef Bacik 419813e6c37bSJosef Bacik iput(fs_info->btree_inode); 4199d6bfde87SChris Mason 420021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 42010b246afaSJeff Mahoney if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) 42022ff7e61eSJeff Mahoney btrfsic_unmount(fs_info->fs_devices); 420321adbd5cSStefan Behrens #endif 420421adbd5cSStefan Behrens 42050b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 420668c94e55SNikolay Borisov btrfs_close_devices(fs_info->fs_devices); 4207eb60ceacSChris Mason } 4208eb60ceacSChris Mason 4209b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4210b9fab919SChris Mason int atomic) 4211ccd467d6SChris Mason { 42121259ab75SChris Mason int ret; 4213727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 42141259ab75SChris Mason 42150b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 42161259ab75SChris Mason if (!ret) 42171259ab75SChris Mason return ret; 42181259ab75SChris Mason 42191259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 4220b9fab919SChris Mason parent_transid, atomic); 4221b9fab919SChris Mason if (ret == -EAGAIN) 4222b9fab919SChris Mason return ret; 42231259ab75SChris Mason return !ret; 42245f39d397SChris Mason } 42256702ed49SChris Mason 42265f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 42275f39d397SChris Mason { 42282f4d60dfSQu Wenruo struct btrfs_fs_info *fs_info = buf->fs_info; 42295f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 4230b9473439SChris Mason int was_dirty; 4231b4ce94deSChris Mason 423206ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 423306ea65a3SJosef Bacik /* 423406ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 423552042d8eSAndrea Gelmini * enabled. Normal people shouldn't be using unmapped buffers as dirty 423606ea65a3SJosef Bacik * outside of the sanity tests. 423706ea65a3SJosef Bacik */ 4238b0132a3bSNikolay Borisov if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) 423906ea65a3SJosef Bacik return; 424006ea65a3SJosef Bacik #endif 4241b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 42420b246afaSJeff Mahoney if (transid != fs_info->generation) 42435d163e0eSJeff Mahoney WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 42440b246afaSJeff Mahoney buf->start, transid, fs_info->generation); 42450b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 4246e2d84521SMiao Xie if (!was_dirty) 4247104b4e51SNikolay Borisov percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, 4248e2d84521SMiao Xie buf->len, 42490b246afaSJeff Mahoney fs_info->dirty_metadata_batch); 42501f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 425169fc6cbbSQu Wenruo /* 425269fc6cbbSQu Wenruo * Since btrfs_mark_buffer_dirty() can be called with item pointer set 425369fc6cbbSQu Wenruo * but item data not updated. 425469fc6cbbSQu Wenruo * So here we should only check item pointers, not item data. 425569fc6cbbSQu Wenruo */ 425669fc6cbbSQu Wenruo if (btrfs_header_level(buf) == 0 && 4257cfdaad5eSDavid Sterba btrfs_check_leaf_relaxed(buf)) { 4258a4f78750SDavid Sterba btrfs_print_leaf(buf); 42591f21ef0aSFilipe Manana ASSERT(0); 42601f21ef0aSFilipe Manana } 42611f21ef0aSFilipe Manana #endif 4262eb60ceacSChris Mason } 4263eb60ceacSChris Mason 42642ff7e61eSJeff Mahoney static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, 4265b53d3f5dSLiu Bo int flush_delayed) 426635b7e476SChris Mason { 4267188de649SChris Mason /* 4268188de649SChris Mason * looks as though older kernels can get into trouble with 4269188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 4270188de649SChris Mason */ 4271e2d84521SMiao Xie int ret; 4272d6bfde87SChris Mason 42736933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 4274d6bfde87SChris Mason return; 4275d6bfde87SChris Mason 4276b53d3f5dSLiu Bo if (flush_delayed) 42772ff7e61eSJeff Mahoney btrfs_balance_delayed_items(fs_info); 427816cdcec7SMiao Xie 4279d814a491SEthan Lien ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, 4280d814a491SEthan Lien BTRFS_DIRTY_METADATA_THRESH, 4281d814a491SEthan Lien fs_info->dirty_metadata_batch); 4282e2d84521SMiao Xie if (ret > 0) { 42830b246afaSJeff Mahoney balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); 428416cdcec7SMiao Xie } 428516cdcec7SMiao Xie } 428616cdcec7SMiao Xie 42872ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) 428816cdcec7SMiao Xie { 42892ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 1); 429035b7e476SChris Mason } 4291b53d3f5dSLiu Bo 42922ff7e61eSJeff Mahoney void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) 4293b53d3f5dSLiu Bo { 42942ff7e61eSJeff Mahoney __btrfs_btree_balance_dirty(fs_info, 0); 4295d6bfde87SChris Mason } 42966b80053dSChris Mason 4297581c1760SQu Wenruo int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid, int level, 4298581c1760SQu Wenruo struct btrfs_key *first_key) 42996b80053dSChris Mason { 43005ab12d1fSDavid Sterba return btree_read_extent_buffer_pages(buf, parent_transid, 4301581c1760SQu Wenruo level, first_key); 43026b80053dSChris Mason } 43030da5468fSChris Mason 43042ff7e61eSJeff Mahoney static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) 4305acce952bSliubo { 4306fe816d0fSNikolay Borisov /* cleanup FS via transaction */ 4307fe816d0fSNikolay Borisov btrfs_cleanup_transaction(fs_info); 4308fe816d0fSNikolay Borisov 43090b246afaSJeff Mahoney mutex_lock(&fs_info->cleaner_mutex); 43102ff7e61eSJeff Mahoney btrfs_run_delayed_iputs(fs_info); 43110b246afaSJeff Mahoney mutex_unlock(&fs_info->cleaner_mutex); 4312acce952bSliubo 43130b246afaSJeff Mahoney down_write(&fs_info->cleanup_work_sem); 43140b246afaSJeff Mahoney up_write(&fs_info->cleanup_work_sem); 4315acce952bSliubo } 4316acce952bSliubo 4317ef67963dSJosef Bacik static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) 4318ef67963dSJosef Bacik { 4319ef67963dSJosef Bacik struct btrfs_root *gang[8]; 4320ef67963dSJosef Bacik u64 root_objectid = 0; 4321ef67963dSJosef Bacik int ret; 4322ef67963dSJosef Bacik 4323ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4324ef67963dSJosef Bacik while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 4325ef67963dSJosef Bacik (void **)gang, root_objectid, 4326ef67963dSJosef Bacik ARRAY_SIZE(gang))) != 0) { 4327ef67963dSJosef Bacik int i; 4328ef67963dSJosef Bacik 4329ef67963dSJosef Bacik for (i = 0; i < ret; i++) 4330ef67963dSJosef Bacik gang[i] = btrfs_grab_root(gang[i]); 4331ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4332ef67963dSJosef Bacik 4333ef67963dSJosef Bacik for (i = 0; i < ret; i++) { 4334ef67963dSJosef Bacik if (!gang[i]) 4335ef67963dSJosef Bacik continue; 4336ef67963dSJosef Bacik root_objectid = gang[i]->root_key.objectid; 4337ef67963dSJosef Bacik btrfs_free_log(NULL, gang[i]); 4338ef67963dSJosef Bacik btrfs_put_root(gang[i]); 4339ef67963dSJosef Bacik } 4340ef67963dSJosef Bacik root_objectid++; 4341ef67963dSJosef Bacik spin_lock(&fs_info->fs_roots_radix_lock); 4342ef67963dSJosef Bacik } 4343ef67963dSJosef Bacik spin_unlock(&fs_info->fs_roots_radix_lock); 4344ef67963dSJosef Bacik btrfs_free_log_root_tree(NULL, fs_info); 4345ef67963dSJosef Bacik } 4346ef67963dSJosef Bacik 4347143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4348acce952bSliubo { 4349acce952bSliubo struct btrfs_ordered_extent *ordered; 4350acce952bSliubo 4351199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 4352779880efSJosef Bacik /* 4353779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 4354779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 4355779880efSJosef Bacik */ 4356199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 4357779880efSJosef Bacik root_extent_list) 4358779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 4359199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 4360199c2a9cSMiao Xie } 4361199c2a9cSMiao Xie 4362199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 4363199c2a9cSMiao Xie { 4364199c2a9cSMiao Xie struct btrfs_root *root; 4365199c2a9cSMiao Xie struct list_head splice; 4366199c2a9cSMiao Xie 4367199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 4368199c2a9cSMiao Xie 4369199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 4370199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 4371199c2a9cSMiao Xie while (!list_empty(&splice)) { 4372199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4373199c2a9cSMiao Xie ordered_root); 43741de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 43751de2cfdeSJosef Bacik &fs_info->ordered_roots); 4376199c2a9cSMiao Xie 43772a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 4378199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 4379199c2a9cSMiao Xie 43802a85d9caSLiu Bo cond_resched(); 43812a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 4382199c2a9cSMiao Xie } 4383199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 438474d5d229SJosef Bacik 438574d5d229SJosef Bacik /* 438674d5d229SJosef Bacik * We need this here because if we've been flipped read-only we won't 438774d5d229SJosef Bacik * get sync() from the umount, so we need to make sure any ordered 438874d5d229SJosef Bacik * extents that haven't had their dirty pages IO start writeout yet 438974d5d229SJosef Bacik * actually get run and error out properly. 439074d5d229SJosef Bacik */ 439174d5d229SJosef Bacik btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 4392acce952bSliubo } 4393acce952bSliubo 439435a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 43952ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4396acce952bSliubo { 4397acce952bSliubo struct rb_node *node; 4398acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 4399acce952bSliubo struct btrfs_delayed_ref_node *ref; 4400acce952bSliubo int ret = 0; 4401acce952bSliubo 4402acce952bSliubo delayed_refs = &trans->delayed_refs; 4403acce952bSliubo 4404acce952bSliubo spin_lock(&delayed_refs->lock); 4405d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 4406cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 4407b79ce3ddSDavid Sterba btrfs_debug(fs_info, "delayed_refs has NO entry"); 4408acce952bSliubo return ret; 4409acce952bSliubo } 4410acce952bSliubo 44115c9d028bSLiu Bo while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { 4412d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 44130e0adbcfSJosef Bacik struct rb_node *n; 4414e78417d1SJosef Bacik bool pin_bytes = false; 4415acce952bSliubo 4416d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 4417d7df2c79SJosef Bacik href_node); 44183069bd26SJosef Bacik if (btrfs_delayed_ref_lock(delayed_refs, head)) 4419b939d1abSJosef Bacik continue; 44203069bd26SJosef Bacik 4421d7df2c79SJosef Bacik spin_lock(&head->lock); 4422e3d03965SLiu Bo while ((n = rb_first_cached(&head->ref_tree)) != NULL) { 44230e0adbcfSJosef Bacik ref = rb_entry(n, struct btrfs_delayed_ref_node, 44240e0adbcfSJosef Bacik ref_node); 4425d7df2c79SJosef Bacik ref->in_tree = 0; 4426e3d03965SLiu Bo rb_erase_cached(&ref->ref_node, &head->ref_tree); 44270e0adbcfSJosef Bacik RB_CLEAR_NODE(&ref->ref_node); 44281d57ee94SWang Xiaoguang if (!list_empty(&ref->add_list)) 44291d57ee94SWang Xiaoguang list_del(&ref->add_list); 4430d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 4431d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 4432d7df2c79SJosef Bacik } 443354067ae9SJosef Bacik if (head->must_insert_reserved) 4434e78417d1SJosef Bacik pin_bytes = true; 443578a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 4436fa781ceaSJosef Bacik btrfs_delete_ref_head(delayed_refs, head); 4437d7df2c79SJosef Bacik spin_unlock(&head->lock); 4438acce952bSliubo spin_unlock(&delayed_refs->lock); 4439e78417d1SJosef Bacik mutex_unlock(&head->mutex); 4440acce952bSliubo 4441f603bb94SNikolay Borisov if (pin_bytes) { 4442f603bb94SNikolay Borisov struct btrfs_block_group *cache; 4443f603bb94SNikolay Borisov 4444f603bb94SNikolay Borisov cache = btrfs_lookup_block_group(fs_info, head->bytenr); 4445f603bb94SNikolay Borisov BUG_ON(!cache); 4446f603bb94SNikolay Borisov 4447f603bb94SNikolay Borisov spin_lock(&cache->space_info->lock); 4448f603bb94SNikolay Borisov spin_lock(&cache->lock); 4449f603bb94SNikolay Borisov cache->pinned += head->num_bytes; 4450f603bb94SNikolay Borisov btrfs_space_info_update_bytes_pinned(fs_info, 4451f603bb94SNikolay Borisov cache->space_info, head->num_bytes); 4452f603bb94SNikolay Borisov cache->reserved -= head->num_bytes; 4453f603bb94SNikolay Borisov cache->space_info->bytes_reserved -= head->num_bytes; 4454f603bb94SNikolay Borisov spin_unlock(&cache->lock); 4455f603bb94SNikolay Borisov spin_unlock(&cache->space_info->lock); 4456f603bb94SNikolay Borisov percpu_counter_add_batch( 4457f603bb94SNikolay Borisov &cache->space_info->total_bytes_pinned, 4458f603bb94SNikolay Borisov head->num_bytes, BTRFS_TOTAL_BYTES_PINNED_BATCH); 4459f603bb94SNikolay Borisov 4460f603bb94SNikolay Borisov btrfs_put_block_group(cache); 4461f603bb94SNikolay Borisov 4462f603bb94SNikolay Borisov btrfs_error_unpin_extent_range(fs_info, head->bytenr, 4463f603bb94SNikolay Borisov head->bytenr + head->num_bytes - 1); 4464f603bb94SNikolay Borisov } 446531890da0SJosef Bacik btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 4466d278850eSJosef Bacik btrfs_put_delayed_ref_head(head); 4467acce952bSliubo cond_resched(); 4468acce952bSliubo spin_lock(&delayed_refs->lock); 4469acce952bSliubo } 447081f7eb00SJeff Mahoney btrfs_qgroup_destroy_extent_records(trans); 4471acce952bSliubo 4472acce952bSliubo spin_unlock(&delayed_refs->lock); 4473acce952bSliubo 4474acce952bSliubo return ret; 4475acce952bSliubo } 4476acce952bSliubo 4477143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 4478acce952bSliubo { 4479acce952bSliubo struct btrfs_inode *btrfs_inode; 4480acce952bSliubo struct list_head splice; 4481acce952bSliubo 4482acce952bSliubo INIT_LIST_HEAD(&splice); 4483acce952bSliubo 4484eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4485eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 4486acce952bSliubo 4487acce952bSliubo while (!list_empty(&splice)) { 4488fe816d0fSNikolay Borisov struct inode *inode = NULL; 4489eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 4490acce952bSliubo delalloc_inodes); 4491fe816d0fSNikolay Borisov __btrfs_del_delalloc_inode(root, btrfs_inode); 4492eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4493acce952bSliubo 4494fe816d0fSNikolay Borisov /* 4495fe816d0fSNikolay Borisov * Make sure we get a live inode and that it'll not disappear 4496fe816d0fSNikolay Borisov * meanwhile. 4497fe816d0fSNikolay Borisov */ 4498fe816d0fSNikolay Borisov inode = igrab(&btrfs_inode->vfs_inode); 4499fe816d0fSNikolay Borisov if (inode) { 4500fe816d0fSNikolay Borisov invalidate_inode_pages2(inode->i_mapping); 4501fe816d0fSNikolay Borisov iput(inode); 4502fe816d0fSNikolay Borisov } 4503eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 4504acce952bSliubo } 4505eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 4506eb73c1b7SMiao Xie } 4507eb73c1b7SMiao Xie 4508eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 4509eb73c1b7SMiao Xie { 4510eb73c1b7SMiao Xie struct btrfs_root *root; 4511eb73c1b7SMiao Xie struct list_head splice; 4512eb73c1b7SMiao Xie 4513eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 4514eb73c1b7SMiao Xie 4515eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4516eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 4517eb73c1b7SMiao Xie while (!list_empty(&splice)) { 4518eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 4519eb73c1b7SMiao Xie delalloc_root); 452000246528SJosef Bacik root = btrfs_grab_root(root); 4521eb73c1b7SMiao Xie BUG_ON(!root); 4522eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4523eb73c1b7SMiao Xie 4524eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 452500246528SJosef Bacik btrfs_put_root(root); 4526eb73c1b7SMiao Xie 4527eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 4528eb73c1b7SMiao Xie } 4529eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 4530acce952bSliubo } 4531acce952bSliubo 45322ff7e61eSJeff Mahoney static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, 4533acce952bSliubo struct extent_io_tree *dirty_pages, 4534acce952bSliubo int mark) 4535acce952bSliubo { 4536acce952bSliubo int ret; 4537acce952bSliubo struct extent_buffer *eb; 4538acce952bSliubo u64 start = 0; 4539acce952bSliubo u64 end; 4540acce952bSliubo 4541acce952bSliubo while (1) { 4542acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 4543e6138876SJosef Bacik mark, NULL); 4544acce952bSliubo if (ret) 4545acce952bSliubo break; 4546acce952bSliubo 454791166212SDavid Sterba clear_extent_bits(dirty_pages, start, end, mark); 4548acce952bSliubo while (start <= end) { 45490b246afaSJeff Mahoney eb = find_extent_buffer(fs_info, start); 45500b246afaSJeff Mahoney start += fs_info->nodesize; 4551fd8b2b61SJosef Bacik if (!eb) 4552acce952bSliubo continue; 4553fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4554acce952bSliubo 4555fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4556fd8b2b61SJosef Bacik &eb->bflags)) 4557fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4558fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4559acce952bSliubo } 4560acce952bSliubo } 4561acce952bSliubo 4562acce952bSliubo return ret; 4563acce952bSliubo } 4564acce952bSliubo 45652ff7e61eSJeff Mahoney static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, 4566fe119a6eSNikolay Borisov struct extent_io_tree *unpin) 4567acce952bSliubo { 4568acce952bSliubo u64 start; 4569acce952bSliubo u64 end; 4570acce952bSliubo int ret; 4571acce952bSliubo 4572acce952bSliubo while (1) { 45730e6ec385SFilipe Manana struct extent_state *cached_state = NULL; 45740e6ec385SFilipe Manana 4575fcd5e742SLu Fengqi /* 4576fcd5e742SLu Fengqi * The btrfs_finish_extent_commit() may get the same range as 4577fcd5e742SLu Fengqi * ours between find_first_extent_bit and clear_extent_dirty. 4578fcd5e742SLu Fengqi * Hence, hold the unused_bg_unpin_mutex to avoid double unpin 4579fcd5e742SLu Fengqi * the same extent range. 4580fcd5e742SLu Fengqi */ 4581fcd5e742SLu Fengqi mutex_lock(&fs_info->unused_bg_unpin_mutex); 4582acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 45830e6ec385SFilipe Manana EXTENT_DIRTY, &cached_state); 4584fcd5e742SLu Fengqi if (ret) { 4585fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4586acce952bSliubo break; 4587fcd5e742SLu Fengqi } 4588acce952bSliubo 45890e6ec385SFilipe Manana clear_extent_dirty(unpin, start, end, &cached_state); 45900e6ec385SFilipe Manana free_extent_state(cached_state); 45912ff7e61eSJeff Mahoney btrfs_error_unpin_extent_range(fs_info, start, end); 4592fcd5e742SLu Fengqi mutex_unlock(&fs_info->unused_bg_unpin_mutex); 4593acce952bSliubo cond_resched(); 4594acce952bSliubo } 4595acce952bSliubo 4596acce952bSliubo return 0; 4597acce952bSliubo } 4598acce952bSliubo 459932da5386SDavid Sterba static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) 4600c79a1751SLiu Bo { 4601c79a1751SLiu Bo struct inode *inode; 4602c79a1751SLiu Bo 4603c79a1751SLiu Bo inode = cache->io_ctl.inode; 4604c79a1751SLiu Bo if (inode) { 4605c79a1751SLiu Bo invalidate_inode_pages2(inode->i_mapping); 4606c79a1751SLiu Bo BTRFS_I(inode)->generation = 0; 4607c79a1751SLiu Bo cache->io_ctl.inode = NULL; 4608c79a1751SLiu Bo iput(inode); 4609c79a1751SLiu Bo } 4610bbc37d6eSFilipe Manana ASSERT(cache->io_ctl.pages == NULL); 4611c79a1751SLiu Bo btrfs_put_block_group(cache); 4612c79a1751SLiu Bo } 4613c79a1751SLiu Bo 4614c79a1751SLiu Bo void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 46152ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 4616c79a1751SLiu Bo { 461732da5386SDavid Sterba struct btrfs_block_group *cache; 4618c79a1751SLiu Bo 4619c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4620c79a1751SLiu Bo while (!list_empty(&cur_trans->dirty_bgs)) { 4621c79a1751SLiu Bo cache = list_first_entry(&cur_trans->dirty_bgs, 462232da5386SDavid Sterba struct btrfs_block_group, 4623c79a1751SLiu Bo dirty_list); 4624c79a1751SLiu Bo 4625c79a1751SLiu Bo if (!list_empty(&cache->io_list)) { 4626c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4627c79a1751SLiu Bo list_del_init(&cache->io_list); 4628c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4629c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4630c79a1751SLiu Bo } 4631c79a1751SLiu Bo 4632c79a1751SLiu Bo list_del_init(&cache->dirty_list); 4633c79a1751SLiu Bo spin_lock(&cache->lock); 4634c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4635c79a1751SLiu Bo spin_unlock(&cache->lock); 4636c79a1751SLiu Bo 4637c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4638c79a1751SLiu Bo btrfs_put_block_group(cache); 4639ba2c4d4eSJosef Bacik btrfs_delayed_refs_rsv_release(fs_info, 1); 4640c79a1751SLiu Bo spin_lock(&cur_trans->dirty_bgs_lock); 4641c79a1751SLiu Bo } 4642c79a1751SLiu Bo spin_unlock(&cur_trans->dirty_bgs_lock); 4643c79a1751SLiu Bo 464445ae2c18SNikolay Borisov /* 464545ae2c18SNikolay Borisov * Refer to the definition of io_bgs member for details why it's safe 464645ae2c18SNikolay Borisov * to use it without any locking 464745ae2c18SNikolay Borisov */ 4648c79a1751SLiu Bo while (!list_empty(&cur_trans->io_bgs)) { 4649c79a1751SLiu Bo cache = list_first_entry(&cur_trans->io_bgs, 465032da5386SDavid Sterba struct btrfs_block_group, 4651c79a1751SLiu Bo io_list); 4652c79a1751SLiu Bo 4653c79a1751SLiu Bo list_del_init(&cache->io_list); 4654c79a1751SLiu Bo spin_lock(&cache->lock); 4655c79a1751SLiu Bo cache->disk_cache_state = BTRFS_DC_ERROR; 4656c79a1751SLiu Bo spin_unlock(&cache->lock); 4657c79a1751SLiu Bo btrfs_cleanup_bg_io(cache); 4658c79a1751SLiu Bo } 4659c79a1751SLiu Bo } 4660c79a1751SLiu Bo 466149b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 46622ff7e61eSJeff Mahoney struct btrfs_fs_info *fs_info) 466349b25e05SJeff Mahoney { 4664bbbf7243SNikolay Borisov struct btrfs_device *dev, *tmp; 4665bbbf7243SNikolay Borisov 46662ff7e61eSJeff Mahoney btrfs_cleanup_dirty_bgs(cur_trans, fs_info); 4667c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->dirty_bgs)); 4668c79a1751SLiu Bo ASSERT(list_empty(&cur_trans->io_bgs)); 4669c79a1751SLiu Bo 4670bbbf7243SNikolay Borisov list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, 4671bbbf7243SNikolay Borisov post_commit_list) { 4672bbbf7243SNikolay Borisov list_del_init(&dev->post_commit_list); 4673bbbf7243SNikolay Borisov } 4674bbbf7243SNikolay Borisov 46752ff7e61eSJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, fs_info); 467649b25e05SJeff Mahoney 46774a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 46780b246afaSJeff Mahoney wake_up(&fs_info->transaction_blocked_wait); 467949b25e05SJeff Mahoney 46804a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 46810b246afaSJeff Mahoney wake_up(&fs_info->transaction_wait); 468249b25e05SJeff Mahoney 4683ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 468467cde344SMiao Xie 46852ff7e61eSJeff Mahoney btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, 468649b25e05SJeff Mahoney EXTENT_DIRTY); 4687fe119a6eSNikolay Borisov btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); 468849b25e05SJeff Mahoney 46894a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 46904a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 469149b25e05SJeff Mahoney } 469249b25e05SJeff Mahoney 46932ff7e61eSJeff Mahoney static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) 4694acce952bSliubo { 4695acce952bSliubo struct btrfs_transaction *t; 4696acce952bSliubo 46970b246afaSJeff Mahoney mutex_lock(&fs_info->transaction_kthread_mutex); 4698acce952bSliubo 46990b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 47000b246afaSJeff Mahoney while (!list_empty(&fs_info->trans_list)) { 47010b246afaSJeff Mahoney t = list_first_entry(&fs_info->trans_list, 4702724e2315SJosef Bacik struct btrfs_transaction, list); 4703724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 47049b64f57dSElena Reshetova refcount_inc(&t->use_count); 47050b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 47062ff7e61eSJeff Mahoney btrfs_wait_for_commit(fs_info, t->transid); 4707724e2315SJosef Bacik btrfs_put_transaction(t); 47080b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4709724e2315SJosef Bacik continue; 4710724e2315SJosef Bacik } 47110b246afaSJeff Mahoney if (t == fs_info->running_transaction) { 4712724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 47130b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4714724e2315SJosef Bacik /* 4715724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4716724e2315SJosef Bacik * handle open currently for this transaction. 4717724e2315SJosef Bacik */ 4718724e2315SJosef Bacik wait_event(t->writer_wait, 4719724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4720724e2315SJosef Bacik } else { 47210b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4722724e2315SJosef Bacik } 47232ff7e61eSJeff Mahoney btrfs_cleanup_one_transaction(t, fs_info); 4724724e2315SJosef Bacik 47250b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 47260b246afaSJeff Mahoney if (t == fs_info->running_transaction) 47270b246afaSJeff Mahoney fs_info->running_transaction = NULL; 4728724e2315SJosef Bacik list_del_init(&t->list); 47290b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 4730a4abeea4SJosef Bacik 4731724e2315SJosef Bacik btrfs_put_transaction(t); 47322ff7e61eSJeff Mahoney trace_btrfs_transaction_commit(fs_info->tree_root); 47330b246afaSJeff Mahoney spin_lock(&fs_info->trans_lock); 4734724e2315SJosef Bacik } 47350b246afaSJeff Mahoney spin_unlock(&fs_info->trans_lock); 47360b246afaSJeff Mahoney btrfs_destroy_all_ordered_extents(fs_info); 4737ccdf9b30SJeff Mahoney btrfs_destroy_delayed_inodes(fs_info); 4738ccdf9b30SJeff Mahoney btrfs_assert_delayed_root_empty(fs_info); 47390b246afaSJeff Mahoney btrfs_destroy_all_delalloc_inodes(fs_info); 4740ef67963dSJosef Bacik btrfs_drop_all_logs(fs_info); 47410b246afaSJeff Mahoney mutex_unlock(&fs_info->transaction_kthread_mutex); 4742acce952bSliubo 4743acce952bSliubo return 0; 4744acce952bSliubo } 4745ec7d6dfdSNikolay Borisov 4746ec7d6dfdSNikolay Borisov int btrfs_find_highest_objectid(struct btrfs_root *root, u64 *objectid) 4747ec7d6dfdSNikolay Borisov { 4748ec7d6dfdSNikolay Borisov struct btrfs_path *path; 4749ec7d6dfdSNikolay Borisov int ret; 4750ec7d6dfdSNikolay Borisov struct extent_buffer *l; 4751ec7d6dfdSNikolay Borisov struct btrfs_key search_key; 4752ec7d6dfdSNikolay Borisov struct btrfs_key found_key; 4753ec7d6dfdSNikolay Borisov int slot; 4754ec7d6dfdSNikolay Borisov 4755ec7d6dfdSNikolay Borisov path = btrfs_alloc_path(); 4756ec7d6dfdSNikolay Borisov if (!path) 4757ec7d6dfdSNikolay Borisov return -ENOMEM; 4758ec7d6dfdSNikolay Borisov 4759ec7d6dfdSNikolay Borisov search_key.objectid = BTRFS_LAST_FREE_OBJECTID; 4760ec7d6dfdSNikolay Borisov search_key.type = -1; 4761ec7d6dfdSNikolay Borisov search_key.offset = (u64)-1; 4762ec7d6dfdSNikolay Borisov ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); 4763ec7d6dfdSNikolay Borisov if (ret < 0) 4764ec7d6dfdSNikolay Borisov goto error; 4765ec7d6dfdSNikolay Borisov BUG_ON(ret == 0); /* Corruption */ 4766ec7d6dfdSNikolay Borisov if (path->slots[0] > 0) { 4767ec7d6dfdSNikolay Borisov slot = path->slots[0] - 1; 4768ec7d6dfdSNikolay Borisov l = path->nodes[0]; 4769ec7d6dfdSNikolay Borisov btrfs_item_key_to_cpu(l, &found_key, slot); 4770ec7d6dfdSNikolay Borisov *objectid = max_t(u64, found_key.objectid, 4771ec7d6dfdSNikolay Borisov BTRFS_FIRST_FREE_OBJECTID - 1); 4772ec7d6dfdSNikolay Borisov } else { 4773ec7d6dfdSNikolay Borisov *objectid = BTRFS_FIRST_FREE_OBJECTID - 1; 4774ec7d6dfdSNikolay Borisov } 4775ec7d6dfdSNikolay Borisov ret = 0; 4776ec7d6dfdSNikolay Borisov error: 4777ec7d6dfdSNikolay Borisov btrfs_free_path(path); 4778ec7d6dfdSNikolay Borisov return ret; 4779ec7d6dfdSNikolay Borisov } 4780ec7d6dfdSNikolay Borisov 4781ec7d6dfdSNikolay Borisov int btrfs_find_free_objectid(struct btrfs_root *root, u64 *objectid) 4782ec7d6dfdSNikolay Borisov { 4783ec7d6dfdSNikolay Borisov int ret; 4784ec7d6dfdSNikolay Borisov mutex_lock(&root->objectid_mutex); 4785ec7d6dfdSNikolay Borisov 4786ec7d6dfdSNikolay Borisov if (unlikely(root->highest_objectid >= BTRFS_LAST_FREE_OBJECTID)) { 4787ec7d6dfdSNikolay Borisov btrfs_warn(root->fs_info, 4788ec7d6dfdSNikolay Borisov "the objectid of root %llu reaches its highest value", 4789ec7d6dfdSNikolay Borisov root->root_key.objectid); 4790ec7d6dfdSNikolay Borisov ret = -ENOSPC; 4791ec7d6dfdSNikolay Borisov goto out; 4792ec7d6dfdSNikolay Borisov } 4793ec7d6dfdSNikolay Borisov 4794ec7d6dfdSNikolay Borisov *objectid = ++root->highest_objectid; 4795ec7d6dfdSNikolay Borisov ret = 0; 4796ec7d6dfdSNikolay Borisov out: 4797ec7d6dfdSNikolay Borisov mutex_unlock(&root->objectid_mutex); 4798ec7d6dfdSNikolay Borisov return ret; 4799ec7d6dfdSNikolay Borisov } 4800