16cbd5570SChris Mason /* 26cbd5570SChris Mason * Copyright (C) 2007 Oracle. All rights reserved. 36cbd5570SChris Mason * 46cbd5570SChris Mason * This program is free software; you can redistribute it and/or 56cbd5570SChris Mason * modify it under the terms of the GNU General Public 66cbd5570SChris Mason * License v2 as published by the Free Software Foundation. 76cbd5570SChris Mason * 86cbd5570SChris Mason * This program is distributed in the hope that it will be useful, 96cbd5570SChris Mason * but WITHOUT ANY WARRANTY; without even the implied warranty of 106cbd5570SChris Mason * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 116cbd5570SChris Mason * General Public License for more details. 126cbd5570SChris Mason * 136cbd5570SChris Mason * You should have received a copy of the GNU General Public 146cbd5570SChris Mason * License along with this program; if not, write to the 156cbd5570SChris Mason * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 166cbd5570SChris Mason * Boston, MA 021110-1307, USA. 176cbd5570SChris Mason */ 186cbd5570SChris Mason 19e20d96d6SChris Mason #include <linux/fs.h> 20d98237b3SChris Mason #include <linux/blkdev.h> 2187cbda5cSChris Mason #include <linux/scatterlist.h> 2222b0ebdaSChris Mason #include <linux/swap.h> 230f7d52f4SChris Mason #include <linux/radix-tree.h> 2435b7e476SChris Mason #include <linux/writeback.h> 25d397712bSChris Mason #include <linux/buffer_head.h> 26ce9adaa5SChris Mason #include <linux/workqueue.h> 27a74a4b97SChris Mason #include <linux/kthread.h> 28a74a4b97SChris Mason #include <linux/freezer.h> 295a0e3ad6STejun Heo #include <linux/slab.h> 30784b4e29SChris Mason #include <linux/migrate.h> 317a36ddecSDavid Sterba #include <linux/ratelimit.h> 326463fe58SStefan Behrens #include <linux/uuid.h> 33803b2f54SStefan Behrens #include <linux/semaphore.h> 347e75bf3fSDavid Sterba #include <asm/unaligned.h> 35eb60ceacSChris Mason #include "ctree.h" 36eb60ceacSChris Mason #include "disk-io.h" 370b947affSFilipe David Borba Manana #include "hash.h" 38e089f05cSChris Mason #include "transaction.h" 390f7d52f4SChris Mason #include "btrfs_inode.h" 400b86a832SChris Mason #include "volumes.h" 41db94535dSChris Mason #include "print-tree.h" 42925baeddSChris Mason #include "locking.h" 43e02119d5SChris Mason #include "tree-log.h" 44fa9c0d79SChris Mason #include "free-space-cache.h" 45581bb050SLi Zefan #include "inode-map.h" 4621adbd5cSStefan Behrens #include "check-integrity.h" 47606686eeSJosef Bacik #include "rcu-string.h" 488dabb742SStefan Behrens #include "dev-replace.h" 4953b381b3SDavid Woodhouse #include "raid56.h" 505ac1d209SJeff Mahoney #include "sysfs.h" 51fcebe456SJosef Bacik #include "qgroup.h" 52eb60ceacSChris Mason 53de0022b9SJosef Bacik #ifdef CONFIG_X86 54de0022b9SJosef Bacik #include <asm/cpufeature.h> 55de0022b9SJosef Bacik #endif 56de0022b9SJosef Bacik 57d1310b2eSChris Mason static struct extent_io_ops btree_extent_io_ops; 588b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work); 594df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root); 60fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 61acce952bSliubo int read_only); 62143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 63acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 64acce952bSliubo struct btrfs_root *root); 65143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 66acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 67acce952bSliubo struct extent_io_tree *dirty_pages, 68acce952bSliubo int mark); 69acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 70acce952bSliubo struct extent_io_tree *pinned_extents); 7148a3b636SEric Sandeen static int btrfs_cleanup_transaction(struct btrfs_root *root); 7248a3b636SEric Sandeen static void btrfs_error_commit_super(struct btrfs_root *root); 73ce9adaa5SChris Mason 74d352ac68SChris Mason /* 75d352ac68SChris Mason * end_io_wq structs are used to do processing in task context when an IO is 76d352ac68SChris Mason * complete. This is used during reads to verify checksums, and it is used 77d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 78d352ac68SChris Mason */ 79ce9adaa5SChris Mason struct end_io_wq { 80ce9adaa5SChris Mason struct bio *bio; 81ce9adaa5SChris Mason bio_end_io_t *end_io; 82ce9adaa5SChris Mason void *private; 83ce9adaa5SChris Mason struct btrfs_fs_info *info; 84ce9adaa5SChris Mason int error; 8522c59948SChris Mason int metadata; 86ce9adaa5SChris Mason struct list_head list; 878b712842SChris Mason struct btrfs_work work; 88ce9adaa5SChris Mason }; 890da5468fSChris Mason 90d352ac68SChris Mason /* 91d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 92d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 93d352ac68SChris Mason * just before they are sent down the IO stack. 94d352ac68SChris Mason */ 9544b8bd7eSChris Mason struct async_submit_bio { 9644b8bd7eSChris Mason struct inode *inode; 9744b8bd7eSChris Mason struct bio *bio; 9844b8bd7eSChris Mason struct list_head list; 994a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start; 1004a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done; 10144b8bd7eSChris Mason int rw; 10244b8bd7eSChris Mason int mirror_num; 103c8b97818SChris Mason unsigned long bio_flags; 104eaf25d93SChris Mason /* 105eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 106eaf25d93SChris Mason * can't tell us where in the file the bio should go 107eaf25d93SChris Mason */ 108eaf25d93SChris Mason u64 bio_offset; 1098b712842SChris Mason struct btrfs_work work; 11079787eaaSJeff Mahoney int error; 11144b8bd7eSChris Mason }; 11244b8bd7eSChris Mason 11385d4e461SChris Mason /* 11485d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 11585d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 11685d4e461SChris Mason * the level the eb occupies in the tree. 1174008c04aSChris Mason * 11885d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 11985d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 12085d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 12185d4e461SChris Mason * by btrfs_root->objectid. This ensures that all special purpose roots 12285d4e461SChris Mason * have separate keysets. 1234008c04aSChris Mason * 12485d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 12585d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 12685d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1274008c04aSChris Mason * 12885d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 12985d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 13085d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 13185d4e461SChris Mason * 13285d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 13385d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 13485d4e461SChris Mason * needs update as well. 1354008c04aSChris Mason */ 1364008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1374008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1384008c04aSChris Mason # error 1394008c04aSChris Mason # endif 14085d4e461SChris Mason 14185d4e461SChris Mason static struct btrfs_lockdep_keyset { 14285d4e461SChris Mason u64 id; /* root objectid */ 14385d4e461SChris Mason const char *name_stem; /* lock name stem */ 14485d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 14585d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 14685d4e461SChris Mason } btrfs_lockdep_keysets[] = { 14785d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 14885d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 14985d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 15085d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 15185d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 15285d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 15360b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 15485d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 15585d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 15685d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 15713fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 15885d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1594008c04aSChris Mason }; 16085d4e461SChris Mason 16185d4e461SChris Mason void __init btrfs_init_lockdep(void) 16285d4e461SChris Mason { 16385d4e461SChris Mason int i, j; 16485d4e461SChris Mason 16585d4e461SChris Mason /* initialize lockdep class names */ 16685d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 16785d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 16885d4e461SChris Mason 16985d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 17085d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 17185d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 17285d4e461SChris Mason } 17385d4e461SChris Mason } 17485d4e461SChris Mason 17585d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 17685d4e461SChris Mason int level) 17785d4e461SChris Mason { 17885d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 17985d4e461SChris Mason 18085d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 18185d4e461SChris Mason 18285d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 18385d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 18485d4e461SChris Mason if (ks->id == objectid) 18585d4e461SChris Mason break; 18685d4e461SChris Mason 18785d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 18885d4e461SChris Mason &ks->keys[level], ks->names[level]); 18985d4e461SChris Mason } 19085d4e461SChris Mason 1914008c04aSChris Mason #endif 1924008c04aSChris Mason 193d352ac68SChris Mason /* 194d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 195d352ac68SChris Mason * that covers the entire device 196d352ac68SChris Mason */ 197b2950863SChristoph Hellwig static struct extent_map *btree_get_extent(struct inode *inode, 198306e16ceSDavid Sterba struct page *page, size_t pg_offset, u64 start, u64 len, 1995f39d397SChris Mason int create) 2005f39d397SChris Mason { 2015f39d397SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 2025f39d397SChris Mason struct extent_map *em; 2035f39d397SChris Mason int ret; 2045f39d397SChris Mason 205890871beSChris Mason read_lock(&em_tree->lock); 206d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 207a061fc8dSChris Mason if (em) { 208a061fc8dSChris Mason em->bdev = 209a061fc8dSChris Mason BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 210890871beSChris Mason read_unlock(&em_tree->lock); 2115f39d397SChris Mason goto out; 212a061fc8dSChris Mason } 213890871beSChris Mason read_unlock(&em_tree->lock); 2147b13b7b1SChris Mason 215172ddd60SDavid Sterba em = alloc_extent_map(); 2165f39d397SChris Mason if (!em) { 2175f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2185f39d397SChris Mason goto out; 2195f39d397SChris Mason } 2205f39d397SChris Mason em->start = 0; 2210afbaf8cSChris Mason em->len = (u64)-1; 222c8b97818SChris Mason em->block_len = (u64)-1; 2235f39d397SChris Mason em->block_start = 0; 224a061fc8dSChris Mason em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 225d1310b2eSChris Mason 226890871beSChris Mason write_lock(&em_tree->lock); 22709a2a8f9SJosef Bacik ret = add_extent_mapping(em_tree, em, 0); 2285f39d397SChris Mason if (ret == -EEXIST) { 2295f39d397SChris Mason free_extent_map(em); 2307b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 231b4f359abSTsutomu Itoh if (!em) 2320433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2335f39d397SChris Mason } else if (ret) { 2347b13b7b1SChris Mason free_extent_map(em); 2350433f20dSTsutomu Itoh em = ERR_PTR(ret); 2365f39d397SChris Mason } 237890871beSChris Mason write_unlock(&em_tree->lock); 2387b13b7b1SChris Mason 2395f39d397SChris Mason out: 2405f39d397SChris Mason return em; 2415f39d397SChris Mason } 2425f39d397SChris Mason 243b0496686SLiu Bo u32 btrfs_csum_data(char *data, u32 seed, size_t len) 24419c00ddcSChris Mason { 2450b947affSFilipe David Borba Manana return btrfs_crc32c(seed, data, len); 24619c00ddcSChris Mason } 24719c00ddcSChris Mason 24819c00ddcSChris Mason void btrfs_csum_final(u32 crc, char *result) 24919c00ddcSChris Mason { 2507e75bf3fSDavid Sterba put_unaligned_le32(~crc, result); 25119c00ddcSChris Mason } 25219c00ddcSChris Mason 253d352ac68SChris Mason /* 254d352ac68SChris Mason * compute the csum for a btree block, and either verify it or write it 255d352ac68SChris Mason * into the csum field of the block. 256d352ac68SChris Mason */ 25719c00ddcSChris Mason static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, 25819c00ddcSChris Mason int verify) 25919c00ddcSChris Mason { 2606c41761fSDavid Sterba u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); 261607d432dSJosef Bacik char *result = NULL; 26219c00ddcSChris Mason unsigned long len; 26319c00ddcSChris Mason unsigned long cur_len; 26419c00ddcSChris Mason unsigned long offset = BTRFS_CSUM_SIZE; 26519c00ddcSChris Mason char *kaddr; 26619c00ddcSChris Mason unsigned long map_start; 26719c00ddcSChris Mason unsigned long map_len; 26819c00ddcSChris Mason int err; 26919c00ddcSChris Mason u32 crc = ~(u32)0; 270607d432dSJosef Bacik unsigned long inline_result; 27119c00ddcSChris Mason 27219c00ddcSChris Mason len = buf->len - offset; 27319c00ddcSChris Mason while (len > 0) { 27419c00ddcSChris Mason err = map_private_extent_buffer(buf, offset, 32, 275a6591715SChris Mason &kaddr, &map_start, &map_len); 276d397712bSChris Mason if (err) 27719c00ddcSChris Mason return 1; 27819c00ddcSChris Mason cur_len = min(len, map_len - (offset - map_start)); 279b0496686SLiu Bo crc = btrfs_csum_data(kaddr + offset - map_start, 28019c00ddcSChris Mason crc, cur_len); 28119c00ddcSChris Mason len -= cur_len; 28219c00ddcSChris Mason offset += cur_len; 28319c00ddcSChris Mason } 284607d432dSJosef Bacik if (csum_size > sizeof(inline_result)) { 285607d432dSJosef Bacik result = kzalloc(csum_size * sizeof(char), GFP_NOFS); 286607d432dSJosef Bacik if (!result) 287607d432dSJosef Bacik return 1; 288607d432dSJosef Bacik } else { 289607d432dSJosef Bacik result = (char *)&inline_result; 290607d432dSJosef Bacik } 291607d432dSJosef Bacik 29219c00ddcSChris Mason btrfs_csum_final(crc, result); 29319c00ddcSChris Mason 29419c00ddcSChris Mason if (verify) { 295607d432dSJosef Bacik if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 296e4204dedSChris Mason u32 val; 297e4204dedSChris Mason u32 found = 0; 298607d432dSJosef Bacik memcpy(&found, result, csum_size); 299e4204dedSChris Mason 300607d432dSJosef Bacik read_extent_buffer(buf, &val, 0, csum_size); 301efe120a0SFrank Holton printk_ratelimited(KERN_INFO 302efe120a0SFrank Holton "BTRFS: %s checksum verify failed on %llu wanted %X found %X " 303193f284dSChris Mason "level %d\n", 304c1c9ff7cSGeert Uytterhoeven root->fs_info->sb->s_id, buf->start, 305c1c9ff7cSGeert Uytterhoeven val, found, btrfs_header_level(buf)); 306607d432dSJosef Bacik if (result != (char *)&inline_result) 307607d432dSJosef Bacik kfree(result); 30819c00ddcSChris Mason return 1; 30919c00ddcSChris Mason } 31019c00ddcSChris Mason } else { 311607d432dSJosef Bacik write_extent_buffer(buf, result, 0, csum_size); 31219c00ddcSChris Mason } 313607d432dSJosef Bacik if (result != (char *)&inline_result) 314607d432dSJosef Bacik kfree(result); 31519c00ddcSChris Mason return 0; 31619c00ddcSChris Mason } 31719c00ddcSChris Mason 318d352ac68SChris Mason /* 319d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 320d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 321d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 322d352ac68SChris Mason * in the wrong place. 323d352ac68SChris Mason */ 3241259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 325b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 326b9fab919SChris Mason int atomic) 3271259ab75SChris Mason { 3282ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 3291259ab75SChris Mason int ret; 330a26e8c9fSJosef Bacik bool need_lock = (current->journal_info == 331a26e8c9fSJosef Bacik (void *)BTRFS_SEND_TRANS_STUB); 3321259ab75SChris Mason 3331259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 3341259ab75SChris Mason return 0; 3351259ab75SChris Mason 336b9fab919SChris Mason if (atomic) 337b9fab919SChris Mason return -EAGAIN; 338b9fab919SChris Mason 339a26e8c9fSJosef Bacik if (need_lock) { 340a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 341a26e8c9fSJosef Bacik btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); 342a26e8c9fSJosef Bacik } 343a26e8c9fSJosef Bacik 3442ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 345d0082371SJeff Mahoney 0, &cached_state); 3460b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3471259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3481259ab75SChris Mason ret = 0; 3491259ab75SChris Mason goto out; 3501259ab75SChris Mason } 35129549aecSWang Shilong printk_ratelimited(KERN_INFO "BTRFS (device %s): parent transid verify failed on %llu wanted %llu found %llu\n", 35229549aecSWang Shilong eb->fs_info->sb->s_id, eb->start, 35329549aecSWang Shilong parent_transid, btrfs_header_generation(eb)); 3541259ab75SChris Mason ret = 1; 355a26e8c9fSJosef Bacik 356a26e8c9fSJosef Bacik /* 357a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 358a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 359a26e8c9fSJosef Bacik * block that has been free'd and re-allocated. So don't clear uptodate 360a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 361a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 362a26e8c9fSJosef Bacik * making everybody very sad. 363a26e8c9fSJosef Bacik */ 364a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 3650b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 36633958dc6SChris Mason out: 3672ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 3682ac55d41SJosef Bacik &cached_state, GFP_NOFS); 369472b909fSJosef Bacik if (need_lock) 370a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 3711259ab75SChris Mason return ret; 3721259ab75SChris Mason } 3731259ab75SChris Mason 374d352ac68SChris Mason /* 3751104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 3761104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 3771104a885SDavid Sterba */ 3781104a885SDavid Sterba static int btrfs_check_super_csum(char *raw_disk_sb) 3791104a885SDavid Sterba { 3801104a885SDavid Sterba struct btrfs_super_block *disk_sb = 3811104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 3821104a885SDavid Sterba u16 csum_type = btrfs_super_csum_type(disk_sb); 3831104a885SDavid Sterba int ret = 0; 3841104a885SDavid Sterba 3851104a885SDavid Sterba if (csum_type == BTRFS_CSUM_TYPE_CRC32) { 3861104a885SDavid Sterba u32 crc = ~(u32)0; 3871104a885SDavid Sterba const int csum_size = sizeof(crc); 3881104a885SDavid Sterba char result[csum_size]; 3891104a885SDavid Sterba 3901104a885SDavid Sterba /* 3911104a885SDavid Sterba * The super_block structure does not span the whole 3921104a885SDavid Sterba * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space 3931104a885SDavid Sterba * is filled with zeros and is included in the checkum. 3941104a885SDavid Sterba */ 3951104a885SDavid Sterba crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, 3961104a885SDavid Sterba crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 3971104a885SDavid Sterba btrfs_csum_final(crc, result); 3981104a885SDavid Sterba 3991104a885SDavid Sterba if (memcmp(raw_disk_sb, result, csum_size)) 4001104a885SDavid Sterba ret = 1; 401667e7d94SChris Mason 402667e7d94SChris Mason if (ret && btrfs_super_generation(disk_sb) < 10) { 403efe120a0SFrank Holton printk(KERN_WARNING 404efe120a0SFrank Holton "BTRFS: super block crcs don't match, older mkfs detected\n"); 405667e7d94SChris Mason ret = 0; 406667e7d94SChris Mason } 4071104a885SDavid Sterba } 4081104a885SDavid Sterba 4091104a885SDavid Sterba if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { 410efe120a0SFrank Holton printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", 4111104a885SDavid Sterba csum_type); 4121104a885SDavid Sterba ret = 1; 4131104a885SDavid Sterba } 4141104a885SDavid Sterba 4151104a885SDavid Sterba return ret; 4161104a885SDavid Sterba } 4171104a885SDavid Sterba 4181104a885SDavid Sterba /* 419d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 420d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 421d352ac68SChris Mason */ 422f188591eSChris Mason static int btree_read_extent_buffer_pages(struct btrfs_root *root, 423f188591eSChris Mason struct extent_buffer *eb, 424ca7a79adSChris Mason u64 start, u64 parent_transid) 425f188591eSChris Mason { 426f188591eSChris Mason struct extent_io_tree *io_tree; 427ea466794SJosef Bacik int failed = 0; 428f188591eSChris Mason int ret; 429f188591eSChris Mason int num_copies = 0; 430f188591eSChris Mason int mirror_num = 0; 431ea466794SJosef Bacik int failed_mirror = 0; 432f188591eSChris Mason 433a826d6dcSJosef Bacik clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 434f188591eSChris Mason io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; 435f188591eSChris Mason while (1) { 436bb82ab88SArne Jansen ret = read_extent_buffer_pages(io_tree, eb, start, 437bb82ab88SArne Jansen WAIT_COMPLETE, 438f188591eSChris Mason btree_get_extent, mirror_num); 439256dd1bbSStefan Behrens if (!ret) { 440256dd1bbSStefan Behrens if (!verify_parent_transid(io_tree, eb, 441b9fab919SChris Mason parent_transid, 0)) 442ea466794SJosef Bacik break; 443256dd1bbSStefan Behrens else 444256dd1bbSStefan Behrens ret = -EIO; 445256dd1bbSStefan Behrens } 446d397712bSChris Mason 447a826d6dcSJosef Bacik /* 448a826d6dcSJosef Bacik * This buffer's crc is fine, but its contents are corrupted, so 449a826d6dcSJosef Bacik * there is no reason to read the other copies, they won't be 450a826d6dcSJosef Bacik * any less wrong. 451a826d6dcSJosef Bacik */ 452a826d6dcSJosef Bacik if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) 453ea466794SJosef Bacik break; 454ea466794SJosef Bacik 4555d964051SStefan Behrens num_copies = btrfs_num_copies(root->fs_info, 456f188591eSChris Mason eb->start, eb->len); 4574235298eSChris Mason if (num_copies == 1) 458ea466794SJosef Bacik break; 4594235298eSChris Mason 4605cf1ab56SJosef Bacik if (!failed_mirror) { 4615cf1ab56SJosef Bacik failed = 1; 4625cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4635cf1ab56SJosef Bacik } 4645cf1ab56SJosef Bacik 465f188591eSChris Mason mirror_num++; 466ea466794SJosef Bacik if (mirror_num == failed_mirror) 467ea466794SJosef Bacik mirror_num++; 468ea466794SJosef Bacik 4694235298eSChris Mason if (mirror_num > num_copies) 470ea466794SJosef Bacik break; 471f188591eSChris Mason } 472ea466794SJosef Bacik 473c0901581SStefan Behrens if (failed && !ret && failed_mirror) 474ea466794SJosef Bacik repair_eb_io_failure(root, eb, failed_mirror); 475ea466794SJosef Bacik 476ea466794SJosef Bacik return ret; 477f188591eSChris Mason } 47819c00ddcSChris Mason 479d352ac68SChris Mason /* 480d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 481d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 482d352ac68SChris Mason */ 483d397712bSChris Mason 484b2950863SChristoph Hellwig static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) 48519c00ddcSChris Mason { 4864eee4fa4SMiao Xie u64 start = page_offset(page); 48719c00ddcSChris Mason u64 found_start; 48819c00ddcSChris Mason struct extent_buffer *eb; 489f188591eSChris Mason 4904f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4914f2de97aSJosef Bacik if (page != eb->pages[0]) 4924f2de97aSJosef Bacik return 0; 49319c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 494fae7f21cSDulshani Gunawardhana if (WARN_ON(found_start != start || !PageUptodate(page))) 4954f2de97aSJosef Bacik return 0; 49619c00ddcSChris Mason csum_tree_block(root, eb, 0); 49719c00ddcSChris Mason return 0; 49819c00ddcSChris Mason } 49919c00ddcSChris Mason 5002b82032cSYan Zheng static int check_tree_block_fsid(struct btrfs_root *root, 5012b82032cSYan Zheng struct extent_buffer *eb) 5022b82032cSYan Zheng { 5032b82032cSYan Zheng struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 5042b82032cSYan Zheng u8 fsid[BTRFS_UUID_SIZE]; 5052b82032cSYan Zheng int ret = 1; 5062b82032cSYan Zheng 5070a4e5586SRoss Kirk read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 5082b82032cSYan Zheng while (fs_devices) { 5092b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 5102b82032cSYan Zheng ret = 0; 5112b82032cSYan Zheng break; 5122b82032cSYan Zheng } 5132b82032cSYan Zheng fs_devices = fs_devices->seed; 5142b82032cSYan Zheng } 5152b82032cSYan Zheng return ret; 5162b82032cSYan Zheng } 5172b82032cSYan Zheng 518a826d6dcSJosef Bacik #define CORRUPT(reason, eb, root, slot) \ 519efe120a0SFrank Holton btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ 520efe120a0SFrank Holton "root=%llu, slot=%d", reason, \ 521c1c9ff7cSGeert Uytterhoeven btrfs_header_bytenr(eb), root->objectid, slot) 522a826d6dcSJosef Bacik 523a826d6dcSJosef Bacik static noinline int check_leaf(struct btrfs_root *root, 524a826d6dcSJosef Bacik struct extent_buffer *leaf) 525a826d6dcSJosef Bacik { 526a826d6dcSJosef Bacik struct btrfs_key key; 527a826d6dcSJosef Bacik struct btrfs_key leaf_key; 528a826d6dcSJosef Bacik u32 nritems = btrfs_header_nritems(leaf); 529a826d6dcSJosef Bacik int slot; 530a826d6dcSJosef Bacik 531a826d6dcSJosef Bacik if (nritems == 0) 532a826d6dcSJosef Bacik return 0; 533a826d6dcSJosef Bacik 534a826d6dcSJosef Bacik /* Check the 0 item */ 535a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != 536a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 537a826d6dcSJosef Bacik CORRUPT("invalid item offset size pair", leaf, root, 0); 538a826d6dcSJosef Bacik return -EIO; 539a826d6dcSJosef Bacik } 540a826d6dcSJosef Bacik 541a826d6dcSJosef Bacik /* 542a826d6dcSJosef Bacik * Check to make sure each items keys are in the correct order and their 543a826d6dcSJosef Bacik * offsets make sense. We only have to loop through nritems-1 because 544a826d6dcSJosef Bacik * we check the current slot against the next slot, which verifies the 545a826d6dcSJosef Bacik * next slot's offset+size makes sense and that the current's slot 546a826d6dcSJosef Bacik * offset is correct. 547a826d6dcSJosef Bacik */ 548a826d6dcSJosef Bacik for (slot = 0; slot < nritems - 1; slot++) { 549a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &leaf_key, slot); 550a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &key, slot + 1); 551a826d6dcSJosef Bacik 552a826d6dcSJosef Bacik /* Make sure the keys are in the right order */ 553a826d6dcSJosef Bacik if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { 554a826d6dcSJosef Bacik CORRUPT("bad key order", leaf, root, slot); 555a826d6dcSJosef Bacik return -EIO; 556a826d6dcSJosef Bacik } 557a826d6dcSJosef Bacik 558a826d6dcSJosef Bacik /* 559a826d6dcSJosef Bacik * Make sure the offset and ends are right, remember that the 560a826d6dcSJosef Bacik * item data starts at the end of the leaf and grows towards the 561a826d6dcSJosef Bacik * front. 562a826d6dcSJosef Bacik */ 563a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, slot) != 564a826d6dcSJosef Bacik btrfs_item_end_nr(leaf, slot + 1)) { 565a826d6dcSJosef Bacik CORRUPT("slot offset bad", leaf, root, slot); 566a826d6dcSJosef Bacik return -EIO; 567a826d6dcSJosef Bacik } 568a826d6dcSJosef Bacik 569a826d6dcSJosef Bacik /* 570a826d6dcSJosef Bacik * Check to make sure that we don't point outside of the leaf, 571a826d6dcSJosef Bacik * just incase all the items are consistent to eachother, but 572a826d6dcSJosef Bacik * all point outside of the leaf. 573a826d6dcSJosef Bacik */ 574a826d6dcSJosef Bacik if (btrfs_item_end_nr(leaf, slot) > 575a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 576a826d6dcSJosef Bacik CORRUPT("slot end outside of leaf", leaf, root, slot); 577a826d6dcSJosef Bacik return -EIO; 578a826d6dcSJosef Bacik } 579a826d6dcSJosef Bacik } 580a826d6dcSJosef Bacik 581a826d6dcSJosef Bacik return 0; 582a826d6dcSJosef Bacik } 583a826d6dcSJosef Bacik 584facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 585facc8a22SMiao Xie u64 phy_offset, struct page *page, 586facc8a22SMiao Xie u64 start, u64 end, int mirror) 587ce9adaa5SChris Mason { 588ce9adaa5SChris Mason u64 found_start; 589ce9adaa5SChris Mason int found_level; 590ce9adaa5SChris Mason struct extent_buffer *eb; 591ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 592f188591eSChris Mason int ret = 0; 593727011e0SChris Mason int reads_done; 594ce9adaa5SChris Mason 595ce9adaa5SChris Mason if (!page->private) 596ce9adaa5SChris Mason goto out; 597d397712bSChris Mason 5984f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 599d397712bSChris Mason 6000b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 6010b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 6020b32f4bbSJosef Bacik */ 6030b32f4bbSJosef Bacik extent_buffer_get(eb); 6040b32f4bbSJosef Bacik 6050b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 606727011e0SChris Mason if (!reads_done) 607727011e0SChris Mason goto err; 608f188591eSChris Mason 6095cf1ab56SJosef Bacik eb->read_mirror = mirror; 610ea466794SJosef Bacik if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { 611ea466794SJosef Bacik ret = -EIO; 612ea466794SJosef Bacik goto err; 613ea466794SJosef Bacik } 614ea466794SJosef Bacik 615ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 616727011e0SChris Mason if (found_start != eb->start) { 61729549aecSWang Shilong printk_ratelimited(KERN_INFO "BTRFS (device %s): bad tree block start " 618193f284dSChris Mason "%llu %llu\n", 61929549aecSWang Shilong eb->fs_info->sb->s_id, found_start, eb->start); 620f188591eSChris Mason ret = -EIO; 621ce9adaa5SChris Mason goto err; 622ce9adaa5SChris Mason } 6232b82032cSYan Zheng if (check_tree_block_fsid(root, eb)) { 62429549aecSWang Shilong printk_ratelimited(KERN_INFO "BTRFS (device %s): bad fsid on block %llu\n", 62529549aecSWang Shilong eb->fs_info->sb->s_id, eb->start); 6261259ab75SChris Mason ret = -EIO; 6271259ab75SChris Mason goto err; 6281259ab75SChris Mason } 629ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6301c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 631efe120a0SFrank Holton btrfs_info(root->fs_info, "bad tree block level %d", 6321c24c3ceSJosef Bacik (int)btrfs_header_level(eb)); 6331c24c3ceSJosef Bacik ret = -EIO; 6341c24c3ceSJosef Bacik goto err; 6351c24c3ceSJosef Bacik } 636ce9adaa5SChris Mason 63785d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 63885d4e461SChris Mason eb, found_level); 6394008c04aSChris Mason 640ce9adaa5SChris Mason ret = csum_tree_block(root, eb, 1); 641a826d6dcSJosef Bacik if (ret) { 642f188591eSChris Mason ret = -EIO; 643a826d6dcSJosef Bacik goto err; 644a826d6dcSJosef Bacik } 645a826d6dcSJosef Bacik 646a826d6dcSJosef Bacik /* 647a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 648a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 649a826d6dcSJosef Bacik * return -EIO. 650a826d6dcSJosef Bacik */ 651a826d6dcSJosef Bacik if (found_level == 0 && check_leaf(root, eb)) { 652a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 653a826d6dcSJosef Bacik ret = -EIO; 654a826d6dcSJosef Bacik } 655ce9adaa5SChris Mason 6560b32f4bbSJosef Bacik if (!ret) 6570b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 658ce9adaa5SChris Mason err: 65979fb65a1SJosef Bacik if (reads_done && 66079fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 6614bb31e92SArne Jansen btree_readahead_hook(root, eb, eb->start, ret); 6624bb31e92SArne Jansen 66353b381b3SDavid Woodhouse if (ret) { 66453b381b3SDavid Woodhouse /* 66553b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 66653b381b3SDavid Woodhouse * again, we have to make sure it has something 66753b381b3SDavid Woodhouse * to decrement 66853b381b3SDavid Woodhouse */ 66953b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6700b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 67153b381b3SDavid Woodhouse } 6720b32f4bbSJosef Bacik free_extent_buffer(eb); 673ce9adaa5SChris Mason out: 674f188591eSChris Mason return ret; 675ce9adaa5SChris Mason } 676ce9adaa5SChris Mason 677ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 6784bb31e92SArne Jansen { 6794bb31e92SArne Jansen struct extent_buffer *eb; 6804bb31e92SArne Jansen struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 6814bb31e92SArne Jansen 6824f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 683ea466794SJosef Bacik set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); 6845cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 68553b381b3SDavid Woodhouse atomic_dec(&eb->io_pages); 686ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 6874bb31e92SArne Jansen btree_readahead_hook(root, eb, eb->start, -EIO); 6884bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 6894bb31e92SArne Jansen } 6904bb31e92SArne Jansen 691ce9adaa5SChris Mason static void end_workqueue_bio(struct bio *bio, int err) 692ce9adaa5SChris Mason { 693ce9adaa5SChris Mason struct end_io_wq *end_io_wq = bio->bi_private; 694ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 6959e0af237SLiu Bo struct btrfs_workqueue *wq; 6969e0af237SLiu Bo btrfs_work_func_t func; 697ce9adaa5SChris Mason 698ce9adaa5SChris Mason fs_info = end_io_wq->info; 699ce9adaa5SChris Mason end_io_wq->error = err; 700d20f7043SChris Mason 7017b6d91daSChristoph Hellwig if (bio->bi_rw & REQ_WRITE) { 7029e0af237SLiu Bo if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) { 7039e0af237SLiu Bo wq = fs_info->endio_meta_write_workers; 7049e0af237SLiu Bo func = btrfs_endio_meta_write_helper; 7059e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) { 7069e0af237SLiu Bo wq = fs_info->endio_freespace_worker; 7079e0af237SLiu Bo func = btrfs_freespace_write_helper; 7089e0af237SLiu Bo } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7099e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7109e0af237SLiu Bo func = btrfs_endio_raid56_helper; 711d20f7043SChris Mason } else { 7129e0af237SLiu Bo wq = fs_info->endio_write_workers; 7139e0af237SLiu Bo func = btrfs_endio_write_helper; 714d20f7043SChris Mason } 7159e0af237SLiu Bo } else { 7169e0af237SLiu Bo if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) { 7179e0af237SLiu Bo wq = fs_info->endio_raid56_workers; 7189e0af237SLiu Bo func = btrfs_endio_raid56_helper; 7199e0af237SLiu Bo } else if (end_io_wq->metadata) { 7209e0af237SLiu Bo wq = fs_info->endio_meta_workers; 7219e0af237SLiu Bo func = btrfs_endio_meta_helper; 7229e0af237SLiu Bo } else { 7239e0af237SLiu Bo wq = fs_info->endio_workers; 7249e0af237SLiu Bo func = btrfs_endio_helper; 7259e0af237SLiu Bo } 7269e0af237SLiu Bo } 7279e0af237SLiu Bo 7289e0af237SLiu Bo btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL); 7299e0af237SLiu Bo btrfs_queue_work(wq, &end_io_wq->work); 730ce9adaa5SChris Mason } 731ce9adaa5SChris Mason 7320cb59c99SJosef Bacik /* 7330cb59c99SJosef Bacik * For the metadata arg you want 7340cb59c99SJosef Bacik * 7350cb59c99SJosef Bacik * 0 - if data 7360cb59c99SJosef Bacik * 1 - if normal metadta 7370cb59c99SJosef Bacik * 2 - if writing to the free space cache area 73853b381b3SDavid Woodhouse * 3 - raid parity work 7390cb59c99SJosef Bacik */ 74022c59948SChris Mason int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 74122c59948SChris Mason int metadata) 7420b86a832SChris Mason { 743ce9adaa5SChris Mason struct end_io_wq *end_io_wq; 744ce9adaa5SChris Mason end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); 745ce9adaa5SChris Mason if (!end_io_wq) 746ce9adaa5SChris Mason return -ENOMEM; 747ce9adaa5SChris Mason 748ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 749ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 75022c59948SChris Mason end_io_wq->info = info; 751ce9adaa5SChris Mason end_io_wq->error = 0; 752ce9adaa5SChris Mason end_io_wq->bio = bio; 75322c59948SChris Mason end_io_wq->metadata = metadata; 754ce9adaa5SChris Mason 755ce9adaa5SChris Mason bio->bi_private = end_io_wq; 756ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 75722c59948SChris Mason return 0; 75822c59948SChris Mason } 75922c59948SChris Mason 760b64a2851SChris Mason unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) 7614854ddd0SChris Mason { 7624854ddd0SChris Mason unsigned long limit = min_t(unsigned long, 7635cdc7ad3SQu Wenruo info->thread_pool_size, 7644854ddd0SChris Mason info->fs_devices->open_devices); 7654854ddd0SChris Mason return 256 * limit; 7664854ddd0SChris Mason } 7674854ddd0SChris Mason 7684a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7694a69a410SChris Mason { 7704a69a410SChris Mason struct async_submit_bio *async; 77179787eaaSJeff Mahoney int ret; 7724a69a410SChris Mason 7734a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 77479787eaaSJeff Mahoney ret = async->submit_bio_start(async->inode, async->rw, async->bio, 775eaf25d93SChris Mason async->mirror_num, async->bio_flags, 776eaf25d93SChris Mason async->bio_offset); 77779787eaaSJeff Mahoney if (ret) 77879787eaaSJeff Mahoney async->error = ret; 7794a69a410SChris Mason } 7804a69a410SChris Mason 7814a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7828b712842SChris Mason { 7838b712842SChris Mason struct btrfs_fs_info *fs_info; 7848b712842SChris Mason struct async_submit_bio *async; 7854854ddd0SChris Mason int limit; 7868b712842SChris Mason 7878b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 7888b712842SChris Mason fs_info = BTRFS_I(async->inode)->root->fs_info; 7894854ddd0SChris Mason 790b64a2851SChris Mason limit = btrfs_async_submit_limit(fs_info); 7914854ddd0SChris Mason limit = limit * 2 / 3; 7924854ddd0SChris Mason 79366657b31SJosef Bacik if (atomic_dec_return(&fs_info->nr_async_submits) < limit && 794b64a2851SChris Mason waitqueue_active(&fs_info->async_submit_wait)) 7954854ddd0SChris Mason wake_up(&fs_info->async_submit_wait); 7964854ddd0SChris Mason 79779787eaaSJeff Mahoney /* If an error occured we just want to clean up the bio and move on */ 79879787eaaSJeff Mahoney if (async->error) { 79979787eaaSJeff Mahoney bio_endio(async->bio, async->error); 80079787eaaSJeff Mahoney return; 80179787eaaSJeff Mahoney } 80279787eaaSJeff Mahoney 8034a69a410SChris Mason async->submit_bio_done(async->inode, async->rw, async->bio, 804eaf25d93SChris Mason async->mirror_num, async->bio_flags, 805eaf25d93SChris Mason async->bio_offset); 8064a69a410SChris Mason } 8074a69a410SChris Mason 8084a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 8094a69a410SChris Mason { 8104a69a410SChris Mason struct async_submit_bio *async; 8114a69a410SChris Mason 8124a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 8138b712842SChris Mason kfree(async); 8148b712842SChris Mason } 8158b712842SChris Mason 81644b8bd7eSChris Mason int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, 81744b8bd7eSChris Mason int rw, struct bio *bio, int mirror_num, 818c8b97818SChris Mason unsigned long bio_flags, 819eaf25d93SChris Mason u64 bio_offset, 8204a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start, 8214a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done) 82244b8bd7eSChris Mason { 82344b8bd7eSChris Mason struct async_submit_bio *async; 82444b8bd7eSChris Mason 82544b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 82644b8bd7eSChris Mason if (!async) 82744b8bd7eSChris Mason return -ENOMEM; 82844b8bd7eSChris Mason 82944b8bd7eSChris Mason async->inode = inode; 83044b8bd7eSChris Mason async->rw = rw; 83144b8bd7eSChris Mason async->bio = bio; 83244b8bd7eSChris Mason async->mirror_num = mirror_num; 8334a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8344a69a410SChris Mason async->submit_bio_done = submit_bio_done; 8354a69a410SChris Mason 8369e0af237SLiu Bo btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start, 8375cdc7ad3SQu Wenruo run_one_async_done, run_one_async_free); 8384a69a410SChris Mason 839c8b97818SChris Mason async->bio_flags = bio_flags; 840eaf25d93SChris Mason async->bio_offset = bio_offset; 8418c8bee1dSChris Mason 84279787eaaSJeff Mahoney async->error = 0; 84379787eaaSJeff Mahoney 844cb03c743SChris Mason atomic_inc(&fs_info->nr_async_submits); 845d313d7a3SChris Mason 8467b6d91daSChristoph Hellwig if (rw & REQ_SYNC) 8475cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 848d313d7a3SChris Mason 8495cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 8509473f16cSChris Mason 851771ed689SChris Mason while (atomic_read(&fs_info->async_submit_draining) && 852771ed689SChris Mason atomic_read(&fs_info->nr_async_submits)) { 853771ed689SChris Mason wait_event(fs_info->async_submit_wait, 854771ed689SChris Mason (atomic_read(&fs_info->nr_async_submits) == 0)); 855771ed689SChris Mason } 856771ed689SChris Mason 85744b8bd7eSChris Mason return 0; 85844b8bd7eSChris Mason } 85944b8bd7eSChris Mason 860ce3ed71aSChris Mason static int btree_csum_one_bio(struct bio *bio) 861ce3ed71aSChris Mason { 8622c30c71bSKent Overstreet struct bio_vec *bvec; 863ce3ed71aSChris Mason struct btrfs_root *root; 8642c30c71bSKent Overstreet int i, ret = 0; 865ce3ed71aSChris Mason 8662c30c71bSKent Overstreet bio_for_each_segment_all(bvec, bio, i) { 867ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 86879787eaaSJeff Mahoney ret = csum_dirty_buffer(root, bvec->bv_page); 86979787eaaSJeff Mahoney if (ret) 87079787eaaSJeff Mahoney break; 871ce3ed71aSChris Mason } 8722c30c71bSKent Overstreet 87379787eaaSJeff Mahoney return ret; 874ce3ed71aSChris Mason } 875ce3ed71aSChris Mason 8764a69a410SChris Mason static int __btree_submit_bio_start(struct inode *inode, int rw, 8774a69a410SChris Mason struct bio *bio, int mirror_num, 878eaf25d93SChris Mason unsigned long bio_flags, 879eaf25d93SChris Mason u64 bio_offset) 88022c59948SChris Mason { 8818b712842SChris Mason /* 8828b712842SChris Mason * when we're called for a write, we're already in the async 8835443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8848b712842SChris Mason */ 88579787eaaSJeff Mahoney return btree_csum_one_bio(bio); 88622c59948SChris Mason } 88722c59948SChris Mason 8884a69a410SChris Mason static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, 889eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 890eaf25d93SChris Mason u64 bio_offset) 8914a69a410SChris Mason { 89261891923SStefan Behrens int ret; 89361891923SStefan Behrens 8948b712842SChris Mason /* 8954a69a410SChris Mason * when we're called for a write, we're already in the async 8964a69a410SChris Mason * submission context. Just jump into btrfs_map_bio 8978b712842SChris Mason */ 89861891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); 89961891923SStefan Behrens if (ret) 90061891923SStefan Behrens bio_endio(bio, ret); 90161891923SStefan Behrens return ret; 9020b86a832SChris Mason } 9030b86a832SChris Mason 904de0022b9SJosef Bacik static int check_async_write(struct inode *inode, unsigned long bio_flags) 905de0022b9SJosef Bacik { 906de0022b9SJosef Bacik if (bio_flags & EXTENT_BIO_TREE_LOG) 907de0022b9SJosef Bacik return 0; 908de0022b9SJosef Bacik #ifdef CONFIG_X86 909de0022b9SJosef Bacik if (cpu_has_xmm4_2) 910de0022b9SJosef Bacik return 0; 911de0022b9SJosef Bacik #endif 912de0022b9SJosef Bacik return 1; 913de0022b9SJosef Bacik } 914de0022b9SJosef Bacik 91544b8bd7eSChris Mason static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, 916eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 917eaf25d93SChris Mason u64 bio_offset) 91844b8bd7eSChris Mason { 919de0022b9SJosef Bacik int async = check_async_write(inode, bio_flags); 9204a69a410SChris Mason int ret; 921cad321adSChris Mason 9227b6d91daSChristoph Hellwig if (!(rw & REQ_WRITE)) { 923cad321adSChris Mason /* 924cad321adSChris Mason * called for a read, do the setup so that checksum validation 925cad321adSChris Mason * can happen in the async kernel threads 926cad321adSChris Mason */ 927f3f266abSChris Mason ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, 928f3f266abSChris Mason bio, 1); 9291d4284bdSChris Mason if (ret) 93061891923SStefan Behrens goto out_w_error; 93161891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 9326f3577bdSChris Mason mirror_num, 0); 933de0022b9SJosef Bacik } else if (!async) { 934de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 935de0022b9SJosef Bacik if (ret) 93661891923SStefan Behrens goto out_w_error; 93761891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 938de0022b9SJosef Bacik mirror_num, 0); 93961891923SStefan Behrens } else { 940cad321adSChris Mason /* 94161891923SStefan Behrens * kthread helpers are used to submit writes so that 94261891923SStefan Behrens * checksumming can happen in parallel across all CPUs 943cad321adSChris Mason */ 94461891923SStefan Behrens ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, 945c8b97818SChris Mason inode, rw, bio, mirror_num, 0, 946eaf25d93SChris Mason bio_offset, 9474a69a410SChris Mason __btree_submit_bio_start, 9484a69a410SChris Mason __btree_submit_bio_done); 94944b8bd7eSChris Mason } 95044b8bd7eSChris Mason 95161891923SStefan Behrens if (ret) { 95261891923SStefan Behrens out_w_error: 95361891923SStefan Behrens bio_endio(bio, ret); 95461891923SStefan Behrens } 95561891923SStefan Behrens return ret; 95661891923SStefan Behrens } 95761891923SStefan Behrens 9583dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 959784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 960a6bc32b8SMel Gorman struct page *newpage, struct page *page, 961a6bc32b8SMel Gorman enum migrate_mode mode) 962784b4e29SChris Mason { 963784b4e29SChris Mason /* 964784b4e29SChris Mason * we can't safely write a btree page from here, 965784b4e29SChris Mason * we haven't done the locking hook 966784b4e29SChris Mason */ 967784b4e29SChris Mason if (PageDirty(page)) 968784b4e29SChris Mason return -EAGAIN; 969784b4e29SChris Mason /* 970784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 971784b4e29SChris Mason * We must have no buffers or drop them. 972784b4e29SChris Mason */ 973784b4e29SChris Mason if (page_has_private(page) && 974784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 975784b4e29SChris Mason return -EAGAIN; 976a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 977784b4e29SChris Mason } 9783dd1462eSJan Beulich #endif 979784b4e29SChris Mason 9800da5468fSChris Mason 9810da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9820da5468fSChris Mason struct writeback_control *wbc) 9830da5468fSChris Mason { 984e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 985e2d84521SMiao Xie int ret; 986e2d84521SMiao Xie 987d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 988448d640bSChris Mason 989448d640bSChris Mason if (wbc->for_kupdate) 990448d640bSChris Mason return 0; 991448d640bSChris Mason 992e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 993b9473439SChris Mason /* this is a bit racy, but that's ok */ 994e2d84521SMiao Xie ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 995e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 996e2d84521SMiao Xie if (ret < 0) 997793955bcSChris Mason return 0; 998793955bcSChris Mason } 9990b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 10000da5468fSChris Mason } 10010da5468fSChris Mason 1002b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 10035f39d397SChris Mason { 1004d1310b2eSChris Mason struct extent_io_tree *tree; 1005d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10068ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 10075f39d397SChris Mason } 10085f39d397SChris Mason 100970dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 10105f39d397SChris Mason { 101198509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 101298509cfcSChris Mason return 0; 10130c4e538bSDavid Sterba 1014f7a52a40SDavid Sterba return try_release_extent_buffer(page); 1015d98237b3SChris Mason } 1016d98237b3SChris Mason 1017d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 1018d47992f8SLukas Czerner unsigned int length) 1019d98237b3SChris Mason { 1020d1310b2eSChris Mason struct extent_io_tree *tree; 1021d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10225f39d397SChris Mason extent_invalidatepage(tree, page, offset); 10235f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 10249ad6b7bcSChris Mason if (PagePrivate(page)) { 1025efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 1026efe120a0SFrank Holton "page private not zero on page %llu", 1027efe120a0SFrank Holton (unsigned long long)page_offset(page)); 10289ad6b7bcSChris Mason ClearPagePrivate(page); 10299ad6b7bcSChris Mason set_page_private(page, 0); 10309ad6b7bcSChris Mason page_cache_release(page); 10319ad6b7bcSChris Mason } 1032d98237b3SChris Mason } 1033d98237b3SChris Mason 10340b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 10350b32f4bbSJosef Bacik { 1036bb146eb2SJosef Bacik #ifdef DEBUG 10370b32f4bbSJosef Bacik struct extent_buffer *eb; 10380b32f4bbSJosef Bacik 10390b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 10400b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 10410b32f4bbSJosef Bacik BUG_ON(!eb); 10420b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10430b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 10440b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 1045bb146eb2SJosef Bacik #endif 10460b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10470b32f4bbSJosef Bacik } 10480b32f4bbSJosef Bacik 10497f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 1050d98237b3SChris Mason .readpage = btree_readpage, 10510da5468fSChris Mason .writepages = btree_writepages, 10525f39d397SChris Mason .releasepage = btree_releasepage, 10535f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10545a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1055784b4e29SChris Mason .migratepage = btree_migratepage, 10565a92bc88SChris Mason #endif 10570b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1058d98237b3SChris Mason }; 1059123abc88SChris Mason 1060ca7a79adSChris Mason int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1061ca7a79adSChris Mason u64 parent_transid) 1062090d1875SChris Mason { 10635f39d397SChris Mason struct extent_buffer *buf = NULL; 10645f39d397SChris Mason struct inode *btree_inode = root->fs_info->btree_inode; 1065de428b63SChris Mason int ret = 0; 1066090d1875SChris Mason 1067db94535dSChris Mason buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 10685f39d397SChris Mason if (!buf) 1069090d1875SChris Mason return 0; 1070d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 1071bb82ab88SArne Jansen buf, 0, WAIT_NONE, btree_get_extent, 0); 10725f39d397SChris Mason free_extent_buffer(buf); 1073de428b63SChris Mason return ret; 1074090d1875SChris Mason } 1075090d1875SChris Mason 1076ab0fff03SArne Jansen int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1077ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1078ab0fff03SArne Jansen { 1079ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 1080ab0fff03SArne Jansen struct inode *btree_inode = root->fs_info->btree_inode; 1081ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1082ab0fff03SArne Jansen int ret; 1083ab0fff03SArne Jansen 1084ab0fff03SArne Jansen buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 1085ab0fff03SArne Jansen if (!buf) 1086ab0fff03SArne Jansen return 0; 1087ab0fff03SArne Jansen 1088ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1089ab0fff03SArne Jansen 1090ab0fff03SArne Jansen ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, 1091ab0fff03SArne Jansen btree_get_extent, mirror_num); 1092ab0fff03SArne Jansen if (ret) { 1093ab0fff03SArne Jansen free_extent_buffer(buf); 1094ab0fff03SArne Jansen return ret; 1095ab0fff03SArne Jansen } 1096ab0fff03SArne Jansen 1097ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1098ab0fff03SArne Jansen free_extent_buffer(buf); 1099ab0fff03SArne Jansen return -EIO; 11000b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1101ab0fff03SArne Jansen *eb = buf; 1102ab0fff03SArne Jansen } else { 1103ab0fff03SArne Jansen free_extent_buffer(buf); 1104ab0fff03SArne Jansen } 1105ab0fff03SArne Jansen return 0; 1106ab0fff03SArne Jansen } 1107ab0fff03SArne Jansen 11080999df54SChris Mason struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, 11090999df54SChris Mason u64 bytenr, u32 blocksize) 11100999df54SChris Mason { 1111f28491e0SJosef Bacik return find_extent_buffer(root->fs_info, bytenr); 11120999df54SChris Mason } 11130999df54SChris Mason 11140999df54SChris Mason struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, 11150999df54SChris Mason u64 bytenr, u32 blocksize) 11160999df54SChris Mason { 1117faa2dbf0SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 1118faa2dbf0SJosef Bacik if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state))) 1119faa2dbf0SJosef Bacik return alloc_test_extent_buffer(root->fs_info, bytenr, 1120faa2dbf0SJosef Bacik blocksize); 1121faa2dbf0SJosef Bacik #endif 1122f28491e0SJosef Bacik return alloc_extent_buffer(root->fs_info, bytenr, blocksize); 11230999df54SChris Mason } 11240999df54SChris Mason 11250999df54SChris Mason 1126e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1127e02119d5SChris Mason { 1128727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 11298aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1130e02119d5SChris Mason } 1131e02119d5SChris Mason 1132e02119d5SChris Mason int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1133e02119d5SChris Mason { 1134727011e0SChris Mason return filemap_fdatawait_range(buf->pages[0]->mapping, 11358aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1136e02119d5SChris Mason } 1137e02119d5SChris Mason 1138db94535dSChris Mason struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, 1139ca7a79adSChris Mason u32 blocksize, u64 parent_transid) 1140e20d96d6SChris Mason { 11415f39d397SChris Mason struct extent_buffer *buf = NULL; 114219c00ddcSChris Mason int ret; 114319c00ddcSChris Mason 1144db94535dSChris Mason buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 11455f39d397SChris Mason if (!buf) 1146d98237b3SChris Mason return NULL; 1147e4204dedSChris Mason 1148ca7a79adSChris Mason ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 11490f0fe8f7SFilipe David Borba Manana if (ret) { 11500f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 11510f0fe8f7SFilipe David Borba Manana return NULL; 11520f0fe8f7SFilipe David Borba Manana } 11535f39d397SChris Mason return buf; 1154ce9adaa5SChris Mason 1155eb60ceacSChris Mason } 1156eb60ceacSChris Mason 1157d5c13f92SJeff Mahoney void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, 11585f39d397SChris Mason struct extent_buffer *buf) 1159ed2ff2cbSChris Mason { 1160e2d84521SMiao Xie struct btrfs_fs_info *fs_info = root->fs_info; 1161e2d84521SMiao Xie 116255c69072SChris Mason if (btrfs_header_generation(buf) == 1163e2d84521SMiao Xie fs_info->running_transaction->transid) { 1164b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1165b4ce94deSChris Mason 1166b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1167e2d84521SMiao Xie __percpu_counter_add(&fs_info->dirty_metadata_bytes, 1168e2d84521SMiao Xie -buf->len, 1169e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1170b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1171b9473439SChris Mason btrfs_set_lock_blocking(buf); 11720b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1173925baeddSChris Mason } 11745f39d397SChris Mason } 1175ed7b63ebSJosef Bacik } 11765f39d397SChris Mason 11778257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 11788257b2dcSMiao Xie { 11798257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 11808257b2dcSMiao Xie int ret; 11818257b2dcSMiao Xie 11828257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 11838257b2dcSMiao Xie if (!writers) 11848257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 11858257b2dcSMiao Xie 11868257b2dcSMiao Xie ret = percpu_counter_init(&writers->counter, 0); 11878257b2dcSMiao Xie if (ret < 0) { 11888257b2dcSMiao Xie kfree(writers); 11898257b2dcSMiao Xie return ERR_PTR(ret); 11908257b2dcSMiao Xie } 11918257b2dcSMiao Xie 11928257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 11938257b2dcSMiao Xie return writers; 11948257b2dcSMiao Xie } 11958257b2dcSMiao Xie 11968257b2dcSMiao Xie static void 11978257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 11988257b2dcSMiao Xie { 11998257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 12008257b2dcSMiao Xie kfree(writers); 12018257b2dcSMiao Xie } 12028257b2dcSMiao Xie 1203707e8a07SDavid Sterba static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize, 1204707e8a07SDavid Sterba struct btrfs_root *root, struct btrfs_fs_info *fs_info, 1205e20d96d6SChris Mason u64 objectid) 1206d97e63b6SChris Mason { 1207cfaa7295SChris Mason root->node = NULL; 1208a28ec197SChris Mason root->commit_root = NULL; 1209db94535dSChris Mason root->sectorsize = sectorsize; 1210db94535dSChris Mason root->nodesize = nodesize; 121187ee04ebSChris Mason root->stripesize = stripesize; 121227cdeb70SMiao Xie root->state = 0; 1213d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 12140b86a832SChris Mason 12150f7d52f4SChris Mason root->objectid = objectid; 12160f7d52f4SChris Mason root->last_trans = 0; 121713a8a7c8SYan, Zheng root->highest_objectid = 0; 1218eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1219199c2a9cSMiao Xie root->nr_ordered_extents = 0; 122058176a96SJosef Bacik root->name = NULL; 12216bef4d31SEric Paris root->inode_tree = RB_ROOT; 122216cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1223f0486c68SYan, Zheng root->block_rsv = NULL; 1224d68fc57bSYan, Zheng root->orphan_block_rsv = NULL; 12250b86a832SChris Mason 12260b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 12275d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1228eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1229eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1230199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1231199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 12322ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 12332ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 1234d68fc57bSYan, Zheng spin_lock_init(&root->orphan_lock); 12355d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1236eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1237199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1238f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 12392ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 12402ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 1241a2135011SChris Mason mutex_init(&root->objectid_mutex); 1242e02119d5SChris Mason mutex_init(&root->log_mutex); 124331f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1244573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 12457237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 12467237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 12477237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 12488b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 12498b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 12507237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 12517237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 12527237f183SYan Zheng atomic_set(&root->log_writers, 0); 12532ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 12548a35d95fSJosef Bacik atomic_set(&root->orphan_inodes, 0); 1255b0feb9d9SMiao Xie atomic_set(&root->refs, 1); 12568257b2dcSMiao Xie atomic_set(&root->will_be_snapshoted, 0); 12577237f183SYan Zheng root->log_transid = 0; 1258d1433debSMiao Xie root->log_transid_committed = -1; 1259257c62e1SChris Mason root->last_log_commit = 0; 126006ea65a3SJosef Bacik if (fs_info) 1261d0c803c4SChris Mason extent_io_tree_init(&root->dirty_log_pages, 1262f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 1263017e5369SChris Mason 12643768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 12653768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 12666702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 126758176a96SJosef Bacik memset(&root->root_kobj, 0, sizeof(root->root_kobj)); 126806ea65a3SJosef Bacik if (fs_info) 12693f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 127006ea65a3SJosef Bacik else 127106ea65a3SJosef Bacik root->defrag_trans_start = 0; 127258176a96SJosef Bacik init_completion(&root->kobj_unregister); 12734d775673SChris Mason root->root_key.objectid = objectid; 12740ee5dc67SAl Viro root->anon_dev = 0; 12758ea05e3aSAlexander Block 12765f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 12773768f368SChris Mason } 12783768f368SChris Mason 1279f84a8bd6SAl Viro static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) 12806f07e42eSAl Viro { 12816f07e42eSAl Viro struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); 12826f07e42eSAl Viro if (root) 12836f07e42eSAl Viro root->fs_info = fs_info; 12846f07e42eSAl Viro return root; 12856f07e42eSAl Viro } 12866f07e42eSAl Viro 128706ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 128806ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 128906ea65a3SJosef Bacik struct btrfs_root *btrfs_alloc_dummy_root(void) 129006ea65a3SJosef Bacik { 129106ea65a3SJosef Bacik struct btrfs_root *root; 129206ea65a3SJosef Bacik 129306ea65a3SJosef Bacik root = btrfs_alloc_root(NULL); 129406ea65a3SJosef Bacik if (!root) 129506ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 1296707e8a07SDavid Sterba __setup_root(4096, 4096, 4096, root, NULL, 1); 129727cdeb70SMiao Xie set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state); 1298faa2dbf0SJosef Bacik root->alloc_bytenr = 0; 129906ea65a3SJosef Bacik 130006ea65a3SJosef Bacik return root; 130106ea65a3SJosef Bacik } 130206ea65a3SJosef Bacik #endif 130306ea65a3SJosef Bacik 130420897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 130520897f5cSArne Jansen struct btrfs_fs_info *fs_info, 130620897f5cSArne Jansen u64 objectid) 130720897f5cSArne Jansen { 130820897f5cSArne Jansen struct extent_buffer *leaf; 130920897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 131020897f5cSArne Jansen struct btrfs_root *root; 131120897f5cSArne Jansen struct btrfs_key key; 131220897f5cSArne Jansen int ret = 0; 13136463fe58SStefan Behrens uuid_le uuid; 131420897f5cSArne Jansen 131520897f5cSArne Jansen root = btrfs_alloc_root(fs_info); 131620897f5cSArne Jansen if (!root) 131720897f5cSArne Jansen return ERR_PTR(-ENOMEM); 131820897f5cSArne Jansen 1319707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1320707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, objectid); 132120897f5cSArne Jansen root->root_key.objectid = objectid; 132220897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 132320897f5cSArne Jansen root->root_key.offset = 0; 132420897f5cSArne Jansen 1325707e8a07SDavid Sterba leaf = btrfs_alloc_free_block(trans, root, root->nodesize, 132620897f5cSArne Jansen 0, objectid, NULL, 0, 0, 0); 132720897f5cSArne Jansen if (IS_ERR(leaf)) { 132820897f5cSArne Jansen ret = PTR_ERR(leaf); 13291dd05682STsutomu Itoh leaf = NULL; 133020897f5cSArne Jansen goto fail; 133120897f5cSArne Jansen } 133220897f5cSArne Jansen 133320897f5cSArne Jansen memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 133420897f5cSArne Jansen btrfs_set_header_bytenr(leaf, leaf->start); 133520897f5cSArne Jansen btrfs_set_header_generation(leaf, trans->transid); 133620897f5cSArne Jansen btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 133720897f5cSArne Jansen btrfs_set_header_owner(leaf, objectid); 133820897f5cSArne Jansen root->node = leaf; 133920897f5cSArne Jansen 13400a4e5586SRoss Kirk write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), 134120897f5cSArne Jansen BTRFS_FSID_SIZE); 134220897f5cSArne Jansen write_extent_buffer(leaf, fs_info->chunk_tree_uuid, 1343b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(leaf), 134420897f5cSArne Jansen BTRFS_UUID_SIZE); 134520897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 134620897f5cSArne Jansen 134720897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 134827cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 134920897f5cSArne Jansen 135020897f5cSArne Jansen root->root_item.flags = 0; 135120897f5cSArne Jansen root->root_item.byte_limit = 0; 135220897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 135320897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 135420897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 135520897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 135620897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 135720897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 135820897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 13596463fe58SStefan Behrens uuid_le_gen(&uuid); 13606463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 136120897f5cSArne Jansen root->root_item.drop_level = 0; 136220897f5cSArne Jansen 136320897f5cSArne Jansen key.objectid = objectid; 136420897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 136520897f5cSArne Jansen key.offset = 0; 136620897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 136720897f5cSArne Jansen if (ret) 136820897f5cSArne Jansen goto fail; 136920897f5cSArne Jansen 137020897f5cSArne Jansen btrfs_tree_unlock(leaf); 137120897f5cSArne Jansen 137220897f5cSArne Jansen return root; 13731dd05682STsutomu Itoh 13741dd05682STsutomu Itoh fail: 13751dd05682STsutomu Itoh if (leaf) { 13761dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 137759885b39STsutomu Itoh free_extent_buffer(root->commit_root); 13781dd05682STsutomu Itoh free_extent_buffer(leaf); 13791dd05682STsutomu Itoh } 13801dd05682STsutomu Itoh kfree(root); 13811dd05682STsutomu Itoh 13821dd05682STsutomu Itoh return ERR_PTR(ret); 138320897f5cSArne Jansen } 138420897f5cSArne Jansen 13857237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1386e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13870f7d52f4SChris Mason { 13880f7d52f4SChris Mason struct btrfs_root *root; 13890f7d52f4SChris Mason struct btrfs_root *tree_root = fs_info->tree_root; 13907237f183SYan Zheng struct extent_buffer *leaf; 1391e02119d5SChris Mason 13926f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1393e02119d5SChris Mason if (!root) 13947237f183SYan Zheng return ERR_PTR(-ENOMEM); 1395e02119d5SChris Mason 1396707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1397707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, 1398707e8a07SDavid Sterba BTRFS_TREE_LOG_OBJECTID); 1399e02119d5SChris Mason 1400e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1401e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1402e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 140327cdeb70SMiao Xie 14047237f183SYan Zheng /* 140527cdeb70SMiao Xie * DON'T set REF_COWS for log trees 140627cdeb70SMiao Xie * 14077237f183SYan Zheng * log trees do not get reference counted because they go away 14087237f183SYan Zheng * before a real commit is actually done. They do store pointers 14097237f183SYan Zheng * to file data extents, and those reference counts still get 14107237f183SYan Zheng * updated (along with back refs to the log tree). 14117237f183SYan Zheng */ 1412e02119d5SChris Mason 1413707e8a07SDavid Sterba leaf = btrfs_alloc_free_block(trans, root, root->nodesize, 0, 141466d7e7f0SArne Jansen BTRFS_TREE_LOG_OBJECTID, NULL, 14155581a51aSJan Schmidt 0, 0, 0); 14167237f183SYan Zheng if (IS_ERR(leaf)) { 14177237f183SYan Zheng kfree(root); 14187237f183SYan Zheng return ERR_CAST(leaf); 14197237f183SYan Zheng } 1420e02119d5SChris Mason 14215d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 14225d4f98a2SYan Zheng btrfs_set_header_bytenr(leaf, leaf->start); 14235d4f98a2SYan Zheng btrfs_set_header_generation(leaf, trans->transid); 14245d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 14255d4f98a2SYan Zheng btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 14267237f183SYan Zheng root->node = leaf; 1427e02119d5SChris Mason 1428e02119d5SChris Mason write_extent_buffer(root->node, root->fs_info->fsid, 14290a4e5586SRoss Kirk btrfs_header_fsid(), BTRFS_FSID_SIZE); 1430e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1431e02119d5SChris Mason btrfs_tree_unlock(root->node); 14327237f183SYan Zheng return root; 14337237f183SYan Zheng } 14347237f183SYan Zheng 14357237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 14367237f183SYan Zheng struct btrfs_fs_info *fs_info) 14377237f183SYan Zheng { 14387237f183SYan Zheng struct btrfs_root *log_root; 14397237f183SYan Zheng 14407237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 14417237f183SYan Zheng if (IS_ERR(log_root)) 14427237f183SYan Zheng return PTR_ERR(log_root); 14437237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 14447237f183SYan Zheng fs_info->log_root_tree = log_root; 14457237f183SYan Zheng return 0; 14467237f183SYan Zheng } 14477237f183SYan Zheng 14487237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 14497237f183SYan Zheng struct btrfs_root *root) 14507237f183SYan Zheng { 14517237f183SYan Zheng struct btrfs_root *log_root; 14527237f183SYan Zheng struct btrfs_inode_item *inode_item; 14537237f183SYan Zheng 14547237f183SYan Zheng log_root = alloc_log_tree(trans, root->fs_info); 14557237f183SYan Zheng if (IS_ERR(log_root)) 14567237f183SYan Zheng return PTR_ERR(log_root); 14577237f183SYan Zheng 14587237f183SYan Zheng log_root->last_trans = trans->transid; 14597237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14607237f183SYan Zheng 14617237f183SYan Zheng inode_item = &log_root->root_item.inode; 14623cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14633cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14643cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 1465707e8a07SDavid Sterba btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); 14663cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14677237f183SYan Zheng 14685d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14697237f183SYan Zheng 14707237f183SYan Zheng WARN_ON(root->log_root); 14717237f183SYan Zheng root->log_root = log_root; 14727237f183SYan Zheng root->log_transid = 0; 1473d1433debSMiao Xie root->log_transid_committed = -1; 1474257c62e1SChris Mason root->last_log_commit = 0; 1475e02119d5SChris Mason return 0; 1476e02119d5SChris Mason } 1477e02119d5SChris Mason 147835a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1479cb517eabSMiao Xie struct btrfs_key *key) 1480e02119d5SChris Mason { 1481e02119d5SChris Mason struct btrfs_root *root; 1482e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 14830f7d52f4SChris Mason struct btrfs_path *path; 148484234f3aSYan Zheng u64 generation; 1485db94535dSChris Mason u32 blocksize; 1486cb517eabSMiao Xie int ret; 1487cb517eabSMiao Xie 1488cb517eabSMiao Xie path = btrfs_alloc_path(); 1489cb517eabSMiao Xie if (!path) 1490cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 14910f7d52f4SChris Mason 14926f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1493cb517eabSMiao Xie if (!root) { 1494cb517eabSMiao Xie ret = -ENOMEM; 1495cb517eabSMiao Xie goto alloc_fail; 14960f7d52f4SChris Mason } 14970f7d52f4SChris Mason 1498707e8a07SDavid Sterba __setup_root(tree_root->nodesize, tree_root->sectorsize, 1499707e8a07SDavid Sterba tree_root->stripesize, root, fs_info, key->objectid); 15000f7d52f4SChris Mason 1501cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1502cb517eabSMiao Xie &root->root_item, &root->root_key); 15030f7d52f4SChris Mason if (ret) { 150413a8a7c8SYan, Zheng if (ret > 0) 150513a8a7c8SYan, Zheng ret = -ENOENT; 1506cb517eabSMiao Xie goto find_fail; 15070f7d52f4SChris Mason } 150813a8a7c8SYan, Zheng 150984234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1510707e8a07SDavid Sterba blocksize = root->nodesize; 1511db94535dSChris Mason root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 151284234f3aSYan Zheng blocksize, generation); 1513cb517eabSMiao Xie if (!root->node) { 1514cb517eabSMiao Xie ret = -ENOMEM; 1515cb517eabSMiao Xie goto find_fail; 1516cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1517cb517eabSMiao Xie ret = -EIO; 1518cb517eabSMiao Xie goto read_fail; 1519416bc658SJosef Bacik } 15205d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 152113a8a7c8SYan, Zheng out: 1522cb517eabSMiao Xie btrfs_free_path(path); 1523cb517eabSMiao Xie return root; 1524cb517eabSMiao Xie 1525cb517eabSMiao Xie read_fail: 1526cb517eabSMiao Xie free_extent_buffer(root->node); 1527cb517eabSMiao Xie find_fail: 1528cb517eabSMiao Xie kfree(root); 1529cb517eabSMiao Xie alloc_fail: 1530cb517eabSMiao Xie root = ERR_PTR(ret); 1531cb517eabSMiao Xie goto out; 1532cb517eabSMiao Xie } 1533cb517eabSMiao Xie 1534cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1535cb517eabSMiao Xie struct btrfs_key *location) 1536cb517eabSMiao Xie { 1537cb517eabSMiao Xie struct btrfs_root *root; 1538cb517eabSMiao Xie 1539cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1540cb517eabSMiao Xie if (IS_ERR(root)) 1541cb517eabSMiao Xie return root; 1542cb517eabSMiao Xie 1543cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 154427cdeb70SMiao Xie set_bit(BTRFS_ROOT_REF_COWS, &root->state); 154508fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 154608fe4db1SLi Zefan } 154713a8a7c8SYan, Zheng 15485eda7b5eSChris Mason return root; 15495eda7b5eSChris Mason } 15505eda7b5eSChris Mason 1551cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1552cb517eabSMiao Xie { 1553cb517eabSMiao Xie int ret; 15548257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1555cb517eabSMiao Xie 1556cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1557cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1558cb517eabSMiao Xie GFP_NOFS); 1559cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1560cb517eabSMiao Xie ret = -ENOMEM; 1561cb517eabSMiao Xie goto fail; 1562cb517eabSMiao Xie } 1563cb517eabSMiao Xie 15648257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 15658257b2dcSMiao Xie if (IS_ERR(writers)) { 15668257b2dcSMiao Xie ret = PTR_ERR(writers); 15678257b2dcSMiao Xie goto fail; 15688257b2dcSMiao Xie } 15698257b2dcSMiao Xie root->subv_writers = writers; 15708257b2dcSMiao Xie 1571cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 157257cdc8dbSDavid Sterba spin_lock_init(&root->ino_cache_lock); 157357cdc8dbSDavid Sterba init_waitqueue_head(&root->ino_cache_wait); 1574cb517eabSMiao Xie 1575cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1576cb517eabSMiao Xie if (ret) 15778257b2dcSMiao Xie goto free_writers; 1578cb517eabSMiao Xie return 0; 15798257b2dcSMiao Xie 15808257b2dcSMiao Xie free_writers: 15818257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 1582cb517eabSMiao Xie fail: 1583cb517eabSMiao Xie kfree(root->free_ino_ctl); 1584cb517eabSMiao Xie kfree(root->free_ino_pinned); 1585cb517eabSMiao Xie return ret; 1586cb517eabSMiao Xie } 1587cb517eabSMiao Xie 1588171170c1SSergei Trofimovich static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1589cb517eabSMiao Xie u64 root_id) 1590cb517eabSMiao Xie { 1591cb517eabSMiao Xie struct btrfs_root *root; 1592cb517eabSMiao Xie 1593cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1594cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1595cb517eabSMiao Xie (unsigned long)root_id); 1596cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1597cb517eabSMiao Xie return root; 1598cb517eabSMiao Xie } 1599cb517eabSMiao Xie 1600cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1601cb517eabSMiao Xie struct btrfs_root *root) 1602cb517eabSMiao Xie { 1603cb517eabSMiao Xie int ret; 1604cb517eabSMiao Xie 1605cb517eabSMiao Xie ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); 1606cb517eabSMiao Xie if (ret) 1607cb517eabSMiao Xie return ret; 1608cb517eabSMiao Xie 1609cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1610cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1611cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1612cb517eabSMiao Xie root); 1613cb517eabSMiao Xie if (ret == 0) 161427cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1615cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1616cb517eabSMiao Xie radix_tree_preload_end(); 1617cb517eabSMiao Xie 1618cb517eabSMiao Xie return ret; 1619cb517eabSMiao Xie } 1620cb517eabSMiao Xie 1621c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1622c00869f1SMiao Xie struct btrfs_key *location, 1623c00869f1SMiao Xie bool check_ref) 16245eda7b5eSChris Mason { 16255eda7b5eSChris Mason struct btrfs_root *root; 16265eda7b5eSChris Mason int ret; 16275eda7b5eSChris Mason 1628edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1629edbd8d4eSChris Mason return fs_info->tree_root; 1630edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1631edbd8d4eSChris Mason return fs_info->extent_root; 16328f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 16338f18cf13SChris Mason return fs_info->chunk_root; 16348f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 16358f18cf13SChris Mason return fs_info->dev_root; 16360403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 16370403e47eSYan Zheng return fs_info->csum_root; 1638bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1639bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1640bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1641f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1642f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1643f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 16444df27c4dSYan, Zheng again: 1645cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 164648475471SStefan Behrens if (root) { 1647c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 164848475471SStefan Behrens return ERR_PTR(-ENOENT); 16495eda7b5eSChris Mason return root; 165048475471SStefan Behrens } 16515eda7b5eSChris Mason 1652cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 16535eda7b5eSChris Mason if (IS_ERR(root)) 16545eda7b5eSChris Mason return root; 16553394e160SChris Mason 1656c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1657d68fc57bSYan, Zheng ret = -ENOENT; 1658d68fc57bSYan, Zheng goto fail; 1659d68fc57bSYan, Zheng } 1660d68fc57bSYan, Zheng 1661cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1662cb517eabSMiao Xie if (ret) 1663cb517eabSMiao Xie goto fail; 1664cb517eabSMiao Xie 16653f870c28SKelley Nielsen ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID, 16663f870c28SKelley Nielsen location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL); 1667d68fc57bSYan, Zheng if (ret < 0) 1668d68fc57bSYan, Zheng goto fail; 1669d68fc57bSYan, Zheng if (ret == 0) 167027cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1671d68fc57bSYan, Zheng 1672cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 16730f7d52f4SChris Mason if (ret) { 16744df27c4dSYan, Zheng if (ret == -EEXIST) { 16754df27c4dSYan, Zheng free_fs_root(root); 16764df27c4dSYan, Zheng goto again; 16770f7d52f4SChris Mason } 16784df27c4dSYan, Zheng goto fail; 16794df27c4dSYan, Zheng } 1680edbd8d4eSChris Mason return root; 16814df27c4dSYan, Zheng fail: 16824df27c4dSYan, Zheng free_fs_root(root); 16834df27c4dSYan, Zheng return ERR_PTR(ret); 1684edbd8d4eSChris Mason } 1685edbd8d4eSChris Mason 168604160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 168704160088SChris Mason { 168804160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 168904160088SChris Mason int ret = 0; 169004160088SChris Mason struct btrfs_device *device; 169104160088SChris Mason struct backing_dev_info *bdi; 1692b7967db7SChris Mason 16931f78160cSXiao Guangrong rcu_read_lock(); 16941f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1695dfe25020SChris Mason if (!device->bdev) 1696dfe25020SChris Mason continue; 169704160088SChris Mason bdi = blk_get_backing_dev_info(device->bdev); 169804160088SChris Mason if (bdi && bdi_congested(bdi, bdi_bits)) { 169904160088SChris Mason ret = 1; 170004160088SChris Mason break; 170104160088SChris Mason } 170204160088SChris Mason } 17031f78160cSXiao Guangrong rcu_read_unlock(); 170404160088SChris Mason return ret; 170504160088SChris Mason } 170604160088SChris Mason 170738b66988SChris Mason /* 1708ad081f14SJens Axboe * If this fails, caller must call bdi_destroy() to get rid of the 1709ad081f14SJens Axboe * bdi again. 1710ad081f14SJens Axboe */ 171104160088SChris Mason static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) 171204160088SChris Mason { 1713ad081f14SJens Axboe int err; 1714ad081f14SJens Axboe 1715ad081f14SJens Axboe bdi->capabilities = BDI_CAP_MAP_COPY; 1716e6d086d8SJens Axboe err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); 1717ad081f14SJens Axboe if (err) 1718ad081f14SJens Axboe return err; 1719ad081f14SJens Axboe 17204575c9ccSChris Mason bdi->ra_pages = default_backing_dev_info.ra_pages; 172104160088SChris Mason bdi->congested_fn = btrfs_congested_fn; 172204160088SChris Mason bdi->congested_data = info; 172304160088SChris Mason return 0; 172404160088SChris Mason } 172504160088SChris Mason 17268b712842SChris Mason /* 17278b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 17288b712842SChris Mason * functions. This is where read checksum verification actually happens 17298b712842SChris Mason */ 17308b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1731ce9adaa5SChris Mason { 1732ce9adaa5SChris Mason struct bio *bio; 17338b712842SChris Mason struct end_io_wq *end_io_wq; 1734ce9adaa5SChris Mason int error; 1735ce9adaa5SChris Mason 17368b712842SChris Mason end_io_wq = container_of(work, struct end_io_wq, work); 1737ce9adaa5SChris Mason bio = end_io_wq->bio; 17388b712842SChris Mason 1739ce9adaa5SChris Mason error = end_io_wq->error; 1740ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1741ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 1742ce9adaa5SChris Mason kfree(end_io_wq); 1743bc1e79acSKent Overstreet bio_endio_nodec(bio, error); 1744ce9adaa5SChris Mason } 174544b8bd7eSChris Mason 1746a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1747a74a4b97SChris Mason { 1748a74a4b97SChris Mason struct btrfs_root *root = arg; 1749d0278245SMiao Xie int again; 1750a74a4b97SChris Mason 1751a74a4b97SChris Mason do { 1752d0278245SMiao Xie again = 0; 17539d1a2a3aSDavid Sterba 1754d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 1755babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) 1756d0278245SMiao Xie goto sleep; 1757d0278245SMiao Xie 1758d0278245SMiao Xie if (!mutex_trylock(&root->fs_info->cleaner_mutex)) 1759d0278245SMiao Xie goto sleep; 1760d0278245SMiao Xie 1761dc7f370cSMiao Xie /* 1762dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1763dc7f370cSMiao Xie * during the above check and trylock. 1764dc7f370cSMiao Xie */ 1765babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) { 1766dc7f370cSMiao Xie mutex_unlock(&root->fs_info->cleaner_mutex); 1767dc7f370cSMiao Xie goto sleep; 1768dc7f370cSMiao Xie } 1769dc7f370cSMiao Xie 177024bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 17719d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 1772a74a4b97SChris Mason mutex_unlock(&root->fs_info->cleaner_mutex); 1773a74a4b97SChris Mason 1774d0278245SMiao Xie /* 177505323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 177605323cd1SMiao Xie * needn't do anything special here. 1777d0278245SMiao Xie */ 1778d0278245SMiao Xie btrfs_run_defrag_inodes(root->fs_info); 1779d0278245SMiao Xie sleep: 17809d1a2a3aSDavid Sterba if (!try_to_freeze() && !again) { 1781a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 17828929ecfaSYan, Zheng if (!kthread_should_stop()) 1783a74a4b97SChris Mason schedule(); 1784a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1785a74a4b97SChris Mason } 1786a74a4b97SChris Mason } while (!kthread_should_stop()); 1787a74a4b97SChris Mason return 0; 1788a74a4b97SChris Mason } 1789a74a4b97SChris Mason 1790a74a4b97SChris Mason static int transaction_kthread(void *arg) 1791a74a4b97SChris Mason { 1792a74a4b97SChris Mason struct btrfs_root *root = arg; 1793a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1794a74a4b97SChris Mason struct btrfs_transaction *cur; 17958929ecfaSYan, Zheng u64 transid; 1796a74a4b97SChris Mason unsigned long now; 1797a74a4b97SChris Mason unsigned long delay; 1798914b2007SJan Kara bool cannot_commit; 1799a74a4b97SChris Mason 1800a74a4b97SChris Mason do { 1801914b2007SJan Kara cannot_commit = false; 18028b87dc17SDavid Sterba delay = HZ * root->fs_info->commit_interval; 1803a74a4b97SChris Mason mutex_lock(&root->fs_info->transaction_kthread_mutex); 1804a74a4b97SChris Mason 1805a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 1806a74a4b97SChris Mason cur = root->fs_info->running_transaction; 1807a74a4b97SChris Mason if (!cur) { 1808a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1809a74a4b97SChris Mason goto sleep; 1810a74a4b97SChris Mason } 181131153d81SYan Zheng 1812a74a4b97SChris Mason now = get_seconds(); 18134a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 18148b87dc17SDavid Sterba (now < cur->start_time || 18158b87dc17SDavid Sterba now - cur->start_time < root->fs_info->commit_interval)) { 1816a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1817a74a4b97SChris Mason delay = HZ * 5; 1818a74a4b97SChris Mason goto sleep; 1819a74a4b97SChris Mason } 18208929ecfaSYan, Zheng transid = cur->transid; 1821a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 182256bec294SChris Mason 182379787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1824354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1825914b2007SJan Kara if (IS_ERR(trans)) { 1826354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1827914b2007SJan Kara cannot_commit = true; 182879787eaaSJeff Mahoney goto sleep; 1829914b2007SJan Kara } 18308929ecfaSYan, Zheng if (transid == trans->transid) { 183179787eaaSJeff Mahoney btrfs_commit_transaction(trans, root); 18328929ecfaSYan, Zheng } else { 18338929ecfaSYan, Zheng btrfs_end_transaction(trans, root); 18348929ecfaSYan, Zheng } 1835a74a4b97SChris Mason sleep: 1836a74a4b97SChris Mason wake_up_process(root->fs_info->cleaner_kthread); 1837a74a4b97SChris Mason mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1838a74a4b97SChris Mason 18394e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 18404e121c06SJosef Bacik &root->fs_info->fs_state))) 18414e121c06SJosef Bacik btrfs_cleanup_transaction(root); 1842a0acae0eSTejun Heo if (!try_to_freeze()) { 1843a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 18448929ecfaSYan, Zheng if (!kthread_should_stop() && 1845914b2007SJan Kara (!btrfs_transaction_blocked(root->fs_info) || 1846914b2007SJan Kara cannot_commit)) 1847a74a4b97SChris Mason schedule_timeout(delay); 1848a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1849a74a4b97SChris Mason } 1850a74a4b97SChris Mason } while (!kthread_should_stop()); 1851a74a4b97SChris Mason return 0; 1852a74a4b97SChris Mason } 1853a74a4b97SChris Mason 1854af31f5e5SChris Mason /* 1855af31f5e5SChris Mason * this will find the highest generation in the array of 1856af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1857af31f5e5SChris Mason * or -1 if we can't find anything. 1858af31f5e5SChris Mason * 1859af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1860af31f5e5SChris Mason * generation of the latest root in the array with the generation 1861af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1862af31f5e5SChris Mason */ 1863af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1864af31f5e5SChris Mason { 1865af31f5e5SChris Mason u64 cur; 1866af31f5e5SChris Mason int newest_index = -1; 1867af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1868af31f5e5SChris Mason int i; 1869af31f5e5SChris Mason 1870af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1871af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1872af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1873af31f5e5SChris Mason if (cur == newest_gen) 1874af31f5e5SChris Mason newest_index = i; 1875af31f5e5SChris Mason } 1876af31f5e5SChris Mason 1877af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1878af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1879af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1880af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1881af31f5e5SChris Mason if (cur == newest_gen) 1882af31f5e5SChris Mason newest_index = 0; 1883af31f5e5SChris Mason } 1884af31f5e5SChris Mason return newest_index; 1885af31f5e5SChris Mason } 1886af31f5e5SChris Mason 1887af31f5e5SChris Mason 1888af31f5e5SChris Mason /* 1889af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1890af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1891af31f5e5SChris Mason * field in the fs_info struct 1892af31f5e5SChris Mason */ 1893af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1894af31f5e5SChris Mason u64 newest_gen) 1895af31f5e5SChris Mason { 1896af31f5e5SChris Mason int newest_index = -1; 1897af31f5e5SChris Mason 1898af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1899af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1900af31f5e5SChris Mason if (newest_index == -1) { 1901af31f5e5SChris Mason info->backup_root_index = 0; 1902af31f5e5SChris Mason } else { 1903af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1904af31f5e5SChris Mason } 1905af31f5e5SChris Mason } 1906af31f5e5SChris Mason 1907af31f5e5SChris Mason /* 1908af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1909af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1910af31f5e5SChris Mason * done 1911af31f5e5SChris Mason */ 1912af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1913af31f5e5SChris Mason { 1914af31f5e5SChris Mason int next_backup; 1915af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1916af31f5e5SChris Mason int last_backup; 1917af31f5e5SChris Mason 1918af31f5e5SChris Mason next_backup = info->backup_root_index; 1919af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1920af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1921af31f5e5SChris Mason 1922af31f5e5SChris Mason /* 1923af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 1924af31f5e5SChris Mason * this happens only at umount 1925af31f5e5SChris Mason */ 1926af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 1927af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 1928af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 1929af31f5e5SChris Mason next_backup = last_backup; 1930af31f5e5SChris Mason 1931af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1932af31f5e5SChris Mason 1933af31f5e5SChris Mason /* 1934af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1935af31f5e5SChris Mason * regardless of which ones we use today 1936af31f5e5SChris Mason */ 1937af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1938af31f5e5SChris Mason 1939af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1940af31f5e5SChris Mason 1941af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1942af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1943af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1944af31f5e5SChris Mason 1945af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1946af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1947af31f5e5SChris Mason 1948af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1949af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1950af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1951af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1952af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1953af31f5e5SChris Mason 1954af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1955af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1956af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1957af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1958af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1959af31f5e5SChris Mason 19607c7e82a7SChris Mason /* 19617c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 19627c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 19637c7e82a7SChris Mason */ 19647c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 19657c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 19667c7e82a7SChris Mason info->fs_root->node->start); 1967af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1968af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1969af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1970af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19717c7e82a7SChris Mason } 1972af31f5e5SChris Mason 1973af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1974af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1975af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1976af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1977af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1978af31f5e5SChris Mason 1979af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1980af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1981af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1982af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1983af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1984af31f5e5SChris Mason 1985af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1986af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1987af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1988af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1989af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1990af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1991af31f5e5SChris Mason 1992af31f5e5SChris Mason /* 1993af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1994af31f5e5SChris Mason * for the next commit 1995af31f5e5SChris Mason */ 1996af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1997af31f5e5SChris Mason &info->super_for_commit->super_roots, 1998af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1999af31f5e5SChris Mason } 2000af31f5e5SChris Mason 2001af31f5e5SChris Mason /* 2002af31f5e5SChris Mason * this copies info out of the root backup array and back into 2003af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 2004af31f5e5SChris Mason * the array, so you send it the number of backups you've already 2005af31f5e5SChris Mason * tried and the last backup index you used. 2006af31f5e5SChris Mason * 2007af31f5e5SChris Mason * this returns -1 when it has tried all the backups 2008af31f5e5SChris Mason */ 2009af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 2010af31f5e5SChris Mason struct btrfs_super_block *super, 2011af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 2012af31f5e5SChris Mason { 2013af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2014af31f5e5SChris Mason int newest = *backup_index; 2015af31f5e5SChris Mason 2016af31f5e5SChris Mason if (*num_backups_tried == 0) { 2017af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 2018af31f5e5SChris Mason 2019af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 2020af31f5e5SChris Mason if (newest == -1) 2021af31f5e5SChris Mason return -1; 2022af31f5e5SChris Mason 2023af31f5e5SChris Mason *backup_index = newest; 2024af31f5e5SChris Mason *num_backups_tried = 1; 2025af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 2026af31f5e5SChris Mason /* we've tried all the backups, all done */ 2027af31f5e5SChris Mason return -1; 2028af31f5e5SChris Mason } else { 2029af31f5e5SChris Mason /* jump to the next oldest backup */ 2030af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 2031af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 2032af31f5e5SChris Mason *backup_index = newest; 2033af31f5e5SChris Mason *num_backups_tried += 1; 2034af31f5e5SChris Mason } 2035af31f5e5SChris Mason root_backup = super->super_roots + newest; 2036af31f5e5SChris Mason 2037af31f5e5SChris Mason btrfs_set_super_generation(super, 2038af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 2039af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2040af31f5e5SChris Mason btrfs_set_super_root_level(super, 2041af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 2042af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2043af31f5e5SChris Mason 2044af31f5e5SChris Mason /* 2045af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 2046af31f5e5SChris Mason * need a fsck 2047af31f5e5SChris Mason */ 2048af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2049af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2050af31f5e5SChris Mason return 0; 2051af31f5e5SChris Mason } 2052af31f5e5SChris Mason 20537abadb64SLiu Bo /* helper to cleanup workers */ 20547abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20557abadb64SLiu Bo { 2056dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2057afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 20585cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2059fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2060fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2061fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 2062d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2063fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 2064fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2065fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2066a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 20675b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2068e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2069736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2070a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2071fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2072a79b7d4bSChris Mason btrfs_destroy_workqueue(fs_info->extent_workers); 20737abadb64SLiu Bo } 20747abadb64SLiu Bo 20752e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20762e9f5954SRashika { 20772e9f5954SRashika if (root) { 20782e9f5954SRashika free_extent_buffer(root->node); 20792e9f5954SRashika free_extent_buffer(root->commit_root); 20802e9f5954SRashika root->node = NULL; 20812e9f5954SRashika root->commit_root = NULL; 20822e9f5954SRashika } 20832e9f5954SRashika } 20842e9f5954SRashika 2085af31f5e5SChris Mason /* helper to cleanup tree roots */ 2086af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2087af31f5e5SChris Mason { 20882e9f5954SRashika free_root_extent_buffers(info->tree_root); 2089655b09feSJosef Bacik 20902e9f5954SRashika free_root_extent_buffers(info->dev_root); 20912e9f5954SRashika free_root_extent_buffers(info->extent_root); 20922e9f5954SRashika free_root_extent_buffers(info->csum_root); 20932e9f5954SRashika free_root_extent_buffers(info->quota_root); 20942e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20952e9f5954SRashika if (chunk_root) 20962e9f5954SRashika free_root_extent_buffers(info->chunk_root); 2097af31f5e5SChris Mason } 2098af31f5e5SChris Mason 2099faa2dbf0SJosef Bacik void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) 2100171f6537SJosef Bacik { 2101171f6537SJosef Bacik int ret; 2102171f6537SJosef Bacik struct btrfs_root *gang[8]; 2103171f6537SJosef Bacik int i; 2104171f6537SJosef Bacik 2105171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2106171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2107171f6537SJosef Bacik struct btrfs_root, root_list); 2108171f6537SJosef Bacik list_del(&gang[0]->root_list); 2109171f6537SJosef Bacik 211027cdeb70SMiao Xie if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { 2111cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2112171f6537SJosef Bacik } else { 2113171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2114171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2115b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2116171f6537SJosef Bacik } 2117171f6537SJosef Bacik } 2118171f6537SJosef Bacik 2119171f6537SJosef Bacik while (1) { 2120171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2121171f6537SJosef Bacik (void **)gang, 0, 2122171f6537SJosef Bacik ARRAY_SIZE(gang)); 2123171f6537SJosef Bacik if (!ret) 2124171f6537SJosef Bacik break; 2125171f6537SJosef Bacik for (i = 0; i < ret; i++) 2126cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2127171f6537SJosef Bacik } 21281a4319ccSLiu Bo 21291a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 21301a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 21311a4319ccSLiu Bo btrfs_destroy_pinned_extent(fs_info->tree_root, 21321a4319ccSLiu Bo fs_info->pinned_extents); 21331a4319ccSLiu Bo } 2134171f6537SJosef Bacik } 2135af31f5e5SChris Mason 2136ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2137dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2138dfe25020SChris Mason char *options) 2139eb60ceacSChris Mason { 2140db94535dSChris Mason u32 sectorsize; 2141db94535dSChris Mason u32 nodesize; 2142db94535dSChris Mason u32 blocksize; 214387ee04ebSChris Mason u32 stripesize; 214484234f3aSYan Zheng u64 generation; 2145f2b636e8SJosef Bacik u64 features; 21463de4586cSChris Mason struct btrfs_key location; 2147a061fc8dSChris Mason struct buffer_head *bh; 21484d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2149815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2150f84a8bd6SAl Viro struct btrfs_root *tree_root; 21514d34b278SIlya Dryomov struct btrfs_root *extent_root; 21524d34b278SIlya Dryomov struct btrfs_root *csum_root; 21534d34b278SIlya Dryomov struct btrfs_root *chunk_root; 21544d34b278SIlya Dryomov struct btrfs_root *dev_root; 2155bcef60f2SArne Jansen struct btrfs_root *quota_root; 2156f7a81ea4SStefan Behrens struct btrfs_root *uuid_root; 2157e02119d5SChris Mason struct btrfs_root *log_tree_root; 2158eb60ceacSChris Mason int ret; 2159e58ca020SYan int err = -EINVAL; 2160af31f5e5SChris Mason int num_backups_tried = 0; 2161af31f5e5SChris Mason int backup_index = 0; 21625cdc7ad3SQu Wenruo int max_active; 21635cdc7ad3SQu Wenruo int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 216470f80175SStefan Behrens bool create_uuid_tree; 216570f80175SStefan Behrens bool check_uuid_tree; 21664543df7eSChris Mason 2167f84a8bd6SAl Viro tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); 21686f07e42eSAl Viro chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); 2169cb517eabSMiao Xie if (!tree_root || !chunk_root) { 217039279cc3SChris Mason err = -ENOMEM; 217139279cc3SChris Mason goto fail; 217239279cc3SChris Mason } 217376dda93cSYan, Zheng 217476dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 217576dda93cSYan, Zheng if (ret) { 217676dda93cSYan, Zheng err = ret; 217776dda93cSYan, Zheng goto fail; 217876dda93cSYan, Zheng } 217976dda93cSYan, Zheng 218076dda93cSYan, Zheng ret = setup_bdi(fs_info, &fs_info->bdi); 218176dda93cSYan, Zheng if (ret) { 218276dda93cSYan, Zheng err = ret; 218376dda93cSYan, Zheng goto fail_srcu; 218476dda93cSYan, Zheng } 218576dda93cSYan, Zheng 2186e2d84521SMiao Xie ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0); 2187e2d84521SMiao Xie if (ret) { 2188e2d84521SMiao Xie err = ret; 2189e2d84521SMiao Xie goto fail_bdi; 2190e2d84521SMiao Xie } 2191e2d84521SMiao Xie fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * 2192e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2193e2d84521SMiao Xie 2194963d678bSMiao Xie ret = percpu_counter_init(&fs_info->delalloc_bytes, 0); 2195963d678bSMiao Xie if (ret) { 2196963d678bSMiao Xie err = ret; 2197963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2198963d678bSMiao Xie } 2199963d678bSMiao Xie 2200c404e0dcSMiao Xie ret = percpu_counter_init(&fs_info->bio_counter, 0); 2201c404e0dcSMiao Xie if (ret) { 2202c404e0dcSMiao Xie err = ret; 2203c404e0dcSMiao Xie goto fail_delalloc_bytes; 2204c404e0dcSMiao Xie } 2205c404e0dcSMiao Xie 220676dda93cSYan, Zheng fs_info->btree_inode = new_inode(sb); 220776dda93cSYan, Zheng if (!fs_info->btree_inode) { 220876dda93cSYan, Zheng err = -ENOMEM; 2209c404e0dcSMiao Xie goto fail_bio_counter; 221076dda93cSYan, Zheng } 221176dda93cSYan, Zheng 2212a6591715SChris Mason mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 22131561dedaSMiao Xie 221476dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2215f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 22168fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2217facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 221824bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2219eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 222011833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2221eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2222a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 222376dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 222424bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 22254cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 22262bf64758SJosef Bacik spin_lock_init(&fs_info->free_chunk_lock); 2227f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2228ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2229fcebe456SJosef Bacik spin_lock_init(&fs_info->qgroup_op_lock); 2230f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 2231f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 22327585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2233573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2234de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 223519c00ddcSChris Mason 223658176a96SJosef Bacik init_completion(&fs_info->kobj_unregister); 22370b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 22386324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2239f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 22400b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 224166d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 224266d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 224366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, 224466d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELALLOC); 224566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 224666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 224766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 224866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 224966d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2250cb03c743SChris Mason atomic_set(&fs_info->nr_async_submits, 0); 2251771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 22528c8bee1dSChris Mason atomic_set(&fs_info->async_submit_draining, 0); 22530986fe9eSChris Mason atomic_set(&fs_info->nr_async_bios, 0); 22544cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2255fcebe456SJosef Bacik atomic_set(&fs_info->qgroup_op_seq, 0); 2256fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2257e20d96d6SChris Mason fs_info->sb = sb; 22586f568d35SChris Mason fs_info->max_inline = 8192 * 1024; 22599ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 22604cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 22612bf64758SJosef Bacik fs_info->free_chunk_space = 0; 2262f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 22638b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 22640a2b2a84SJosef Bacik fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64); 226590519d66SArne Jansen /* readahead state */ 226690519d66SArne Jansen INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); 226790519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2268c8b97818SChris Mason 2269b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2270b34b086cSChris Mason num_online_cpus() + 2, 8); 22710afbaf8cSChris Mason 2272199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2273199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 227416cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 227516cdcec7SMiao Xie GFP_NOFS); 227616cdcec7SMiao Xie if (!fs_info->delayed_root) { 227716cdcec7SMiao Xie err = -ENOMEM; 227816cdcec7SMiao Xie goto fail_iput; 227916cdcec7SMiao Xie } 228016cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 22813eaa2885SChris Mason 2282a2de733cSArne Jansen mutex_init(&fs_info->scrub_lock); 2283a2de733cSArne Jansen atomic_set(&fs_info->scrubs_running, 0); 2284a2de733cSArne Jansen atomic_set(&fs_info->scrub_pause_req, 0); 2285a2de733cSArne Jansen atomic_set(&fs_info->scrubs_paused, 0); 2286a2de733cSArne Jansen atomic_set(&fs_info->scrub_cancel_req, 0); 2287c404e0dcSMiao Xie init_waitqueue_head(&fs_info->replace_wait); 2288a2de733cSArne Jansen init_waitqueue_head(&fs_info->scrub_pause_wait); 2289a2de733cSArne Jansen fs_info->scrub_workers_refcnt = 0; 229021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 229121adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 229221adbd5cSStefan Behrens #endif 2293a2de733cSArne Jansen 2294c9e9f97bSIlya Dryomov spin_lock_init(&fs_info->balance_lock); 2295c9e9f97bSIlya Dryomov mutex_init(&fs_info->balance_mutex); 2296837d5b6eSIlya Dryomov atomic_set(&fs_info->balance_running, 0); 2297837d5b6eSIlya Dryomov atomic_set(&fs_info->balance_pause_req, 0); 2298a7e99c69SIlya Dryomov atomic_set(&fs_info->balance_cancel_req, 0); 2299c9e9f97bSIlya Dryomov fs_info->balance_ctl = NULL; 2300837d5b6eSIlya Dryomov init_waitqueue_head(&fs_info->balance_wait_q); 230121c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2302a061fc8dSChris Mason 23030afbaf8cSChris Mason sb->s_blocksize = 4096; 23040afbaf8cSChris Mason sb->s_blocksize_bits = blksize_bits(4096); 230532a88aa1SJens Axboe sb->s_bdi = &fs_info->bdi; 23060afbaf8cSChris Mason 230776dda93cSYan, Zheng fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 2308bfe86848SMiklos Szeredi set_nlink(fs_info->btree_inode, 1); 23090afbaf8cSChris Mason /* 23100afbaf8cSChris Mason * we set the i_size on the btree inode to the max possible int. 23110afbaf8cSChris Mason * the real end of the address space is determined by all of 23120afbaf8cSChris Mason * the devices in the system 23130afbaf8cSChris Mason */ 23140afbaf8cSChris Mason fs_info->btree_inode->i_size = OFFSET_MAX; 2315d98237b3SChris Mason fs_info->btree_inode->i_mapping->a_ops = &btree_aops; 231604160088SChris Mason fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; 231704160088SChris Mason 23185d4f98a2SYan Zheng RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); 2319d1310b2eSChris Mason extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, 2320f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 23210b32f4bbSJosef Bacik BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; 2322a8067e02SDavid Sterba extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); 23230da5468fSChris Mason 2324d1310b2eSChris Mason BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; 2325d1310b2eSChris Mason 23260f7d52f4SChris Mason BTRFS_I(fs_info->btree_inode)->root = tree_root; 23270f7d52f4SChris Mason memset(&BTRFS_I(fs_info->btree_inode)->location, 0, 23280f7d52f4SChris Mason sizeof(struct btrfs_key)); 232972ac3c0dSJosef Bacik set_bit(BTRFS_INODE_DUMMY, 233072ac3c0dSJosef Bacik &BTRFS_I(fs_info->btree_inode)->runtime_flags); 2331778ba82bSFilipe David Borba Manana btrfs_insert_inode_hash(fs_info->btree_inode); 233239279cc3SChris Mason 2333e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 23346bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2335a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2336925baeddSChris Mason 233711833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[0], 2338f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 233911833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[1], 2340f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 234111833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2342e6dcd2dcSChris Mason fs_info->do_barriers = 1; 2343f9295749SChris Mason 2344d98237b3SChris Mason 23455a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 23469ffba8cdSJosef Bacik mutex_init(&fs_info->ordered_extent_flush_mutex); 234730ae8467SChris Mason mutex_init(&fs_info->tree_log_mutex); 2348925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2349925baeddSChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2350a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 2351a74a4b97SChris Mason mutex_init(&fs_info->volume_mutex); 23529e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2353c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 235476dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2355803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2356e922e087SStefan Behrens fs_info->dev_replace.lock_owner = 0; 2357e922e087SStefan Behrens atomic_set(&fs_info->dev_replace.nesting_level, 0); 2358e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2359e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock_management_lock); 2360e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock); 2361fa9c0d79SChris Mason 2362416ac51dSArne Jansen spin_lock_init(&fs_info->qgroup_lock); 2363f2f6ed3dSWang Shilong mutex_init(&fs_info->qgroup_ioctl_lock); 2364416ac51dSArne Jansen fs_info->qgroup_tree = RB_ROOT; 2365fcebe456SJosef Bacik fs_info->qgroup_op_tree = RB_ROOT; 2366416ac51dSArne Jansen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2367416ac51dSArne Jansen fs_info->qgroup_seq = 1; 2368416ac51dSArne Jansen fs_info->quota_enabled = 0; 2369416ac51dSArne Jansen fs_info->pending_quota_state = 0; 23701e8f9158SWang Shilong fs_info->qgroup_ulist = NULL; 23712f232036SJan Schmidt mutex_init(&fs_info->qgroup_rescan_lock); 2372416ac51dSArne Jansen 2373fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2374fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2375fa9c0d79SChris Mason 23767d9eb12cSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 23773768f368SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2378bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 23794854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 23809a8dd150SChris Mason 238153b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 238253b381b3SDavid Woodhouse if (ret) { 238383c8266aSDavid Sterba err = ret; 238453b381b3SDavid Woodhouse goto fail_alloc; 238553b381b3SDavid Woodhouse } 238653b381b3SDavid Woodhouse 2387707e8a07SDavid Sterba __setup_root(4096, 4096, 4096, tree_root, 23882c90e5d6SChris Mason fs_info, BTRFS_ROOT_TREE_OBJECTID); 23897eccb903SChris Mason 23903c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 23911104a885SDavid Sterba 23921104a885SDavid Sterba /* 23931104a885SDavid Sterba * Read super block and check the signature bytes only 23941104a885SDavid Sterba */ 2395a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 239620b45077SDave Young if (!bh) { 239720b45077SDave Young err = -EINVAL; 239816cdcec7SMiao Xie goto fail_alloc; 239920b45077SDave Young } 240039279cc3SChris Mason 24011104a885SDavid Sterba /* 24021104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 24031104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 24041104a885SDavid Sterba */ 24051104a885SDavid Sterba if (btrfs_check_super_csum(bh->b_data)) { 2406efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); 24071104a885SDavid Sterba err = -EINVAL; 24081104a885SDavid Sterba goto fail_alloc; 24091104a885SDavid Sterba } 24101104a885SDavid Sterba 24111104a885SDavid Sterba /* 24121104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 24131104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 24141104a885SDavid Sterba * the whole block of INFO_SIZE 24151104a885SDavid Sterba */ 24166c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 24176c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 24186c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2419a061fc8dSChris Mason brelse(bh); 24205f39d397SChris Mason 24216c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 24220b86a832SChris Mason 24231104a885SDavid Sterba ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); 24241104a885SDavid Sterba if (ret) { 2425efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); 24261104a885SDavid Sterba err = -EINVAL; 24271104a885SDavid Sterba goto fail_alloc; 24281104a885SDavid Sterba } 24291104a885SDavid Sterba 24306c41761fSDavid Sterba disk_super = fs_info->super_copy; 24310f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 243216cdcec7SMiao Xie goto fail_alloc; 24330f7d52f4SChris Mason 2434acce952bSliubo /* check FS state, whether FS is broken. */ 243587533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 243687533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2437acce952bSliubo 243875e7cb7fSLiu Bo /* 2439af31f5e5SChris Mason * run through our array of backup supers and setup 2440af31f5e5SChris Mason * our ring pointer to the oldest one 2441af31f5e5SChris Mason */ 2442af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2443af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2444af31f5e5SChris Mason 2445af31f5e5SChris Mason /* 244675e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 244775e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 244875e7cb7fSLiu Bo */ 244975e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 245075e7cb7fSLiu Bo 24512b82032cSYan Zheng ret = btrfs_parse_options(tree_root, options); 24522b82032cSYan Zheng if (ret) { 24532b82032cSYan Zheng err = ret; 245416cdcec7SMiao Xie goto fail_alloc; 24552b82032cSYan Zheng } 2456dfe25020SChris Mason 2457f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2458f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2459f2b636e8SJosef Bacik if (features) { 2460f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount because of " 2461f2b636e8SJosef Bacik "unsupported optional features (%Lx).\n", 2462c1c9ff7cSGeert Uytterhoeven features); 2463f2b636e8SJosef Bacik err = -EINVAL; 246416cdcec7SMiao Xie goto fail_alloc; 2465f2b636e8SJosef Bacik } 2466f2b636e8SJosef Bacik 2467707e8a07SDavid Sterba /* 2468707e8a07SDavid Sterba * Leafsize and nodesize were always equal, this is only a sanity check. 2469707e8a07SDavid Sterba */ 2470707e8a07SDavid Sterba if (le32_to_cpu(disk_super->__unused_leafsize) != 2471727011e0SChris Mason btrfs_super_nodesize(disk_super)) { 2472727011e0SChris Mason printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2473727011e0SChris Mason "blocksizes don't match. node %d leaf %d\n", 2474727011e0SChris Mason btrfs_super_nodesize(disk_super), 2475707e8a07SDavid Sterba le32_to_cpu(disk_super->__unused_leafsize)); 2476727011e0SChris Mason err = -EINVAL; 2477727011e0SChris Mason goto fail_alloc; 2478727011e0SChris Mason } 2479707e8a07SDavid Sterba if (btrfs_super_nodesize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { 2480727011e0SChris Mason printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2481727011e0SChris Mason "blocksize (%d) was too large\n", 2482707e8a07SDavid Sterba btrfs_super_nodesize(disk_super)); 2483727011e0SChris Mason err = -EINVAL; 2484727011e0SChris Mason goto fail_alloc; 2485727011e0SChris Mason } 2486727011e0SChris Mason 24875d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 24885d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 248969e380d1SLi Zefan if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 2490a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 2491727011e0SChris Mason 24923173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 2493efe120a0SFrank Holton printk(KERN_ERR "BTRFS: has skinny extents\n"); 24943173a18fSJosef Bacik 2495727011e0SChris Mason /* 2496727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2497727011e0SChris Mason * they are bigger than the page size 2498727011e0SChris Mason */ 2499707e8a07SDavid Sterba if (btrfs_super_nodesize(disk_super) > PAGE_CACHE_SIZE) { 2500727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 2501efe120a0SFrank Holton printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); 2502727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2503727011e0SChris Mason } 2504727011e0SChris Mason 2505bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2506bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2507bc3f116fSChris Mason stripesize = btrfs_super_stripesize(disk_super); 2508707e8a07SDavid Sterba fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 2509963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2510bc3f116fSChris Mason 2511bc3f116fSChris Mason /* 2512bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2513bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2514bc3f116fSChris Mason */ 2515bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2516707e8a07SDavid Sterba (sectorsize != nodesize)) { 2517efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes " 2518bc3f116fSChris Mason "are not allowed for mixed block groups on %s\n", 2519bc3f116fSChris Mason sb->s_id); 2520bc3f116fSChris Mason goto fail_alloc; 2521bc3f116fSChris Mason } 2522bc3f116fSChris Mason 2523ceda0864SMiao Xie /* 2524ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2525ceda0864SMiao Xie * update the flag. 2526ceda0864SMiao Xie */ 25275d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 25285d4f98a2SYan Zheng 2529f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2530f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2531f2b636e8SJosef Bacik if (!(sb->s_flags & MS_RDONLY) && features) { 2532f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " 2533f2b636e8SJosef Bacik "unsupported option features (%Lx).\n", 2534c1c9ff7cSGeert Uytterhoeven features); 2535f2b636e8SJosef Bacik err = -EINVAL; 253616cdcec7SMiao Xie goto fail_alloc; 2537f2b636e8SJosef Bacik } 253861d92c32SChris Mason 25395cdc7ad3SQu Wenruo max_active = fs_info->thread_pool_size; 254061d92c32SChris Mason 25415cdc7ad3SQu Wenruo fs_info->workers = 25425cdc7ad3SQu Wenruo btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI, 25435cdc7ad3SQu Wenruo max_active, 16); 2544c8b97818SChris Mason 2545afe3d242SQu Wenruo fs_info->delalloc_workers = 2546afe3d242SQu Wenruo btrfs_alloc_workqueue("delalloc", flags, max_active, 2); 2547771ed689SChris Mason 2548a44903abSQu Wenruo fs_info->flush_workers = 2549a44903abSQu Wenruo btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0); 25508ccf6f19SMiao Xie 2551e66f0bb1SQu Wenruo fs_info->caching_workers = 2552e66f0bb1SQu Wenruo btrfs_alloc_workqueue("cache", flags, max_active, 0); 255361b49440SChris Mason 2554a8c93d4eSQu Wenruo /* 2555a8c93d4eSQu Wenruo * a higher idle thresh on the submit workers makes it much more 255661b49440SChris Mason * likely that bios will be send down in a sane order to the 255761b49440SChris Mason * devices 255861b49440SChris Mason */ 2559a8c93d4eSQu Wenruo fs_info->submit_workers = 2560a8c93d4eSQu Wenruo btrfs_alloc_workqueue("submit", flags, 2561a8c93d4eSQu Wenruo min_t(u64, fs_devices->num_devices, 2562a8c93d4eSQu Wenruo max_active), 64); 256353863232SChris Mason 2564dc6e3209SQu Wenruo fs_info->fixup_workers = 2565dc6e3209SQu Wenruo btrfs_alloc_workqueue("fixup", flags, 1, 0); 256661b49440SChris Mason 256761b49440SChris Mason /* 256861b49440SChris Mason * endios are largely parallel and should have a very 256961b49440SChris Mason * low idle thresh 257061b49440SChris Mason */ 2571fccb5d86SQu Wenruo fs_info->endio_workers = 2572fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio", flags, max_active, 4); 2573fccb5d86SQu Wenruo fs_info->endio_meta_workers = 2574fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-meta", flags, max_active, 4); 2575fccb5d86SQu Wenruo fs_info->endio_meta_write_workers = 2576fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2); 2577fccb5d86SQu Wenruo fs_info->endio_raid56_workers = 2578fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4); 2579d05a33acSQu Wenruo fs_info->rmw_workers = 2580d05a33acSQu Wenruo btrfs_alloc_workqueue("rmw", flags, max_active, 2); 2581fccb5d86SQu Wenruo fs_info->endio_write_workers = 2582fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-write", flags, max_active, 2); 2583fccb5d86SQu Wenruo fs_info->endio_freespace_worker = 2584fccb5d86SQu Wenruo btrfs_alloc_workqueue("freespace-write", flags, max_active, 0); 25855b3bc44eSQu Wenruo fs_info->delayed_workers = 25865b3bc44eSQu Wenruo btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0); 2587736cfa15SQu Wenruo fs_info->readahead_workers = 2588736cfa15SQu Wenruo btrfs_alloc_workqueue("readahead", flags, max_active, 2); 2589fc97fab0SQu Wenruo fs_info->qgroup_rescan_workers = 2590fc97fab0SQu Wenruo btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0); 2591a79b7d4bSChris Mason fs_info->extent_workers = 2592a79b7d4bSChris Mason btrfs_alloc_workqueue("extent-refs", flags, 2593a79b7d4bSChris Mason min_t(u64, fs_devices->num_devices, 2594a79b7d4bSChris Mason max_active), 8); 2595b51912c9SChris Mason 2596a8c93d4eSQu Wenruo if (!(fs_info->workers && fs_info->delalloc_workers && 2597fccb5d86SQu Wenruo fs_info->submit_workers && fs_info->flush_workers && 2598fccb5d86SQu Wenruo fs_info->endio_workers && fs_info->endio_meta_workers && 2599fccb5d86SQu Wenruo fs_info->endio_meta_write_workers && 2600fccb5d86SQu Wenruo fs_info->endio_write_workers && fs_info->endio_raid56_workers && 2601e66f0bb1SQu Wenruo fs_info->endio_freespace_worker && fs_info->rmw_workers && 2602dc6e3209SQu Wenruo fs_info->caching_workers && fs_info->readahead_workers && 2603fc97fab0SQu Wenruo fs_info->fixup_workers && fs_info->delayed_workers && 2604a79b7d4bSChris Mason fs_info->fixup_workers && fs_info->extent_workers && 2605fc97fab0SQu Wenruo fs_info->qgroup_rescan_workers)) { 2606fed425c7SIlya Dryomov err = -ENOMEM; 26070dc3b84aSJosef Bacik goto fail_sb_buffer; 26080dc3b84aSJosef Bacik } 26094543df7eSChris Mason 26104575c9ccSChris Mason fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); 2611c8b97818SChris Mason fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 2612c8b97818SChris Mason 4 * 1024 * 1024 / PAGE_CACHE_SIZE); 26134575c9ccSChris Mason 2614db94535dSChris Mason tree_root->nodesize = nodesize; 2615db94535dSChris Mason tree_root->sectorsize = sectorsize; 261687ee04ebSChris Mason tree_root->stripesize = stripesize; 2617a061fc8dSChris Mason 2618a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2619a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2620db94535dSChris Mason 26213cae210fSQu Wenruo if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) { 2622efe120a0SFrank Holton printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id); 262339279cc3SChris Mason goto fail_sb_buffer; 262439279cc3SChris Mason } 262519c00ddcSChris Mason 26268d082fb7SLiu Bo if (sectorsize != PAGE_SIZE) { 2627efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) " 26288d082fb7SLiu Bo "found on %s\n", (unsigned long)sectorsize, sb->s_id); 2629941b2ddfSKeith Mannthey goto fail_sb_buffer; 2630941b2ddfSKeith Mannthey } 2631941b2ddfSKeith Mannthey 2632925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 2633e4404d6eSYan Zheng ret = btrfs_read_sys_array(tree_root); 2634925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 263584eed90fSChris Mason if (ret) { 2636efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read the system " 2637d397712bSChris Mason "array on %s\n", sb->s_id); 26385d4f98a2SYan Zheng goto fail_sb_buffer; 263984eed90fSChris Mason } 26400b86a832SChris Mason 2641707e8a07SDavid Sterba blocksize = tree_root->nodesize; 264284234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 26430b86a832SChris Mason 2644707e8a07SDavid Sterba __setup_root(nodesize, sectorsize, stripesize, chunk_root, 2645707e8a07SDavid Sterba fs_info, BTRFS_CHUNK_TREE_OBJECTID); 26460b86a832SChris Mason 26470b86a832SChris Mason chunk_root->node = read_tree_block(chunk_root, 26480b86a832SChris Mason btrfs_super_chunk_root(disk_super), 264984234f3aSYan Zheng blocksize, generation); 2650416bc658SJosef Bacik if (!chunk_root->node || 2651416bc658SJosef Bacik !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { 2652efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n", 265383121942SDavid Woodhouse sb->s_id); 2654af31f5e5SChris Mason goto fail_tree_roots; 265583121942SDavid Woodhouse } 26565d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 26575d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 26580b86a832SChris Mason 2659e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2660b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2661e17cade2SChris Mason 26620b86a832SChris Mason ret = btrfs_read_chunk_tree(chunk_root); 26632b82032cSYan Zheng if (ret) { 2664efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n", 2665d397712bSChris Mason sb->s_id); 2666af31f5e5SChris Mason goto fail_tree_roots; 26672b82032cSYan Zheng } 26680b86a832SChris Mason 26698dabb742SStefan Behrens /* 26708dabb742SStefan Behrens * keep the device that is marked to be the target device for the 26718dabb742SStefan Behrens * dev_replace procedure 26728dabb742SStefan Behrens */ 26738dabb742SStefan Behrens btrfs_close_extra_devices(fs_info, fs_devices, 0); 2674dfe25020SChris Mason 2675a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 2676efe120a0SFrank Holton printk(KERN_CRIT "BTRFS: failed to read devices on %s\n", 2677a6b0d5c8SChris Mason sb->s_id); 2678a6b0d5c8SChris Mason goto fail_tree_roots; 2679a6b0d5c8SChris Mason } 2680a6b0d5c8SChris Mason 2681af31f5e5SChris Mason retry_root_backup: 2682707e8a07SDavid Sterba blocksize = tree_root->nodesize; 268384234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 26840b86a832SChris Mason 2685e20d96d6SChris Mason tree_root->node = read_tree_block(tree_root, 2686db94535dSChris Mason btrfs_super_root(disk_super), 268784234f3aSYan Zheng blocksize, generation); 2688af31f5e5SChris Mason if (!tree_root->node || 2689af31f5e5SChris Mason !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { 2690efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", 269183121942SDavid Woodhouse sb->s_id); 2692af31f5e5SChris Mason 2693af31f5e5SChris Mason goto recovery_tree_root; 269483121942SDavid Woodhouse } 2695af31f5e5SChris Mason 26965d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 26975d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 269869e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 2699db94535dSChris Mason 2700cb517eabSMiao Xie location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 2701cb517eabSMiao Xie location.type = BTRFS_ROOT_ITEM_KEY; 2702cb517eabSMiao Xie location.offset = 0; 2703cb517eabSMiao Xie 2704cb517eabSMiao Xie extent_root = btrfs_read_tree_root(tree_root, &location); 2705cb517eabSMiao Xie if (IS_ERR(extent_root)) { 2706cb517eabSMiao Xie ret = PTR_ERR(extent_root); 2707af31f5e5SChris Mason goto recovery_tree_root; 2708cb517eabSMiao Xie } 270927cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state); 2710cb517eabSMiao Xie fs_info->extent_root = extent_root; 27110b86a832SChris Mason 2712cb517eabSMiao Xie location.objectid = BTRFS_DEV_TREE_OBJECTID; 2713cb517eabSMiao Xie dev_root = btrfs_read_tree_root(tree_root, &location); 2714cb517eabSMiao Xie if (IS_ERR(dev_root)) { 2715cb517eabSMiao Xie ret = PTR_ERR(dev_root); 2716af31f5e5SChris Mason goto recovery_tree_root; 2717cb517eabSMiao Xie } 271827cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state); 2719cb517eabSMiao Xie fs_info->dev_root = dev_root; 2720cb517eabSMiao Xie btrfs_init_devices_late(fs_info); 27213768f368SChris Mason 2722cb517eabSMiao Xie location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2723cb517eabSMiao Xie csum_root = btrfs_read_tree_root(tree_root, &location); 2724cb517eabSMiao Xie if (IS_ERR(csum_root)) { 2725cb517eabSMiao Xie ret = PTR_ERR(csum_root); 2726af31f5e5SChris Mason goto recovery_tree_root; 2727cb517eabSMiao Xie } 272827cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state); 2729cb517eabSMiao Xie fs_info->csum_root = csum_root; 2730d20f7043SChris Mason 2731cb517eabSMiao Xie location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2732cb517eabSMiao Xie quota_root = btrfs_read_tree_root(tree_root, &location); 2733cb517eabSMiao Xie if (!IS_ERR(quota_root)) { 273427cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, "a_root->state); 2735bcef60f2SArne Jansen fs_info->quota_enabled = 1; 2736bcef60f2SArne Jansen fs_info->pending_quota_state = 1; 2737cb517eabSMiao Xie fs_info->quota_root = quota_root; 2738bcef60f2SArne Jansen } 2739bcef60f2SArne Jansen 2740f7a81ea4SStefan Behrens location.objectid = BTRFS_UUID_TREE_OBJECTID; 2741f7a81ea4SStefan Behrens uuid_root = btrfs_read_tree_root(tree_root, &location); 2742f7a81ea4SStefan Behrens if (IS_ERR(uuid_root)) { 2743f7a81ea4SStefan Behrens ret = PTR_ERR(uuid_root); 2744f7a81ea4SStefan Behrens if (ret != -ENOENT) 2745f7a81ea4SStefan Behrens goto recovery_tree_root; 2746f7a81ea4SStefan Behrens create_uuid_tree = true; 274770f80175SStefan Behrens check_uuid_tree = false; 2748f7a81ea4SStefan Behrens } else { 274927cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state); 2750f7a81ea4SStefan Behrens fs_info->uuid_root = uuid_root; 275170f80175SStefan Behrens create_uuid_tree = false; 275270f80175SStefan Behrens check_uuid_tree = 275370f80175SStefan Behrens generation != btrfs_super_uuid_tree_generation(disk_super); 2754f7a81ea4SStefan Behrens } 2755f7a81ea4SStefan Behrens 27568929ecfaSYan, Zheng fs_info->generation = generation; 27578929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 27588929ecfaSYan, Zheng 275968310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 276068310a5eSIlya Dryomov if (ret) { 2761efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to recover balance\n"); 276268310a5eSIlya Dryomov goto fail_block_groups; 276368310a5eSIlya Dryomov } 276468310a5eSIlya Dryomov 2765733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 2766733f4fbbSStefan Behrens if (ret) { 2767efe120a0SFrank Holton printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", 2768733f4fbbSStefan Behrens ret); 2769733f4fbbSStefan Behrens goto fail_block_groups; 2770733f4fbbSStefan Behrens } 2771733f4fbbSStefan Behrens 27728dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 27738dabb742SStefan Behrens if (ret) { 2774efe120a0SFrank Holton pr_err("BTRFS: failed to init dev_replace: %d\n", ret); 27758dabb742SStefan Behrens goto fail_block_groups; 27768dabb742SStefan Behrens } 27778dabb742SStefan Behrens 27788dabb742SStefan Behrens btrfs_close_extra_devices(fs_info, fs_devices, 1); 27798dabb742SStefan Behrens 27805ac1d209SJeff Mahoney ret = btrfs_sysfs_add_one(fs_info); 27815ac1d209SJeff Mahoney if (ret) { 2782efe120a0SFrank Holton pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); 27835ac1d209SJeff Mahoney goto fail_block_groups; 27845ac1d209SJeff Mahoney } 27855ac1d209SJeff Mahoney 2786c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 2787c59021f8Sliubo if (ret) { 2788efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); 27892365dd3cSAnand Jain goto fail_sysfs; 2790c59021f8Sliubo } 2791c59021f8Sliubo 27921b1d1f66SJosef Bacik ret = btrfs_read_block_groups(extent_root); 27931b1d1f66SJosef Bacik if (ret) { 2794efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); 27952365dd3cSAnand Jain goto fail_sysfs; 27961b1d1f66SJosef Bacik } 27975af3e8ccSStefan Behrens fs_info->num_tolerated_disk_barrier_failures = 27985af3e8ccSStefan Behrens btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); 2799292fd7fcSStefan Behrens if (fs_info->fs_devices->missing_devices > 2800292fd7fcSStefan Behrens fs_info->num_tolerated_disk_barrier_failures && 2801292fd7fcSStefan Behrens !(sb->s_flags & MS_RDONLY)) { 2802efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: " 2803efe120a0SFrank Holton "too many missing devices, writeable mount is not allowed\n"); 28042365dd3cSAnand Jain goto fail_sysfs; 2805292fd7fcSStefan Behrens } 28069078a3e1SChris Mason 2807a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 2808a74a4b97SChris Mason "btrfs-cleaner"); 280957506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 28102365dd3cSAnand Jain goto fail_sysfs; 2811a74a4b97SChris Mason 2812a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 2813a74a4b97SChris Mason tree_root, 2814a74a4b97SChris Mason "btrfs-transaction"); 281557506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 28163f157a2fSChris Mason goto fail_cleaner; 2817a74a4b97SChris Mason 2818c289811cSChris Mason if (!btrfs_test_opt(tree_root, SSD) && 2819c289811cSChris Mason !btrfs_test_opt(tree_root, NOSSD) && 2820c289811cSChris Mason !fs_info->fs_devices->rotating) { 2821efe120a0SFrank Holton printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " 2822c289811cSChris Mason "mode\n"); 2823c289811cSChris Mason btrfs_set_opt(fs_info->mount_opt, SSD); 2824c289811cSChris Mason } 2825c289811cSChris Mason 28263818aea2SQu Wenruo /* Set the real inode map cache flag */ 28273818aea2SQu Wenruo if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE)) 28283818aea2SQu Wenruo btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE); 28293818aea2SQu Wenruo 283021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 283121adbd5cSStefan Behrens if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { 283221adbd5cSStefan Behrens ret = btrfsic_mount(tree_root, fs_devices, 283321adbd5cSStefan Behrens btrfs_test_opt(tree_root, 283421adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 283521adbd5cSStefan Behrens 1 : 0, 283621adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 283721adbd5cSStefan Behrens if (ret) 2838efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to initialize" 283921adbd5cSStefan Behrens " integrity check module %s\n", sb->s_id); 284021adbd5cSStefan Behrens } 284121adbd5cSStefan Behrens #endif 2842bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 2843bcef60f2SArne Jansen if (ret) 2844bcef60f2SArne Jansen goto fail_trans_kthread; 284521adbd5cSStefan Behrens 2846acce952bSliubo /* do not make disk changes in broken FS */ 284768ce9682SStefan Behrens if (btrfs_super_log_root(disk_super) != 0) { 2848e02119d5SChris Mason u64 bytenr = btrfs_super_log_root(disk_super); 2849d18a2c44SChris Mason 28507c2ca468SChris Mason if (fs_devices->rw_devices == 0) { 2851efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: log replay required " 2852d397712bSChris Mason "on RO media\n"); 28537c2ca468SChris Mason err = -EIO; 2854bcef60f2SArne Jansen goto fail_qgroup; 28557c2ca468SChris Mason } 2856707e8a07SDavid Sterba blocksize = tree_root->nodesize; 2857e02119d5SChris Mason 28586f07e42eSAl Viro log_tree_root = btrfs_alloc_root(fs_info); 2859676e4c86SDan Carpenter if (!log_tree_root) { 2860676e4c86SDan Carpenter err = -ENOMEM; 2861bcef60f2SArne Jansen goto fail_qgroup; 2862676e4c86SDan Carpenter } 2863e02119d5SChris Mason 2864707e8a07SDavid Sterba __setup_root(nodesize, sectorsize, stripesize, 2865e02119d5SChris Mason log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 2866e02119d5SChris Mason 2867e02119d5SChris Mason log_tree_root->node = read_tree_block(tree_root, bytenr, 286884234f3aSYan Zheng blocksize, 286984234f3aSYan Zheng generation + 1); 2870416bc658SJosef Bacik if (!log_tree_root->node || 2871416bc658SJosef Bacik !extent_buffer_uptodate(log_tree_root->node)) { 2872efe120a0SFrank Holton printk(KERN_ERR "BTRFS: failed to read log tree\n"); 2873416bc658SJosef Bacik free_extent_buffer(log_tree_root->node); 2874416bc658SJosef Bacik kfree(log_tree_root); 287528c16cbbSWang Shilong goto fail_qgroup; 2876416bc658SJosef Bacik } 287779787eaaSJeff Mahoney /* returns with log_tree_root freed on success */ 2878e02119d5SChris Mason ret = btrfs_recover_log_trees(log_tree_root); 287979787eaaSJeff Mahoney if (ret) { 288079787eaaSJeff Mahoney btrfs_error(tree_root->fs_info, ret, 288179787eaaSJeff Mahoney "Failed to recover log tree"); 288279787eaaSJeff Mahoney free_extent_buffer(log_tree_root->node); 288379787eaaSJeff Mahoney kfree(log_tree_root); 288428c16cbbSWang Shilong goto fail_qgroup; 288579787eaaSJeff Mahoney } 2886e556ce2cSYan Zheng 2887e556ce2cSYan Zheng if (sb->s_flags & MS_RDONLY) { 2888e556ce2cSYan Zheng ret = btrfs_commit_super(tree_root); 288979787eaaSJeff Mahoney if (ret) 289028c16cbbSWang Shilong goto fail_qgroup; 2891e556ce2cSYan Zheng } 2892e02119d5SChris Mason } 28931a40e23bSZheng Yan 289476dda93cSYan, Zheng ret = btrfs_find_orphan_roots(tree_root); 289579787eaaSJeff Mahoney if (ret) 289628c16cbbSWang Shilong goto fail_qgroup; 289776dda93cSYan, Zheng 28987c2ca468SChris Mason if (!(sb->s_flags & MS_RDONLY)) { 2899d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 290044c44af2SIlya Dryomov if (ret) 290128c16cbbSWang Shilong goto fail_qgroup; 2902d68fc57bSYan, Zheng 29035f316481SWang Shilong mutex_lock(&fs_info->cleaner_mutex); 29045d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 29055f316481SWang Shilong mutex_unlock(&fs_info->cleaner_mutex); 2906d7ce5843SMiao Xie if (ret < 0) { 2907d7ce5843SMiao Xie printk(KERN_WARNING 2908efe120a0SFrank Holton "BTRFS: failed to recover relocation\n"); 2909d7ce5843SMiao Xie err = -EINVAL; 2910bcef60f2SArne Jansen goto fail_qgroup; 2911d7ce5843SMiao Xie } 29127c2ca468SChris Mason } 29131a40e23bSZheng Yan 29143de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 29153de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 2916cb517eabSMiao Xie location.offset = 0; 29173de4586cSChris Mason 29183de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 29193140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 29203140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 2921bcef60f2SArne Jansen goto fail_qgroup; 29223140c9a3SDan Carpenter } 2923c289811cSChris Mason 29242b6ba629SIlya Dryomov if (sb->s_flags & MS_RDONLY) 29252b6ba629SIlya Dryomov return 0; 29262b6ba629SIlya Dryomov 2927e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 29282b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29292b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29302b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 29312b6ba629SIlya Dryomov close_ctree(tree_root); 29322b6ba629SIlya Dryomov return ret; 29332b6ba629SIlya Dryomov } 2934e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 293559641015SIlya Dryomov 29362b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 29372b6ba629SIlya Dryomov if (ret) { 2938efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to resume balance\n"); 293966b4ffd1SJosef Bacik close_ctree(tree_root); 29402b6ba629SIlya Dryomov return ret; 2941e3acc2a6SJosef Bacik } 2942e3acc2a6SJosef Bacik 29438dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 29448dabb742SStefan Behrens if (ret) { 2945efe120a0SFrank Holton pr_warn("BTRFS: failed to resume dev_replace\n"); 29468dabb742SStefan Behrens close_ctree(tree_root); 29478dabb742SStefan Behrens return ret; 29488dabb742SStefan Behrens } 29498dabb742SStefan Behrens 2950b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 2951b382a324SJan Schmidt 2952f7a81ea4SStefan Behrens if (create_uuid_tree) { 2953efe120a0SFrank Holton pr_info("BTRFS: creating UUID tree\n"); 2954f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 2955f7a81ea4SStefan Behrens if (ret) { 2956efe120a0SFrank Holton pr_warn("BTRFS: failed to create the UUID tree %d\n", 2957f7a81ea4SStefan Behrens ret); 2958f7a81ea4SStefan Behrens close_ctree(tree_root); 2959f7a81ea4SStefan Behrens return ret; 2960f7a81ea4SStefan Behrens } 2961f420ee1eSStefan Behrens } else if (check_uuid_tree || 2962f420ee1eSStefan Behrens btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) { 2963efe120a0SFrank Holton pr_info("BTRFS: checking UUID tree\n"); 296470f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 296570f80175SStefan Behrens if (ret) { 2966efe120a0SFrank Holton pr_warn("BTRFS: failed to check the UUID tree %d\n", 296770f80175SStefan Behrens ret); 296870f80175SStefan Behrens close_ctree(tree_root); 296970f80175SStefan Behrens return ret; 297070f80175SStefan Behrens } 297170f80175SStefan Behrens } else { 297270f80175SStefan Behrens fs_info->update_uuid_tree_gen = 1; 2973f7a81ea4SStefan Behrens } 2974f7a81ea4SStefan Behrens 2975ad2b2c80SAl Viro return 0; 297639279cc3SChris Mason 2977bcef60f2SArne Jansen fail_qgroup: 2978bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 29797c2ca468SChris Mason fail_trans_kthread: 29807c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 298154067ae9SJosef Bacik btrfs_cleanup_transaction(fs_info->tree_root); 2982faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 29833f157a2fSChris Mason fail_cleaner: 2984a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 29857c2ca468SChris Mason 29867c2ca468SChris Mason /* 29877c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 29887c2ca468SChris Mason * kthreads 29897c2ca468SChris Mason */ 29907c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 29917c2ca468SChris Mason 29922365dd3cSAnand Jain fail_sysfs: 29932365dd3cSAnand Jain btrfs_sysfs_remove_one(fs_info); 29942365dd3cSAnand Jain 29951b1d1f66SJosef Bacik fail_block_groups: 299654067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 29971b1d1f66SJosef Bacik btrfs_free_block_groups(fs_info); 2998af31f5e5SChris Mason 2999af31f5e5SChris Mason fail_tree_roots: 3000af31f5e5SChris Mason free_root_pointers(fs_info, 1); 30012b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3002af31f5e5SChris Mason 300339279cc3SChris Mason fail_sb_buffer: 30047abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 300516cdcec7SMiao Xie fail_alloc: 30064543df7eSChris Mason fail_iput: 3007586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 3008586e46e2SIlya Dryomov 30094543df7eSChris Mason iput(fs_info->btree_inode); 3010c404e0dcSMiao Xie fail_bio_counter: 3011c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 3012963d678bSMiao Xie fail_delalloc_bytes: 3013963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3014e2d84521SMiao Xie fail_dirty_metadata_bytes: 3015e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3016ad081f14SJens Axboe fail_bdi: 30177e662854SQinghuang Feng bdi_destroy(&fs_info->bdi); 301876dda93cSYan, Zheng fail_srcu: 301976dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 30207e662854SQinghuang Feng fail: 302153b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3022586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3023ad2b2c80SAl Viro return err; 3024af31f5e5SChris Mason 3025af31f5e5SChris Mason recovery_tree_root: 3026af31f5e5SChris Mason if (!btrfs_test_opt(tree_root, RECOVERY)) 3027af31f5e5SChris Mason goto fail_tree_roots; 3028af31f5e5SChris Mason 3029af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3030af31f5e5SChris Mason 3031af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3032af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3033af31f5e5SChris Mason 3034af31f5e5SChris Mason /* we can't trust the free space cache either */ 3035af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3036af31f5e5SChris Mason 3037af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3038af31f5e5SChris Mason &num_backups_tried, &backup_index); 3039af31f5e5SChris Mason if (ret == -1) 3040af31f5e5SChris Mason goto fail_block_groups; 3041af31f5e5SChris Mason goto retry_root_backup; 3042eb60ceacSChris Mason } 3043eb60ceacSChris Mason 3044f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3045f2984462SChris Mason { 3046f2984462SChris Mason if (uptodate) { 3047f2984462SChris Mason set_buffer_uptodate(bh); 3048f2984462SChris Mason } else { 3049442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3050442a4f63SStefan Behrens bh->b_private; 3051442a4f63SStefan Behrens 3052efe120a0SFrank Holton printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to " 3053606686eeSJosef Bacik "I/O error on %s\n", 3054606686eeSJosef Bacik rcu_str_deref(device->name)); 30551259ab75SChris Mason /* note, we dont' set_buffer_write_io_error because we have 30561259ab75SChris Mason * our own ways of dealing with the IO errors 30571259ab75SChris Mason */ 3058f2984462SChris Mason clear_buffer_uptodate(bh); 3059442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3060f2984462SChris Mason } 3061f2984462SChris Mason unlock_buffer(bh); 3062f2984462SChris Mason put_bh(bh); 3063f2984462SChris Mason } 3064f2984462SChris Mason 3065a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3066a512bbf8SYan Zheng { 3067a512bbf8SYan Zheng struct buffer_head *bh; 3068a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3069a512bbf8SYan Zheng struct btrfs_super_block *super; 3070a512bbf8SYan Zheng int i; 3071a512bbf8SYan Zheng u64 transid = 0; 3072a512bbf8SYan Zheng u64 bytenr; 3073a512bbf8SYan Zheng 3074a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3075a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3076a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3077a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3078a512bbf8SYan Zheng */ 3079a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3080a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 30818068a47eSAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= 30828068a47eSAnand Jain i_size_read(bdev->bd_inode)) 3083a512bbf8SYan Zheng break; 30848068a47eSAnand Jain bh = __bread(bdev, bytenr / 4096, 30858068a47eSAnand Jain BTRFS_SUPER_INFO_SIZE); 3086a512bbf8SYan Zheng if (!bh) 3087a512bbf8SYan Zheng continue; 3088a512bbf8SYan Zheng 3089a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3090a512bbf8SYan Zheng if (btrfs_super_bytenr(super) != bytenr || 30913cae210fSQu Wenruo btrfs_super_magic(super) != BTRFS_MAGIC) { 3092a512bbf8SYan Zheng brelse(bh); 3093a512bbf8SYan Zheng continue; 3094a512bbf8SYan Zheng } 3095a512bbf8SYan Zheng 3096a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3097a512bbf8SYan Zheng brelse(latest); 3098a512bbf8SYan Zheng latest = bh; 3099a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3100a512bbf8SYan Zheng } else { 3101a512bbf8SYan Zheng brelse(bh); 3102a512bbf8SYan Zheng } 3103a512bbf8SYan Zheng } 3104a512bbf8SYan Zheng return latest; 3105a512bbf8SYan Zheng } 3106a512bbf8SYan Zheng 31074eedeb75SHisashi Hifumi /* 31084eedeb75SHisashi Hifumi * this should be called twice, once with wait == 0 and 31094eedeb75SHisashi Hifumi * once with wait == 1. When wait == 0 is done, all the buffer heads 31104eedeb75SHisashi Hifumi * we write are pinned. 31114eedeb75SHisashi Hifumi * 31124eedeb75SHisashi Hifumi * They are released when wait == 1 is done. 31134eedeb75SHisashi Hifumi * max_mirrors must be the same for both runs, and it indicates how 31144eedeb75SHisashi Hifumi * many supers on this one device should be written. 31154eedeb75SHisashi Hifumi * 31164eedeb75SHisashi Hifumi * max_mirrors == 0 means to write them all. 31174eedeb75SHisashi Hifumi */ 3118a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3119a512bbf8SYan Zheng struct btrfs_super_block *sb, 3120a512bbf8SYan Zheng int do_barriers, int wait, int max_mirrors) 3121a512bbf8SYan Zheng { 3122a512bbf8SYan Zheng struct buffer_head *bh; 3123a512bbf8SYan Zheng int i; 3124a512bbf8SYan Zheng int ret; 3125a512bbf8SYan Zheng int errors = 0; 3126a512bbf8SYan Zheng u32 crc; 3127a512bbf8SYan Zheng u64 bytenr; 3128a512bbf8SYan Zheng 3129a512bbf8SYan Zheng if (max_mirrors == 0) 3130a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3131a512bbf8SYan Zheng 3132a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3133a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3134a512bbf8SYan Zheng if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) 3135a512bbf8SYan Zheng break; 3136a512bbf8SYan Zheng 3137a512bbf8SYan Zheng if (wait) { 3138a512bbf8SYan Zheng bh = __find_get_block(device->bdev, bytenr / 4096, 3139a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3140634554dcSJosef Bacik if (!bh) { 3141634554dcSJosef Bacik errors++; 3142634554dcSJosef Bacik continue; 3143634554dcSJosef Bacik } 3144a512bbf8SYan Zheng wait_on_buffer(bh); 31454eedeb75SHisashi Hifumi if (!buffer_uptodate(bh)) 31464eedeb75SHisashi Hifumi errors++; 31474eedeb75SHisashi Hifumi 31484eedeb75SHisashi Hifumi /* drop our reference */ 31494eedeb75SHisashi Hifumi brelse(bh); 31504eedeb75SHisashi Hifumi 31514eedeb75SHisashi Hifumi /* drop the reference from the wait == 0 run */ 3152a512bbf8SYan Zheng brelse(bh); 3153a512bbf8SYan Zheng continue; 3154a512bbf8SYan Zheng } else { 3155a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3156a512bbf8SYan Zheng 3157a512bbf8SYan Zheng crc = ~(u32)0; 3158b0496686SLiu Bo crc = btrfs_csum_data((char *)sb + 3159a512bbf8SYan Zheng BTRFS_CSUM_SIZE, crc, 3160a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE - 3161a512bbf8SYan Zheng BTRFS_CSUM_SIZE); 3162a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3163a512bbf8SYan Zheng 31644eedeb75SHisashi Hifumi /* 31654eedeb75SHisashi Hifumi * one reference for us, and we leave it for the 31664eedeb75SHisashi Hifumi * caller 31674eedeb75SHisashi Hifumi */ 3168a512bbf8SYan Zheng bh = __getblk(device->bdev, bytenr / 4096, 3169a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3170634554dcSJosef Bacik if (!bh) { 3171efe120a0SFrank Holton printk(KERN_ERR "BTRFS: couldn't get super " 3172634554dcSJosef Bacik "buffer head for bytenr %Lu\n", bytenr); 3173634554dcSJosef Bacik errors++; 3174634554dcSJosef Bacik continue; 3175634554dcSJosef Bacik } 3176634554dcSJosef Bacik 3177a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3178a512bbf8SYan Zheng 31794eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3180a512bbf8SYan Zheng get_bh(bh); 31814eedeb75SHisashi Hifumi 31824eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3183a512bbf8SYan Zheng lock_buffer(bh); 3184a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3185442a4f63SStefan Behrens bh->b_private = device; 3186a512bbf8SYan Zheng } 3187a512bbf8SYan Zheng 3188387125fcSChris Mason /* 3189387125fcSChris Mason * we fua the first super. The others we allow 3190387125fcSChris Mason * to go down lazy. 3191387125fcSChris Mason */ 3192e8117c26SWang Shilong if (i == 0) 319321adbd5cSStefan Behrens ret = btrfsic_submit_bh(WRITE_FUA, bh); 3194e8117c26SWang Shilong else 3195e8117c26SWang Shilong ret = btrfsic_submit_bh(WRITE_SYNC, bh); 31964eedeb75SHisashi Hifumi if (ret) 3197a512bbf8SYan Zheng errors++; 3198a512bbf8SYan Zheng } 3199a512bbf8SYan Zheng return errors < i ? 0 : -1; 3200a512bbf8SYan Zheng } 3201a512bbf8SYan Zheng 3202387125fcSChris Mason /* 3203387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3204387125fcSChris Mason * for the barrier when it is done 3205387125fcSChris Mason */ 3206387125fcSChris Mason static void btrfs_end_empty_barrier(struct bio *bio, int err) 3207387125fcSChris Mason { 3208387125fcSChris Mason if (err) { 3209387125fcSChris Mason if (err == -EOPNOTSUPP) 3210387125fcSChris Mason set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); 3211387125fcSChris Mason clear_bit(BIO_UPTODATE, &bio->bi_flags); 3212387125fcSChris Mason } 3213387125fcSChris Mason if (bio->bi_private) 3214387125fcSChris Mason complete(bio->bi_private); 3215387125fcSChris Mason bio_put(bio); 3216387125fcSChris Mason } 3217387125fcSChris Mason 3218387125fcSChris Mason /* 3219387125fcSChris Mason * trigger flushes for one the devices. If you pass wait == 0, the flushes are 3220387125fcSChris Mason * sent down. With wait == 1, it waits for the previous flush. 3221387125fcSChris Mason * 3222387125fcSChris Mason * any device where the flush fails with eopnotsupp are flagged as not-barrier 3223387125fcSChris Mason * capable 3224387125fcSChris Mason */ 3225387125fcSChris Mason static int write_dev_flush(struct btrfs_device *device, int wait) 3226387125fcSChris Mason { 3227387125fcSChris Mason struct bio *bio; 3228387125fcSChris Mason int ret = 0; 3229387125fcSChris Mason 3230387125fcSChris Mason if (device->nobarriers) 3231387125fcSChris Mason return 0; 3232387125fcSChris Mason 3233387125fcSChris Mason if (wait) { 3234387125fcSChris Mason bio = device->flush_bio; 3235387125fcSChris Mason if (!bio) 3236387125fcSChris Mason return 0; 3237387125fcSChris Mason 3238387125fcSChris Mason wait_for_completion(&device->flush_wait); 3239387125fcSChris Mason 3240387125fcSChris Mason if (bio_flagged(bio, BIO_EOPNOTSUPP)) { 3241efe120a0SFrank Holton printk_in_rcu("BTRFS: disabling barriers on dev %s\n", 3242606686eeSJosef Bacik rcu_str_deref(device->name)); 3243387125fcSChris Mason device->nobarriers = 1; 32445af3e8ccSStefan Behrens } else if (!bio_flagged(bio, BIO_UPTODATE)) { 3245387125fcSChris Mason ret = -EIO; 3246442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, 3247442a4f63SStefan Behrens BTRFS_DEV_STAT_FLUSH_ERRS); 3248387125fcSChris Mason } 3249387125fcSChris Mason 3250387125fcSChris Mason /* drop the reference from the wait == 0 run */ 3251387125fcSChris Mason bio_put(bio); 3252387125fcSChris Mason device->flush_bio = NULL; 3253387125fcSChris Mason 3254387125fcSChris Mason return ret; 3255387125fcSChris Mason } 3256387125fcSChris Mason 3257387125fcSChris Mason /* 3258387125fcSChris Mason * one reference for us, and we leave it for the 3259387125fcSChris Mason * caller 3260387125fcSChris Mason */ 32619c017abcSJesper Juhl device->flush_bio = NULL; 32629be3395bSChris Mason bio = btrfs_io_bio_alloc(GFP_NOFS, 0); 3263387125fcSChris Mason if (!bio) 3264387125fcSChris Mason return -ENOMEM; 3265387125fcSChris Mason 3266387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3267387125fcSChris Mason bio->bi_bdev = device->bdev; 3268387125fcSChris Mason init_completion(&device->flush_wait); 3269387125fcSChris Mason bio->bi_private = &device->flush_wait; 3270387125fcSChris Mason device->flush_bio = bio; 3271387125fcSChris Mason 3272387125fcSChris Mason bio_get(bio); 327321adbd5cSStefan Behrens btrfsic_submit_bio(WRITE_FLUSH, bio); 3274387125fcSChris Mason 3275387125fcSChris Mason return 0; 3276387125fcSChris Mason } 3277387125fcSChris Mason 3278387125fcSChris Mason /* 3279387125fcSChris Mason * send an empty flush down to each device in parallel, 3280387125fcSChris Mason * then wait for them 3281387125fcSChris Mason */ 3282387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3283387125fcSChris Mason { 3284387125fcSChris Mason struct list_head *head; 3285387125fcSChris Mason struct btrfs_device *dev; 32865af3e8ccSStefan Behrens int errors_send = 0; 32875af3e8ccSStefan Behrens int errors_wait = 0; 3288387125fcSChris Mason int ret; 3289387125fcSChris Mason 3290387125fcSChris Mason /* send down all the barriers */ 3291387125fcSChris Mason head = &info->fs_devices->devices; 3292387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3293f88ba6a2SHidetoshi Seto if (dev->missing) 3294f88ba6a2SHidetoshi Seto continue; 3295387125fcSChris Mason if (!dev->bdev) { 32965af3e8ccSStefan Behrens errors_send++; 3297387125fcSChris Mason continue; 3298387125fcSChris Mason } 3299387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3300387125fcSChris Mason continue; 3301387125fcSChris Mason 3302387125fcSChris Mason ret = write_dev_flush(dev, 0); 3303387125fcSChris Mason if (ret) 33045af3e8ccSStefan Behrens errors_send++; 3305387125fcSChris Mason } 3306387125fcSChris Mason 3307387125fcSChris Mason /* wait for all the barriers */ 3308387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3309f88ba6a2SHidetoshi Seto if (dev->missing) 3310f88ba6a2SHidetoshi Seto continue; 3311387125fcSChris Mason if (!dev->bdev) { 33125af3e8ccSStefan Behrens errors_wait++; 3313387125fcSChris Mason continue; 3314387125fcSChris Mason } 3315387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3316387125fcSChris Mason continue; 3317387125fcSChris Mason 3318387125fcSChris Mason ret = write_dev_flush(dev, 1); 3319387125fcSChris Mason if (ret) 33205af3e8ccSStefan Behrens errors_wait++; 3321387125fcSChris Mason } 33225af3e8ccSStefan Behrens if (errors_send > info->num_tolerated_disk_barrier_failures || 33235af3e8ccSStefan Behrens errors_wait > info->num_tolerated_disk_barrier_failures) 3324387125fcSChris Mason return -EIO; 3325387125fcSChris Mason return 0; 3326387125fcSChris Mason } 3327387125fcSChris Mason 33285af3e8ccSStefan Behrens int btrfs_calc_num_tolerated_disk_barrier_failures( 33295af3e8ccSStefan Behrens struct btrfs_fs_info *fs_info) 33305af3e8ccSStefan Behrens { 33315af3e8ccSStefan Behrens struct btrfs_ioctl_space_info space; 33325af3e8ccSStefan Behrens struct btrfs_space_info *sinfo; 33335af3e8ccSStefan Behrens u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 33345af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_SYSTEM, 33355af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_METADATA, 33365af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 33375af3e8ccSStefan Behrens int num_types = 4; 33385af3e8ccSStefan Behrens int i; 33395af3e8ccSStefan Behrens int c; 33405af3e8ccSStefan Behrens int num_tolerated_disk_barrier_failures = 33415af3e8ccSStefan Behrens (int)fs_info->fs_devices->num_devices; 33425af3e8ccSStefan Behrens 33435af3e8ccSStefan Behrens for (i = 0; i < num_types; i++) { 33445af3e8ccSStefan Behrens struct btrfs_space_info *tmp; 33455af3e8ccSStefan Behrens 33465af3e8ccSStefan Behrens sinfo = NULL; 33475af3e8ccSStefan Behrens rcu_read_lock(); 33485af3e8ccSStefan Behrens list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { 33495af3e8ccSStefan Behrens if (tmp->flags == types[i]) { 33505af3e8ccSStefan Behrens sinfo = tmp; 33515af3e8ccSStefan Behrens break; 33525af3e8ccSStefan Behrens } 33535af3e8ccSStefan Behrens } 33545af3e8ccSStefan Behrens rcu_read_unlock(); 33555af3e8ccSStefan Behrens 33565af3e8ccSStefan Behrens if (!sinfo) 33575af3e8ccSStefan Behrens continue; 33585af3e8ccSStefan Behrens 33595af3e8ccSStefan Behrens down_read(&sinfo->groups_sem); 33605af3e8ccSStefan Behrens for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 33615af3e8ccSStefan Behrens if (!list_empty(&sinfo->block_groups[c])) { 33625af3e8ccSStefan Behrens u64 flags; 33635af3e8ccSStefan Behrens 33645af3e8ccSStefan Behrens btrfs_get_block_group_info( 33655af3e8ccSStefan Behrens &sinfo->block_groups[c], &space); 33665af3e8ccSStefan Behrens if (space.total_bytes == 0 || 33675af3e8ccSStefan Behrens space.used_bytes == 0) 33685af3e8ccSStefan Behrens continue; 33695af3e8ccSStefan Behrens flags = space.flags; 33705af3e8ccSStefan Behrens /* 33715af3e8ccSStefan Behrens * return 33725af3e8ccSStefan Behrens * 0: if dup, single or RAID0 is configured for 33735af3e8ccSStefan Behrens * any of metadata, system or data, else 33745af3e8ccSStefan Behrens * 1: if RAID5 is configured, or if RAID1 or 33755af3e8ccSStefan Behrens * RAID10 is configured and only two mirrors 33765af3e8ccSStefan Behrens * are used, else 33775af3e8ccSStefan Behrens * 2: if RAID6 is configured, else 33785af3e8ccSStefan Behrens * num_mirrors - 1: if RAID1 or RAID10 is 33795af3e8ccSStefan Behrens * configured and more than 33805af3e8ccSStefan Behrens * 2 mirrors are used. 33815af3e8ccSStefan Behrens */ 33825af3e8ccSStefan Behrens if (num_tolerated_disk_barrier_failures > 0 && 33835af3e8ccSStefan Behrens ((flags & (BTRFS_BLOCK_GROUP_DUP | 33845af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_RAID0)) || 33855af3e8ccSStefan Behrens ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) 33865af3e8ccSStefan Behrens == 0))) 33875af3e8ccSStefan Behrens num_tolerated_disk_barrier_failures = 0; 338853b381b3SDavid Woodhouse else if (num_tolerated_disk_barrier_failures > 1) { 338953b381b3SDavid Woodhouse if (flags & (BTRFS_BLOCK_GROUP_RAID1 | 339053b381b3SDavid Woodhouse BTRFS_BLOCK_GROUP_RAID5 | 339153b381b3SDavid Woodhouse BTRFS_BLOCK_GROUP_RAID10)) { 33925af3e8ccSStefan Behrens num_tolerated_disk_barrier_failures = 1; 339353b381b3SDavid Woodhouse } else if (flags & 339415b0a89dSHenrik Nordvik BTRFS_BLOCK_GROUP_RAID6) { 339553b381b3SDavid Woodhouse num_tolerated_disk_barrier_failures = 2; 339653b381b3SDavid Woodhouse } 339753b381b3SDavid Woodhouse } 33985af3e8ccSStefan Behrens } 33995af3e8ccSStefan Behrens } 34005af3e8ccSStefan Behrens up_read(&sinfo->groups_sem); 34015af3e8ccSStefan Behrens } 34025af3e8ccSStefan Behrens 34035af3e8ccSStefan Behrens return num_tolerated_disk_barrier_failures; 34045af3e8ccSStefan Behrens } 34055af3e8ccSStefan Behrens 340648a3b636SEric Sandeen static int write_all_supers(struct btrfs_root *root, int max_mirrors) 3407f2984462SChris Mason { 3408e5e9a520SChris Mason struct list_head *head; 3409f2984462SChris Mason struct btrfs_device *dev; 3410a061fc8dSChris Mason struct btrfs_super_block *sb; 3411f2984462SChris Mason struct btrfs_dev_item *dev_item; 3412f2984462SChris Mason int ret; 3413f2984462SChris Mason int do_barriers; 3414a236aed1SChris Mason int max_errors; 3415a236aed1SChris Mason int total_errors = 0; 3416a061fc8dSChris Mason u64 flags; 3417f2984462SChris Mason 3418f2984462SChris Mason do_barriers = !btrfs_test_opt(root, NOBARRIER); 3419af31f5e5SChris Mason backup_super_roots(root->fs_info); 3420f2984462SChris Mason 34216c41761fSDavid Sterba sb = root->fs_info->super_for_commit; 3422a061fc8dSChris Mason dev_item = &sb->dev_item; 3423e5e9a520SChris Mason 3424174ba509SChris Mason mutex_lock(&root->fs_info->fs_devices->device_list_mutex); 3425e5e9a520SChris Mason head = &root->fs_info->fs_devices->devices; 3426d7306801SFilipe David Borba Manana max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; 3427387125fcSChris Mason 34285af3e8ccSStefan Behrens if (do_barriers) { 34295af3e8ccSStefan Behrens ret = barrier_all_devices(root->fs_info); 34305af3e8ccSStefan Behrens if (ret) { 34315af3e8ccSStefan Behrens mutex_unlock( 34325af3e8ccSStefan Behrens &root->fs_info->fs_devices->device_list_mutex); 34335af3e8ccSStefan Behrens btrfs_error(root->fs_info, ret, 34345af3e8ccSStefan Behrens "errors while submitting device barriers."); 34355af3e8ccSStefan Behrens return ret; 34365af3e8ccSStefan Behrens } 34375af3e8ccSStefan Behrens } 3438387125fcSChris Mason 34391f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3440dfe25020SChris Mason if (!dev->bdev) { 3441dfe25020SChris Mason total_errors++; 3442dfe25020SChris Mason continue; 3443dfe25020SChris Mason } 34442b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3445dfe25020SChris Mason continue; 3446dfe25020SChris Mason 34472b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3448a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3449a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 34507df69d3eSMiao Xie btrfs_set_stack_device_total_bytes(dev_item, 34517df69d3eSMiao Xie dev->disk_total_bytes); 3452a061fc8dSChris Mason btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); 3453a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3454a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3455a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3456a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 34572b82032cSYan Zheng memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 3458a512bbf8SYan Zheng 3459a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3460a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3461f2984462SChris Mason 3462a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); 3463a236aed1SChris Mason if (ret) 3464a236aed1SChris Mason total_errors++; 3465f2984462SChris Mason } 3466a236aed1SChris Mason if (total_errors > max_errors) { 3467efe120a0SFrank Holton btrfs_err(root->fs_info, "%d errors while writing supers", 3468d397712bSChris Mason total_errors); 3469a724b436SStefan Behrens mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 347079787eaaSJeff Mahoney 34719d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 34729d565ba4SStefan Behrens btrfs_error(root->fs_info, -EIO, 34739d565ba4SStefan Behrens "%d errors while writing supers", total_errors); 34749d565ba4SStefan Behrens return -EIO; 3475a236aed1SChris Mason } 3476f2984462SChris Mason 3477a512bbf8SYan Zheng total_errors = 0; 34781f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3479dfe25020SChris Mason if (!dev->bdev) 3480dfe25020SChris Mason continue; 34812b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3482dfe25020SChris Mason continue; 3483dfe25020SChris Mason 3484a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); 3485a512bbf8SYan Zheng if (ret) 34861259ab75SChris Mason total_errors++; 3487f2984462SChris Mason } 3488174ba509SChris Mason mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3489a236aed1SChris Mason if (total_errors > max_errors) { 349079787eaaSJeff Mahoney btrfs_error(root->fs_info, -EIO, 349179787eaaSJeff Mahoney "%d errors while writing supers", total_errors); 349279787eaaSJeff Mahoney return -EIO; 3493a236aed1SChris Mason } 3494f2984462SChris Mason return 0; 3495f2984462SChris Mason } 3496f2984462SChris Mason 3497a512bbf8SYan Zheng int write_ctree_super(struct btrfs_trans_handle *trans, 3498a512bbf8SYan Zheng struct btrfs_root *root, int max_mirrors) 3499cfaa7295SChris Mason { 3500f570e757SRashika return write_all_supers(root, max_mirrors); 3501cfaa7295SChris Mason } 3502cfaa7295SChris Mason 3503cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3504cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3505cb517eabSMiao Xie struct btrfs_root *root) 35062619ba1fSChris Mason { 35074df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 35082619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 35092619ba1fSChris Mason (unsigned long)root->root_key.objectid); 35104df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 351176dda93cSYan, Zheng 351276dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 351376dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 351476dda93cSYan, Zheng 35151a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 35163321719eSLiu Bo btrfs_free_log(NULL, root); 35173321719eSLiu Bo 3518faa2dbf0SJosef Bacik if (root->free_ino_pinned) 3519581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3520faa2dbf0SJosef Bacik if (root->free_ino_ctl) 3521581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 35224df27c4dSYan, Zheng free_fs_root(root); 35234df27c4dSYan, Zheng } 35244df27c4dSYan, Zheng 35254df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root) 35264df27c4dSYan, Zheng { 352757cdc8dbSDavid Sterba iput(root->ino_cache_inode); 35284df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 35291cb048f5SFilipe David Borba Manana btrfs_free_block_rsv(root, root->orphan_block_rsv); 35301cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 35310ee5dc67SAl Viro if (root->anon_dev) 35320ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 35338257b2dcSMiao Xie if (root->subv_writers) 35348257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 35355f39d397SChris Mason free_extent_buffer(root->node); 35365f39d397SChris Mason free_extent_buffer(root->commit_root); 3537581bb050SLi Zefan kfree(root->free_ino_ctl); 3538581bb050SLi Zefan kfree(root->free_ino_pinned); 353958176a96SJosef Bacik kfree(root->name); 3540b0feb9d9SMiao Xie btrfs_put_fs_root(root); 35412619ba1fSChris Mason } 35422619ba1fSChris Mason 3543cb517eabSMiao Xie void btrfs_free_fs_root(struct btrfs_root *root) 3544cb517eabSMiao Xie { 3545cb517eabSMiao Xie free_fs_root(root); 3546eb60ceacSChris Mason } 3547eb60ceacSChris Mason 3548c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3549c146afadSYan Zheng { 3550c146afadSYan Zheng u64 root_objectid = 0; 3551c146afadSYan Zheng struct btrfs_root *gang[8]; 355265d33fd7SQu Wenruo int i = 0; 355365d33fd7SQu Wenruo int err = 0; 355465d33fd7SQu Wenruo unsigned int ret = 0; 355565d33fd7SQu Wenruo int index; 3556c146afadSYan Zheng 3557c146afadSYan Zheng while (1) { 355865d33fd7SQu Wenruo index = srcu_read_lock(&fs_info->subvol_srcu); 3559c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3560c146afadSYan Zheng (void **)gang, root_objectid, 3561c146afadSYan Zheng ARRAY_SIZE(gang)); 356265d33fd7SQu Wenruo if (!ret) { 356365d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 3564c146afadSYan Zheng break; 356565d33fd7SQu Wenruo } 35665d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 356766b4ffd1SJosef Bacik 356865d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 356965d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 357065d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 357165d33fd7SQu Wenruo gang[i] = NULL; 357265d33fd7SQu Wenruo continue; 357365d33fd7SQu Wenruo } 357465d33fd7SQu Wenruo /* grab all the search result for later use */ 357565d33fd7SQu Wenruo gang[i] = btrfs_grab_fs_root(gang[i]); 357665d33fd7SQu Wenruo } 357765d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 357865d33fd7SQu Wenruo 357965d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 358065d33fd7SQu Wenruo if (!gang[i]) 358165d33fd7SQu Wenruo continue; 3582c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 358366b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 358466b4ffd1SJosef Bacik if (err) 358565d33fd7SQu Wenruo break; 358665d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 3587c146afadSYan Zheng } 3588c146afadSYan Zheng root_objectid++; 3589c146afadSYan Zheng } 359065d33fd7SQu Wenruo 359165d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 359265d33fd7SQu Wenruo for (; i < ret; i++) { 359365d33fd7SQu Wenruo if (gang[i]) 359465d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 359565d33fd7SQu Wenruo } 359665d33fd7SQu Wenruo return err; 3597c146afadSYan Zheng } 3598c146afadSYan Zheng 3599c146afadSYan Zheng int btrfs_commit_super(struct btrfs_root *root) 3600c146afadSYan Zheng { 3601c146afadSYan Zheng struct btrfs_trans_handle *trans; 3602c146afadSYan Zheng 3603c146afadSYan Zheng mutex_lock(&root->fs_info->cleaner_mutex); 360424bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 3605c146afadSYan Zheng mutex_unlock(&root->fs_info->cleaner_mutex); 36069d1a2a3aSDavid Sterba wake_up_process(root->fs_info->cleaner_kthread); 3607c71bf099SYan, Zheng 3608c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 3609c71bf099SYan, Zheng down_write(&root->fs_info->cleanup_work_sem); 3610c71bf099SYan, Zheng up_write(&root->fs_info->cleanup_work_sem); 3611c71bf099SYan, Zheng 36127a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 36133612b495STsutomu Itoh if (IS_ERR(trans)) 36143612b495STsutomu Itoh return PTR_ERR(trans); 3615d52c1bccSLiu Bo return btrfs_commit_transaction(trans, root); 3616c146afadSYan Zheng } 3617c146afadSYan Zheng 36183abdbd78SDavid Sterba void close_ctree(struct btrfs_root *root) 3619eb60ceacSChris Mason { 36200f7d52f4SChris Mason struct btrfs_fs_info *fs_info = root->fs_info; 3621c146afadSYan Zheng int ret; 3622e089f05cSChris Mason 3623facda1e7SChris Mason fs_info->closing = 1; 3624a2135011SChris Mason smp_mb(); 3625a2135011SChris Mason 3626803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3627803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3628803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3629803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3630803b2f54SStefan Behrens 3631837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3632aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3633837d5b6eSIlya Dryomov 36348dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 36358dabb742SStefan Behrens 3636aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 36374cb5300bSChris Mason 36384cb5300bSChris Mason /* wait for any defraggers to finish */ 36394cb5300bSChris Mason wait_event(fs_info->transaction_wait, 36404cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 36414cb5300bSChris Mason 36424cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 364326176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 36444cb5300bSChris Mason 364521c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 364621c7e756SMiao Xie 3647c146afadSYan Zheng if (!(fs_info->sb->s_flags & MS_RDONLY)) { 3648c146afadSYan Zheng ret = btrfs_commit_super(root); 3649d397712bSChris Mason if (ret) 3650efe120a0SFrank Holton btrfs_err(root->fs_info, "commit super ret %d", ret); 3651c146afadSYan Zheng } 3652ed2ff2cbSChris Mason 365387533c47SMiao Xie if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 365468ce9682SStefan Behrens btrfs_error_commit_super(root); 3655acce952bSliubo 3656e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3657e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 36588929ecfaSYan, Zheng 3659f25784b3SYan Zheng fs_info->closing = 2; 3660f25784b3SYan Zheng smp_mb(); 3661f25784b3SYan Zheng 3662bcef60f2SArne Jansen btrfs_free_qgroup_config(root->fs_info); 3663bcef60f2SArne Jansen 3664963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 3665efe120a0SFrank Holton btrfs_info(root->fs_info, "at unmount delalloc count %lld", 3666963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3667b0c68f8bSChris Mason } 366831153d81SYan Zheng 36695ac1d209SJeff Mahoney btrfs_sysfs_remove_one(fs_info); 36705ac1d209SJeff Mahoney 3671faa2dbf0SJosef Bacik btrfs_free_fs_roots(fs_info); 3672d10c5f31SChris Mason 36731a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 36741a4319ccSLiu Bo 36752b1360daSJosef Bacik btrfs_free_block_groups(fs_info); 36762b1360daSJosef Bacik 3677de348ee0SWang Shilong /* 3678de348ee0SWang Shilong * we must make sure there is not any read request to 3679de348ee0SWang Shilong * submit after we stopping all workers. 3680de348ee0SWang Shilong */ 3681de348ee0SWang Shilong invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 368296192499SJosef Bacik btrfs_stop_all_workers(fs_info); 368396192499SJosef Bacik 368413e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 36859ad6b7bcSChris Mason 368613e6c37bSJosef Bacik iput(fs_info->btree_inode); 3687d6bfde87SChris Mason 368821adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 368921adbd5cSStefan Behrens if (btrfs_test_opt(root, CHECK_INTEGRITY)) 369021adbd5cSStefan Behrens btrfsic_unmount(root, fs_info->fs_devices); 369121adbd5cSStefan Behrens #endif 369221adbd5cSStefan Behrens 3693dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 36940b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3695b248a415SChris Mason 3696e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3697963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3698c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 369904160088SChris Mason bdi_destroy(&fs_info->bdi); 370076dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 37010b86a832SChris Mason 370253b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 370353b381b3SDavid Woodhouse 37041cb048f5SFilipe David Borba Manana btrfs_free_block_rsv(root, root->orphan_block_rsv); 37051cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 3706eb60ceacSChris Mason } 3707eb60ceacSChris Mason 3708b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3709b9fab919SChris Mason int atomic) 3710ccd467d6SChris Mason { 37111259ab75SChris Mason int ret; 3712727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 37131259ab75SChris Mason 37140b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 37151259ab75SChris Mason if (!ret) 37161259ab75SChris Mason return ret; 37171259ab75SChris Mason 37181259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3719b9fab919SChris Mason parent_transid, atomic); 3720b9fab919SChris Mason if (ret == -EAGAIN) 3721b9fab919SChris Mason return ret; 37221259ab75SChris Mason return !ret; 37235f39d397SChris Mason } 37246702ed49SChris Mason 37255f39d397SChris Mason int btrfs_set_buffer_uptodate(struct extent_buffer *buf) 37265f39d397SChris Mason { 37270b32f4bbSJosef Bacik return set_extent_buffer_uptodate(buf); 37285f39d397SChris Mason } 37295f39d397SChris Mason 37305f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 37315f39d397SChris Mason { 373206ea65a3SJosef Bacik struct btrfs_root *root; 37335f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 3734b9473439SChris Mason int was_dirty; 3735b4ce94deSChris Mason 373606ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 373706ea65a3SJosef Bacik /* 373806ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 373906ea65a3SJosef Bacik * enabled. Normal people shouldn't be marking dummy buffers as dirty 374006ea65a3SJosef Bacik * outside of the sanity tests. 374106ea65a3SJosef Bacik */ 374206ea65a3SJosef Bacik if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) 374306ea65a3SJosef Bacik return; 374406ea65a3SJosef Bacik #endif 374506ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 3746b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 374731b1a2bdSJulia Lawall if (transid != root->fs_info->generation) 374831b1a2bdSJulia Lawall WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " 3749d397712bSChris Mason "found %llu running %llu\n", 3750c1c9ff7cSGeert Uytterhoeven buf->start, transid, root->fs_info->generation); 37510b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 3752e2d84521SMiao Xie if (!was_dirty) 3753e2d84521SMiao Xie __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, 3754e2d84521SMiao Xie buf->len, 3755e2d84521SMiao Xie root->fs_info->dirty_metadata_batch); 37561f21ef0aSFilipe Manana #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 37571f21ef0aSFilipe Manana if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { 37581f21ef0aSFilipe Manana btrfs_print_leaf(root, buf); 37591f21ef0aSFilipe Manana ASSERT(0); 37601f21ef0aSFilipe Manana } 37611f21ef0aSFilipe Manana #endif 3762eb60ceacSChris Mason } 3763eb60ceacSChris Mason 3764b53d3f5dSLiu Bo static void __btrfs_btree_balance_dirty(struct btrfs_root *root, 3765b53d3f5dSLiu Bo int flush_delayed) 376635b7e476SChris Mason { 3767188de649SChris Mason /* 3768188de649SChris Mason * looks as though older kernels can get into trouble with 3769188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 3770188de649SChris Mason */ 3771e2d84521SMiao Xie int ret; 3772d6bfde87SChris Mason 37736933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 3774d6bfde87SChris Mason return; 3775d6bfde87SChris Mason 3776b53d3f5dSLiu Bo if (flush_delayed) 377716cdcec7SMiao Xie btrfs_balance_delayed_items(root); 377816cdcec7SMiao Xie 3779e2d84521SMiao Xie ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, 3780e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 3781e2d84521SMiao Xie if (ret > 0) { 3782d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited( 3783d0e1d66bSNamjae Jeon root->fs_info->btree_inode->i_mapping); 378416cdcec7SMiao Xie } 378516cdcec7SMiao Xie return; 378616cdcec7SMiao Xie } 378716cdcec7SMiao Xie 3788b53d3f5dSLiu Bo void btrfs_btree_balance_dirty(struct btrfs_root *root) 378916cdcec7SMiao Xie { 3790b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 1); 379135b7e476SChris Mason } 3792b53d3f5dSLiu Bo 3793b53d3f5dSLiu Bo void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) 3794b53d3f5dSLiu Bo { 3795b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 0); 3796d6bfde87SChris Mason } 37976b80053dSChris Mason 3798ca7a79adSChris Mason int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 37996b80053dSChris Mason { 3800727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 38010b32f4bbSJosef Bacik return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 38026b80053dSChris Mason } 38030da5468fSChris Mason 3804fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 3805acce952bSliubo int read_only) 3806acce952bSliubo { 38071104a885SDavid Sterba /* 38081104a885SDavid Sterba * Placeholder for checks 38091104a885SDavid Sterba */ 3810fcd1f065SDavid Sterba return 0; 3811fcd1f065SDavid Sterba } 3812fcd1f065SDavid Sterba 381348a3b636SEric Sandeen static void btrfs_error_commit_super(struct btrfs_root *root) 3814acce952bSliubo { 3815acce952bSliubo mutex_lock(&root->fs_info->cleaner_mutex); 3816acce952bSliubo btrfs_run_delayed_iputs(root); 3817acce952bSliubo mutex_unlock(&root->fs_info->cleaner_mutex); 3818acce952bSliubo 3819acce952bSliubo down_write(&root->fs_info->cleanup_work_sem); 3820acce952bSliubo up_write(&root->fs_info->cleanup_work_sem); 3821acce952bSliubo 3822acce952bSliubo /* cleanup FS via transaction */ 3823acce952bSliubo btrfs_cleanup_transaction(root); 3824acce952bSliubo } 3825acce952bSliubo 3826143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 3827acce952bSliubo { 3828acce952bSliubo struct btrfs_ordered_extent *ordered; 3829acce952bSliubo 3830199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 3831779880efSJosef Bacik /* 3832779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 3833779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 3834779880efSJosef Bacik */ 3835199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 3836779880efSJosef Bacik root_extent_list) 3837779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 3838199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 3839199c2a9cSMiao Xie } 3840199c2a9cSMiao Xie 3841199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 3842199c2a9cSMiao Xie { 3843199c2a9cSMiao Xie struct btrfs_root *root; 3844199c2a9cSMiao Xie struct list_head splice; 3845199c2a9cSMiao Xie 3846199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 3847199c2a9cSMiao Xie 3848199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 3849199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 3850199c2a9cSMiao Xie while (!list_empty(&splice)) { 3851199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 3852199c2a9cSMiao Xie ordered_root); 38531de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 38541de2cfdeSJosef Bacik &fs_info->ordered_roots); 3855199c2a9cSMiao Xie 38562a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 3857199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 3858199c2a9cSMiao Xie 38592a85d9caSLiu Bo cond_resched(); 38602a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 3861199c2a9cSMiao Xie } 3862199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 3863acce952bSliubo } 3864acce952bSliubo 386535a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 3866acce952bSliubo struct btrfs_root *root) 3867acce952bSliubo { 3868acce952bSliubo struct rb_node *node; 3869acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 3870acce952bSliubo struct btrfs_delayed_ref_node *ref; 3871acce952bSliubo int ret = 0; 3872acce952bSliubo 3873acce952bSliubo delayed_refs = &trans->delayed_refs; 3874acce952bSliubo 3875acce952bSliubo spin_lock(&delayed_refs->lock); 3876d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 3877cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 3878efe120a0SFrank Holton btrfs_info(root->fs_info, "delayed_refs has NO entry"); 3879acce952bSliubo return ret; 3880acce952bSliubo } 3881acce952bSliubo 3882d7df2c79SJosef Bacik while ((node = rb_first(&delayed_refs->href_root)) != NULL) { 3883d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 3884e78417d1SJosef Bacik bool pin_bytes = false; 3885acce952bSliubo 3886d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 3887d7df2c79SJosef Bacik href_node); 3888b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 3889d7df2c79SJosef Bacik atomic_inc(&head->node.refs); 389079787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 3891b939d1abSJosef Bacik 3892acce952bSliubo mutex_lock(&head->mutex); 3893b939d1abSJosef Bacik mutex_unlock(&head->mutex); 3894d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 3895e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 3896b939d1abSJosef Bacik continue; 3897b939d1abSJosef Bacik } 3898d7df2c79SJosef Bacik spin_lock(&head->lock); 3899d7df2c79SJosef Bacik while ((node = rb_first(&head->ref_root)) != NULL) { 3900d7df2c79SJosef Bacik ref = rb_entry(node, struct btrfs_delayed_ref_node, 3901d7df2c79SJosef Bacik rb_node); 3902d7df2c79SJosef Bacik ref->in_tree = 0; 3903d7df2c79SJosef Bacik rb_erase(&ref->rb_node, &head->ref_root); 3904d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 3905d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 3906d7df2c79SJosef Bacik } 390754067ae9SJosef Bacik if (head->must_insert_reserved) 3908e78417d1SJosef Bacik pin_bytes = true; 390978a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 3910acce952bSliubo delayed_refs->num_heads--; 3911d7df2c79SJosef Bacik if (head->processing == 0) 3912acce952bSliubo delayed_refs->num_heads_ready--; 3913d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 3914d7df2c79SJosef Bacik head->node.in_tree = 0; 3915c46effa6SLiu Bo rb_erase(&head->href_node, &delayed_refs->href_root); 3916d7df2c79SJosef Bacik spin_unlock(&head->lock); 3917acce952bSliubo spin_unlock(&delayed_refs->lock); 3918e78417d1SJosef Bacik mutex_unlock(&head->mutex); 3919acce952bSliubo 3920d7df2c79SJosef Bacik if (pin_bytes) 3921d7df2c79SJosef Bacik btrfs_pin_extent(root, head->node.bytenr, 3922d7df2c79SJosef Bacik head->node.num_bytes, 1); 3923d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 3924acce952bSliubo cond_resched(); 3925acce952bSliubo spin_lock(&delayed_refs->lock); 3926acce952bSliubo } 3927acce952bSliubo 3928acce952bSliubo spin_unlock(&delayed_refs->lock); 3929acce952bSliubo 3930acce952bSliubo return ret; 3931acce952bSliubo } 3932acce952bSliubo 3933143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 3934acce952bSliubo { 3935acce952bSliubo struct btrfs_inode *btrfs_inode; 3936acce952bSliubo struct list_head splice; 3937acce952bSliubo 3938acce952bSliubo INIT_LIST_HEAD(&splice); 3939acce952bSliubo 3940eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 3941eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 3942acce952bSliubo 3943acce952bSliubo while (!list_empty(&splice)) { 3944eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 3945acce952bSliubo delalloc_inodes); 3946acce952bSliubo 3947acce952bSliubo list_del_init(&btrfs_inode->delalloc_inodes); 3948df0af1a5SMiao Xie clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, 3949df0af1a5SMiao Xie &btrfs_inode->runtime_flags); 3950eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 3951acce952bSliubo 3952acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 3953b216cbfbSMiao Xie 3954eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 3955acce952bSliubo } 3956acce952bSliubo 3957eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 3958eb73c1b7SMiao Xie } 3959eb73c1b7SMiao Xie 3960eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 3961eb73c1b7SMiao Xie { 3962eb73c1b7SMiao Xie struct btrfs_root *root; 3963eb73c1b7SMiao Xie struct list_head splice; 3964eb73c1b7SMiao Xie 3965eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 3966eb73c1b7SMiao Xie 3967eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 3968eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 3969eb73c1b7SMiao Xie while (!list_empty(&splice)) { 3970eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 3971eb73c1b7SMiao Xie delalloc_root); 3972eb73c1b7SMiao Xie list_del_init(&root->delalloc_root); 3973eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 3974eb73c1b7SMiao Xie BUG_ON(!root); 3975eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 3976eb73c1b7SMiao Xie 3977eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 3978eb73c1b7SMiao Xie btrfs_put_fs_root(root); 3979eb73c1b7SMiao Xie 3980eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 3981eb73c1b7SMiao Xie } 3982eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 3983acce952bSliubo } 3984acce952bSliubo 3985acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 3986acce952bSliubo struct extent_io_tree *dirty_pages, 3987acce952bSliubo int mark) 3988acce952bSliubo { 3989acce952bSliubo int ret; 3990acce952bSliubo struct extent_buffer *eb; 3991acce952bSliubo u64 start = 0; 3992acce952bSliubo u64 end; 3993acce952bSliubo 3994acce952bSliubo while (1) { 3995acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 3996e6138876SJosef Bacik mark, NULL); 3997acce952bSliubo if (ret) 3998acce952bSliubo break; 3999acce952bSliubo 4000acce952bSliubo clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); 4001acce952bSliubo while (start <= end) { 4002fd8b2b61SJosef Bacik eb = btrfs_find_tree_block(root, start, 4003707e8a07SDavid Sterba root->nodesize); 4004707e8a07SDavid Sterba start += root->nodesize; 4005fd8b2b61SJosef Bacik if (!eb) 4006acce952bSliubo continue; 4007fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4008acce952bSliubo 4009fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4010fd8b2b61SJosef Bacik &eb->bflags)) 4011fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4012fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4013acce952bSliubo } 4014acce952bSliubo } 4015acce952bSliubo 4016acce952bSliubo return ret; 4017acce952bSliubo } 4018acce952bSliubo 4019acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 4020acce952bSliubo struct extent_io_tree *pinned_extents) 4021acce952bSliubo { 4022acce952bSliubo struct extent_io_tree *unpin; 4023acce952bSliubo u64 start; 4024acce952bSliubo u64 end; 4025acce952bSliubo int ret; 4026ed0eaa14SLiu Bo bool loop = true; 4027acce952bSliubo 4028acce952bSliubo unpin = pinned_extents; 4029ed0eaa14SLiu Bo again: 4030acce952bSliubo while (1) { 4031acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4032e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4033acce952bSliubo if (ret) 4034acce952bSliubo break; 4035acce952bSliubo 4036acce952bSliubo /* opt_discard */ 40375378e607SLi Dongyang if (btrfs_test_opt(root, DISCARD)) 40385378e607SLi Dongyang ret = btrfs_error_discard_extent(root, start, 40395378e607SLi Dongyang end + 1 - start, 40405378e607SLi Dongyang NULL); 4041acce952bSliubo 4042acce952bSliubo clear_extent_dirty(unpin, start, end, GFP_NOFS); 4043acce952bSliubo btrfs_error_unpin_extent_range(root, start, end); 4044acce952bSliubo cond_resched(); 4045acce952bSliubo } 4046acce952bSliubo 4047ed0eaa14SLiu Bo if (loop) { 4048ed0eaa14SLiu Bo if (unpin == &root->fs_info->freed_extents[0]) 4049ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[1]; 4050ed0eaa14SLiu Bo else 4051ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[0]; 4052ed0eaa14SLiu Bo loop = false; 4053ed0eaa14SLiu Bo goto again; 4054ed0eaa14SLiu Bo } 4055ed0eaa14SLiu Bo 4056acce952bSliubo return 0; 4057acce952bSliubo } 4058acce952bSliubo 405949b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 406049b25e05SJeff Mahoney struct btrfs_root *root) 406149b25e05SJeff Mahoney { 406249b25e05SJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, root); 406349b25e05SJeff Mahoney 40644a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 406549b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_blocked_wait); 406649b25e05SJeff Mahoney 40674a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 406849b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_wait); 406949b25e05SJeff Mahoney 407067cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 407167cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 407267cde344SMiao Xie 407349b25e05SJeff Mahoney btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, 407449b25e05SJeff Mahoney EXTENT_DIRTY); 40756e841e32SLiu Bo btrfs_destroy_pinned_extent(root, 40766e841e32SLiu Bo root->fs_info->pinned_extents); 407749b25e05SJeff Mahoney 40784a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 40794a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 40804a9d8bdeSMiao Xie 408149b25e05SJeff Mahoney /* 408249b25e05SJeff Mahoney memset(cur_trans, 0, sizeof(*cur_trans)); 408349b25e05SJeff Mahoney kmem_cache_free(btrfs_transaction_cachep, cur_trans); 408449b25e05SJeff Mahoney */ 408549b25e05SJeff Mahoney } 408649b25e05SJeff Mahoney 408748a3b636SEric Sandeen static int btrfs_cleanup_transaction(struct btrfs_root *root) 4088acce952bSliubo { 4089acce952bSliubo struct btrfs_transaction *t; 4090acce952bSliubo 4091acce952bSliubo mutex_lock(&root->fs_info->transaction_kthread_mutex); 4092acce952bSliubo 4093a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4094724e2315SJosef Bacik while (!list_empty(&root->fs_info->trans_list)) { 4095724e2315SJosef Bacik t = list_first_entry(&root->fs_info->trans_list, 4096724e2315SJosef Bacik struct btrfs_transaction, list); 4097724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 4098724e2315SJosef Bacik atomic_inc(&t->use_count); 4099724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4100724e2315SJosef Bacik btrfs_wait_for_commit(root, t->transid); 4101724e2315SJosef Bacik btrfs_put_transaction(t); 4102724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4103724e2315SJosef Bacik continue; 4104724e2315SJosef Bacik } 4105724e2315SJosef Bacik if (t == root->fs_info->running_transaction) { 4106724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 4107724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4108724e2315SJosef Bacik /* 4109724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4110724e2315SJosef Bacik * handle open currently for this transaction. 4111724e2315SJosef Bacik */ 4112724e2315SJosef Bacik wait_event(t->writer_wait, 4113724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4114724e2315SJosef Bacik } else { 4115724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4116724e2315SJosef Bacik } 4117724e2315SJosef Bacik btrfs_cleanup_one_transaction(t, root); 4118724e2315SJosef Bacik 4119724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4120724e2315SJosef Bacik if (t == root->fs_info->running_transaction) 4121ac673879SMiao Xie root->fs_info->running_transaction = NULL; 4122724e2315SJosef Bacik list_del_init(&t->list); 4123a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4124a4abeea4SJosef Bacik 4125724e2315SJosef Bacik btrfs_put_transaction(t); 4126724e2315SJosef Bacik trace_btrfs_transaction_commit(root); 4127724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4128724e2315SJosef Bacik } 4129724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4130199c2a9cSMiao Xie btrfs_destroy_all_ordered_extents(root->fs_info); 413167cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 413267cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 4133724e2315SJosef Bacik btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); 4134eb73c1b7SMiao Xie btrfs_destroy_all_delalloc_inodes(root->fs_info); 4135acce952bSliubo mutex_unlock(&root->fs_info->transaction_kthread_mutex); 4136acce952bSliubo 4137acce952bSliubo return 0; 4138acce952bSliubo } 4139acce952bSliubo 4140d1310b2eSChris Mason static struct extent_io_ops btree_extent_io_ops = { 4141ce9adaa5SChris Mason .readpage_end_io_hook = btree_readpage_end_io_hook, 41424bb31e92SArne Jansen .readpage_io_failed_hook = btree_io_failed_hook, 41430b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 4144239b14b3SChris Mason /* note we're sharing with inode.c for the merge bio hook */ 4145239b14b3SChris Mason .merge_bio_hook = btrfs_merge_bio_hook, 41460da5468fSChris Mason }; 4147