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" 428b712842SChris Mason #include "async-thread.h" 43925baeddSChris Mason #include "locking.h" 44e02119d5SChris Mason #include "tree-log.h" 45fa9c0d79SChris Mason #include "free-space-cache.h" 46581bb050SLi Zefan #include "inode-map.h" 4721adbd5cSStefan Behrens #include "check-integrity.h" 48606686eeSJosef Bacik #include "rcu-string.h" 498dabb742SStefan Behrens #include "dev-replace.h" 5053b381b3SDavid Woodhouse #include "raid56.h" 515ac1d209SJeff Mahoney #include "sysfs.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); 62569e0f35SJosef Bacik static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, 63569e0f35SJosef Bacik struct btrfs_root *root); 64143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 65acce952bSliubo static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 66acce952bSliubo struct btrfs_root *root); 67143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 68acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 69acce952bSliubo struct extent_io_tree *dirty_pages, 70acce952bSliubo int mark); 71acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 72acce952bSliubo struct extent_io_tree *pinned_extents); 7348a3b636SEric Sandeen static int btrfs_cleanup_transaction(struct btrfs_root *root); 7448a3b636SEric Sandeen static void btrfs_error_commit_super(struct btrfs_root *root); 75ce9adaa5SChris Mason 76d352ac68SChris Mason /* 77d352ac68SChris Mason * end_io_wq structs are used to do processing in task context when an IO is 78d352ac68SChris Mason * complete. This is used during reads to verify checksums, and it is used 79d352ac68SChris Mason * by writes to insert metadata for new file extents after IO is complete. 80d352ac68SChris Mason */ 81ce9adaa5SChris Mason struct end_io_wq { 82ce9adaa5SChris Mason struct bio *bio; 83ce9adaa5SChris Mason bio_end_io_t *end_io; 84ce9adaa5SChris Mason void *private; 85ce9adaa5SChris Mason struct btrfs_fs_info *info; 86ce9adaa5SChris Mason int error; 8722c59948SChris Mason int metadata; 88ce9adaa5SChris Mason struct list_head list; 898b712842SChris Mason struct btrfs_work work; 90ce9adaa5SChris Mason }; 910da5468fSChris Mason 92d352ac68SChris Mason /* 93d352ac68SChris Mason * async submit bios are used to offload expensive checksumming 94d352ac68SChris Mason * onto the worker threads. They checksum file and metadata bios 95d352ac68SChris Mason * just before they are sent down the IO stack. 96d352ac68SChris Mason */ 9744b8bd7eSChris Mason struct async_submit_bio { 9844b8bd7eSChris Mason struct inode *inode; 9944b8bd7eSChris Mason struct bio *bio; 10044b8bd7eSChris Mason struct list_head list; 1014a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start; 1024a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done; 10344b8bd7eSChris Mason int rw; 10444b8bd7eSChris Mason int mirror_num; 105c8b97818SChris Mason unsigned long bio_flags; 106eaf25d93SChris Mason /* 107eaf25d93SChris Mason * bio_offset is optional, can be used if the pages in the bio 108eaf25d93SChris Mason * can't tell us where in the file the bio should go 109eaf25d93SChris Mason */ 110eaf25d93SChris Mason u64 bio_offset; 1118b712842SChris Mason struct btrfs_work work; 11279787eaaSJeff Mahoney int error; 11344b8bd7eSChris Mason }; 11444b8bd7eSChris Mason 11585d4e461SChris Mason /* 11685d4e461SChris Mason * Lockdep class keys for extent_buffer->lock's in this root. For a given 11785d4e461SChris Mason * eb, the lockdep key is determined by the btrfs_root it belongs to and 11885d4e461SChris Mason * the level the eb occupies in the tree. 1194008c04aSChris Mason * 12085d4e461SChris Mason * Different roots are used for different purposes and may nest inside each 12185d4e461SChris Mason * other and they require separate keysets. As lockdep keys should be 12285d4e461SChris Mason * static, assign keysets according to the purpose of the root as indicated 12385d4e461SChris Mason * by btrfs_root->objectid. This ensures that all special purpose roots 12485d4e461SChris Mason * have separate keysets. 1254008c04aSChris Mason * 12685d4e461SChris Mason * Lock-nesting across peer nodes is always done with the immediate parent 12785d4e461SChris Mason * node locked thus preventing deadlock. As lockdep doesn't know this, use 12885d4e461SChris Mason * subclass to avoid triggering lockdep warning in such cases. 1294008c04aSChris Mason * 13085d4e461SChris Mason * The key is set by the readpage_end_io_hook after the buffer has passed 13185d4e461SChris Mason * csum validation but before the pages are unlocked. It is also set by 13285d4e461SChris Mason * btrfs_init_new_buffer on freshly allocated blocks. 13385d4e461SChris Mason * 13485d4e461SChris Mason * We also add a check to make sure the highest level of the tree is the 13585d4e461SChris Mason * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code 13685d4e461SChris Mason * needs update as well. 1374008c04aSChris Mason */ 1384008c04aSChris Mason #ifdef CONFIG_DEBUG_LOCK_ALLOC 1394008c04aSChris Mason # if BTRFS_MAX_LEVEL != 8 1404008c04aSChris Mason # error 1414008c04aSChris Mason # endif 14285d4e461SChris Mason 14385d4e461SChris Mason static struct btrfs_lockdep_keyset { 14485d4e461SChris Mason u64 id; /* root objectid */ 14585d4e461SChris Mason const char *name_stem; /* lock name stem */ 14685d4e461SChris Mason char names[BTRFS_MAX_LEVEL + 1][20]; 14785d4e461SChris Mason struct lock_class_key keys[BTRFS_MAX_LEVEL + 1]; 14885d4e461SChris Mason } btrfs_lockdep_keysets[] = { 14985d4e461SChris Mason { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" }, 15085d4e461SChris Mason { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" }, 15185d4e461SChris Mason { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" }, 15285d4e461SChris Mason { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" }, 15385d4e461SChris Mason { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" }, 15485d4e461SChris Mason { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" }, 15560b62978SDavid Sterba { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" }, 15685d4e461SChris Mason { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" }, 15785d4e461SChris Mason { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" }, 15885d4e461SChris Mason { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" }, 15913fd8da9SDavid Sterba { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" }, 16085d4e461SChris Mason { .id = 0, .name_stem = "tree" }, 1614008c04aSChris Mason }; 16285d4e461SChris Mason 16385d4e461SChris Mason void __init btrfs_init_lockdep(void) 16485d4e461SChris Mason { 16585d4e461SChris Mason int i, j; 16685d4e461SChris Mason 16785d4e461SChris Mason /* initialize lockdep class names */ 16885d4e461SChris Mason for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) { 16985d4e461SChris Mason struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i]; 17085d4e461SChris Mason 17185d4e461SChris Mason for (j = 0; j < ARRAY_SIZE(ks->names); j++) 17285d4e461SChris Mason snprintf(ks->names[j], sizeof(ks->names[j]), 17385d4e461SChris Mason "btrfs-%s-%02d", ks->name_stem, j); 17485d4e461SChris Mason } 17585d4e461SChris Mason } 17685d4e461SChris Mason 17785d4e461SChris Mason void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb, 17885d4e461SChris Mason int level) 17985d4e461SChris Mason { 18085d4e461SChris Mason struct btrfs_lockdep_keyset *ks; 18185d4e461SChris Mason 18285d4e461SChris Mason BUG_ON(level >= ARRAY_SIZE(ks->keys)); 18385d4e461SChris Mason 18485d4e461SChris Mason /* find the matching keyset, id 0 is the default entry */ 18585d4e461SChris Mason for (ks = btrfs_lockdep_keysets; ks->id; ks++) 18685d4e461SChris Mason if (ks->id == objectid) 18785d4e461SChris Mason break; 18885d4e461SChris Mason 18985d4e461SChris Mason lockdep_set_class_and_name(&eb->lock, 19085d4e461SChris Mason &ks->keys[level], ks->names[level]); 19185d4e461SChris Mason } 19285d4e461SChris Mason 1934008c04aSChris Mason #endif 1944008c04aSChris Mason 195d352ac68SChris Mason /* 196d352ac68SChris Mason * extents on the btree inode are pretty simple, there's one extent 197d352ac68SChris Mason * that covers the entire device 198d352ac68SChris Mason */ 199b2950863SChristoph Hellwig static struct extent_map *btree_get_extent(struct inode *inode, 200306e16ceSDavid Sterba struct page *page, size_t pg_offset, u64 start, u64 len, 2015f39d397SChris Mason int create) 2025f39d397SChris Mason { 2035f39d397SChris Mason struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 2045f39d397SChris Mason struct extent_map *em; 2055f39d397SChris Mason int ret; 2065f39d397SChris Mason 207890871beSChris Mason read_lock(&em_tree->lock); 208d1310b2eSChris Mason em = lookup_extent_mapping(em_tree, start, len); 209a061fc8dSChris Mason if (em) { 210a061fc8dSChris Mason em->bdev = 211a061fc8dSChris Mason BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 212890871beSChris Mason read_unlock(&em_tree->lock); 2135f39d397SChris Mason goto out; 214a061fc8dSChris Mason } 215890871beSChris Mason read_unlock(&em_tree->lock); 2167b13b7b1SChris Mason 217172ddd60SDavid Sterba em = alloc_extent_map(); 2185f39d397SChris Mason if (!em) { 2195f39d397SChris Mason em = ERR_PTR(-ENOMEM); 2205f39d397SChris Mason goto out; 2215f39d397SChris Mason } 2225f39d397SChris Mason em->start = 0; 2230afbaf8cSChris Mason em->len = (u64)-1; 224c8b97818SChris Mason em->block_len = (u64)-1; 2255f39d397SChris Mason em->block_start = 0; 226a061fc8dSChris Mason em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 227d1310b2eSChris Mason 228890871beSChris Mason write_lock(&em_tree->lock); 22909a2a8f9SJosef Bacik ret = add_extent_mapping(em_tree, em, 0); 2305f39d397SChris Mason if (ret == -EEXIST) { 2315f39d397SChris Mason free_extent_map(em); 2327b13b7b1SChris Mason em = lookup_extent_mapping(em_tree, start, len); 233b4f359abSTsutomu Itoh if (!em) 2340433f20dSTsutomu Itoh em = ERR_PTR(-EIO); 2355f39d397SChris Mason } else if (ret) { 2367b13b7b1SChris Mason free_extent_map(em); 2370433f20dSTsutomu Itoh em = ERR_PTR(ret); 2385f39d397SChris Mason } 239890871beSChris Mason write_unlock(&em_tree->lock); 2407b13b7b1SChris Mason 2415f39d397SChris Mason out: 2425f39d397SChris Mason return em; 2435f39d397SChris Mason } 2445f39d397SChris Mason 245b0496686SLiu Bo u32 btrfs_csum_data(char *data, u32 seed, size_t len) 24619c00ddcSChris Mason { 2470b947affSFilipe David Borba Manana return btrfs_crc32c(seed, data, len); 24819c00ddcSChris Mason } 24919c00ddcSChris Mason 25019c00ddcSChris Mason void btrfs_csum_final(u32 crc, char *result) 25119c00ddcSChris Mason { 2527e75bf3fSDavid Sterba put_unaligned_le32(~crc, result); 25319c00ddcSChris Mason } 25419c00ddcSChris Mason 255d352ac68SChris Mason /* 256d352ac68SChris Mason * compute the csum for a btree block, and either verify it or write it 257d352ac68SChris Mason * into the csum field of the block. 258d352ac68SChris Mason */ 25919c00ddcSChris Mason static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, 26019c00ddcSChris Mason int verify) 26119c00ddcSChris Mason { 2626c41761fSDavid Sterba u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy); 263607d432dSJosef Bacik char *result = NULL; 26419c00ddcSChris Mason unsigned long len; 26519c00ddcSChris Mason unsigned long cur_len; 26619c00ddcSChris Mason unsigned long offset = BTRFS_CSUM_SIZE; 26719c00ddcSChris Mason char *kaddr; 26819c00ddcSChris Mason unsigned long map_start; 26919c00ddcSChris Mason unsigned long map_len; 27019c00ddcSChris Mason int err; 27119c00ddcSChris Mason u32 crc = ~(u32)0; 272607d432dSJosef Bacik unsigned long inline_result; 27319c00ddcSChris Mason 27419c00ddcSChris Mason len = buf->len - offset; 27519c00ddcSChris Mason while (len > 0) { 27619c00ddcSChris Mason err = map_private_extent_buffer(buf, offset, 32, 277a6591715SChris Mason &kaddr, &map_start, &map_len); 278d397712bSChris Mason if (err) 27919c00ddcSChris Mason return 1; 28019c00ddcSChris Mason cur_len = min(len, map_len - (offset - map_start)); 281b0496686SLiu Bo crc = btrfs_csum_data(kaddr + offset - map_start, 28219c00ddcSChris Mason crc, cur_len); 28319c00ddcSChris Mason len -= cur_len; 28419c00ddcSChris Mason offset += cur_len; 28519c00ddcSChris Mason } 286607d432dSJosef Bacik if (csum_size > sizeof(inline_result)) { 287607d432dSJosef Bacik result = kzalloc(csum_size * sizeof(char), GFP_NOFS); 288607d432dSJosef Bacik if (!result) 289607d432dSJosef Bacik return 1; 290607d432dSJosef Bacik } else { 291607d432dSJosef Bacik result = (char *)&inline_result; 292607d432dSJosef Bacik } 293607d432dSJosef Bacik 29419c00ddcSChris Mason btrfs_csum_final(crc, result); 29519c00ddcSChris Mason 29619c00ddcSChris Mason if (verify) { 297607d432dSJosef Bacik if (memcmp_extent_buffer(buf, result, 0, csum_size)) { 298e4204dedSChris Mason u32 val; 299e4204dedSChris Mason u32 found = 0; 300607d432dSJosef Bacik memcpy(&found, result, csum_size); 301e4204dedSChris Mason 302607d432dSJosef Bacik read_extent_buffer(buf, &val, 0, csum_size); 303efe120a0SFrank Holton printk_ratelimited(KERN_INFO 304efe120a0SFrank Holton "BTRFS: %s checksum verify failed on %llu wanted %X found %X " 305193f284dSChris Mason "level %d\n", 306c1c9ff7cSGeert Uytterhoeven root->fs_info->sb->s_id, buf->start, 307c1c9ff7cSGeert Uytterhoeven val, found, btrfs_header_level(buf)); 308607d432dSJosef Bacik if (result != (char *)&inline_result) 309607d432dSJosef Bacik kfree(result); 31019c00ddcSChris Mason return 1; 31119c00ddcSChris Mason } 31219c00ddcSChris Mason } else { 313607d432dSJosef Bacik write_extent_buffer(buf, result, 0, csum_size); 31419c00ddcSChris Mason } 315607d432dSJosef Bacik if (result != (char *)&inline_result) 316607d432dSJosef Bacik kfree(result); 31719c00ddcSChris Mason return 0; 31819c00ddcSChris Mason } 31919c00ddcSChris Mason 320d352ac68SChris Mason /* 321d352ac68SChris Mason * we can't consider a given block up to date unless the transid of the 322d352ac68SChris Mason * block matches the transid in the parent node's pointer. This is how we 323d352ac68SChris Mason * detect blocks that either didn't get written at all or got written 324d352ac68SChris Mason * in the wrong place. 325d352ac68SChris Mason */ 3261259ab75SChris Mason static int verify_parent_transid(struct extent_io_tree *io_tree, 327b9fab919SChris Mason struct extent_buffer *eb, u64 parent_transid, 328b9fab919SChris Mason int atomic) 3291259ab75SChris Mason { 3302ac55d41SJosef Bacik struct extent_state *cached_state = NULL; 3311259ab75SChris Mason int ret; 332a26e8c9fSJosef Bacik bool need_lock = (current->journal_info == 333a26e8c9fSJosef Bacik (void *)BTRFS_SEND_TRANS_STUB); 3341259ab75SChris Mason 3351259ab75SChris Mason if (!parent_transid || btrfs_header_generation(eb) == parent_transid) 3361259ab75SChris Mason return 0; 3371259ab75SChris Mason 338b9fab919SChris Mason if (atomic) 339b9fab919SChris Mason return -EAGAIN; 340b9fab919SChris Mason 341a26e8c9fSJosef Bacik if (need_lock) { 342a26e8c9fSJosef Bacik btrfs_tree_read_lock(eb); 343a26e8c9fSJosef Bacik btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK); 344a26e8c9fSJosef Bacik } 345a26e8c9fSJosef Bacik 3462ac55d41SJosef Bacik lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1, 347d0082371SJeff Mahoney 0, &cached_state); 3480b32f4bbSJosef Bacik if (extent_buffer_uptodate(eb) && 3491259ab75SChris Mason btrfs_header_generation(eb) == parent_transid) { 3501259ab75SChris Mason ret = 0; 3511259ab75SChris Mason goto out; 3521259ab75SChris Mason } 3537a36ddecSDavid Sterba printk_ratelimited("parent transid verify failed on %llu wanted %llu " 354193f284dSChris Mason "found %llu\n", 355c1c9ff7cSGeert Uytterhoeven eb->start, parent_transid, btrfs_header_generation(eb)); 3561259ab75SChris Mason ret = 1; 357a26e8c9fSJosef Bacik 358a26e8c9fSJosef Bacik /* 359a26e8c9fSJosef Bacik * Things reading via commit roots that don't have normal protection, 360a26e8c9fSJosef Bacik * like send, can have a really old block in cache that may point at a 361a26e8c9fSJosef Bacik * block that has been free'd and re-allocated. So don't clear uptodate 362a26e8c9fSJosef Bacik * if we find an eb that is under IO (dirty/writeback) because we could 363a26e8c9fSJosef Bacik * end up reading in the stale data and then writing it back out and 364a26e8c9fSJosef Bacik * making everybody very sad. 365a26e8c9fSJosef Bacik */ 366a26e8c9fSJosef Bacik if (!extent_buffer_under_io(eb)) 3670b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 36833958dc6SChris Mason out: 3692ac55d41SJosef Bacik unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1, 3702ac55d41SJosef Bacik &cached_state, GFP_NOFS); 371a26e8c9fSJosef Bacik btrfs_tree_read_unlock_blocking(eb); 3721259ab75SChris Mason return ret; 3731259ab75SChris Mason } 3741259ab75SChris Mason 375d352ac68SChris Mason /* 3761104a885SDavid Sterba * Return 0 if the superblock checksum type matches the checksum value of that 3771104a885SDavid Sterba * algorithm. Pass the raw disk superblock data. 3781104a885SDavid Sterba */ 3791104a885SDavid Sterba static int btrfs_check_super_csum(char *raw_disk_sb) 3801104a885SDavid Sterba { 3811104a885SDavid Sterba struct btrfs_super_block *disk_sb = 3821104a885SDavid Sterba (struct btrfs_super_block *)raw_disk_sb; 3831104a885SDavid Sterba u16 csum_type = btrfs_super_csum_type(disk_sb); 3841104a885SDavid Sterba int ret = 0; 3851104a885SDavid Sterba 3861104a885SDavid Sterba if (csum_type == BTRFS_CSUM_TYPE_CRC32) { 3871104a885SDavid Sterba u32 crc = ~(u32)0; 3881104a885SDavid Sterba const int csum_size = sizeof(crc); 3891104a885SDavid Sterba char result[csum_size]; 3901104a885SDavid Sterba 3911104a885SDavid Sterba /* 3921104a885SDavid Sterba * The super_block structure does not span the whole 3931104a885SDavid Sterba * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space 3941104a885SDavid Sterba * is filled with zeros and is included in the checkum. 3951104a885SDavid Sterba */ 3961104a885SDavid Sterba crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE, 3971104a885SDavid Sterba crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); 3981104a885SDavid Sterba btrfs_csum_final(crc, result); 3991104a885SDavid Sterba 4001104a885SDavid Sterba if (memcmp(raw_disk_sb, result, csum_size)) 4011104a885SDavid Sterba ret = 1; 402667e7d94SChris Mason 403667e7d94SChris Mason if (ret && btrfs_super_generation(disk_sb) < 10) { 404efe120a0SFrank Holton printk(KERN_WARNING 405efe120a0SFrank Holton "BTRFS: super block crcs don't match, older mkfs detected\n"); 406667e7d94SChris Mason ret = 0; 407667e7d94SChris Mason } 4081104a885SDavid Sterba } 4091104a885SDavid Sterba 4101104a885SDavid Sterba if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) { 411efe120a0SFrank Holton printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n", 4121104a885SDavid Sterba csum_type); 4131104a885SDavid Sterba ret = 1; 4141104a885SDavid Sterba } 4151104a885SDavid Sterba 4161104a885SDavid Sterba return ret; 4171104a885SDavid Sterba } 4181104a885SDavid Sterba 4191104a885SDavid Sterba /* 420d352ac68SChris Mason * helper to read a given tree block, doing retries as required when 421d352ac68SChris Mason * the checksums don't match and we have alternate mirrors to try. 422d352ac68SChris Mason */ 423f188591eSChris Mason static int btree_read_extent_buffer_pages(struct btrfs_root *root, 424f188591eSChris Mason struct extent_buffer *eb, 425ca7a79adSChris Mason u64 start, u64 parent_transid) 426f188591eSChris Mason { 427f188591eSChris Mason struct extent_io_tree *io_tree; 428ea466794SJosef Bacik int failed = 0; 429f188591eSChris Mason int ret; 430f188591eSChris Mason int num_copies = 0; 431f188591eSChris Mason int mirror_num = 0; 432ea466794SJosef Bacik int failed_mirror = 0; 433f188591eSChris Mason 434a826d6dcSJosef Bacik clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 435f188591eSChris Mason io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; 436f188591eSChris Mason while (1) { 437bb82ab88SArne Jansen ret = read_extent_buffer_pages(io_tree, eb, start, 438bb82ab88SArne Jansen WAIT_COMPLETE, 439f188591eSChris Mason btree_get_extent, mirror_num); 440256dd1bbSStefan Behrens if (!ret) { 441256dd1bbSStefan Behrens if (!verify_parent_transid(io_tree, eb, 442b9fab919SChris Mason parent_transid, 0)) 443ea466794SJosef Bacik break; 444256dd1bbSStefan Behrens else 445256dd1bbSStefan Behrens ret = -EIO; 446256dd1bbSStefan Behrens } 447d397712bSChris Mason 448a826d6dcSJosef Bacik /* 449a826d6dcSJosef Bacik * This buffer's crc is fine, but its contents are corrupted, so 450a826d6dcSJosef Bacik * there is no reason to read the other copies, they won't be 451a826d6dcSJosef Bacik * any less wrong. 452a826d6dcSJosef Bacik */ 453a826d6dcSJosef Bacik if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) 454ea466794SJosef Bacik break; 455ea466794SJosef Bacik 4565d964051SStefan Behrens num_copies = btrfs_num_copies(root->fs_info, 457f188591eSChris Mason eb->start, eb->len); 4584235298eSChris Mason if (num_copies == 1) 459ea466794SJosef Bacik break; 4604235298eSChris Mason 4615cf1ab56SJosef Bacik if (!failed_mirror) { 4625cf1ab56SJosef Bacik failed = 1; 4635cf1ab56SJosef Bacik failed_mirror = eb->read_mirror; 4645cf1ab56SJosef Bacik } 4655cf1ab56SJosef Bacik 466f188591eSChris Mason mirror_num++; 467ea466794SJosef Bacik if (mirror_num == failed_mirror) 468ea466794SJosef Bacik mirror_num++; 469ea466794SJosef Bacik 4704235298eSChris Mason if (mirror_num > num_copies) 471ea466794SJosef Bacik break; 472f188591eSChris Mason } 473ea466794SJosef Bacik 474c0901581SStefan Behrens if (failed && !ret && failed_mirror) 475ea466794SJosef Bacik repair_eb_io_failure(root, eb, failed_mirror); 476ea466794SJosef Bacik 477ea466794SJosef Bacik return ret; 478f188591eSChris Mason } 47919c00ddcSChris Mason 480d352ac68SChris Mason /* 481d397712bSChris Mason * checksum a dirty tree block before IO. This has extra checks to make sure 482d397712bSChris Mason * we only fill in the checksum field in the first page of a multi-page block 483d352ac68SChris Mason */ 484d397712bSChris Mason 485b2950863SChristoph Hellwig static int csum_dirty_buffer(struct btrfs_root *root, struct page *page) 48619c00ddcSChris Mason { 4874eee4fa4SMiao Xie u64 start = page_offset(page); 48819c00ddcSChris Mason u64 found_start; 48919c00ddcSChris Mason struct extent_buffer *eb; 490f188591eSChris Mason 4914f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 4924f2de97aSJosef Bacik if (page != eb->pages[0]) 4934f2de97aSJosef Bacik return 0; 49419c00ddcSChris Mason found_start = btrfs_header_bytenr(eb); 495fae7f21cSDulshani Gunawardhana if (WARN_ON(found_start != start || !PageUptodate(page))) 4964f2de97aSJosef Bacik return 0; 49719c00ddcSChris Mason csum_tree_block(root, eb, 0); 49819c00ddcSChris Mason return 0; 49919c00ddcSChris Mason } 50019c00ddcSChris Mason 5012b82032cSYan Zheng static int check_tree_block_fsid(struct btrfs_root *root, 5022b82032cSYan Zheng struct extent_buffer *eb) 5032b82032cSYan Zheng { 5042b82032cSYan Zheng struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; 5052b82032cSYan Zheng u8 fsid[BTRFS_UUID_SIZE]; 5062b82032cSYan Zheng int ret = 1; 5072b82032cSYan Zheng 5080a4e5586SRoss Kirk read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE); 5092b82032cSYan Zheng while (fs_devices) { 5102b82032cSYan Zheng if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { 5112b82032cSYan Zheng ret = 0; 5122b82032cSYan Zheng break; 5132b82032cSYan Zheng } 5142b82032cSYan Zheng fs_devices = fs_devices->seed; 5152b82032cSYan Zheng } 5162b82032cSYan Zheng return ret; 5172b82032cSYan Zheng } 5182b82032cSYan Zheng 519a826d6dcSJosef Bacik #define CORRUPT(reason, eb, root, slot) \ 520efe120a0SFrank Holton btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \ 521efe120a0SFrank Holton "root=%llu, slot=%d", reason, \ 522c1c9ff7cSGeert Uytterhoeven btrfs_header_bytenr(eb), root->objectid, slot) 523a826d6dcSJosef Bacik 524a826d6dcSJosef Bacik static noinline int check_leaf(struct btrfs_root *root, 525a826d6dcSJosef Bacik struct extent_buffer *leaf) 526a826d6dcSJosef Bacik { 527a826d6dcSJosef Bacik struct btrfs_key key; 528a826d6dcSJosef Bacik struct btrfs_key leaf_key; 529a826d6dcSJosef Bacik u32 nritems = btrfs_header_nritems(leaf); 530a826d6dcSJosef Bacik int slot; 531a826d6dcSJosef Bacik 532a826d6dcSJosef Bacik if (nritems == 0) 533a826d6dcSJosef Bacik return 0; 534a826d6dcSJosef Bacik 535a826d6dcSJosef Bacik /* Check the 0 item */ 536a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != 537a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 538a826d6dcSJosef Bacik CORRUPT("invalid item offset size pair", leaf, root, 0); 539a826d6dcSJosef Bacik return -EIO; 540a826d6dcSJosef Bacik } 541a826d6dcSJosef Bacik 542a826d6dcSJosef Bacik /* 543a826d6dcSJosef Bacik * Check to make sure each items keys are in the correct order and their 544a826d6dcSJosef Bacik * offsets make sense. We only have to loop through nritems-1 because 545a826d6dcSJosef Bacik * we check the current slot against the next slot, which verifies the 546a826d6dcSJosef Bacik * next slot's offset+size makes sense and that the current's slot 547a826d6dcSJosef Bacik * offset is correct. 548a826d6dcSJosef Bacik */ 549a826d6dcSJosef Bacik for (slot = 0; slot < nritems - 1; slot++) { 550a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &leaf_key, slot); 551a826d6dcSJosef Bacik btrfs_item_key_to_cpu(leaf, &key, slot + 1); 552a826d6dcSJosef Bacik 553a826d6dcSJosef Bacik /* Make sure the keys are in the right order */ 554a826d6dcSJosef Bacik if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) { 555a826d6dcSJosef Bacik CORRUPT("bad key order", leaf, root, slot); 556a826d6dcSJosef Bacik return -EIO; 557a826d6dcSJosef Bacik } 558a826d6dcSJosef Bacik 559a826d6dcSJosef Bacik /* 560a826d6dcSJosef Bacik * Make sure the offset and ends are right, remember that the 561a826d6dcSJosef Bacik * item data starts at the end of the leaf and grows towards the 562a826d6dcSJosef Bacik * front. 563a826d6dcSJosef Bacik */ 564a826d6dcSJosef Bacik if (btrfs_item_offset_nr(leaf, slot) != 565a826d6dcSJosef Bacik btrfs_item_end_nr(leaf, slot + 1)) { 566a826d6dcSJosef Bacik CORRUPT("slot offset bad", leaf, root, slot); 567a826d6dcSJosef Bacik return -EIO; 568a826d6dcSJosef Bacik } 569a826d6dcSJosef Bacik 570a826d6dcSJosef Bacik /* 571a826d6dcSJosef Bacik * Check to make sure that we don't point outside of the leaf, 572a826d6dcSJosef Bacik * just incase all the items are consistent to eachother, but 573a826d6dcSJosef Bacik * all point outside of the leaf. 574a826d6dcSJosef Bacik */ 575a826d6dcSJosef Bacik if (btrfs_item_end_nr(leaf, slot) > 576a826d6dcSJosef Bacik BTRFS_LEAF_DATA_SIZE(root)) { 577a826d6dcSJosef Bacik CORRUPT("slot end outside of leaf", leaf, root, slot); 578a826d6dcSJosef Bacik return -EIO; 579a826d6dcSJosef Bacik } 580a826d6dcSJosef Bacik } 581a826d6dcSJosef Bacik 582a826d6dcSJosef Bacik return 0; 583a826d6dcSJosef Bacik } 584a826d6dcSJosef Bacik 585facc8a22SMiao Xie static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio, 586facc8a22SMiao Xie u64 phy_offset, struct page *page, 587facc8a22SMiao Xie u64 start, u64 end, int mirror) 588ce9adaa5SChris Mason { 589ce9adaa5SChris Mason u64 found_start; 590ce9adaa5SChris Mason int found_level; 591ce9adaa5SChris Mason struct extent_buffer *eb; 592ce9adaa5SChris Mason struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 593f188591eSChris Mason int ret = 0; 594727011e0SChris Mason int reads_done; 595ce9adaa5SChris Mason 596ce9adaa5SChris Mason if (!page->private) 597ce9adaa5SChris Mason goto out; 598d397712bSChris Mason 5994f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 600d397712bSChris Mason 6010b32f4bbSJosef Bacik /* the pending IO might have been the only thing that kept this buffer 6020b32f4bbSJosef Bacik * in memory. Make sure we have a ref for all this other checks 6030b32f4bbSJosef Bacik */ 6040b32f4bbSJosef Bacik extent_buffer_get(eb); 6050b32f4bbSJosef Bacik 6060b32f4bbSJosef Bacik reads_done = atomic_dec_and_test(&eb->io_pages); 607727011e0SChris Mason if (!reads_done) 608727011e0SChris Mason goto err; 609f188591eSChris Mason 6105cf1ab56SJosef Bacik eb->read_mirror = mirror; 611ea466794SJosef Bacik if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) { 612ea466794SJosef Bacik ret = -EIO; 613ea466794SJosef Bacik goto err; 614ea466794SJosef Bacik } 615ea466794SJosef Bacik 616ce9adaa5SChris Mason found_start = btrfs_header_bytenr(eb); 617727011e0SChris Mason if (found_start != eb->start) { 618efe120a0SFrank Holton printk_ratelimited(KERN_INFO "BTRFS: bad tree block start " 619193f284dSChris Mason "%llu %llu\n", 620c1c9ff7cSGeert Uytterhoeven found_start, eb->start); 621f188591eSChris Mason ret = -EIO; 622ce9adaa5SChris Mason goto err; 623ce9adaa5SChris Mason } 6242b82032cSYan Zheng if (check_tree_block_fsid(root, eb)) { 625efe120a0SFrank Holton printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n", 626c1c9ff7cSGeert Uytterhoeven eb->start); 6271259ab75SChris Mason ret = -EIO; 6281259ab75SChris Mason goto err; 6291259ab75SChris Mason } 630ce9adaa5SChris Mason found_level = btrfs_header_level(eb); 6311c24c3ceSJosef Bacik if (found_level >= BTRFS_MAX_LEVEL) { 632efe120a0SFrank Holton btrfs_info(root->fs_info, "bad tree block level %d", 6331c24c3ceSJosef Bacik (int)btrfs_header_level(eb)); 6341c24c3ceSJosef Bacik ret = -EIO; 6351c24c3ceSJosef Bacik goto err; 6361c24c3ceSJosef Bacik } 637ce9adaa5SChris Mason 63885d4e461SChris Mason btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), 63985d4e461SChris Mason eb, found_level); 6404008c04aSChris Mason 641ce9adaa5SChris Mason ret = csum_tree_block(root, eb, 1); 642a826d6dcSJosef Bacik if (ret) { 643f188591eSChris Mason ret = -EIO; 644a826d6dcSJosef Bacik goto err; 645a826d6dcSJosef Bacik } 646a826d6dcSJosef Bacik 647a826d6dcSJosef Bacik /* 648a826d6dcSJosef Bacik * If this is a leaf block and it is corrupt, set the corrupt bit so 649a826d6dcSJosef Bacik * that we don't try and read the other copies of this block, just 650a826d6dcSJosef Bacik * return -EIO. 651a826d6dcSJosef Bacik */ 652a826d6dcSJosef Bacik if (found_level == 0 && check_leaf(root, eb)) { 653a826d6dcSJosef Bacik set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 654a826d6dcSJosef Bacik ret = -EIO; 655a826d6dcSJosef Bacik } 656ce9adaa5SChris Mason 6570b32f4bbSJosef Bacik if (!ret) 6580b32f4bbSJosef Bacik set_extent_buffer_uptodate(eb); 659ce9adaa5SChris Mason err: 66079fb65a1SJosef Bacik if (reads_done && 66179fb65a1SJosef Bacik test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 6624bb31e92SArne Jansen btree_readahead_hook(root, eb, eb->start, ret); 6634bb31e92SArne Jansen 66453b381b3SDavid Woodhouse if (ret) { 66553b381b3SDavid Woodhouse /* 66653b381b3SDavid Woodhouse * our io error hook is going to dec the io pages 66753b381b3SDavid Woodhouse * again, we have to make sure it has something 66853b381b3SDavid Woodhouse * to decrement 66953b381b3SDavid Woodhouse */ 67053b381b3SDavid Woodhouse atomic_inc(&eb->io_pages); 6710b32f4bbSJosef Bacik clear_extent_buffer_uptodate(eb); 67253b381b3SDavid Woodhouse } 6730b32f4bbSJosef Bacik free_extent_buffer(eb); 674ce9adaa5SChris Mason out: 675f188591eSChris Mason return ret; 676ce9adaa5SChris Mason } 677ce9adaa5SChris Mason 678ea466794SJosef Bacik static int btree_io_failed_hook(struct page *page, int failed_mirror) 6794bb31e92SArne Jansen { 6804bb31e92SArne Jansen struct extent_buffer *eb; 6814bb31e92SArne Jansen struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; 6824bb31e92SArne Jansen 6834f2de97aSJosef Bacik eb = (struct extent_buffer *)page->private; 684ea466794SJosef Bacik set_bit(EXTENT_BUFFER_IOERR, &eb->bflags); 6855cf1ab56SJosef Bacik eb->read_mirror = failed_mirror; 68653b381b3SDavid Woodhouse atomic_dec(&eb->io_pages); 687ea466794SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 6884bb31e92SArne Jansen btree_readahead_hook(root, eb, eb->start, -EIO); 6894bb31e92SArne Jansen return -EIO; /* we fixed nothing */ 6904bb31e92SArne Jansen } 6914bb31e92SArne Jansen 692ce9adaa5SChris Mason static void end_workqueue_bio(struct bio *bio, int err) 693ce9adaa5SChris Mason { 694ce9adaa5SChris Mason struct end_io_wq *end_io_wq = bio->bi_private; 695ce9adaa5SChris Mason struct btrfs_fs_info *fs_info; 696ce9adaa5SChris Mason 697ce9adaa5SChris Mason fs_info = end_io_wq->info; 698ce9adaa5SChris Mason end_io_wq->error = err; 699fccb5d86SQu Wenruo btrfs_init_work(&end_io_wq->work, end_workqueue_fn, NULL, NULL); 700d20f7043SChris Mason 7017b6d91daSChristoph Hellwig if (bio->bi_rw & REQ_WRITE) { 70253b381b3SDavid Woodhouse if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) 703fccb5d86SQu Wenruo btrfs_queue_work(fs_info->endio_meta_write_workers, 704cad321adSChris Mason &end_io_wq->work); 70553b381b3SDavid Woodhouse else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) 706fccb5d86SQu Wenruo btrfs_queue_work(fs_info->endio_freespace_worker, 7070cb59c99SJosef Bacik &end_io_wq->work); 70853b381b3SDavid Woodhouse else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 709fccb5d86SQu Wenruo btrfs_queue_work(fs_info->endio_raid56_workers, 71053b381b3SDavid Woodhouse &end_io_wq->work); 711cad321adSChris Mason else 712fccb5d86SQu Wenruo btrfs_queue_work(fs_info->endio_write_workers, 713e6dcd2dcSChris Mason &end_io_wq->work); 714d20f7043SChris Mason } else { 71553b381b3SDavid Woodhouse if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) 716fccb5d86SQu Wenruo btrfs_queue_work(fs_info->endio_raid56_workers, 71753b381b3SDavid Woodhouse &end_io_wq->work); 71853b381b3SDavid Woodhouse else if (end_io_wq->metadata) 719fccb5d86SQu Wenruo btrfs_queue_work(fs_info->endio_meta_workers, 720d20f7043SChris Mason &end_io_wq->work); 721e6dcd2dcSChris Mason else 722fccb5d86SQu Wenruo btrfs_queue_work(fs_info->endio_workers, 723d20f7043SChris Mason &end_io_wq->work); 724d20f7043SChris Mason } 725ce9adaa5SChris Mason } 726ce9adaa5SChris Mason 7270cb59c99SJosef Bacik /* 7280cb59c99SJosef Bacik * For the metadata arg you want 7290cb59c99SJosef Bacik * 7300cb59c99SJosef Bacik * 0 - if data 7310cb59c99SJosef Bacik * 1 - if normal metadta 7320cb59c99SJosef Bacik * 2 - if writing to the free space cache area 73353b381b3SDavid Woodhouse * 3 - raid parity work 7340cb59c99SJosef Bacik */ 73522c59948SChris Mason int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 73622c59948SChris Mason int metadata) 7370b86a832SChris Mason { 738ce9adaa5SChris Mason struct end_io_wq *end_io_wq; 739ce9adaa5SChris Mason end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); 740ce9adaa5SChris Mason if (!end_io_wq) 741ce9adaa5SChris Mason return -ENOMEM; 742ce9adaa5SChris Mason 743ce9adaa5SChris Mason end_io_wq->private = bio->bi_private; 744ce9adaa5SChris Mason end_io_wq->end_io = bio->bi_end_io; 74522c59948SChris Mason end_io_wq->info = info; 746ce9adaa5SChris Mason end_io_wq->error = 0; 747ce9adaa5SChris Mason end_io_wq->bio = bio; 74822c59948SChris Mason end_io_wq->metadata = metadata; 749ce9adaa5SChris Mason 750ce9adaa5SChris Mason bio->bi_private = end_io_wq; 751ce9adaa5SChris Mason bio->bi_end_io = end_workqueue_bio; 75222c59948SChris Mason return 0; 75322c59948SChris Mason } 75422c59948SChris Mason 755b64a2851SChris Mason unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) 7564854ddd0SChris Mason { 7574854ddd0SChris Mason unsigned long limit = min_t(unsigned long, 7585cdc7ad3SQu Wenruo info->thread_pool_size, 7594854ddd0SChris Mason info->fs_devices->open_devices); 7604854ddd0SChris Mason return 256 * limit; 7614854ddd0SChris Mason } 7624854ddd0SChris Mason 7634a69a410SChris Mason static void run_one_async_start(struct btrfs_work *work) 7644a69a410SChris Mason { 7654a69a410SChris Mason struct async_submit_bio *async; 76679787eaaSJeff Mahoney int ret; 7674a69a410SChris Mason 7684a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 76979787eaaSJeff Mahoney ret = async->submit_bio_start(async->inode, async->rw, async->bio, 770eaf25d93SChris Mason async->mirror_num, async->bio_flags, 771eaf25d93SChris Mason async->bio_offset); 77279787eaaSJeff Mahoney if (ret) 77379787eaaSJeff Mahoney async->error = ret; 7744a69a410SChris Mason } 7754a69a410SChris Mason 7764a69a410SChris Mason static void run_one_async_done(struct btrfs_work *work) 7778b712842SChris Mason { 7788b712842SChris Mason struct btrfs_fs_info *fs_info; 7798b712842SChris Mason struct async_submit_bio *async; 7804854ddd0SChris Mason int limit; 7818b712842SChris Mason 7828b712842SChris Mason async = container_of(work, struct async_submit_bio, work); 7838b712842SChris Mason fs_info = BTRFS_I(async->inode)->root->fs_info; 7844854ddd0SChris Mason 785b64a2851SChris Mason limit = btrfs_async_submit_limit(fs_info); 7864854ddd0SChris Mason limit = limit * 2 / 3; 7874854ddd0SChris Mason 78866657b31SJosef Bacik if (atomic_dec_return(&fs_info->nr_async_submits) < limit && 789b64a2851SChris Mason waitqueue_active(&fs_info->async_submit_wait)) 7904854ddd0SChris Mason wake_up(&fs_info->async_submit_wait); 7914854ddd0SChris Mason 79279787eaaSJeff Mahoney /* If an error occured we just want to clean up the bio and move on */ 79379787eaaSJeff Mahoney if (async->error) { 79479787eaaSJeff Mahoney bio_endio(async->bio, async->error); 79579787eaaSJeff Mahoney return; 79679787eaaSJeff Mahoney } 79779787eaaSJeff Mahoney 7984a69a410SChris Mason async->submit_bio_done(async->inode, async->rw, async->bio, 799eaf25d93SChris Mason async->mirror_num, async->bio_flags, 800eaf25d93SChris Mason async->bio_offset); 8014a69a410SChris Mason } 8024a69a410SChris Mason 8034a69a410SChris Mason static void run_one_async_free(struct btrfs_work *work) 8044a69a410SChris Mason { 8054a69a410SChris Mason struct async_submit_bio *async; 8064a69a410SChris Mason 8074a69a410SChris Mason async = container_of(work, struct async_submit_bio, work); 8088b712842SChris Mason kfree(async); 8098b712842SChris Mason } 8108b712842SChris Mason 81144b8bd7eSChris Mason int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, 81244b8bd7eSChris Mason int rw, struct bio *bio, int mirror_num, 813c8b97818SChris Mason unsigned long bio_flags, 814eaf25d93SChris Mason u64 bio_offset, 8154a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_start, 8164a69a410SChris Mason extent_submit_bio_hook_t *submit_bio_done) 81744b8bd7eSChris Mason { 81844b8bd7eSChris Mason struct async_submit_bio *async; 81944b8bd7eSChris Mason 82044b8bd7eSChris Mason async = kmalloc(sizeof(*async), GFP_NOFS); 82144b8bd7eSChris Mason if (!async) 82244b8bd7eSChris Mason return -ENOMEM; 82344b8bd7eSChris Mason 82444b8bd7eSChris Mason async->inode = inode; 82544b8bd7eSChris Mason async->rw = rw; 82644b8bd7eSChris Mason async->bio = bio; 82744b8bd7eSChris Mason async->mirror_num = mirror_num; 8284a69a410SChris Mason async->submit_bio_start = submit_bio_start; 8294a69a410SChris Mason async->submit_bio_done = submit_bio_done; 8304a69a410SChris Mason 8315cdc7ad3SQu Wenruo btrfs_init_work(&async->work, run_one_async_start, 8325cdc7ad3SQu Wenruo run_one_async_done, run_one_async_free); 8334a69a410SChris Mason 834c8b97818SChris Mason async->bio_flags = bio_flags; 835eaf25d93SChris Mason async->bio_offset = bio_offset; 8368c8bee1dSChris Mason 83779787eaaSJeff Mahoney async->error = 0; 83879787eaaSJeff Mahoney 839cb03c743SChris Mason atomic_inc(&fs_info->nr_async_submits); 840d313d7a3SChris Mason 8417b6d91daSChristoph Hellwig if (rw & REQ_SYNC) 8425cdc7ad3SQu Wenruo btrfs_set_work_high_priority(&async->work); 843d313d7a3SChris Mason 8445cdc7ad3SQu Wenruo btrfs_queue_work(fs_info->workers, &async->work); 8459473f16cSChris Mason 846771ed689SChris Mason while (atomic_read(&fs_info->async_submit_draining) && 847771ed689SChris Mason atomic_read(&fs_info->nr_async_submits)) { 848771ed689SChris Mason wait_event(fs_info->async_submit_wait, 849771ed689SChris Mason (atomic_read(&fs_info->nr_async_submits) == 0)); 850771ed689SChris Mason } 851771ed689SChris Mason 85244b8bd7eSChris Mason return 0; 85344b8bd7eSChris Mason } 85444b8bd7eSChris Mason 855ce3ed71aSChris Mason static int btree_csum_one_bio(struct bio *bio) 856ce3ed71aSChris Mason { 8572c30c71bSKent Overstreet struct bio_vec *bvec; 858ce3ed71aSChris Mason struct btrfs_root *root; 8592c30c71bSKent Overstreet int i, ret = 0; 860ce3ed71aSChris Mason 8612c30c71bSKent Overstreet bio_for_each_segment_all(bvec, bio, i) { 862ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 86379787eaaSJeff Mahoney ret = csum_dirty_buffer(root, bvec->bv_page); 86479787eaaSJeff Mahoney if (ret) 86579787eaaSJeff Mahoney break; 866ce3ed71aSChris Mason } 8672c30c71bSKent Overstreet 86879787eaaSJeff Mahoney return ret; 869ce3ed71aSChris Mason } 870ce3ed71aSChris Mason 8714a69a410SChris Mason static int __btree_submit_bio_start(struct inode *inode, int rw, 8724a69a410SChris Mason struct bio *bio, int mirror_num, 873eaf25d93SChris Mason unsigned long bio_flags, 874eaf25d93SChris Mason u64 bio_offset) 87522c59948SChris Mason { 8768b712842SChris Mason /* 8778b712842SChris Mason * when we're called for a write, we're already in the async 8785443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8798b712842SChris Mason */ 88079787eaaSJeff Mahoney return btree_csum_one_bio(bio); 88122c59948SChris Mason } 88222c59948SChris Mason 8834a69a410SChris Mason static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, 884eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 885eaf25d93SChris Mason u64 bio_offset) 8864a69a410SChris Mason { 88761891923SStefan Behrens int ret; 88861891923SStefan Behrens 8898b712842SChris Mason /* 8904a69a410SChris Mason * when we're called for a write, we're already in the async 8914a69a410SChris Mason * submission context. Just jump into btrfs_map_bio 8928b712842SChris Mason */ 89361891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); 89461891923SStefan Behrens if (ret) 89561891923SStefan Behrens bio_endio(bio, ret); 89661891923SStefan Behrens return ret; 8970b86a832SChris Mason } 8980b86a832SChris Mason 899de0022b9SJosef Bacik static int check_async_write(struct inode *inode, unsigned long bio_flags) 900de0022b9SJosef Bacik { 901de0022b9SJosef Bacik if (bio_flags & EXTENT_BIO_TREE_LOG) 902de0022b9SJosef Bacik return 0; 903de0022b9SJosef Bacik #ifdef CONFIG_X86 904de0022b9SJosef Bacik if (cpu_has_xmm4_2) 905de0022b9SJosef Bacik return 0; 906de0022b9SJosef Bacik #endif 907de0022b9SJosef Bacik return 1; 908de0022b9SJosef Bacik } 909de0022b9SJosef Bacik 91044b8bd7eSChris Mason static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, 911eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 912eaf25d93SChris Mason u64 bio_offset) 91344b8bd7eSChris Mason { 914de0022b9SJosef Bacik int async = check_async_write(inode, bio_flags); 9154a69a410SChris Mason int ret; 916cad321adSChris Mason 9177b6d91daSChristoph Hellwig if (!(rw & REQ_WRITE)) { 918cad321adSChris Mason /* 919cad321adSChris Mason * called for a read, do the setup so that checksum validation 920cad321adSChris Mason * can happen in the async kernel threads 921cad321adSChris Mason */ 922f3f266abSChris Mason ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, 923f3f266abSChris Mason bio, 1); 9241d4284bdSChris Mason if (ret) 92561891923SStefan Behrens goto out_w_error; 92661891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 9276f3577bdSChris Mason mirror_num, 0); 928de0022b9SJosef Bacik } else if (!async) { 929de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 930de0022b9SJosef Bacik if (ret) 93161891923SStefan Behrens goto out_w_error; 93261891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 933de0022b9SJosef Bacik mirror_num, 0); 93461891923SStefan Behrens } else { 935cad321adSChris Mason /* 93661891923SStefan Behrens * kthread helpers are used to submit writes so that 93761891923SStefan Behrens * checksumming can happen in parallel across all CPUs 938cad321adSChris Mason */ 93961891923SStefan Behrens ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, 940c8b97818SChris Mason inode, rw, bio, mirror_num, 0, 941eaf25d93SChris Mason bio_offset, 9424a69a410SChris Mason __btree_submit_bio_start, 9434a69a410SChris Mason __btree_submit_bio_done); 94444b8bd7eSChris Mason } 94544b8bd7eSChris Mason 94661891923SStefan Behrens if (ret) { 94761891923SStefan Behrens out_w_error: 94861891923SStefan Behrens bio_endio(bio, ret); 94961891923SStefan Behrens } 95061891923SStefan Behrens return ret; 95161891923SStefan Behrens } 95261891923SStefan Behrens 9533dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 954784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 955a6bc32b8SMel Gorman struct page *newpage, struct page *page, 956a6bc32b8SMel Gorman enum migrate_mode mode) 957784b4e29SChris Mason { 958784b4e29SChris Mason /* 959784b4e29SChris Mason * we can't safely write a btree page from here, 960784b4e29SChris Mason * we haven't done the locking hook 961784b4e29SChris Mason */ 962784b4e29SChris Mason if (PageDirty(page)) 963784b4e29SChris Mason return -EAGAIN; 964784b4e29SChris Mason /* 965784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 966784b4e29SChris Mason * We must have no buffers or drop them. 967784b4e29SChris Mason */ 968784b4e29SChris Mason if (page_has_private(page) && 969784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 970784b4e29SChris Mason return -EAGAIN; 971a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 972784b4e29SChris Mason } 9733dd1462eSJan Beulich #endif 974784b4e29SChris Mason 9750da5468fSChris Mason 9760da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9770da5468fSChris Mason struct writeback_control *wbc) 9780da5468fSChris Mason { 979e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 980e2d84521SMiao Xie int ret; 981e2d84521SMiao Xie 982d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 983448d640bSChris Mason 984448d640bSChris Mason if (wbc->for_kupdate) 985448d640bSChris Mason return 0; 986448d640bSChris Mason 987e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 988b9473439SChris Mason /* this is a bit racy, but that's ok */ 989e2d84521SMiao Xie ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 990e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 991e2d84521SMiao Xie if (ret < 0) 992793955bcSChris Mason return 0; 993793955bcSChris Mason } 9940b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9950da5468fSChris Mason } 9960da5468fSChris Mason 997b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 9985f39d397SChris Mason { 999d1310b2eSChris Mason struct extent_io_tree *tree; 1000d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10018ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 10025f39d397SChris Mason } 10035f39d397SChris Mason 100470dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 10055f39d397SChris Mason { 100698509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 100798509cfcSChris Mason return 0; 10080c4e538bSDavid Sterba 1009f7a52a40SDavid Sterba return try_release_extent_buffer(page); 1010d98237b3SChris Mason } 1011d98237b3SChris Mason 1012d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 1013d47992f8SLukas Czerner unsigned int length) 1014d98237b3SChris Mason { 1015d1310b2eSChris Mason struct extent_io_tree *tree; 1016d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10175f39d397SChris Mason extent_invalidatepage(tree, page, offset); 10185f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 10199ad6b7bcSChris Mason if (PagePrivate(page)) { 1020efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 1021efe120a0SFrank Holton "page private not zero on page %llu", 1022efe120a0SFrank Holton (unsigned long long)page_offset(page)); 10239ad6b7bcSChris Mason ClearPagePrivate(page); 10249ad6b7bcSChris Mason set_page_private(page, 0); 10259ad6b7bcSChris Mason page_cache_release(page); 10269ad6b7bcSChris Mason } 1027d98237b3SChris Mason } 1028d98237b3SChris Mason 10290b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 10300b32f4bbSJosef Bacik { 1031bb146eb2SJosef Bacik #ifdef DEBUG 10320b32f4bbSJosef Bacik struct extent_buffer *eb; 10330b32f4bbSJosef Bacik 10340b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 10350b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 10360b32f4bbSJosef Bacik BUG_ON(!eb); 10370b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10380b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 10390b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 1040bb146eb2SJosef Bacik #endif 10410b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10420b32f4bbSJosef Bacik } 10430b32f4bbSJosef Bacik 10447f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 1045d98237b3SChris Mason .readpage = btree_readpage, 10460da5468fSChris Mason .writepages = btree_writepages, 10475f39d397SChris Mason .releasepage = btree_releasepage, 10485f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10495a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1050784b4e29SChris Mason .migratepage = btree_migratepage, 10515a92bc88SChris Mason #endif 10520b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1053d98237b3SChris Mason }; 1054123abc88SChris Mason 1055ca7a79adSChris Mason int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1056ca7a79adSChris Mason u64 parent_transid) 1057090d1875SChris Mason { 10585f39d397SChris Mason struct extent_buffer *buf = NULL; 10595f39d397SChris Mason struct inode *btree_inode = root->fs_info->btree_inode; 1060de428b63SChris Mason int ret = 0; 1061090d1875SChris Mason 1062db94535dSChris Mason buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 10635f39d397SChris Mason if (!buf) 1064090d1875SChris Mason return 0; 1065d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 1066bb82ab88SArne Jansen buf, 0, WAIT_NONE, btree_get_extent, 0); 10675f39d397SChris Mason free_extent_buffer(buf); 1068de428b63SChris Mason return ret; 1069090d1875SChris Mason } 1070090d1875SChris Mason 1071ab0fff03SArne Jansen int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1072ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1073ab0fff03SArne Jansen { 1074ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 1075ab0fff03SArne Jansen struct inode *btree_inode = root->fs_info->btree_inode; 1076ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1077ab0fff03SArne Jansen int ret; 1078ab0fff03SArne Jansen 1079ab0fff03SArne Jansen buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 1080ab0fff03SArne Jansen if (!buf) 1081ab0fff03SArne Jansen return 0; 1082ab0fff03SArne Jansen 1083ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1084ab0fff03SArne Jansen 1085ab0fff03SArne Jansen ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, 1086ab0fff03SArne Jansen btree_get_extent, mirror_num); 1087ab0fff03SArne Jansen if (ret) { 1088ab0fff03SArne Jansen free_extent_buffer(buf); 1089ab0fff03SArne Jansen return ret; 1090ab0fff03SArne Jansen } 1091ab0fff03SArne Jansen 1092ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1093ab0fff03SArne Jansen free_extent_buffer(buf); 1094ab0fff03SArne Jansen return -EIO; 10950b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1096ab0fff03SArne Jansen *eb = buf; 1097ab0fff03SArne Jansen } else { 1098ab0fff03SArne Jansen free_extent_buffer(buf); 1099ab0fff03SArne Jansen } 1100ab0fff03SArne Jansen return 0; 1101ab0fff03SArne Jansen } 1102ab0fff03SArne Jansen 11030999df54SChris Mason struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, 11040999df54SChris Mason u64 bytenr, u32 blocksize) 11050999df54SChris Mason { 1106f28491e0SJosef Bacik return find_extent_buffer(root->fs_info, bytenr); 11070999df54SChris Mason } 11080999df54SChris Mason 11090999df54SChris Mason struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, 11100999df54SChris Mason u64 bytenr, u32 blocksize) 11110999df54SChris Mason { 1112f28491e0SJosef Bacik return alloc_extent_buffer(root->fs_info, bytenr, blocksize); 11130999df54SChris Mason } 11140999df54SChris Mason 11150999df54SChris Mason 1116e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1117e02119d5SChris Mason { 1118727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 11198aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1120e02119d5SChris Mason } 1121e02119d5SChris Mason 1122e02119d5SChris Mason int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1123e02119d5SChris Mason { 1124727011e0SChris Mason return filemap_fdatawait_range(buf->pages[0]->mapping, 11258aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1126e02119d5SChris Mason } 1127e02119d5SChris Mason 1128db94535dSChris Mason struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, 1129ca7a79adSChris Mason u32 blocksize, u64 parent_transid) 1130e20d96d6SChris Mason { 11315f39d397SChris Mason struct extent_buffer *buf = NULL; 113219c00ddcSChris Mason int ret; 113319c00ddcSChris Mason 1134db94535dSChris Mason buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 11355f39d397SChris Mason if (!buf) 1136d98237b3SChris Mason return NULL; 1137e4204dedSChris Mason 1138ca7a79adSChris Mason ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 11390f0fe8f7SFilipe David Borba Manana if (ret) { 11400f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 11410f0fe8f7SFilipe David Borba Manana return NULL; 11420f0fe8f7SFilipe David Borba Manana } 11435f39d397SChris Mason return buf; 1144ce9adaa5SChris Mason 1145eb60ceacSChris Mason } 1146eb60ceacSChris Mason 1147d5c13f92SJeff Mahoney void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, 11485f39d397SChris Mason struct extent_buffer *buf) 1149ed2ff2cbSChris Mason { 1150e2d84521SMiao Xie struct btrfs_fs_info *fs_info = root->fs_info; 1151e2d84521SMiao Xie 115255c69072SChris Mason if (btrfs_header_generation(buf) == 1153e2d84521SMiao Xie fs_info->running_transaction->transid) { 1154b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1155b4ce94deSChris Mason 1156b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1157e2d84521SMiao Xie __percpu_counter_add(&fs_info->dirty_metadata_bytes, 1158e2d84521SMiao Xie -buf->len, 1159e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1160b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1161b9473439SChris Mason btrfs_set_lock_blocking(buf); 11620b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1163925baeddSChris Mason } 11645f39d397SChris Mason } 1165ed7b63ebSJosef Bacik } 11665f39d397SChris Mason 11678257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 11688257b2dcSMiao Xie { 11698257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 11708257b2dcSMiao Xie int ret; 11718257b2dcSMiao Xie 11728257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 11738257b2dcSMiao Xie if (!writers) 11748257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 11758257b2dcSMiao Xie 11768257b2dcSMiao Xie ret = percpu_counter_init(&writers->counter, 0); 11778257b2dcSMiao Xie if (ret < 0) { 11788257b2dcSMiao Xie kfree(writers); 11798257b2dcSMiao Xie return ERR_PTR(ret); 11808257b2dcSMiao Xie } 11818257b2dcSMiao Xie 11828257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 11838257b2dcSMiao Xie return writers; 11848257b2dcSMiao Xie } 11858257b2dcSMiao Xie 11868257b2dcSMiao Xie static void 11878257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 11888257b2dcSMiao Xie { 11898257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 11908257b2dcSMiao Xie kfree(writers); 11918257b2dcSMiao Xie } 11928257b2dcSMiao Xie 1193143bede5SJeff Mahoney static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, 119487ee04ebSChris Mason u32 stripesize, struct btrfs_root *root, 11959f5fae2fSChris Mason struct btrfs_fs_info *fs_info, 1196e20d96d6SChris Mason u64 objectid) 1197d97e63b6SChris Mason { 1198cfaa7295SChris Mason root->node = NULL; 1199a28ec197SChris Mason root->commit_root = NULL; 1200db94535dSChris Mason root->sectorsize = sectorsize; 1201db94535dSChris Mason root->nodesize = nodesize; 1202db94535dSChris Mason root->leafsize = leafsize; 120387ee04ebSChris Mason root->stripesize = stripesize; 120427cdeb70SMiao Xie root->state = 0; 1205d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 12060b86a832SChris Mason 12070f7d52f4SChris Mason root->objectid = objectid; 12080f7d52f4SChris Mason root->last_trans = 0; 120913a8a7c8SYan, Zheng root->highest_objectid = 0; 1210eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1211199c2a9cSMiao Xie root->nr_ordered_extents = 0; 121258176a96SJosef Bacik root->name = NULL; 12136bef4d31SEric Paris root->inode_tree = RB_ROOT; 121416cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1215f0486c68SYan, Zheng root->block_rsv = NULL; 1216d68fc57bSYan, Zheng root->orphan_block_rsv = NULL; 12170b86a832SChris Mason 12180b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 12195d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1220eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1221eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1222199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1223199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 12242ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 12252ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 1226d68fc57bSYan, Zheng spin_lock_init(&root->orphan_lock); 12275d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1228eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1229199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1230f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 12312ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 12322ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 1233a2135011SChris Mason mutex_init(&root->objectid_mutex); 1234e02119d5SChris Mason mutex_init(&root->log_mutex); 123531f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1236573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 12377237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 12387237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 12397237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 12408b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 12418b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 12427237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 12437237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 12447237f183SYan Zheng atomic_set(&root->log_writers, 0); 12452ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 12468a35d95fSJosef Bacik atomic_set(&root->orphan_inodes, 0); 1247b0feb9d9SMiao Xie atomic_set(&root->refs, 1); 12488257b2dcSMiao Xie atomic_set(&root->will_be_snapshoted, 0); 12497237f183SYan Zheng root->log_transid = 0; 1250d1433debSMiao Xie root->log_transid_committed = -1; 1251257c62e1SChris Mason root->last_log_commit = 0; 125206ea65a3SJosef Bacik if (fs_info) 1253d0c803c4SChris Mason extent_io_tree_init(&root->dirty_log_pages, 1254f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 1255017e5369SChris Mason 12563768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 12573768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 12586702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 125958176a96SJosef Bacik memset(&root->root_kobj, 0, sizeof(root->root_kobj)); 126006ea65a3SJosef Bacik if (fs_info) 12613f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 126206ea65a3SJosef Bacik else 126306ea65a3SJosef Bacik root->defrag_trans_start = 0; 126458176a96SJosef Bacik init_completion(&root->kobj_unregister); 12654d775673SChris Mason root->root_key.objectid = objectid; 12660ee5dc67SAl Viro root->anon_dev = 0; 12678ea05e3aSAlexander Block 12685f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 12693768f368SChris Mason } 12703768f368SChris Mason 1271f84a8bd6SAl Viro static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) 12726f07e42eSAl Viro { 12736f07e42eSAl Viro struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); 12746f07e42eSAl Viro if (root) 12756f07e42eSAl Viro root->fs_info = fs_info; 12766f07e42eSAl Viro return root; 12776f07e42eSAl Viro } 12786f07e42eSAl Viro 127906ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 128006ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 128106ea65a3SJosef Bacik struct btrfs_root *btrfs_alloc_dummy_root(void) 128206ea65a3SJosef Bacik { 128306ea65a3SJosef Bacik struct btrfs_root *root; 128406ea65a3SJosef Bacik 128506ea65a3SJosef Bacik root = btrfs_alloc_root(NULL); 128606ea65a3SJosef Bacik if (!root) 128706ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 128806ea65a3SJosef Bacik __setup_root(4096, 4096, 4096, 4096, root, NULL, 1); 128927cdeb70SMiao Xie set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state); 129006ea65a3SJosef Bacik 129106ea65a3SJosef Bacik return root; 129206ea65a3SJosef Bacik } 129306ea65a3SJosef Bacik #endif 129406ea65a3SJosef Bacik 129520897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 129620897f5cSArne Jansen struct btrfs_fs_info *fs_info, 129720897f5cSArne Jansen u64 objectid) 129820897f5cSArne Jansen { 129920897f5cSArne Jansen struct extent_buffer *leaf; 130020897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 130120897f5cSArne Jansen struct btrfs_root *root; 130220897f5cSArne Jansen struct btrfs_key key; 130320897f5cSArne Jansen int ret = 0; 13046463fe58SStefan Behrens uuid_le uuid; 130520897f5cSArne Jansen 130620897f5cSArne Jansen root = btrfs_alloc_root(fs_info); 130720897f5cSArne Jansen if (!root) 130820897f5cSArne Jansen return ERR_PTR(-ENOMEM); 130920897f5cSArne Jansen 131020897f5cSArne Jansen __setup_root(tree_root->nodesize, tree_root->leafsize, 131120897f5cSArne Jansen tree_root->sectorsize, tree_root->stripesize, 131220897f5cSArne Jansen root, fs_info, objectid); 131320897f5cSArne Jansen root->root_key.objectid = objectid; 131420897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 131520897f5cSArne Jansen root->root_key.offset = 0; 131620897f5cSArne Jansen 131720897f5cSArne Jansen leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 131820897f5cSArne Jansen 0, objectid, NULL, 0, 0, 0); 131920897f5cSArne Jansen if (IS_ERR(leaf)) { 132020897f5cSArne Jansen ret = PTR_ERR(leaf); 13211dd05682STsutomu Itoh leaf = NULL; 132220897f5cSArne Jansen goto fail; 132320897f5cSArne Jansen } 132420897f5cSArne Jansen 132520897f5cSArne Jansen memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 132620897f5cSArne Jansen btrfs_set_header_bytenr(leaf, leaf->start); 132720897f5cSArne Jansen btrfs_set_header_generation(leaf, trans->transid); 132820897f5cSArne Jansen btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 132920897f5cSArne Jansen btrfs_set_header_owner(leaf, objectid); 133020897f5cSArne Jansen root->node = leaf; 133120897f5cSArne Jansen 13320a4e5586SRoss Kirk write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), 133320897f5cSArne Jansen BTRFS_FSID_SIZE); 133420897f5cSArne Jansen write_extent_buffer(leaf, fs_info->chunk_tree_uuid, 1335b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(leaf), 133620897f5cSArne Jansen BTRFS_UUID_SIZE); 133720897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 133820897f5cSArne Jansen 133920897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 134027cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); 134120897f5cSArne Jansen 134220897f5cSArne Jansen root->root_item.flags = 0; 134320897f5cSArne Jansen root->root_item.byte_limit = 0; 134420897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 134520897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 134620897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 134720897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 134820897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 134920897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 135020897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 13516463fe58SStefan Behrens uuid_le_gen(&uuid); 13526463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 135320897f5cSArne Jansen root->root_item.drop_level = 0; 135420897f5cSArne Jansen 135520897f5cSArne Jansen key.objectid = objectid; 135620897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 135720897f5cSArne Jansen key.offset = 0; 135820897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 135920897f5cSArne Jansen if (ret) 136020897f5cSArne Jansen goto fail; 136120897f5cSArne Jansen 136220897f5cSArne Jansen btrfs_tree_unlock(leaf); 136320897f5cSArne Jansen 136420897f5cSArne Jansen return root; 13651dd05682STsutomu Itoh 13661dd05682STsutomu Itoh fail: 13671dd05682STsutomu Itoh if (leaf) { 13681dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 13691dd05682STsutomu Itoh free_extent_buffer(leaf); 13701dd05682STsutomu Itoh } 13711dd05682STsutomu Itoh kfree(root); 13721dd05682STsutomu Itoh 13731dd05682STsutomu Itoh return ERR_PTR(ret); 137420897f5cSArne Jansen } 137520897f5cSArne Jansen 13767237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1377e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13780f7d52f4SChris Mason { 13790f7d52f4SChris Mason struct btrfs_root *root; 13800f7d52f4SChris Mason struct btrfs_root *tree_root = fs_info->tree_root; 13817237f183SYan Zheng struct extent_buffer *leaf; 1382e02119d5SChris Mason 13836f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1384e02119d5SChris Mason if (!root) 13857237f183SYan Zheng return ERR_PTR(-ENOMEM); 1386e02119d5SChris Mason 1387e02119d5SChris Mason __setup_root(tree_root->nodesize, tree_root->leafsize, 1388e02119d5SChris Mason tree_root->sectorsize, tree_root->stripesize, 1389e02119d5SChris Mason root, fs_info, BTRFS_TREE_LOG_OBJECTID); 1390e02119d5SChris Mason 1391e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1392e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1393e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 139427cdeb70SMiao Xie 13957237f183SYan Zheng /* 139627cdeb70SMiao Xie * DON'T set REF_COWS for log trees 139727cdeb70SMiao Xie * 13987237f183SYan Zheng * log trees do not get reference counted because they go away 13997237f183SYan Zheng * before a real commit is actually done. They do store pointers 14007237f183SYan Zheng * to file data extents, and those reference counts still get 14017237f183SYan Zheng * updated (along with back refs to the log tree). 14027237f183SYan Zheng */ 1403e02119d5SChris Mason 14045d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 140566d7e7f0SArne Jansen BTRFS_TREE_LOG_OBJECTID, NULL, 14065581a51aSJan Schmidt 0, 0, 0); 14077237f183SYan Zheng if (IS_ERR(leaf)) { 14087237f183SYan Zheng kfree(root); 14097237f183SYan Zheng return ERR_CAST(leaf); 14107237f183SYan Zheng } 1411e02119d5SChris Mason 14125d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 14135d4f98a2SYan Zheng btrfs_set_header_bytenr(leaf, leaf->start); 14145d4f98a2SYan Zheng btrfs_set_header_generation(leaf, trans->transid); 14155d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 14165d4f98a2SYan Zheng btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 14177237f183SYan Zheng root->node = leaf; 1418e02119d5SChris Mason 1419e02119d5SChris Mason write_extent_buffer(root->node, root->fs_info->fsid, 14200a4e5586SRoss Kirk btrfs_header_fsid(), BTRFS_FSID_SIZE); 1421e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1422e02119d5SChris Mason btrfs_tree_unlock(root->node); 14237237f183SYan Zheng return root; 14247237f183SYan Zheng } 14257237f183SYan Zheng 14267237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 14277237f183SYan Zheng struct btrfs_fs_info *fs_info) 14287237f183SYan Zheng { 14297237f183SYan Zheng struct btrfs_root *log_root; 14307237f183SYan Zheng 14317237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 14327237f183SYan Zheng if (IS_ERR(log_root)) 14337237f183SYan Zheng return PTR_ERR(log_root); 14347237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 14357237f183SYan Zheng fs_info->log_root_tree = log_root; 14367237f183SYan Zheng return 0; 14377237f183SYan Zheng } 14387237f183SYan Zheng 14397237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 14407237f183SYan Zheng struct btrfs_root *root) 14417237f183SYan Zheng { 14427237f183SYan Zheng struct btrfs_root *log_root; 14437237f183SYan Zheng struct btrfs_inode_item *inode_item; 14447237f183SYan Zheng 14457237f183SYan Zheng log_root = alloc_log_tree(trans, root->fs_info); 14467237f183SYan Zheng if (IS_ERR(log_root)) 14477237f183SYan Zheng return PTR_ERR(log_root); 14487237f183SYan Zheng 14497237f183SYan Zheng log_root->last_trans = trans->transid; 14507237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14517237f183SYan Zheng 14527237f183SYan Zheng inode_item = &log_root->root_item.inode; 14533cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14543cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14553cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 14563cae210fSQu Wenruo btrfs_set_stack_inode_nbytes(inode_item, root->leafsize); 14573cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14587237f183SYan Zheng 14595d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14607237f183SYan Zheng 14617237f183SYan Zheng WARN_ON(root->log_root); 14627237f183SYan Zheng root->log_root = log_root; 14637237f183SYan Zheng root->log_transid = 0; 1464d1433debSMiao Xie root->log_transid_committed = -1; 1465257c62e1SChris Mason root->last_log_commit = 0; 1466e02119d5SChris Mason return 0; 1467e02119d5SChris Mason } 1468e02119d5SChris Mason 146935a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1470cb517eabSMiao Xie struct btrfs_key *key) 1471e02119d5SChris Mason { 1472e02119d5SChris Mason struct btrfs_root *root; 1473e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 14740f7d52f4SChris Mason struct btrfs_path *path; 147584234f3aSYan Zheng u64 generation; 1476db94535dSChris Mason u32 blocksize; 1477cb517eabSMiao Xie int ret; 1478cb517eabSMiao Xie 1479cb517eabSMiao Xie path = btrfs_alloc_path(); 1480cb517eabSMiao Xie if (!path) 1481cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 14820f7d52f4SChris Mason 14836f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1484cb517eabSMiao Xie if (!root) { 1485cb517eabSMiao Xie ret = -ENOMEM; 1486cb517eabSMiao Xie goto alloc_fail; 14870f7d52f4SChris Mason } 14880f7d52f4SChris Mason 1489db94535dSChris Mason __setup_root(tree_root->nodesize, tree_root->leafsize, 149087ee04ebSChris Mason tree_root->sectorsize, tree_root->stripesize, 1491cb517eabSMiao Xie root, fs_info, key->objectid); 14920f7d52f4SChris Mason 1493cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1494cb517eabSMiao Xie &root->root_item, &root->root_key); 14950f7d52f4SChris Mason if (ret) { 149613a8a7c8SYan, Zheng if (ret > 0) 149713a8a7c8SYan, Zheng ret = -ENOENT; 1498cb517eabSMiao Xie goto find_fail; 14990f7d52f4SChris Mason } 150013a8a7c8SYan, Zheng 150184234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1502db94535dSChris Mason blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); 1503db94535dSChris Mason root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 150484234f3aSYan Zheng blocksize, generation); 1505cb517eabSMiao Xie if (!root->node) { 1506cb517eabSMiao Xie ret = -ENOMEM; 1507cb517eabSMiao Xie goto find_fail; 1508cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1509cb517eabSMiao Xie ret = -EIO; 1510cb517eabSMiao Xie goto read_fail; 1511416bc658SJosef Bacik } 15125d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 151313a8a7c8SYan, Zheng out: 1514cb517eabSMiao Xie btrfs_free_path(path); 1515cb517eabSMiao Xie return root; 1516cb517eabSMiao Xie 1517cb517eabSMiao Xie read_fail: 1518cb517eabSMiao Xie free_extent_buffer(root->node); 1519cb517eabSMiao Xie find_fail: 1520cb517eabSMiao Xie kfree(root); 1521cb517eabSMiao Xie alloc_fail: 1522cb517eabSMiao Xie root = ERR_PTR(ret); 1523cb517eabSMiao Xie goto out; 1524cb517eabSMiao Xie } 1525cb517eabSMiao Xie 1526cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1527cb517eabSMiao Xie struct btrfs_key *location) 1528cb517eabSMiao Xie { 1529cb517eabSMiao Xie struct btrfs_root *root; 1530cb517eabSMiao Xie 1531cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1532cb517eabSMiao Xie if (IS_ERR(root)) 1533cb517eabSMiao Xie return root; 1534cb517eabSMiao Xie 1535cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 153627cdeb70SMiao Xie set_bit(BTRFS_ROOT_REF_COWS, &root->state); 153708fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 153808fe4db1SLi Zefan } 153913a8a7c8SYan, Zheng 15405eda7b5eSChris Mason return root; 15415eda7b5eSChris Mason } 15425eda7b5eSChris Mason 1543cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1544cb517eabSMiao Xie { 1545cb517eabSMiao Xie int ret; 15468257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1547cb517eabSMiao Xie 1548cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1549cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1550cb517eabSMiao Xie GFP_NOFS); 1551cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1552cb517eabSMiao Xie ret = -ENOMEM; 1553cb517eabSMiao Xie goto fail; 1554cb517eabSMiao Xie } 1555cb517eabSMiao Xie 15568257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 15578257b2dcSMiao Xie if (IS_ERR(writers)) { 15588257b2dcSMiao Xie ret = PTR_ERR(writers); 15598257b2dcSMiao Xie goto fail; 15608257b2dcSMiao Xie } 15618257b2dcSMiao Xie root->subv_writers = writers; 15628257b2dcSMiao Xie 1563cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 1564cb517eabSMiao Xie spin_lock_init(&root->cache_lock); 1565cb517eabSMiao Xie init_waitqueue_head(&root->cache_wait); 1566cb517eabSMiao Xie 1567cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1568cb517eabSMiao Xie if (ret) 15698257b2dcSMiao Xie goto free_writers; 1570cb517eabSMiao Xie return 0; 15718257b2dcSMiao Xie 15728257b2dcSMiao Xie free_writers: 15738257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 1574cb517eabSMiao Xie fail: 1575cb517eabSMiao Xie kfree(root->free_ino_ctl); 1576cb517eabSMiao Xie kfree(root->free_ino_pinned); 1577cb517eabSMiao Xie return ret; 1578cb517eabSMiao Xie } 1579cb517eabSMiao Xie 1580171170c1SSergei Trofimovich static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1581cb517eabSMiao Xie u64 root_id) 1582cb517eabSMiao Xie { 1583cb517eabSMiao Xie struct btrfs_root *root; 1584cb517eabSMiao Xie 1585cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1586cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1587cb517eabSMiao Xie (unsigned long)root_id); 1588cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1589cb517eabSMiao Xie return root; 1590cb517eabSMiao Xie } 1591cb517eabSMiao Xie 1592cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1593cb517eabSMiao Xie struct btrfs_root *root) 1594cb517eabSMiao Xie { 1595cb517eabSMiao Xie int ret; 1596cb517eabSMiao Xie 1597cb517eabSMiao Xie ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); 1598cb517eabSMiao Xie if (ret) 1599cb517eabSMiao Xie return ret; 1600cb517eabSMiao Xie 1601cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1602cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1603cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1604cb517eabSMiao Xie root); 1605cb517eabSMiao Xie if (ret == 0) 160627cdeb70SMiao Xie set_bit(BTRFS_ROOT_IN_RADIX, &root->state); 1607cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1608cb517eabSMiao Xie radix_tree_preload_end(); 1609cb517eabSMiao Xie 1610cb517eabSMiao Xie return ret; 1611cb517eabSMiao Xie } 1612cb517eabSMiao Xie 1613c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1614c00869f1SMiao Xie struct btrfs_key *location, 1615c00869f1SMiao Xie bool check_ref) 16165eda7b5eSChris Mason { 16175eda7b5eSChris Mason struct btrfs_root *root; 16185eda7b5eSChris Mason int ret; 16195eda7b5eSChris Mason 1620edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1621edbd8d4eSChris Mason return fs_info->tree_root; 1622edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1623edbd8d4eSChris Mason return fs_info->extent_root; 16248f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 16258f18cf13SChris Mason return fs_info->chunk_root; 16268f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 16278f18cf13SChris Mason return fs_info->dev_root; 16280403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 16290403e47eSYan Zheng return fs_info->csum_root; 1630bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1631bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1632bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1633f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1634f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1635f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 16364df27c4dSYan, Zheng again: 1637cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 163848475471SStefan Behrens if (root) { 1639c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 164048475471SStefan Behrens return ERR_PTR(-ENOENT); 16415eda7b5eSChris Mason return root; 164248475471SStefan Behrens } 16435eda7b5eSChris Mason 1644cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 16455eda7b5eSChris Mason if (IS_ERR(root)) 16465eda7b5eSChris Mason return root; 16473394e160SChris Mason 1648c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1649d68fc57bSYan, Zheng ret = -ENOENT; 1650d68fc57bSYan, Zheng goto fail; 1651d68fc57bSYan, Zheng } 1652d68fc57bSYan, Zheng 1653cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1654cb517eabSMiao Xie if (ret) 1655cb517eabSMiao Xie goto fail; 1656cb517eabSMiao Xie 16573f870c28SKelley Nielsen ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID, 16583f870c28SKelley Nielsen location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL); 1659d68fc57bSYan, Zheng if (ret < 0) 1660d68fc57bSYan, Zheng goto fail; 1661d68fc57bSYan, Zheng if (ret == 0) 166227cdeb70SMiao Xie set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 1663d68fc57bSYan, Zheng 1664cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 16650f7d52f4SChris Mason if (ret) { 16664df27c4dSYan, Zheng if (ret == -EEXIST) { 16674df27c4dSYan, Zheng free_fs_root(root); 16684df27c4dSYan, Zheng goto again; 16690f7d52f4SChris Mason } 16704df27c4dSYan, Zheng goto fail; 16714df27c4dSYan, Zheng } 1672edbd8d4eSChris Mason return root; 16734df27c4dSYan, Zheng fail: 16744df27c4dSYan, Zheng free_fs_root(root); 16754df27c4dSYan, Zheng return ERR_PTR(ret); 1676edbd8d4eSChris Mason } 1677edbd8d4eSChris Mason 167804160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 167904160088SChris Mason { 168004160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 168104160088SChris Mason int ret = 0; 168204160088SChris Mason struct btrfs_device *device; 168304160088SChris Mason struct backing_dev_info *bdi; 1684b7967db7SChris Mason 16851f78160cSXiao Guangrong rcu_read_lock(); 16861f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1687dfe25020SChris Mason if (!device->bdev) 1688dfe25020SChris Mason continue; 168904160088SChris Mason bdi = blk_get_backing_dev_info(device->bdev); 169004160088SChris Mason if (bdi && bdi_congested(bdi, bdi_bits)) { 169104160088SChris Mason ret = 1; 169204160088SChris Mason break; 169304160088SChris Mason } 169404160088SChris Mason } 16951f78160cSXiao Guangrong rcu_read_unlock(); 169604160088SChris Mason return ret; 169704160088SChris Mason } 169804160088SChris Mason 169938b66988SChris Mason /* 1700ad081f14SJens Axboe * If this fails, caller must call bdi_destroy() to get rid of the 1701ad081f14SJens Axboe * bdi again. 1702ad081f14SJens Axboe */ 170304160088SChris Mason static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) 170404160088SChris Mason { 1705ad081f14SJens Axboe int err; 1706ad081f14SJens Axboe 1707ad081f14SJens Axboe bdi->capabilities = BDI_CAP_MAP_COPY; 1708e6d086d8SJens Axboe err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); 1709ad081f14SJens Axboe if (err) 1710ad081f14SJens Axboe return err; 1711ad081f14SJens Axboe 17124575c9ccSChris Mason bdi->ra_pages = default_backing_dev_info.ra_pages; 171304160088SChris Mason bdi->congested_fn = btrfs_congested_fn; 171404160088SChris Mason bdi->congested_data = info; 171504160088SChris Mason return 0; 171604160088SChris Mason } 171704160088SChris Mason 17188b712842SChris Mason /* 17198b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 17208b712842SChris Mason * functions. This is where read checksum verification actually happens 17218b712842SChris Mason */ 17228b712842SChris Mason static void end_workqueue_fn(struct btrfs_work *work) 1723ce9adaa5SChris Mason { 1724ce9adaa5SChris Mason struct bio *bio; 17258b712842SChris Mason struct end_io_wq *end_io_wq; 1726ce9adaa5SChris Mason int error; 1727ce9adaa5SChris Mason 17288b712842SChris Mason end_io_wq = container_of(work, struct end_io_wq, work); 1729ce9adaa5SChris Mason bio = end_io_wq->bio; 17308b712842SChris Mason 1731ce9adaa5SChris Mason error = end_io_wq->error; 1732ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1733ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 1734ce9adaa5SChris Mason kfree(end_io_wq); 1735bc1e79acSKent Overstreet bio_endio_nodec(bio, error); 1736ce9adaa5SChris Mason } 173744b8bd7eSChris Mason 1738a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1739a74a4b97SChris Mason { 1740a74a4b97SChris Mason struct btrfs_root *root = arg; 1741d0278245SMiao Xie int again; 1742a74a4b97SChris Mason 1743a74a4b97SChris Mason do { 1744d0278245SMiao Xie again = 0; 17459d1a2a3aSDavid Sterba 1746d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 1747babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) 1748d0278245SMiao Xie goto sleep; 1749d0278245SMiao Xie 1750d0278245SMiao Xie if (!mutex_trylock(&root->fs_info->cleaner_mutex)) 1751d0278245SMiao Xie goto sleep; 1752d0278245SMiao Xie 1753dc7f370cSMiao Xie /* 1754dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1755dc7f370cSMiao Xie * during the above check and trylock. 1756dc7f370cSMiao Xie */ 1757babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) { 1758dc7f370cSMiao Xie mutex_unlock(&root->fs_info->cleaner_mutex); 1759dc7f370cSMiao Xie goto sleep; 1760dc7f370cSMiao Xie } 1761dc7f370cSMiao Xie 176224bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 17639d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 1764a74a4b97SChris Mason mutex_unlock(&root->fs_info->cleaner_mutex); 1765a74a4b97SChris Mason 1766d0278245SMiao Xie /* 176705323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 176805323cd1SMiao Xie * needn't do anything special here. 1769d0278245SMiao Xie */ 1770d0278245SMiao Xie btrfs_run_defrag_inodes(root->fs_info); 1771d0278245SMiao Xie sleep: 17729d1a2a3aSDavid Sterba if (!try_to_freeze() && !again) { 1773a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 17748929ecfaSYan, Zheng if (!kthread_should_stop()) 1775a74a4b97SChris Mason schedule(); 1776a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1777a74a4b97SChris Mason } 1778a74a4b97SChris Mason } while (!kthread_should_stop()); 1779a74a4b97SChris Mason return 0; 1780a74a4b97SChris Mason } 1781a74a4b97SChris Mason 1782a74a4b97SChris Mason static int transaction_kthread(void *arg) 1783a74a4b97SChris Mason { 1784a74a4b97SChris Mason struct btrfs_root *root = arg; 1785a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1786a74a4b97SChris Mason struct btrfs_transaction *cur; 17878929ecfaSYan, Zheng u64 transid; 1788a74a4b97SChris Mason unsigned long now; 1789a74a4b97SChris Mason unsigned long delay; 1790914b2007SJan Kara bool cannot_commit; 1791a74a4b97SChris Mason 1792a74a4b97SChris Mason do { 1793914b2007SJan Kara cannot_commit = false; 17948b87dc17SDavid Sterba delay = HZ * root->fs_info->commit_interval; 1795a74a4b97SChris Mason mutex_lock(&root->fs_info->transaction_kthread_mutex); 1796a74a4b97SChris Mason 1797a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 1798a74a4b97SChris Mason cur = root->fs_info->running_transaction; 1799a74a4b97SChris Mason if (!cur) { 1800a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1801a74a4b97SChris Mason goto sleep; 1802a74a4b97SChris Mason } 180331153d81SYan Zheng 1804a74a4b97SChris Mason now = get_seconds(); 18054a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 18068b87dc17SDavid Sterba (now < cur->start_time || 18078b87dc17SDavid Sterba now - cur->start_time < root->fs_info->commit_interval)) { 1808a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1809a74a4b97SChris Mason delay = HZ * 5; 1810a74a4b97SChris Mason goto sleep; 1811a74a4b97SChris Mason } 18128929ecfaSYan, Zheng transid = cur->transid; 1813a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 181456bec294SChris Mason 181579787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1816354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1817914b2007SJan Kara if (IS_ERR(trans)) { 1818354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1819914b2007SJan Kara cannot_commit = true; 182079787eaaSJeff Mahoney goto sleep; 1821914b2007SJan Kara } 18228929ecfaSYan, Zheng if (transid == trans->transid) { 182379787eaaSJeff Mahoney btrfs_commit_transaction(trans, root); 18248929ecfaSYan, Zheng } else { 18258929ecfaSYan, Zheng btrfs_end_transaction(trans, root); 18268929ecfaSYan, Zheng } 1827a74a4b97SChris Mason sleep: 1828a74a4b97SChris Mason wake_up_process(root->fs_info->cleaner_kthread); 1829a74a4b97SChris Mason mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1830a74a4b97SChris Mason 18314e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 18324e121c06SJosef Bacik &root->fs_info->fs_state))) 18334e121c06SJosef Bacik btrfs_cleanup_transaction(root); 1834a0acae0eSTejun Heo if (!try_to_freeze()) { 1835a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 18368929ecfaSYan, Zheng if (!kthread_should_stop() && 1837914b2007SJan Kara (!btrfs_transaction_blocked(root->fs_info) || 1838914b2007SJan Kara cannot_commit)) 1839a74a4b97SChris Mason schedule_timeout(delay); 1840a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1841a74a4b97SChris Mason } 1842a74a4b97SChris Mason } while (!kthread_should_stop()); 1843a74a4b97SChris Mason return 0; 1844a74a4b97SChris Mason } 1845a74a4b97SChris Mason 1846af31f5e5SChris Mason /* 1847af31f5e5SChris Mason * this will find the highest generation in the array of 1848af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1849af31f5e5SChris Mason * or -1 if we can't find anything. 1850af31f5e5SChris Mason * 1851af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1852af31f5e5SChris Mason * generation of the latest root in the array with the generation 1853af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1854af31f5e5SChris Mason */ 1855af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1856af31f5e5SChris Mason { 1857af31f5e5SChris Mason u64 cur; 1858af31f5e5SChris Mason int newest_index = -1; 1859af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1860af31f5e5SChris Mason int i; 1861af31f5e5SChris Mason 1862af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1863af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1864af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1865af31f5e5SChris Mason if (cur == newest_gen) 1866af31f5e5SChris Mason newest_index = i; 1867af31f5e5SChris Mason } 1868af31f5e5SChris Mason 1869af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1870af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1871af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1872af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1873af31f5e5SChris Mason if (cur == newest_gen) 1874af31f5e5SChris Mason newest_index = 0; 1875af31f5e5SChris Mason } 1876af31f5e5SChris Mason return newest_index; 1877af31f5e5SChris Mason } 1878af31f5e5SChris Mason 1879af31f5e5SChris Mason 1880af31f5e5SChris Mason /* 1881af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1882af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1883af31f5e5SChris Mason * field in the fs_info struct 1884af31f5e5SChris Mason */ 1885af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1886af31f5e5SChris Mason u64 newest_gen) 1887af31f5e5SChris Mason { 1888af31f5e5SChris Mason int newest_index = -1; 1889af31f5e5SChris Mason 1890af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1891af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1892af31f5e5SChris Mason if (newest_index == -1) { 1893af31f5e5SChris Mason info->backup_root_index = 0; 1894af31f5e5SChris Mason } else { 1895af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1896af31f5e5SChris Mason } 1897af31f5e5SChris Mason } 1898af31f5e5SChris Mason 1899af31f5e5SChris Mason /* 1900af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1901af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1902af31f5e5SChris Mason * done 1903af31f5e5SChris Mason */ 1904af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1905af31f5e5SChris Mason { 1906af31f5e5SChris Mason int next_backup; 1907af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1908af31f5e5SChris Mason int last_backup; 1909af31f5e5SChris Mason 1910af31f5e5SChris Mason next_backup = info->backup_root_index; 1911af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1912af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1913af31f5e5SChris Mason 1914af31f5e5SChris Mason /* 1915af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 1916af31f5e5SChris Mason * this happens only at umount 1917af31f5e5SChris Mason */ 1918af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 1919af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 1920af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 1921af31f5e5SChris Mason next_backup = last_backup; 1922af31f5e5SChris Mason 1923af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1924af31f5e5SChris Mason 1925af31f5e5SChris Mason /* 1926af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1927af31f5e5SChris Mason * regardless of which ones we use today 1928af31f5e5SChris Mason */ 1929af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1930af31f5e5SChris Mason 1931af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1932af31f5e5SChris Mason 1933af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1934af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1935af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1936af31f5e5SChris Mason 1937af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1938af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1939af31f5e5SChris Mason 1940af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1941af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1942af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1943af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1944af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1945af31f5e5SChris Mason 1946af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1947af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1948af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1949af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1950af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1951af31f5e5SChris Mason 19527c7e82a7SChris Mason /* 19537c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 19547c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 19557c7e82a7SChris Mason */ 19567c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 19577c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 19587c7e82a7SChris Mason info->fs_root->node->start); 1959af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1960af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1961af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1962af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19637c7e82a7SChris Mason } 1964af31f5e5SChris Mason 1965af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1966af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1967af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1968af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1969af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1970af31f5e5SChris Mason 1971af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1972af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1973af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1974af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1975af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1976af31f5e5SChris Mason 1977af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1978af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1979af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1980af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1981af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1982af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1983af31f5e5SChris Mason 1984af31f5e5SChris Mason /* 1985af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1986af31f5e5SChris Mason * for the next commit 1987af31f5e5SChris Mason */ 1988af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1989af31f5e5SChris Mason &info->super_for_commit->super_roots, 1990af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1991af31f5e5SChris Mason } 1992af31f5e5SChris Mason 1993af31f5e5SChris Mason /* 1994af31f5e5SChris Mason * this copies info out of the root backup array and back into 1995af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 1996af31f5e5SChris Mason * the array, so you send it the number of backups you've already 1997af31f5e5SChris Mason * tried and the last backup index you used. 1998af31f5e5SChris Mason * 1999af31f5e5SChris Mason * this returns -1 when it has tried all the backups 2000af31f5e5SChris Mason */ 2001af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 2002af31f5e5SChris Mason struct btrfs_super_block *super, 2003af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 2004af31f5e5SChris Mason { 2005af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2006af31f5e5SChris Mason int newest = *backup_index; 2007af31f5e5SChris Mason 2008af31f5e5SChris Mason if (*num_backups_tried == 0) { 2009af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 2010af31f5e5SChris Mason 2011af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 2012af31f5e5SChris Mason if (newest == -1) 2013af31f5e5SChris Mason return -1; 2014af31f5e5SChris Mason 2015af31f5e5SChris Mason *backup_index = newest; 2016af31f5e5SChris Mason *num_backups_tried = 1; 2017af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 2018af31f5e5SChris Mason /* we've tried all the backups, all done */ 2019af31f5e5SChris Mason return -1; 2020af31f5e5SChris Mason } else { 2021af31f5e5SChris Mason /* jump to the next oldest backup */ 2022af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 2023af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 2024af31f5e5SChris Mason *backup_index = newest; 2025af31f5e5SChris Mason *num_backups_tried += 1; 2026af31f5e5SChris Mason } 2027af31f5e5SChris Mason root_backup = super->super_roots + newest; 2028af31f5e5SChris Mason 2029af31f5e5SChris Mason btrfs_set_super_generation(super, 2030af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 2031af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2032af31f5e5SChris Mason btrfs_set_super_root_level(super, 2033af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 2034af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2035af31f5e5SChris Mason 2036af31f5e5SChris Mason /* 2037af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 2038af31f5e5SChris Mason * need a fsck 2039af31f5e5SChris Mason */ 2040af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2041af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2042af31f5e5SChris Mason return 0; 2043af31f5e5SChris Mason } 2044af31f5e5SChris Mason 20457abadb64SLiu Bo /* helper to cleanup workers */ 20467abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20477abadb64SLiu Bo { 2048dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2049afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 20505cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2051fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2052fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2053fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 2054d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2055fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 2056fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2057fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2058a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 20595b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2060e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2061736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2062a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2063fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 20647abadb64SLiu Bo } 20657abadb64SLiu Bo 20662e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20672e9f5954SRashika { 20682e9f5954SRashika if (root) { 20692e9f5954SRashika free_extent_buffer(root->node); 20702e9f5954SRashika free_extent_buffer(root->commit_root); 20712e9f5954SRashika root->node = NULL; 20722e9f5954SRashika root->commit_root = NULL; 20732e9f5954SRashika } 20742e9f5954SRashika } 20752e9f5954SRashika 2076af31f5e5SChris Mason /* helper to cleanup tree roots */ 2077af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2078af31f5e5SChris Mason { 20792e9f5954SRashika free_root_extent_buffers(info->tree_root); 2080655b09feSJosef Bacik 20812e9f5954SRashika free_root_extent_buffers(info->dev_root); 20822e9f5954SRashika free_root_extent_buffers(info->extent_root); 20832e9f5954SRashika free_root_extent_buffers(info->csum_root); 20842e9f5954SRashika free_root_extent_buffers(info->quota_root); 20852e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20862e9f5954SRashika if (chunk_root) 20872e9f5954SRashika free_root_extent_buffers(info->chunk_root); 2088af31f5e5SChris Mason } 2089af31f5e5SChris Mason 2090171f6537SJosef Bacik static void del_fs_roots(struct btrfs_fs_info *fs_info) 2091171f6537SJosef Bacik { 2092171f6537SJosef Bacik int ret; 2093171f6537SJosef Bacik struct btrfs_root *gang[8]; 2094171f6537SJosef Bacik int i; 2095171f6537SJosef Bacik 2096171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2097171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2098171f6537SJosef Bacik struct btrfs_root, root_list); 2099171f6537SJosef Bacik list_del(&gang[0]->root_list); 2100171f6537SJosef Bacik 210127cdeb70SMiao Xie if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) { 2102cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2103171f6537SJosef Bacik } else { 2104171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2105171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2106b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2107171f6537SJosef Bacik } 2108171f6537SJosef Bacik } 2109171f6537SJosef Bacik 2110171f6537SJosef Bacik while (1) { 2111171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2112171f6537SJosef Bacik (void **)gang, 0, 2113171f6537SJosef Bacik ARRAY_SIZE(gang)); 2114171f6537SJosef Bacik if (!ret) 2115171f6537SJosef Bacik break; 2116171f6537SJosef Bacik for (i = 0; i < ret; i++) 2117cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2118171f6537SJosef Bacik } 21191a4319ccSLiu Bo 21201a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 21211a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 21221a4319ccSLiu Bo btrfs_destroy_pinned_extent(fs_info->tree_root, 21231a4319ccSLiu Bo fs_info->pinned_extents); 21241a4319ccSLiu Bo } 2125171f6537SJosef Bacik } 2126af31f5e5SChris Mason 2127ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2128dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2129dfe25020SChris Mason char *options) 2130eb60ceacSChris Mason { 2131db94535dSChris Mason u32 sectorsize; 2132db94535dSChris Mason u32 nodesize; 2133db94535dSChris Mason u32 leafsize; 2134db94535dSChris Mason u32 blocksize; 213587ee04ebSChris Mason u32 stripesize; 213684234f3aSYan Zheng u64 generation; 2137f2b636e8SJosef Bacik u64 features; 21383de4586cSChris Mason struct btrfs_key location; 2139a061fc8dSChris Mason struct buffer_head *bh; 21404d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2141815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2142f84a8bd6SAl Viro struct btrfs_root *tree_root; 21434d34b278SIlya Dryomov struct btrfs_root *extent_root; 21444d34b278SIlya Dryomov struct btrfs_root *csum_root; 21454d34b278SIlya Dryomov struct btrfs_root *chunk_root; 21464d34b278SIlya Dryomov struct btrfs_root *dev_root; 2147bcef60f2SArne Jansen struct btrfs_root *quota_root; 2148f7a81ea4SStefan Behrens struct btrfs_root *uuid_root; 2149e02119d5SChris Mason struct btrfs_root *log_tree_root; 2150eb60ceacSChris Mason int ret; 2151e58ca020SYan int err = -EINVAL; 2152af31f5e5SChris Mason int num_backups_tried = 0; 2153af31f5e5SChris Mason int backup_index = 0; 21545cdc7ad3SQu Wenruo int max_active; 21555cdc7ad3SQu Wenruo int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 215670f80175SStefan Behrens bool create_uuid_tree; 215770f80175SStefan Behrens bool check_uuid_tree; 21584543df7eSChris Mason 2159f84a8bd6SAl Viro tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); 21606f07e42eSAl Viro chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); 2161cb517eabSMiao Xie if (!tree_root || !chunk_root) { 216239279cc3SChris Mason err = -ENOMEM; 216339279cc3SChris Mason goto fail; 216439279cc3SChris Mason } 216576dda93cSYan, Zheng 216676dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 216776dda93cSYan, Zheng if (ret) { 216876dda93cSYan, Zheng err = ret; 216976dda93cSYan, Zheng goto fail; 217076dda93cSYan, Zheng } 217176dda93cSYan, Zheng 217276dda93cSYan, Zheng ret = setup_bdi(fs_info, &fs_info->bdi); 217376dda93cSYan, Zheng if (ret) { 217476dda93cSYan, Zheng err = ret; 217576dda93cSYan, Zheng goto fail_srcu; 217676dda93cSYan, Zheng } 217776dda93cSYan, Zheng 2178e2d84521SMiao Xie ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0); 2179e2d84521SMiao Xie if (ret) { 2180e2d84521SMiao Xie err = ret; 2181e2d84521SMiao Xie goto fail_bdi; 2182e2d84521SMiao Xie } 2183e2d84521SMiao Xie fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * 2184e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2185e2d84521SMiao Xie 2186963d678bSMiao Xie ret = percpu_counter_init(&fs_info->delalloc_bytes, 0); 2187963d678bSMiao Xie if (ret) { 2188963d678bSMiao Xie err = ret; 2189963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2190963d678bSMiao Xie } 2191963d678bSMiao Xie 2192c404e0dcSMiao Xie ret = percpu_counter_init(&fs_info->bio_counter, 0); 2193c404e0dcSMiao Xie if (ret) { 2194c404e0dcSMiao Xie err = ret; 2195c404e0dcSMiao Xie goto fail_delalloc_bytes; 2196c404e0dcSMiao Xie } 2197c404e0dcSMiao Xie 219876dda93cSYan, Zheng fs_info->btree_inode = new_inode(sb); 219976dda93cSYan, Zheng if (!fs_info->btree_inode) { 220076dda93cSYan, Zheng err = -ENOMEM; 2201c404e0dcSMiao Xie goto fail_bio_counter; 220276dda93cSYan, Zheng } 220376dda93cSYan, Zheng 2204a6591715SChris Mason mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 22051561dedaSMiao Xie 220676dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2207f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 22088fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2209facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 221024bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2211eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 221211833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2213eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2214a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 221576dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 221624bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 22174cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 22182bf64758SJosef Bacik spin_lock_init(&fs_info->free_chunk_lock); 2219f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2220ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2221f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 2222f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 22237585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2224573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2225de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 222619c00ddcSChris Mason 222758176a96SJosef Bacik init_completion(&fs_info->kobj_unregister); 22280b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 22296324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2230f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 22310b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 223266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 223366d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 223466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, 223566d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELALLOC); 223666d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 223766d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 223866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 223966d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 224066d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2241cb03c743SChris Mason atomic_set(&fs_info->nr_async_submits, 0); 2242771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 22438c8bee1dSChris Mason atomic_set(&fs_info->async_submit_draining, 0); 22440986fe9eSChris Mason atomic_set(&fs_info->nr_async_bios, 0); 22454cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2246fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2247e20d96d6SChris Mason fs_info->sb = sb; 22486f568d35SChris Mason fs_info->max_inline = 8192 * 1024; 22499ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 22504cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 22512bf64758SJosef Bacik fs_info->free_chunk_space = 0; 2252f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 22538b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 22540a2b2a84SJosef Bacik fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64); 225590519d66SArne Jansen /* readahead state */ 225690519d66SArne Jansen INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); 225790519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2258c8b97818SChris Mason 2259b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2260b34b086cSChris Mason num_online_cpus() + 2, 8); 22610afbaf8cSChris Mason 2262199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2263199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 226416cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 226516cdcec7SMiao Xie GFP_NOFS); 226616cdcec7SMiao Xie if (!fs_info->delayed_root) { 226716cdcec7SMiao Xie err = -ENOMEM; 226816cdcec7SMiao Xie goto fail_iput; 226916cdcec7SMiao Xie } 227016cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 22713eaa2885SChris Mason 2272a2de733cSArne Jansen mutex_init(&fs_info->scrub_lock); 2273a2de733cSArne Jansen atomic_set(&fs_info->scrubs_running, 0); 2274a2de733cSArne Jansen atomic_set(&fs_info->scrub_pause_req, 0); 2275a2de733cSArne Jansen atomic_set(&fs_info->scrubs_paused, 0); 2276a2de733cSArne Jansen atomic_set(&fs_info->scrub_cancel_req, 0); 2277c404e0dcSMiao Xie init_waitqueue_head(&fs_info->replace_wait); 2278a2de733cSArne Jansen init_waitqueue_head(&fs_info->scrub_pause_wait); 2279a2de733cSArne Jansen fs_info->scrub_workers_refcnt = 0; 228021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 228121adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 228221adbd5cSStefan Behrens #endif 2283a2de733cSArne Jansen 2284c9e9f97bSIlya Dryomov spin_lock_init(&fs_info->balance_lock); 2285c9e9f97bSIlya Dryomov mutex_init(&fs_info->balance_mutex); 2286837d5b6eSIlya Dryomov atomic_set(&fs_info->balance_running, 0); 2287837d5b6eSIlya Dryomov atomic_set(&fs_info->balance_pause_req, 0); 2288a7e99c69SIlya Dryomov atomic_set(&fs_info->balance_cancel_req, 0); 2289c9e9f97bSIlya Dryomov fs_info->balance_ctl = NULL; 2290837d5b6eSIlya Dryomov init_waitqueue_head(&fs_info->balance_wait_q); 229121c7e756SMiao Xie btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work); 2292a061fc8dSChris Mason 22930afbaf8cSChris Mason sb->s_blocksize = 4096; 22940afbaf8cSChris Mason sb->s_blocksize_bits = blksize_bits(4096); 229532a88aa1SJens Axboe sb->s_bdi = &fs_info->bdi; 22960afbaf8cSChris Mason 229776dda93cSYan, Zheng fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 2298bfe86848SMiklos Szeredi set_nlink(fs_info->btree_inode, 1); 22990afbaf8cSChris Mason /* 23000afbaf8cSChris Mason * we set the i_size on the btree inode to the max possible int. 23010afbaf8cSChris Mason * the real end of the address space is determined by all of 23020afbaf8cSChris Mason * the devices in the system 23030afbaf8cSChris Mason */ 23040afbaf8cSChris Mason fs_info->btree_inode->i_size = OFFSET_MAX; 2305d98237b3SChris Mason fs_info->btree_inode->i_mapping->a_ops = &btree_aops; 230604160088SChris Mason fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; 230704160088SChris Mason 23085d4f98a2SYan Zheng RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); 2309d1310b2eSChris Mason extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, 2310f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 23110b32f4bbSJosef Bacik BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; 2312a8067e02SDavid Sterba extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); 23130da5468fSChris Mason 2314d1310b2eSChris Mason BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; 2315d1310b2eSChris Mason 23160f7d52f4SChris Mason BTRFS_I(fs_info->btree_inode)->root = tree_root; 23170f7d52f4SChris Mason memset(&BTRFS_I(fs_info->btree_inode)->location, 0, 23180f7d52f4SChris Mason sizeof(struct btrfs_key)); 231972ac3c0dSJosef Bacik set_bit(BTRFS_INODE_DUMMY, 232072ac3c0dSJosef Bacik &BTRFS_I(fs_info->btree_inode)->runtime_flags); 2321778ba82bSFilipe David Borba Manana btrfs_insert_inode_hash(fs_info->btree_inode); 232239279cc3SChris Mason 2323e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 23246bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2325a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2326925baeddSChris Mason 232711833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[0], 2328f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 232911833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[1], 2330f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 233111833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2332e6dcd2dcSChris Mason fs_info->do_barriers = 1; 2333f9295749SChris Mason 2334d98237b3SChris Mason 23355a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 23369ffba8cdSJosef Bacik mutex_init(&fs_info->ordered_extent_flush_mutex); 233730ae8467SChris Mason mutex_init(&fs_info->tree_log_mutex); 2338925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2339925baeddSChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2340a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 2341a74a4b97SChris Mason mutex_init(&fs_info->volume_mutex); 23429e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2343c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 234476dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2345803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2346e922e087SStefan Behrens fs_info->dev_replace.lock_owner = 0; 2347e922e087SStefan Behrens atomic_set(&fs_info->dev_replace.nesting_level, 0); 2348e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2349e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock_management_lock); 2350e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock); 2351fa9c0d79SChris Mason 2352416ac51dSArne Jansen spin_lock_init(&fs_info->qgroup_lock); 2353f2f6ed3dSWang Shilong mutex_init(&fs_info->qgroup_ioctl_lock); 2354416ac51dSArne Jansen fs_info->qgroup_tree = RB_ROOT; 2355416ac51dSArne Jansen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2356416ac51dSArne Jansen fs_info->qgroup_seq = 1; 2357416ac51dSArne Jansen fs_info->quota_enabled = 0; 2358416ac51dSArne Jansen fs_info->pending_quota_state = 0; 23591e8f9158SWang Shilong fs_info->qgroup_ulist = NULL; 23602f232036SJan Schmidt mutex_init(&fs_info->qgroup_rescan_lock); 2361416ac51dSArne Jansen 2362fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2363fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2364fa9c0d79SChris Mason 23657d9eb12cSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 23663768f368SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2367bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 23684854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 23699a8dd150SChris Mason 237053b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 237153b381b3SDavid Woodhouse if (ret) { 237283c8266aSDavid Sterba err = ret; 237353b381b3SDavid Woodhouse goto fail_alloc; 237453b381b3SDavid Woodhouse } 237553b381b3SDavid Woodhouse 23760b86a832SChris Mason __setup_root(4096, 4096, 4096, 4096, tree_root, 23772c90e5d6SChris Mason fs_info, BTRFS_ROOT_TREE_OBJECTID); 23787eccb903SChris Mason 23793c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 23801104a885SDavid Sterba 23811104a885SDavid Sterba /* 23821104a885SDavid Sterba * Read super block and check the signature bytes only 23831104a885SDavid Sterba */ 2384a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 238520b45077SDave Young if (!bh) { 238620b45077SDave Young err = -EINVAL; 238716cdcec7SMiao Xie goto fail_alloc; 238820b45077SDave Young } 238939279cc3SChris Mason 23901104a885SDavid Sterba /* 23911104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 23921104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 23931104a885SDavid Sterba */ 23941104a885SDavid Sterba if (btrfs_check_super_csum(bh->b_data)) { 2395efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); 23961104a885SDavid Sterba err = -EINVAL; 23971104a885SDavid Sterba goto fail_alloc; 23981104a885SDavid Sterba } 23991104a885SDavid Sterba 24001104a885SDavid Sterba /* 24011104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 24021104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 24031104a885SDavid Sterba * the whole block of INFO_SIZE 24041104a885SDavid Sterba */ 24056c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 24066c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 24076c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2408a061fc8dSChris Mason brelse(bh); 24095f39d397SChris Mason 24106c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 24110b86a832SChris Mason 24121104a885SDavid Sterba ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); 24131104a885SDavid Sterba if (ret) { 2414efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); 24151104a885SDavid Sterba err = -EINVAL; 24161104a885SDavid Sterba goto fail_alloc; 24171104a885SDavid Sterba } 24181104a885SDavid Sterba 24196c41761fSDavid Sterba disk_super = fs_info->super_copy; 24200f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 242116cdcec7SMiao Xie goto fail_alloc; 24220f7d52f4SChris Mason 2423acce952bSliubo /* check FS state, whether FS is broken. */ 242487533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 242587533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2426acce952bSliubo 242775e7cb7fSLiu Bo /* 2428af31f5e5SChris Mason * run through our array of backup supers and setup 2429af31f5e5SChris Mason * our ring pointer to the oldest one 2430af31f5e5SChris Mason */ 2431af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2432af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2433af31f5e5SChris Mason 2434af31f5e5SChris Mason /* 243575e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 243675e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 243775e7cb7fSLiu Bo */ 243875e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 243975e7cb7fSLiu Bo 24402b82032cSYan Zheng ret = btrfs_parse_options(tree_root, options); 24412b82032cSYan Zheng if (ret) { 24422b82032cSYan Zheng err = ret; 244316cdcec7SMiao Xie goto fail_alloc; 24442b82032cSYan Zheng } 2445dfe25020SChris Mason 2446f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2447f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2448f2b636e8SJosef Bacik if (features) { 2449f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount because of " 2450f2b636e8SJosef Bacik "unsupported optional features (%Lx).\n", 2451c1c9ff7cSGeert Uytterhoeven features); 2452f2b636e8SJosef Bacik err = -EINVAL; 245316cdcec7SMiao Xie goto fail_alloc; 2454f2b636e8SJosef Bacik } 2455f2b636e8SJosef Bacik 2456727011e0SChris Mason if (btrfs_super_leafsize(disk_super) != 2457727011e0SChris Mason btrfs_super_nodesize(disk_super)) { 2458727011e0SChris Mason printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2459727011e0SChris Mason "blocksizes don't match. node %d leaf %d\n", 2460727011e0SChris Mason btrfs_super_nodesize(disk_super), 2461727011e0SChris Mason btrfs_super_leafsize(disk_super)); 2462727011e0SChris Mason err = -EINVAL; 2463727011e0SChris Mason goto fail_alloc; 2464727011e0SChris Mason } 2465727011e0SChris Mason if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { 2466727011e0SChris Mason printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2467727011e0SChris Mason "blocksize (%d) was too large\n", 2468727011e0SChris Mason btrfs_super_leafsize(disk_super)); 2469727011e0SChris Mason err = -EINVAL; 2470727011e0SChris Mason goto fail_alloc; 2471727011e0SChris Mason } 2472727011e0SChris Mason 24735d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 24745d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 247569e380d1SLi Zefan if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 2476a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 2477727011e0SChris Mason 24783173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 2479efe120a0SFrank Holton printk(KERN_ERR "BTRFS: has skinny extents\n"); 24803173a18fSJosef Bacik 2481727011e0SChris Mason /* 2482727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2483727011e0SChris Mason * they are bigger than the page size 2484727011e0SChris Mason */ 2485727011e0SChris Mason if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { 2486727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 2487efe120a0SFrank Holton printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); 2488727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2489727011e0SChris Mason } 2490727011e0SChris Mason 2491bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2492bc3f116fSChris Mason leafsize = btrfs_super_leafsize(disk_super); 2493bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2494bc3f116fSChris Mason stripesize = btrfs_super_stripesize(disk_super); 2495e2d84521SMiao Xie fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids)); 2496963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2497bc3f116fSChris Mason 2498bc3f116fSChris Mason /* 2499bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2500bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2501bc3f116fSChris Mason */ 2502bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2503bc3f116fSChris Mason (sectorsize != leafsize)) { 2504efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes " 2505bc3f116fSChris Mason "are not allowed for mixed block groups on %s\n", 2506bc3f116fSChris Mason sb->s_id); 2507bc3f116fSChris Mason goto fail_alloc; 2508bc3f116fSChris Mason } 2509bc3f116fSChris Mason 2510ceda0864SMiao Xie /* 2511ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2512ceda0864SMiao Xie * update the flag. 2513ceda0864SMiao Xie */ 25145d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 25155d4f98a2SYan Zheng 2516f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2517f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2518f2b636e8SJosef Bacik if (!(sb->s_flags & MS_RDONLY) && features) { 2519f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " 2520f2b636e8SJosef Bacik "unsupported option features (%Lx).\n", 2521c1c9ff7cSGeert Uytterhoeven features); 2522f2b636e8SJosef Bacik err = -EINVAL; 252316cdcec7SMiao Xie goto fail_alloc; 2524f2b636e8SJosef Bacik } 252561d92c32SChris Mason 25265cdc7ad3SQu Wenruo max_active = fs_info->thread_pool_size; 252761d92c32SChris Mason 25285cdc7ad3SQu Wenruo fs_info->workers = 25295cdc7ad3SQu Wenruo btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI, 25305cdc7ad3SQu Wenruo max_active, 16); 2531c8b97818SChris Mason 2532afe3d242SQu Wenruo fs_info->delalloc_workers = 2533afe3d242SQu Wenruo btrfs_alloc_workqueue("delalloc", flags, max_active, 2); 2534771ed689SChris Mason 2535a44903abSQu Wenruo fs_info->flush_workers = 2536a44903abSQu Wenruo btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0); 25378ccf6f19SMiao Xie 2538e66f0bb1SQu Wenruo fs_info->caching_workers = 2539e66f0bb1SQu Wenruo btrfs_alloc_workqueue("cache", flags, max_active, 0); 254061b49440SChris Mason 2541a8c93d4eSQu Wenruo /* 2542a8c93d4eSQu Wenruo * a higher idle thresh on the submit workers makes it much more 254361b49440SChris Mason * likely that bios will be send down in a sane order to the 254461b49440SChris Mason * devices 254561b49440SChris Mason */ 2546a8c93d4eSQu Wenruo fs_info->submit_workers = 2547a8c93d4eSQu Wenruo btrfs_alloc_workqueue("submit", flags, 2548a8c93d4eSQu Wenruo min_t(u64, fs_devices->num_devices, 2549a8c93d4eSQu Wenruo max_active), 64); 255053863232SChris Mason 2551dc6e3209SQu Wenruo fs_info->fixup_workers = 2552dc6e3209SQu Wenruo btrfs_alloc_workqueue("fixup", flags, 1, 0); 255361b49440SChris Mason 255461b49440SChris Mason /* 255561b49440SChris Mason * endios are largely parallel and should have a very 255661b49440SChris Mason * low idle thresh 255761b49440SChris Mason */ 2558fccb5d86SQu Wenruo fs_info->endio_workers = 2559fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio", flags, max_active, 4); 2560fccb5d86SQu Wenruo fs_info->endio_meta_workers = 2561fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-meta", flags, max_active, 4); 2562fccb5d86SQu Wenruo fs_info->endio_meta_write_workers = 2563fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2); 2564fccb5d86SQu Wenruo fs_info->endio_raid56_workers = 2565fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4); 2566d05a33acSQu Wenruo fs_info->rmw_workers = 2567d05a33acSQu Wenruo btrfs_alloc_workqueue("rmw", flags, max_active, 2); 2568fccb5d86SQu Wenruo fs_info->endio_write_workers = 2569fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-write", flags, max_active, 2); 2570fccb5d86SQu Wenruo fs_info->endio_freespace_worker = 2571fccb5d86SQu Wenruo btrfs_alloc_workqueue("freespace-write", flags, max_active, 0); 25725b3bc44eSQu Wenruo fs_info->delayed_workers = 25735b3bc44eSQu Wenruo btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0); 2574736cfa15SQu Wenruo fs_info->readahead_workers = 2575736cfa15SQu Wenruo btrfs_alloc_workqueue("readahead", flags, max_active, 2); 2576fc97fab0SQu Wenruo fs_info->qgroup_rescan_workers = 2577fc97fab0SQu Wenruo btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0); 2578b51912c9SChris Mason 2579a8c93d4eSQu Wenruo if (!(fs_info->workers && fs_info->delalloc_workers && 2580fccb5d86SQu Wenruo fs_info->submit_workers && fs_info->flush_workers && 2581fccb5d86SQu Wenruo fs_info->endio_workers && fs_info->endio_meta_workers && 2582fccb5d86SQu Wenruo fs_info->endio_meta_write_workers && 2583fccb5d86SQu Wenruo fs_info->endio_write_workers && fs_info->endio_raid56_workers && 2584e66f0bb1SQu Wenruo fs_info->endio_freespace_worker && fs_info->rmw_workers && 2585dc6e3209SQu Wenruo fs_info->caching_workers && fs_info->readahead_workers && 2586fc97fab0SQu Wenruo fs_info->fixup_workers && fs_info->delayed_workers && 2587fc97fab0SQu Wenruo fs_info->qgroup_rescan_workers)) { 2588fed425c7SIlya Dryomov err = -ENOMEM; 25890dc3b84aSJosef Bacik goto fail_sb_buffer; 25900dc3b84aSJosef Bacik } 25914543df7eSChris Mason 25924575c9ccSChris Mason fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); 2593c8b97818SChris Mason fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 2594c8b97818SChris Mason 4 * 1024 * 1024 / PAGE_CACHE_SIZE); 25954575c9ccSChris Mason 2596db94535dSChris Mason tree_root->nodesize = nodesize; 2597db94535dSChris Mason tree_root->leafsize = leafsize; 2598db94535dSChris Mason tree_root->sectorsize = sectorsize; 259987ee04ebSChris Mason tree_root->stripesize = stripesize; 2600a061fc8dSChris Mason 2601a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2602a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2603db94535dSChris Mason 26043cae210fSQu Wenruo if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) { 2605efe120a0SFrank Holton printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id); 260639279cc3SChris Mason goto fail_sb_buffer; 260739279cc3SChris Mason } 260819c00ddcSChris Mason 26098d082fb7SLiu Bo if (sectorsize != PAGE_SIZE) { 2610efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) " 26118d082fb7SLiu Bo "found on %s\n", (unsigned long)sectorsize, sb->s_id); 2612941b2ddfSKeith Mannthey goto fail_sb_buffer; 2613941b2ddfSKeith Mannthey } 2614941b2ddfSKeith Mannthey 2615925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 2616e4404d6eSYan Zheng ret = btrfs_read_sys_array(tree_root); 2617925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 261884eed90fSChris Mason if (ret) { 2619efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read the system " 2620d397712bSChris Mason "array on %s\n", sb->s_id); 26215d4f98a2SYan Zheng goto fail_sb_buffer; 262284eed90fSChris Mason } 26230b86a832SChris Mason 26240b86a832SChris Mason blocksize = btrfs_level_size(tree_root, 26250b86a832SChris Mason btrfs_super_chunk_root_level(disk_super)); 262684234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 26270b86a832SChris Mason 26280b86a832SChris Mason __setup_root(nodesize, leafsize, sectorsize, stripesize, 26290b86a832SChris Mason chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); 26300b86a832SChris Mason 26310b86a832SChris Mason chunk_root->node = read_tree_block(chunk_root, 26320b86a832SChris Mason btrfs_super_chunk_root(disk_super), 263384234f3aSYan Zheng blocksize, generation); 2634416bc658SJosef Bacik if (!chunk_root->node || 2635416bc658SJosef Bacik !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { 2636efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n", 263783121942SDavid Woodhouse sb->s_id); 2638af31f5e5SChris Mason goto fail_tree_roots; 263983121942SDavid Woodhouse } 26405d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 26415d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 26420b86a832SChris Mason 2643e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2644b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2645e17cade2SChris Mason 26460b86a832SChris Mason ret = btrfs_read_chunk_tree(chunk_root); 26472b82032cSYan Zheng if (ret) { 2648efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n", 2649d397712bSChris Mason sb->s_id); 2650af31f5e5SChris Mason goto fail_tree_roots; 26512b82032cSYan Zheng } 26520b86a832SChris Mason 26538dabb742SStefan Behrens /* 26548dabb742SStefan Behrens * keep the device that is marked to be the target device for the 26558dabb742SStefan Behrens * dev_replace procedure 26568dabb742SStefan Behrens */ 26578dabb742SStefan Behrens btrfs_close_extra_devices(fs_info, fs_devices, 0); 2658dfe25020SChris Mason 2659a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 2660efe120a0SFrank Holton printk(KERN_CRIT "BTRFS: failed to read devices on %s\n", 2661a6b0d5c8SChris Mason sb->s_id); 2662a6b0d5c8SChris Mason goto fail_tree_roots; 2663a6b0d5c8SChris Mason } 2664a6b0d5c8SChris Mason 2665af31f5e5SChris Mason retry_root_backup: 2666db94535dSChris Mason blocksize = btrfs_level_size(tree_root, 2667db94535dSChris Mason btrfs_super_root_level(disk_super)); 266884234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 26690b86a832SChris Mason 2670e20d96d6SChris Mason tree_root->node = read_tree_block(tree_root, 2671db94535dSChris Mason btrfs_super_root(disk_super), 267284234f3aSYan Zheng blocksize, generation); 2673af31f5e5SChris Mason if (!tree_root->node || 2674af31f5e5SChris Mason !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { 2675efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", 267683121942SDavid Woodhouse sb->s_id); 2677af31f5e5SChris Mason 2678af31f5e5SChris Mason goto recovery_tree_root; 267983121942SDavid Woodhouse } 2680af31f5e5SChris Mason 26815d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 26825d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 268369e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 2684db94535dSChris Mason 2685cb517eabSMiao Xie location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 2686cb517eabSMiao Xie location.type = BTRFS_ROOT_ITEM_KEY; 2687cb517eabSMiao Xie location.offset = 0; 2688cb517eabSMiao Xie 2689cb517eabSMiao Xie extent_root = btrfs_read_tree_root(tree_root, &location); 2690cb517eabSMiao Xie if (IS_ERR(extent_root)) { 2691cb517eabSMiao Xie ret = PTR_ERR(extent_root); 2692af31f5e5SChris Mason goto recovery_tree_root; 2693cb517eabSMiao Xie } 269427cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state); 2695cb517eabSMiao Xie fs_info->extent_root = extent_root; 26960b86a832SChris Mason 2697cb517eabSMiao Xie location.objectid = BTRFS_DEV_TREE_OBJECTID; 2698cb517eabSMiao Xie dev_root = btrfs_read_tree_root(tree_root, &location); 2699cb517eabSMiao Xie if (IS_ERR(dev_root)) { 2700cb517eabSMiao Xie ret = PTR_ERR(dev_root); 2701af31f5e5SChris Mason goto recovery_tree_root; 2702cb517eabSMiao Xie } 270327cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state); 2704cb517eabSMiao Xie fs_info->dev_root = dev_root; 2705cb517eabSMiao Xie btrfs_init_devices_late(fs_info); 27063768f368SChris Mason 2707cb517eabSMiao Xie location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2708cb517eabSMiao Xie csum_root = btrfs_read_tree_root(tree_root, &location); 2709cb517eabSMiao Xie if (IS_ERR(csum_root)) { 2710cb517eabSMiao Xie ret = PTR_ERR(csum_root); 2711af31f5e5SChris Mason goto recovery_tree_root; 2712cb517eabSMiao Xie } 271327cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state); 2714cb517eabSMiao Xie fs_info->csum_root = csum_root; 2715d20f7043SChris Mason 2716cb517eabSMiao Xie location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2717cb517eabSMiao Xie quota_root = btrfs_read_tree_root(tree_root, &location); 2718cb517eabSMiao Xie if (!IS_ERR(quota_root)) { 271927cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, "a_root->state); 2720bcef60f2SArne Jansen fs_info->quota_enabled = 1; 2721bcef60f2SArne Jansen fs_info->pending_quota_state = 1; 2722cb517eabSMiao Xie fs_info->quota_root = quota_root; 2723bcef60f2SArne Jansen } 2724bcef60f2SArne Jansen 2725f7a81ea4SStefan Behrens location.objectid = BTRFS_UUID_TREE_OBJECTID; 2726f7a81ea4SStefan Behrens uuid_root = btrfs_read_tree_root(tree_root, &location); 2727f7a81ea4SStefan Behrens if (IS_ERR(uuid_root)) { 2728f7a81ea4SStefan Behrens ret = PTR_ERR(uuid_root); 2729f7a81ea4SStefan Behrens if (ret != -ENOENT) 2730f7a81ea4SStefan Behrens goto recovery_tree_root; 2731f7a81ea4SStefan Behrens create_uuid_tree = true; 273270f80175SStefan Behrens check_uuid_tree = false; 2733f7a81ea4SStefan Behrens } else { 273427cdeb70SMiao Xie set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state); 2735f7a81ea4SStefan Behrens fs_info->uuid_root = uuid_root; 273670f80175SStefan Behrens create_uuid_tree = false; 273770f80175SStefan Behrens check_uuid_tree = 273870f80175SStefan Behrens generation != btrfs_super_uuid_tree_generation(disk_super); 2739f7a81ea4SStefan Behrens } 2740f7a81ea4SStefan Behrens 27418929ecfaSYan, Zheng fs_info->generation = generation; 27428929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 27438929ecfaSYan, Zheng 274468310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 274568310a5eSIlya Dryomov if (ret) { 2746efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to recover balance\n"); 274768310a5eSIlya Dryomov goto fail_block_groups; 274868310a5eSIlya Dryomov } 274968310a5eSIlya Dryomov 2750733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 2751733f4fbbSStefan Behrens if (ret) { 2752efe120a0SFrank Holton printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", 2753733f4fbbSStefan Behrens ret); 2754733f4fbbSStefan Behrens goto fail_block_groups; 2755733f4fbbSStefan Behrens } 2756733f4fbbSStefan Behrens 27578dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 27588dabb742SStefan Behrens if (ret) { 2759efe120a0SFrank Holton pr_err("BTRFS: failed to init dev_replace: %d\n", ret); 27608dabb742SStefan Behrens goto fail_block_groups; 27618dabb742SStefan Behrens } 27628dabb742SStefan Behrens 27638dabb742SStefan Behrens btrfs_close_extra_devices(fs_info, fs_devices, 1); 27648dabb742SStefan Behrens 27655ac1d209SJeff Mahoney ret = btrfs_sysfs_add_one(fs_info); 27665ac1d209SJeff Mahoney if (ret) { 2767efe120a0SFrank Holton pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); 27685ac1d209SJeff Mahoney goto fail_block_groups; 27695ac1d209SJeff Mahoney } 27705ac1d209SJeff Mahoney 2771c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 2772c59021f8Sliubo if (ret) { 2773efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); 27742365dd3cSAnand Jain goto fail_sysfs; 2775c59021f8Sliubo } 2776c59021f8Sliubo 27771b1d1f66SJosef Bacik ret = btrfs_read_block_groups(extent_root); 27781b1d1f66SJosef Bacik if (ret) { 2779efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); 27802365dd3cSAnand Jain goto fail_sysfs; 27811b1d1f66SJosef Bacik } 27825af3e8ccSStefan Behrens fs_info->num_tolerated_disk_barrier_failures = 27835af3e8ccSStefan Behrens btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); 2784292fd7fcSStefan Behrens if (fs_info->fs_devices->missing_devices > 2785292fd7fcSStefan Behrens fs_info->num_tolerated_disk_barrier_failures && 2786292fd7fcSStefan Behrens !(sb->s_flags & MS_RDONLY)) { 2787efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: " 2788efe120a0SFrank Holton "too many missing devices, writeable mount is not allowed\n"); 27892365dd3cSAnand Jain goto fail_sysfs; 2790292fd7fcSStefan Behrens } 27919078a3e1SChris Mason 2792a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 2793a74a4b97SChris Mason "btrfs-cleaner"); 279457506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 27952365dd3cSAnand Jain goto fail_sysfs; 2796a74a4b97SChris Mason 2797a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 2798a74a4b97SChris Mason tree_root, 2799a74a4b97SChris Mason "btrfs-transaction"); 280057506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 28013f157a2fSChris Mason goto fail_cleaner; 2802a74a4b97SChris Mason 2803c289811cSChris Mason if (!btrfs_test_opt(tree_root, SSD) && 2804c289811cSChris Mason !btrfs_test_opt(tree_root, NOSSD) && 2805c289811cSChris Mason !fs_info->fs_devices->rotating) { 2806efe120a0SFrank Holton printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " 2807c289811cSChris Mason "mode\n"); 2808c289811cSChris Mason btrfs_set_opt(fs_info->mount_opt, SSD); 2809c289811cSChris Mason } 2810c289811cSChris Mason 28113818aea2SQu Wenruo /* Set the real inode map cache flag */ 28123818aea2SQu Wenruo if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE)) 28133818aea2SQu Wenruo btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE); 28143818aea2SQu Wenruo 281521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 281621adbd5cSStefan Behrens if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { 281721adbd5cSStefan Behrens ret = btrfsic_mount(tree_root, fs_devices, 281821adbd5cSStefan Behrens btrfs_test_opt(tree_root, 281921adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 282021adbd5cSStefan Behrens 1 : 0, 282121adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 282221adbd5cSStefan Behrens if (ret) 2823efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to initialize" 282421adbd5cSStefan Behrens " integrity check module %s\n", sb->s_id); 282521adbd5cSStefan Behrens } 282621adbd5cSStefan Behrens #endif 2827bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 2828bcef60f2SArne Jansen if (ret) 2829bcef60f2SArne Jansen goto fail_trans_kthread; 283021adbd5cSStefan Behrens 2831acce952bSliubo /* do not make disk changes in broken FS */ 283268ce9682SStefan Behrens if (btrfs_super_log_root(disk_super) != 0) { 2833e02119d5SChris Mason u64 bytenr = btrfs_super_log_root(disk_super); 2834d18a2c44SChris Mason 28357c2ca468SChris Mason if (fs_devices->rw_devices == 0) { 2836efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: log replay required " 2837d397712bSChris Mason "on RO media\n"); 28387c2ca468SChris Mason err = -EIO; 2839bcef60f2SArne Jansen goto fail_qgroup; 28407c2ca468SChris Mason } 2841e02119d5SChris Mason blocksize = 2842e02119d5SChris Mason btrfs_level_size(tree_root, 2843e02119d5SChris Mason btrfs_super_log_root_level(disk_super)); 2844e02119d5SChris Mason 28456f07e42eSAl Viro log_tree_root = btrfs_alloc_root(fs_info); 2846676e4c86SDan Carpenter if (!log_tree_root) { 2847676e4c86SDan Carpenter err = -ENOMEM; 2848bcef60f2SArne Jansen goto fail_qgroup; 2849676e4c86SDan Carpenter } 2850e02119d5SChris Mason 2851e02119d5SChris Mason __setup_root(nodesize, leafsize, sectorsize, stripesize, 2852e02119d5SChris Mason log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 2853e02119d5SChris Mason 2854e02119d5SChris Mason log_tree_root->node = read_tree_block(tree_root, bytenr, 285584234f3aSYan Zheng blocksize, 285684234f3aSYan Zheng generation + 1); 2857416bc658SJosef Bacik if (!log_tree_root->node || 2858416bc658SJosef Bacik !extent_buffer_uptodate(log_tree_root->node)) { 2859efe120a0SFrank Holton printk(KERN_ERR "BTRFS: failed to read log tree\n"); 2860416bc658SJosef Bacik free_extent_buffer(log_tree_root->node); 2861416bc658SJosef Bacik kfree(log_tree_root); 286228c16cbbSWang Shilong goto fail_qgroup; 2863416bc658SJosef Bacik } 286479787eaaSJeff Mahoney /* returns with log_tree_root freed on success */ 2865e02119d5SChris Mason ret = btrfs_recover_log_trees(log_tree_root); 286679787eaaSJeff Mahoney if (ret) { 286779787eaaSJeff Mahoney btrfs_error(tree_root->fs_info, ret, 286879787eaaSJeff Mahoney "Failed to recover log tree"); 286979787eaaSJeff Mahoney free_extent_buffer(log_tree_root->node); 287079787eaaSJeff Mahoney kfree(log_tree_root); 287128c16cbbSWang Shilong goto fail_qgroup; 287279787eaaSJeff Mahoney } 2873e556ce2cSYan Zheng 2874e556ce2cSYan Zheng if (sb->s_flags & MS_RDONLY) { 2875e556ce2cSYan Zheng ret = btrfs_commit_super(tree_root); 287679787eaaSJeff Mahoney if (ret) 287728c16cbbSWang Shilong goto fail_qgroup; 2878e556ce2cSYan Zheng } 2879e02119d5SChris Mason } 28801a40e23bSZheng Yan 288176dda93cSYan, Zheng ret = btrfs_find_orphan_roots(tree_root); 288279787eaaSJeff Mahoney if (ret) 288328c16cbbSWang Shilong goto fail_qgroup; 288476dda93cSYan, Zheng 28857c2ca468SChris Mason if (!(sb->s_flags & MS_RDONLY)) { 2886d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 288744c44af2SIlya Dryomov if (ret) 288828c16cbbSWang Shilong goto fail_qgroup; 2889d68fc57bSYan, Zheng 28905d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 2891d7ce5843SMiao Xie if (ret < 0) { 2892d7ce5843SMiao Xie printk(KERN_WARNING 2893efe120a0SFrank Holton "BTRFS: failed to recover relocation\n"); 2894d7ce5843SMiao Xie err = -EINVAL; 2895bcef60f2SArne Jansen goto fail_qgroup; 2896d7ce5843SMiao Xie } 28977c2ca468SChris Mason } 28981a40e23bSZheng Yan 28993de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 29003de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 2901cb517eabSMiao Xie location.offset = 0; 29023de4586cSChris Mason 29033de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 29043140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 29053140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 2906bcef60f2SArne Jansen goto fail_qgroup; 29073140c9a3SDan Carpenter } 2908c289811cSChris Mason 29092b6ba629SIlya Dryomov if (sb->s_flags & MS_RDONLY) 29102b6ba629SIlya Dryomov return 0; 29112b6ba629SIlya Dryomov 2912e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 29132b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29142b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29152b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 29162b6ba629SIlya Dryomov close_ctree(tree_root); 29172b6ba629SIlya Dryomov return ret; 29182b6ba629SIlya Dryomov } 2919e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 292059641015SIlya Dryomov 29212b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 29222b6ba629SIlya Dryomov if (ret) { 2923efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to resume balance\n"); 292466b4ffd1SJosef Bacik close_ctree(tree_root); 29252b6ba629SIlya Dryomov return ret; 2926e3acc2a6SJosef Bacik } 2927e3acc2a6SJosef Bacik 29288dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 29298dabb742SStefan Behrens if (ret) { 2930efe120a0SFrank Holton pr_warn("BTRFS: failed to resume dev_replace\n"); 29318dabb742SStefan Behrens close_ctree(tree_root); 29328dabb742SStefan Behrens return ret; 29338dabb742SStefan Behrens } 29348dabb742SStefan Behrens 2935b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 2936b382a324SJan Schmidt 2937f7a81ea4SStefan Behrens if (create_uuid_tree) { 2938efe120a0SFrank Holton pr_info("BTRFS: creating UUID tree\n"); 2939f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 2940f7a81ea4SStefan Behrens if (ret) { 2941efe120a0SFrank Holton pr_warn("BTRFS: failed to create the UUID tree %d\n", 2942f7a81ea4SStefan Behrens ret); 2943f7a81ea4SStefan Behrens close_ctree(tree_root); 2944f7a81ea4SStefan Behrens return ret; 2945f7a81ea4SStefan Behrens } 2946f420ee1eSStefan Behrens } else if (check_uuid_tree || 2947f420ee1eSStefan Behrens btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) { 2948efe120a0SFrank Holton pr_info("BTRFS: checking UUID tree\n"); 294970f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 295070f80175SStefan Behrens if (ret) { 2951efe120a0SFrank Holton pr_warn("BTRFS: failed to check the UUID tree %d\n", 295270f80175SStefan Behrens ret); 295370f80175SStefan Behrens close_ctree(tree_root); 295470f80175SStefan Behrens return ret; 295570f80175SStefan Behrens } 295670f80175SStefan Behrens } else { 295770f80175SStefan Behrens fs_info->update_uuid_tree_gen = 1; 2958f7a81ea4SStefan Behrens } 2959f7a81ea4SStefan Behrens 2960ad2b2c80SAl Viro return 0; 296139279cc3SChris Mason 2962bcef60f2SArne Jansen fail_qgroup: 2963bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 29647c2ca468SChris Mason fail_trans_kthread: 29657c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 296654067ae9SJosef Bacik btrfs_cleanup_transaction(fs_info->tree_root); 29677b5ff90eSJosef Bacik del_fs_roots(fs_info); 29683f157a2fSChris Mason fail_cleaner: 2969a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 29707c2ca468SChris Mason 29717c2ca468SChris Mason /* 29727c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 29737c2ca468SChris Mason * kthreads 29747c2ca468SChris Mason */ 29757c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 29767c2ca468SChris Mason 29772365dd3cSAnand Jain fail_sysfs: 29782365dd3cSAnand Jain btrfs_sysfs_remove_one(fs_info); 29792365dd3cSAnand Jain 29801b1d1f66SJosef Bacik fail_block_groups: 298154067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 29821b1d1f66SJosef Bacik btrfs_free_block_groups(fs_info); 2983af31f5e5SChris Mason 2984af31f5e5SChris Mason fail_tree_roots: 2985af31f5e5SChris Mason free_root_pointers(fs_info, 1); 29862b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 2987af31f5e5SChris Mason 298839279cc3SChris Mason fail_sb_buffer: 29897abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 299016cdcec7SMiao Xie fail_alloc: 29914543df7eSChris Mason fail_iput: 2992586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 2993586e46e2SIlya Dryomov 29944543df7eSChris Mason iput(fs_info->btree_inode); 2995c404e0dcSMiao Xie fail_bio_counter: 2996c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 2997963d678bSMiao Xie fail_delalloc_bytes: 2998963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 2999e2d84521SMiao Xie fail_dirty_metadata_bytes: 3000e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3001ad081f14SJens Axboe fail_bdi: 30027e662854SQinghuang Feng bdi_destroy(&fs_info->bdi); 300376dda93cSYan, Zheng fail_srcu: 300476dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 30057e662854SQinghuang Feng fail: 300653b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3007586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3008ad2b2c80SAl Viro return err; 3009af31f5e5SChris Mason 3010af31f5e5SChris Mason recovery_tree_root: 3011af31f5e5SChris Mason if (!btrfs_test_opt(tree_root, RECOVERY)) 3012af31f5e5SChris Mason goto fail_tree_roots; 3013af31f5e5SChris Mason 3014af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3015af31f5e5SChris Mason 3016af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3017af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3018af31f5e5SChris Mason 3019af31f5e5SChris Mason /* we can't trust the free space cache either */ 3020af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3021af31f5e5SChris Mason 3022af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3023af31f5e5SChris Mason &num_backups_tried, &backup_index); 3024af31f5e5SChris Mason if (ret == -1) 3025af31f5e5SChris Mason goto fail_block_groups; 3026af31f5e5SChris Mason goto retry_root_backup; 3027eb60ceacSChris Mason } 3028eb60ceacSChris Mason 3029f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3030f2984462SChris Mason { 3031f2984462SChris Mason if (uptodate) { 3032f2984462SChris Mason set_buffer_uptodate(bh); 3033f2984462SChris Mason } else { 3034442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3035442a4f63SStefan Behrens bh->b_private; 3036442a4f63SStefan Behrens 3037efe120a0SFrank Holton printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to " 3038606686eeSJosef Bacik "I/O error on %s\n", 3039606686eeSJosef Bacik rcu_str_deref(device->name)); 30401259ab75SChris Mason /* note, we dont' set_buffer_write_io_error because we have 30411259ab75SChris Mason * our own ways of dealing with the IO errors 30421259ab75SChris Mason */ 3043f2984462SChris Mason clear_buffer_uptodate(bh); 3044442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3045f2984462SChris Mason } 3046f2984462SChris Mason unlock_buffer(bh); 3047f2984462SChris Mason put_bh(bh); 3048f2984462SChris Mason } 3049f2984462SChris Mason 3050a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3051a512bbf8SYan Zheng { 3052a512bbf8SYan Zheng struct buffer_head *bh; 3053a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3054a512bbf8SYan Zheng struct btrfs_super_block *super; 3055a512bbf8SYan Zheng int i; 3056a512bbf8SYan Zheng u64 transid = 0; 3057a512bbf8SYan Zheng u64 bytenr; 3058a512bbf8SYan Zheng 3059a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3060a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3061a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3062a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3063a512bbf8SYan Zheng */ 3064a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3065a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 30668068a47eSAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= 30678068a47eSAnand Jain i_size_read(bdev->bd_inode)) 3068a512bbf8SYan Zheng break; 30698068a47eSAnand Jain bh = __bread(bdev, bytenr / 4096, 30708068a47eSAnand Jain BTRFS_SUPER_INFO_SIZE); 3071a512bbf8SYan Zheng if (!bh) 3072a512bbf8SYan Zheng continue; 3073a512bbf8SYan Zheng 3074a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3075a512bbf8SYan Zheng if (btrfs_super_bytenr(super) != bytenr || 30763cae210fSQu Wenruo btrfs_super_magic(super) != BTRFS_MAGIC) { 3077a512bbf8SYan Zheng brelse(bh); 3078a512bbf8SYan Zheng continue; 3079a512bbf8SYan Zheng } 3080a512bbf8SYan Zheng 3081a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3082a512bbf8SYan Zheng brelse(latest); 3083a512bbf8SYan Zheng latest = bh; 3084a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3085a512bbf8SYan Zheng } else { 3086a512bbf8SYan Zheng brelse(bh); 3087a512bbf8SYan Zheng } 3088a512bbf8SYan Zheng } 3089a512bbf8SYan Zheng return latest; 3090a512bbf8SYan Zheng } 3091a512bbf8SYan Zheng 30924eedeb75SHisashi Hifumi /* 30934eedeb75SHisashi Hifumi * this should be called twice, once with wait == 0 and 30944eedeb75SHisashi Hifumi * once with wait == 1. When wait == 0 is done, all the buffer heads 30954eedeb75SHisashi Hifumi * we write are pinned. 30964eedeb75SHisashi Hifumi * 30974eedeb75SHisashi Hifumi * They are released when wait == 1 is done. 30984eedeb75SHisashi Hifumi * max_mirrors must be the same for both runs, and it indicates how 30994eedeb75SHisashi Hifumi * many supers on this one device should be written. 31004eedeb75SHisashi Hifumi * 31014eedeb75SHisashi Hifumi * max_mirrors == 0 means to write them all. 31024eedeb75SHisashi Hifumi */ 3103a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3104a512bbf8SYan Zheng struct btrfs_super_block *sb, 3105a512bbf8SYan Zheng int do_barriers, int wait, int max_mirrors) 3106a512bbf8SYan Zheng { 3107a512bbf8SYan Zheng struct buffer_head *bh; 3108a512bbf8SYan Zheng int i; 3109a512bbf8SYan Zheng int ret; 3110a512bbf8SYan Zheng int errors = 0; 3111a512bbf8SYan Zheng u32 crc; 3112a512bbf8SYan Zheng u64 bytenr; 3113a512bbf8SYan Zheng 3114a512bbf8SYan Zheng if (max_mirrors == 0) 3115a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3116a512bbf8SYan Zheng 3117a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3118a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3119a512bbf8SYan Zheng if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) 3120a512bbf8SYan Zheng break; 3121a512bbf8SYan Zheng 3122a512bbf8SYan Zheng if (wait) { 3123a512bbf8SYan Zheng bh = __find_get_block(device->bdev, bytenr / 4096, 3124a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3125634554dcSJosef Bacik if (!bh) { 3126634554dcSJosef Bacik errors++; 3127634554dcSJosef Bacik continue; 3128634554dcSJosef Bacik } 3129a512bbf8SYan Zheng wait_on_buffer(bh); 31304eedeb75SHisashi Hifumi if (!buffer_uptodate(bh)) 31314eedeb75SHisashi Hifumi errors++; 31324eedeb75SHisashi Hifumi 31334eedeb75SHisashi Hifumi /* drop our reference */ 31344eedeb75SHisashi Hifumi brelse(bh); 31354eedeb75SHisashi Hifumi 31364eedeb75SHisashi Hifumi /* drop the reference from the wait == 0 run */ 3137a512bbf8SYan Zheng brelse(bh); 3138a512bbf8SYan Zheng continue; 3139a512bbf8SYan Zheng } else { 3140a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3141a512bbf8SYan Zheng 3142a512bbf8SYan Zheng crc = ~(u32)0; 3143b0496686SLiu Bo crc = btrfs_csum_data((char *)sb + 3144a512bbf8SYan Zheng BTRFS_CSUM_SIZE, crc, 3145a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE - 3146a512bbf8SYan Zheng BTRFS_CSUM_SIZE); 3147a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3148a512bbf8SYan Zheng 31494eedeb75SHisashi Hifumi /* 31504eedeb75SHisashi Hifumi * one reference for us, and we leave it for the 31514eedeb75SHisashi Hifumi * caller 31524eedeb75SHisashi Hifumi */ 3153a512bbf8SYan Zheng bh = __getblk(device->bdev, bytenr / 4096, 3154a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3155634554dcSJosef Bacik if (!bh) { 3156efe120a0SFrank Holton printk(KERN_ERR "BTRFS: couldn't get super " 3157634554dcSJosef Bacik "buffer head for bytenr %Lu\n", bytenr); 3158634554dcSJosef Bacik errors++; 3159634554dcSJosef Bacik continue; 3160634554dcSJosef Bacik } 3161634554dcSJosef Bacik 3162a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3163a512bbf8SYan Zheng 31644eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3165a512bbf8SYan Zheng get_bh(bh); 31664eedeb75SHisashi Hifumi 31674eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3168a512bbf8SYan Zheng lock_buffer(bh); 3169a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3170442a4f63SStefan Behrens bh->b_private = device; 3171a512bbf8SYan Zheng } 3172a512bbf8SYan Zheng 3173387125fcSChris Mason /* 3174387125fcSChris Mason * we fua the first super. The others we allow 3175387125fcSChris Mason * to go down lazy. 3176387125fcSChris Mason */ 3177e8117c26SWang Shilong if (i == 0) 317821adbd5cSStefan Behrens ret = btrfsic_submit_bh(WRITE_FUA, bh); 3179e8117c26SWang Shilong else 3180e8117c26SWang Shilong ret = btrfsic_submit_bh(WRITE_SYNC, bh); 31814eedeb75SHisashi Hifumi if (ret) 3182a512bbf8SYan Zheng errors++; 3183a512bbf8SYan Zheng } 3184a512bbf8SYan Zheng return errors < i ? 0 : -1; 3185a512bbf8SYan Zheng } 3186a512bbf8SYan Zheng 3187387125fcSChris Mason /* 3188387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3189387125fcSChris Mason * for the barrier when it is done 3190387125fcSChris Mason */ 3191387125fcSChris Mason static void btrfs_end_empty_barrier(struct bio *bio, int err) 3192387125fcSChris Mason { 3193387125fcSChris Mason if (err) { 3194387125fcSChris Mason if (err == -EOPNOTSUPP) 3195387125fcSChris Mason set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); 3196387125fcSChris Mason clear_bit(BIO_UPTODATE, &bio->bi_flags); 3197387125fcSChris Mason } 3198387125fcSChris Mason if (bio->bi_private) 3199387125fcSChris Mason complete(bio->bi_private); 3200387125fcSChris Mason bio_put(bio); 3201387125fcSChris Mason } 3202387125fcSChris Mason 3203387125fcSChris Mason /* 3204387125fcSChris Mason * trigger flushes for one the devices. If you pass wait == 0, the flushes are 3205387125fcSChris Mason * sent down. With wait == 1, it waits for the previous flush. 3206387125fcSChris Mason * 3207387125fcSChris Mason * any device where the flush fails with eopnotsupp are flagged as not-barrier 3208387125fcSChris Mason * capable 3209387125fcSChris Mason */ 3210387125fcSChris Mason static int write_dev_flush(struct btrfs_device *device, int wait) 3211387125fcSChris Mason { 3212387125fcSChris Mason struct bio *bio; 3213387125fcSChris Mason int ret = 0; 3214387125fcSChris Mason 3215387125fcSChris Mason if (device->nobarriers) 3216387125fcSChris Mason return 0; 3217387125fcSChris Mason 3218387125fcSChris Mason if (wait) { 3219387125fcSChris Mason bio = device->flush_bio; 3220387125fcSChris Mason if (!bio) 3221387125fcSChris Mason return 0; 3222387125fcSChris Mason 3223387125fcSChris Mason wait_for_completion(&device->flush_wait); 3224387125fcSChris Mason 3225387125fcSChris Mason if (bio_flagged(bio, BIO_EOPNOTSUPP)) { 3226efe120a0SFrank Holton printk_in_rcu("BTRFS: disabling barriers on dev %s\n", 3227606686eeSJosef Bacik rcu_str_deref(device->name)); 3228387125fcSChris Mason device->nobarriers = 1; 32295af3e8ccSStefan Behrens } else if (!bio_flagged(bio, BIO_UPTODATE)) { 3230387125fcSChris Mason ret = -EIO; 3231442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, 3232442a4f63SStefan Behrens BTRFS_DEV_STAT_FLUSH_ERRS); 3233387125fcSChris Mason } 3234387125fcSChris Mason 3235387125fcSChris Mason /* drop the reference from the wait == 0 run */ 3236387125fcSChris Mason bio_put(bio); 3237387125fcSChris Mason device->flush_bio = NULL; 3238387125fcSChris Mason 3239387125fcSChris Mason return ret; 3240387125fcSChris Mason } 3241387125fcSChris Mason 3242387125fcSChris Mason /* 3243387125fcSChris Mason * one reference for us, and we leave it for the 3244387125fcSChris Mason * caller 3245387125fcSChris Mason */ 32469c017abcSJesper Juhl device->flush_bio = NULL; 32479be3395bSChris Mason bio = btrfs_io_bio_alloc(GFP_NOFS, 0); 3248387125fcSChris Mason if (!bio) 3249387125fcSChris Mason return -ENOMEM; 3250387125fcSChris Mason 3251387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3252387125fcSChris Mason bio->bi_bdev = device->bdev; 3253387125fcSChris Mason init_completion(&device->flush_wait); 3254387125fcSChris Mason bio->bi_private = &device->flush_wait; 3255387125fcSChris Mason device->flush_bio = bio; 3256387125fcSChris Mason 3257387125fcSChris Mason bio_get(bio); 325821adbd5cSStefan Behrens btrfsic_submit_bio(WRITE_FLUSH, bio); 3259387125fcSChris Mason 3260387125fcSChris Mason return 0; 3261387125fcSChris Mason } 3262387125fcSChris Mason 3263387125fcSChris Mason /* 3264387125fcSChris Mason * send an empty flush down to each device in parallel, 3265387125fcSChris Mason * then wait for them 3266387125fcSChris Mason */ 3267387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3268387125fcSChris Mason { 3269387125fcSChris Mason struct list_head *head; 3270387125fcSChris Mason struct btrfs_device *dev; 32715af3e8ccSStefan Behrens int errors_send = 0; 32725af3e8ccSStefan Behrens int errors_wait = 0; 3273387125fcSChris Mason int ret; 3274387125fcSChris Mason 3275387125fcSChris Mason /* send down all the barriers */ 3276387125fcSChris Mason head = &info->fs_devices->devices; 3277387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3278f88ba6a2SHidetoshi Seto if (dev->missing) 3279f88ba6a2SHidetoshi Seto continue; 3280387125fcSChris Mason if (!dev->bdev) { 32815af3e8ccSStefan Behrens errors_send++; 3282387125fcSChris Mason continue; 3283387125fcSChris Mason } 3284387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3285387125fcSChris Mason continue; 3286387125fcSChris Mason 3287387125fcSChris Mason ret = write_dev_flush(dev, 0); 3288387125fcSChris Mason if (ret) 32895af3e8ccSStefan Behrens errors_send++; 3290387125fcSChris Mason } 3291387125fcSChris Mason 3292387125fcSChris Mason /* wait for all the barriers */ 3293387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3294f88ba6a2SHidetoshi Seto if (dev->missing) 3295f88ba6a2SHidetoshi Seto continue; 3296387125fcSChris Mason if (!dev->bdev) { 32975af3e8ccSStefan Behrens errors_wait++; 3298387125fcSChris Mason continue; 3299387125fcSChris Mason } 3300387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3301387125fcSChris Mason continue; 3302387125fcSChris Mason 3303387125fcSChris Mason ret = write_dev_flush(dev, 1); 3304387125fcSChris Mason if (ret) 33055af3e8ccSStefan Behrens errors_wait++; 3306387125fcSChris Mason } 33075af3e8ccSStefan Behrens if (errors_send > info->num_tolerated_disk_barrier_failures || 33085af3e8ccSStefan Behrens errors_wait > info->num_tolerated_disk_barrier_failures) 3309387125fcSChris Mason return -EIO; 3310387125fcSChris Mason return 0; 3311387125fcSChris Mason } 3312387125fcSChris Mason 33135af3e8ccSStefan Behrens int btrfs_calc_num_tolerated_disk_barrier_failures( 33145af3e8ccSStefan Behrens struct btrfs_fs_info *fs_info) 33155af3e8ccSStefan Behrens { 33165af3e8ccSStefan Behrens struct btrfs_ioctl_space_info space; 33175af3e8ccSStefan Behrens struct btrfs_space_info *sinfo; 33185af3e8ccSStefan Behrens u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 33195af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_SYSTEM, 33205af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_METADATA, 33215af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 33225af3e8ccSStefan Behrens int num_types = 4; 33235af3e8ccSStefan Behrens int i; 33245af3e8ccSStefan Behrens int c; 33255af3e8ccSStefan Behrens int num_tolerated_disk_barrier_failures = 33265af3e8ccSStefan Behrens (int)fs_info->fs_devices->num_devices; 33275af3e8ccSStefan Behrens 33285af3e8ccSStefan Behrens for (i = 0; i < num_types; i++) { 33295af3e8ccSStefan Behrens struct btrfs_space_info *tmp; 33305af3e8ccSStefan Behrens 33315af3e8ccSStefan Behrens sinfo = NULL; 33325af3e8ccSStefan Behrens rcu_read_lock(); 33335af3e8ccSStefan Behrens list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { 33345af3e8ccSStefan Behrens if (tmp->flags == types[i]) { 33355af3e8ccSStefan Behrens sinfo = tmp; 33365af3e8ccSStefan Behrens break; 33375af3e8ccSStefan Behrens } 33385af3e8ccSStefan Behrens } 33395af3e8ccSStefan Behrens rcu_read_unlock(); 33405af3e8ccSStefan Behrens 33415af3e8ccSStefan Behrens if (!sinfo) 33425af3e8ccSStefan Behrens continue; 33435af3e8ccSStefan Behrens 33445af3e8ccSStefan Behrens down_read(&sinfo->groups_sem); 33455af3e8ccSStefan Behrens for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 33465af3e8ccSStefan Behrens if (!list_empty(&sinfo->block_groups[c])) { 33475af3e8ccSStefan Behrens u64 flags; 33485af3e8ccSStefan Behrens 33495af3e8ccSStefan Behrens btrfs_get_block_group_info( 33505af3e8ccSStefan Behrens &sinfo->block_groups[c], &space); 33515af3e8ccSStefan Behrens if (space.total_bytes == 0 || 33525af3e8ccSStefan Behrens space.used_bytes == 0) 33535af3e8ccSStefan Behrens continue; 33545af3e8ccSStefan Behrens flags = space.flags; 33555af3e8ccSStefan Behrens /* 33565af3e8ccSStefan Behrens * return 33575af3e8ccSStefan Behrens * 0: if dup, single or RAID0 is configured for 33585af3e8ccSStefan Behrens * any of metadata, system or data, else 33595af3e8ccSStefan Behrens * 1: if RAID5 is configured, or if RAID1 or 33605af3e8ccSStefan Behrens * RAID10 is configured and only two mirrors 33615af3e8ccSStefan Behrens * are used, else 33625af3e8ccSStefan Behrens * 2: if RAID6 is configured, else 33635af3e8ccSStefan Behrens * num_mirrors - 1: if RAID1 or RAID10 is 33645af3e8ccSStefan Behrens * configured and more than 33655af3e8ccSStefan Behrens * 2 mirrors are used. 33665af3e8ccSStefan Behrens */ 33675af3e8ccSStefan Behrens if (num_tolerated_disk_barrier_failures > 0 && 33685af3e8ccSStefan Behrens ((flags & (BTRFS_BLOCK_GROUP_DUP | 33695af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_RAID0)) || 33705af3e8ccSStefan Behrens ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) 33715af3e8ccSStefan Behrens == 0))) 33725af3e8ccSStefan Behrens num_tolerated_disk_barrier_failures = 0; 337353b381b3SDavid Woodhouse else if (num_tolerated_disk_barrier_failures > 1) { 337453b381b3SDavid Woodhouse if (flags & (BTRFS_BLOCK_GROUP_RAID1 | 337553b381b3SDavid Woodhouse BTRFS_BLOCK_GROUP_RAID5 | 337653b381b3SDavid Woodhouse BTRFS_BLOCK_GROUP_RAID10)) { 33775af3e8ccSStefan Behrens num_tolerated_disk_barrier_failures = 1; 337853b381b3SDavid Woodhouse } else if (flags & 337915b0a89dSHenrik Nordvik BTRFS_BLOCK_GROUP_RAID6) { 338053b381b3SDavid Woodhouse num_tolerated_disk_barrier_failures = 2; 338153b381b3SDavid Woodhouse } 338253b381b3SDavid Woodhouse } 33835af3e8ccSStefan Behrens } 33845af3e8ccSStefan Behrens } 33855af3e8ccSStefan Behrens up_read(&sinfo->groups_sem); 33865af3e8ccSStefan Behrens } 33875af3e8ccSStefan Behrens 33885af3e8ccSStefan Behrens return num_tolerated_disk_barrier_failures; 33895af3e8ccSStefan Behrens } 33905af3e8ccSStefan Behrens 339148a3b636SEric Sandeen static int write_all_supers(struct btrfs_root *root, int max_mirrors) 3392f2984462SChris Mason { 3393e5e9a520SChris Mason struct list_head *head; 3394f2984462SChris Mason struct btrfs_device *dev; 3395a061fc8dSChris Mason struct btrfs_super_block *sb; 3396f2984462SChris Mason struct btrfs_dev_item *dev_item; 3397f2984462SChris Mason int ret; 3398f2984462SChris Mason int do_barriers; 3399a236aed1SChris Mason int max_errors; 3400a236aed1SChris Mason int total_errors = 0; 3401a061fc8dSChris Mason u64 flags; 3402f2984462SChris Mason 3403f2984462SChris Mason do_barriers = !btrfs_test_opt(root, NOBARRIER); 3404af31f5e5SChris Mason backup_super_roots(root->fs_info); 3405f2984462SChris Mason 34066c41761fSDavid Sterba sb = root->fs_info->super_for_commit; 3407a061fc8dSChris Mason dev_item = &sb->dev_item; 3408e5e9a520SChris Mason 3409174ba509SChris Mason mutex_lock(&root->fs_info->fs_devices->device_list_mutex); 3410e5e9a520SChris Mason head = &root->fs_info->fs_devices->devices; 3411d7306801SFilipe David Borba Manana max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; 3412387125fcSChris Mason 34135af3e8ccSStefan Behrens if (do_barriers) { 34145af3e8ccSStefan Behrens ret = barrier_all_devices(root->fs_info); 34155af3e8ccSStefan Behrens if (ret) { 34165af3e8ccSStefan Behrens mutex_unlock( 34175af3e8ccSStefan Behrens &root->fs_info->fs_devices->device_list_mutex); 34185af3e8ccSStefan Behrens btrfs_error(root->fs_info, ret, 34195af3e8ccSStefan Behrens "errors while submitting device barriers."); 34205af3e8ccSStefan Behrens return ret; 34215af3e8ccSStefan Behrens } 34225af3e8ccSStefan Behrens } 3423387125fcSChris Mason 34241f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3425dfe25020SChris Mason if (!dev->bdev) { 3426dfe25020SChris Mason total_errors++; 3427dfe25020SChris Mason continue; 3428dfe25020SChris Mason } 34292b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3430dfe25020SChris Mason continue; 3431dfe25020SChris Mason 34322b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3433a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3434a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 3435a061fc8dSChris Mason btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); 3436a061fc8dSChris Mason btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); 3437a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3438a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3439a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3440a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 34412b82032cSYan Zheng memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 3442a512bbf8SYan Zheng 3443a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3444a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3445f2984462SChris Mason 3446a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); 3447a236aed1SChris Mason if (ret) 3448a236aed1SChris Mason total_errors++; 3449f2984462SChris Mason } 3450a236aed1SChris Mason if (total_errors > max_errors) { 3451efe120a0SFrank Holton btrfs_err(root->fs_info, "%d errors while writing supers", 3452d397712bSChris Mason total_errors); 3453a724b436SStefan Behrens mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 345479787eaaSJeff Mahoney 34559d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 34569d565ba4SStefan Behrens btrfs_error(root->fs_info, -EIO, 34579d565ba4SStefan Behrens "%d errors while writing supers", total_errors); 34589d565ba4SStefan Behrens return -EIO; 3459a236aed1SChris Mason } 3460f2984462SChris Mason 3461a512bbf8SYan Zheng total_errors = 0; 34621f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3463dfe25020SChris Mason if (!dev->bdev) 3464dfe25020SChris Mason continue; 34652b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3466dfe25020SChris Mason continue; 3467dfe25020SChris Mason 3468a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); 3469a512bbf8SYan Zheng if (ret) 34701259ab75SChris Mason total_errors++; 3471f2984462SChris Mason } 3472174ba509SChris Mason mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3473a236aed1SChris Mason if (total_errors > max_errors) { 347479787eaaSJeff Mahoney btrfs_error(root->fs_info, -EIO, 347579787eaaSJeff Mahoney "%d errors while writing supers", total_errors); 347679787eaaSJeff Mahoney return -EIO; 3477a236aed1SChris Mason } 3478f2984462SChris Mason return 0; 3479f2984462SChris Mason } 3480f2984462SChris Mason 3481a512bbf8SYan Zheng int write_ctree_super(struct btrfs_trans_handle *trans, 3482a512bbf8SYan Zheng struct btrfs_root *root, int max_mirrors) 3483cfaa7295SChris Mason { 3484f570e757SRashika return write_all_supers(root, max_mirrors); 3485cfaa7295SChris Mason } 3486cfaa7295SChris Mason 3487cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3488cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3489cb517eabSMiao Xie struct btrfs_root *root) 34902619ba1fSChris Mason { 34914df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 34922619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 34932619ba1fSChris Mason (unsigned long)root->root_key.objectid); 34944df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 349576dda93cSYan, Zheng 349676dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 349776dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 349876dda93cSYan, Zheng 34991a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 35003321719eSLiu Bo btrfs_free_log(NULL, root); 35013321719eSLiu Bo 3502581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3503581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 35044df27c4dSYan, Zheng free_fs_root(root); 35054df27c4dSYan, Zheng } 35064df27c4dSYan, Zheng 35074df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root) 35084df27c4dSYan, Zheng { 350982d5902dSLi Zefan iput(root->cache_inode); 35104df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 35111cb048f5SFilipe David Borba Manana btrfs_free_block_rsv(root, root->orphan_block_rsv); 35121cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 35130ee5dc67SAl Viro if (root->anon_dev) 35140ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 35158257b2dcSMiao Xie if (root->subv_writers) 35168257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 35175f39d397SChris Mason free_extent_buffer(root->node); 35185f39d397SChris Mason free_extent_buffer(root->commit_root); 3519581bb050SLi Zefan kfree(root->free_ino_ctl); 3520581bb050SLi Zefan kfree(root->free_ino_pinned); 352158176a96SJosef Bacik kfree(root->name); 3522b0feb9d9SMiao Xie btrfs_put_fs_root(root); 35232619ba1fSChris Mason } 35242619ba1fSChris Mason 3525cb517eabSMiao Xie void btrfs_free_fs_root(struct btrfs_root *root) 3526cb517eabSMiao Xie { 3527cb517eabSMiao Xie free_fs_root(root); 3528eb60ceacSChris Mason } 3529eb60ceacSChris Mason 3530c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3531c146afadSYan Zheng { 3532c146afadSYan Zheng u64 root_objectid = 0; 3533c146afadSYan Zheng struct btrfs_root *gang[8]; 353465d33fd7SQu Wenruo int i = 0; 353565d33fd7SQu Wenruo int err = 0; 353665d33fd7SQu Wenruo unsigned int ret = 0; 353765d33fd7SQu Wenruo int index; 3538c146afadSYan Zheng 3539c146afadSYan Zheng while (1) { 354065d33fd7SQu Wenruo index = srcu_read_lock(&fs_info->subvol_srcu); 3541c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3542c146afadSYan Zheng (void **)gang, root_objectid, 3543c146afadSYan Zheng ARRAY_SIZE(gang)); 354465d33fd7SQu Wenruo if (!ret) { 354565d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 3546c146afadSYan Zheng break; 354765d33fd7SQu Wenruo } 35485d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 354966b4ffd1SJosef Bacik 355065d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 355165d33fd7SQu Wenruo /* Avoid to grab roots in dead_roots */ 355265d33fd7SQu Wenruo if (btrfs_root_refs(&gang[i]->root_item) == 0) { 355365d33fd7SQu Wenruo gang[i] = NULL; 355465d33fd7SQu Wenruo continue; 355565d33fd7SQu Wenruo } 355665d33fd7SQu Wenruo /* grab all the search result for later use */ 355765d33fd7SQu Wenruo gang[i] = btrfs_grab_fs_root(gang[i]); 355865d33fd7SQu Wenruo } 355965d33fd7SQu Wenruo srcu_read_unlock(&fs_info->subvol_srcu, index); 356065d33fd7SQu Wenruo 356165d33fd7SQu Wenruo for (i = 0; i < ret; i++) { 356265d33fd7SQu Wenruo if (!gang[i]) 356365d33fd7SQu Wenruo continue; 3564c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 356566b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 356666b4ffd1SJosef Bacik if (err) 356765d33fd7SQu Wenruo break; 356865d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 3569c146afadSYan Zheng } 3570c146afadSYan Zheng root_objectid++; 3571c146afadSYan Zheng } 357265d33fd7SQu Wenruo 357365d33fd7SQu Wenruo /* release the uncleaned roots due to error */ 357465d33fd7SQu Wenruo for (; i < ret; i++) { 357565d33fd7SQu Wenruo if (gang[i]) 357665d33fd7SQu Wenruo btrfs_put_fs_root(gang[i]); 357765d33fd7SQu Wenruo } 357865d33fd7SQu Wenruo return err; 3579c146afadSYan Zheng } 3580c146afadSYan Zheng 3581c146afadSYan Zheng int btrfs_commit_super(struct btrfs_root *root) 3582c146afadSYan Zheng { 3583c146afadSYan Zheng struct btrfs_trans_handle *trans; 3584c146afadSYan Zheng 3585c146afadSYan Zheng mutex_lock(&root->fs_info->cleaner_mutex); 358624bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 3587c146afadSYan Zheng mutex_unlock(&root->fs_info->cleaner_mutex); 35889d1a2a3aSDavid Sterba wake_up_process(root->fs_info->cleaner_kthread); 3589c71bf099SYan, Zheng 3590c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 3591c71bf099SYan, Zheng down_write(&root->fs_info->cleanup_work_sem); 3592c71bf099SYan, Zheng up_write(&root->fs_info->cleanup_work_sem); 3593c71bf099SYan, Zheng 35947a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 35953612b495STsutomu Itoh if (IS_ERR(trans)) 35963612b495STsutomu Itoh return PTR_ERR(trans); 3597d52c1bccSLiu Bo return btrfs_commit_transaction(trans, root); 3598c146afadSYan Zheng } 3599c146afadSYan Zheng 3600e20d96d6SChris Mason int close_ctree(struct btrfs_root *root) 3601eb60ceacSChris Mason { 36020f7d52f4SChris Mason struct btrfs_fs_info *fs_info = root->fs_info; 3603c146afadSYan Zheng int ret; 3604e089f05cSChris Mason 3605facda1e7SChris Mason fs_info->closing = 1; 3606a2135011SChris Mason smp_mb(); 3607a2135011SChris Mason 3608803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3609803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3610803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3611803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3612803b2f54SStefan Behrens 3613837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3614aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3615837d5b6eSIlya Dryomov 36168dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 36178dabb742SStefan Behrens 3618aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 36194cb5300bSChris Mason 36204cb5300bSChris Mason /* wait for any defraggers to finish */ 36214cb5300bSChris Mason wait_event(fs_info->transaction_wait, 36224cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 36234cb5300bSChris Mason 36244cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 362526176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 36264cb5300bSChris Mason 362721c7e756SMiao Xie cancel_work_sync(&fs_info->async_reclaim_work); 362821c7e756SMiao Xie 3629c146afadSYan Zheng if (!(fs_info->sb->s_flags & MS_RDONLY)) { 3630c146afadSYan Zheng ret = btrfs_commit_super(root); 3631d397712bSChris Mason if (ret) 3632efe120a0SFrank Holton btrfs_err(root->fs_info, "commit super ret %d", ret); 3633c146afadSYan Zheng } 3634ed2ff2cbSChris Mason 363587533c47SMiao Xie if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 363668ce9682SStefan Behrens btrfs_error_commit_super(root); 3637acce952bSliubo 3638e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3639e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 36408929ecfaSYan, Zheng 3641f25784b3SYan Zheng fs_info->closing = 2; 3642f25784b3SYan Zheng smp_mb(); 3643f25784b3SYan Zheng 3644bcef60f2SArne Jansen btrfs_free_qgroup_config(root->fs_info); 3645bcef60f2SArne Jansen 3646963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 3647efe120a0SFrank Holton btrfs_info(root->fs_info, "at unmount delalloc count %lld", 3648963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3649b0c68f8bSChris Mason } 365031153d81SYan Zheng 36515ac1d209SJeff Mahoney btrfs_sysfs_remove_one(fs_info); 36525ac1d209SJeff Mahoney 36530f7d52f4SChris Mason del_fs_roots(fs_info); 3654d10c5f31SChris Mason 36551a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 36561a4319ccSLiu Bo 36572b1360daSJosef Bacik btrfs_free_block_groups(fs_info); 36582b1360daSJosef Bacik 365996192499SJosef Bacik btrfs_stop_all_workers(fs_info); 366096192499SJosef Bacik 366113e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 36629ad6b7bcSChris Mason 366313e6c37bSJosef Bacik iput(fs_info->btree_inode); 3664d6bfde87SChris Mason 366521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 366621adbd5cSStefan Behrens if (btrfs_test_opt(root, CHECK_INTEGRITY)) 366721adbd5cSStefan Behrens btrfsic_unmount(root, fs_info->fs_devices); 366821adbd5cSStefan Behrens #endif 366921adbd5cSStefan Behrens 3670dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 36710b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3672b248a415SChris Mason 3673e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3674963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3675c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 367604160088SChris Mason bdi_destroy(&fs_info->bdi); 367776dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 36780b86a832SChris Mason 367953b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 368053b381b3SDavid Woodhouse 36811cb048f5SFilipe David Borba Manana btrfs_free_block_rsv(root, root->orphan_block_rsv); 36821cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 36831cb048f5SFilipe David Borba Manana 3684eb60ceacSChris Mason return 0; 3685eb60ceacSChris Mason } 3686eb60ceacSChris Mason 3687b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3688b9fab919SChris Mason int atomic) 3689ccd467d6SChris Mason { 36901259ab75SChris Mason int ret; 3691727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 36921259ab75SChris Mason 36930b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 36941259ab75SChris Mason if (!ret) 36951259ab75SChris Mason return ret; 36961259ab75SChris Mason 36971259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3698b9fab919SChris Mason parent_transid, atomic); 3699b9fab919SChris Mason if (ret == -EAGAIN) 3700b9fab919SChris Mason return ret; 37011259ab75SChris Mason return !ret; 37025f39d397SChris Mason } 37036702ed49SChris Mason 37045f39d397SChris Mason int btrfs_set_buffer_uptodate(struct extent_buffer *buf) 37055f39d397SChris Mason { 37060b32f4bbSJosef Bacik return set_extent_buffer_uptodate(buf); 37075f39d397SChris Mason } 37085f39d397SChris Mason 37095f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 37105f39d397SChris Mason { 371106ea65a3SJosef Bacik struct btrfs_root *root; 37125f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 3713b9473439SChris Mason int was_dirty; 3714b4ce94deSChris Mason 371506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 371606ea65a3SJosef Bacik /* 371706ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 371806ea65a3SJosef Bacik * enabled. Normal people shouldn't be marking dummy buffers as dirty 371906ea65a3SJosef Bacik * outside of the sanity tests. 372006ea65a3SJosef Bacik */ 372106ea65a3SJosef Bacik if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) 372206ea65a3SJosef Bacik return; 372306ea65a3SJosef Bacik #endif 372406ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 3725b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 372631b1a2bdSJulia Lawall if (transid != root->fs_info->generation) 372731b1a2bdSJulia Lawall WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " 3728d397712bSChris Mason "found %llu running %llu\n", 3729c1c9ff7cSGeert Uytterhoeven buf->start, transid, root->fs_info->generation); 37300b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 3731e2d84521SMiao Xie if (!was_dirty) 3732e2d84521SMiao Xie __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, 3733e2d84521SMiao Xie buf->len, 3734e2d84521SMiao Xie root->fs_info->dirty_metadata_batch); 3735eb60ceacSChris Mason } 3736eb60ceacSChris Mason 3737b53d3f5dSLiu Bo static void __btrfs_btree_balance_dirty(struct btrfs_root *root, 3738b53d3f5dSLiu Bo int flush_delayed) 373935b7e476SChris Mason { 3740188de649SChris Mason /* 3741188de649SChris Mason * looks as though older kernels can get into trouble with 3742188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 3743188de649SChris Mason */ 3744e2d84521SMiao Xie int ret; 3745d6bfde87SChris Mason 37466933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 3747d6bfde87SChris Mason return; 3748d6bfde87SChris Mason 3749b53d3f5dSLiu Bo if (flush_delayed) 375016cdcec7SMiao Xie btrfs_balance_delayed_items(root); 375116cdcec7SMiao Xie 3752e2d84521SMiao Xie ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, 3753e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 3754e2d84521SMiao Xie if (ret > 0) { 3755d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited( 3756d0e1d66bSNamjae Jeon root->fs_info->btree_inode->i_mapping); 375716cdcec7SMiao Xie } 375816cdcec7SMiao Xie return; 375916cdcec7SMiao Xie } 376016cdcec7SMiao Xie 3761b53d3f5dSLiu Bo void btrfs_btree_balance_dirty(struct btrfs_root *root) 376216cdcec7SMiao Xie { 3763b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 1); 376435b7e476SChris Mason } 3765b53d3f5dSLiu Bo 3766b53d3f5dSLiu Bo void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) 3767b53d3f5dSLiu Bo { 3768b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 0); 3769d6bfde87SChris Mason } 37706b80053dSChris Mason 3771ca7a79adSChris Mason int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 37726b80053dSChris Mason { 3773727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 37740b32f4bbSJosef Bacik return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 37756b80053dSChris Mason } 37760da5468fSChris Mason 3777fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 3778acce952bSliubo int read_only) 3779acce952bSliubo { 37801104a885SDavid Sterba /* 37811104a885SDavid Sterba * Placeholder for checks 37821104a885SDavid Sterba */ 3783fcd1f065SDavid Sterba return 0; 3784fcd1f065SDavid Sterba } 3785fcd1f065SDavid Sterba 378648a3b636SEric Sandeen static void btrfs_error_commit_super(struct btrfs_root *root) 3787acce952bSliubo { 3788acce952bSliubo mutex_lock(&root->fs_info->cleaner_mutex); 3789acce952bSliubo btrfs_run_delayed_iputs(root); 3790acce952bSliubo mutex_unlock(&root->fs_info->cleaner_mutex); 3791acce952bSliubo 3792acce952bSliubo down_write(&root->fs_info->cleanup_work_sem); 3793acce952bSliubo up_write(&root->fs_info->cleanup_work_sem); 3794acce952bSliubo 3795acce952bSliubo /* cleanup FS via transaction */ 3796acce952bSliubo btrfs_cleanup_transaction(root); 3797acce952bSliubo } 3798acce952bSliubo 3799569e0f35SJosef Bacik static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, 3800569e0f35SJosef Bacik struct btrfs_root *root) 3801acce952bSliubo { 3802acce952bSliubo struct btrfs_inode *btrfs_inode; 3803acce952bSliubo struct list_head splice; 3804acce952bSliubo 3805acce952bSliubo INIT_LIST_HEAD(&splice); 3806acce952bSliubo 3807acce952bSliubo mutex_lock(&root->fs_info->ordered_operations_mutex); 3808199c2a9cSMiao Xie spin_lock(&root->fs_info->ordered_root_lock); 3809acce952bSliubo 3810569e0f35SJosef Bacik list_splice_init(&t->ordered_operations, &splice); 3811acce952bSliubo while (!list_empty(&splice)) { 3812acce952bSliubo btrfs_inode = list_entry(splice.next, struct btrfs_inode, 3813acce952bSliubo ordered_operations); 3814acce952bSliubo 3815acce952bSliubo list_del_init(&btrfs_inode->ordered_operations); 3816199c2a9cSMiao Xie spin_unlock(&root->fs_info->ordered_root_lock); 3817acce952bSliubo 3818acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 3819b216cbfbSMiao Xie 3820199c2a9cSMiao Xie spin_lock(&root->fs_info->ordered_root_lock); 3821acce952bSliubo } 3822acce952bSliubo 3823199c2a9cSMiao Xie spin_unlock(&root->fs_info->ordered_root_lock); 3824acce952bSliubo mutex_unlock(&root->fs_info->ordered_operations_mutex); 3825acce952bSliubo } 3826acce952bSliubo 3827143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 3828acce952bSliubo { 3829acce952bSliubo struct btrfs_ordered_extent *ordered; 3830acce952bSliubo 3831199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 3832779880efSJosef Bacik /* 3833779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 3834779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 3835779880efSJosef Bacik */ 3836199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 3837779880efSJosef Bacik root_extent_list) 3838779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 3839199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 3840199c2a9cSMiao Xie } 3841199c2a9cSMiao Xie 3842199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 3843199c2a9cSMiao Xie { 3844199c2a9cSMiao Xie struct btrfs_root *root; 3845199c2a9cSMiao Xie struct list_head splice; 3846199c2a9cSMiao Xie 3847199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 3848199c2a9cSMiao Xie 3849199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 3850199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 3851199c2a9cSMiao Xie while (!list_empty(&splice)) { 3852199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 3853199c2a9cSMiao Xie ordered_root); 38541de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 38551de2cfdeSJosef Bacik &fs_info->ordered_roots); 3856199c2a9cSMiao Xie 38572a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 3858199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 3859199c2a9cSMiao Xie 38602a85d9caSLiu Bo cond_resched(); 38612a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 3862199c2a9cSMiao Xie } 3863199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 3864acce952bSliubo } 3865acce952bSliubo 386635a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 3867acce952bSliubo struct btrfs_root *root) 3868acce952bSliubo { 3869acce952bSliubo struct rb_node *node; 3870acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 3871acce952bSliubo struct btrfs_delayed_ref_node *ref; 3872acce952bSliubo int ret = 0; 3873acce952bSliubo 3874acce952bSliubo delayed_refs = &trans->delayed_refs; 3875acce952bSliubo 3876acce952bSliubo spin_lock(&delayed_refs->lock); 3877d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 3878cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 3879efe120a0SFrank Holton btrfs_info(root->fs_info, "delayed_refs has NO entry"); 3880acce952bSliubo return ret; 3881acce952bSliubo } 3882acce952bSliubo 3883d7df2c79SJosef Bacik while ((node = rb_first(&delayed_refs->href_root)) != NULL) { 3884d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 3885e78417d1SJosef Bacik bool pin_bytes = false; 3886acce952bSliubo 3887d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 3888d7df2c79SJosef Bacik href_node); 3889b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 3890d7df2c79SJosef Bacik atomic_inc(&head->node.refs); 389179787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 3892b939d1abSJosef Bacik 3893acce952bSliubo mutex_lock(&head->mutex); 3894b939d1abSJosef Bacik mutex_unlock(&head->mutex); 3895d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 3896e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 3897b939d1abSJosef Bacik continue; 3898b939d1abSJosef Bacik } 3899d7df2c79SJosef Bacik spin_lock(&head->lock); 3900d7df2c79SJosef Bacik while ((node = rb_first(&head->ref_root)) != NULL) { 3901d7df2c79SJosef Bacik ref = rb_entry(node, struct btrfs_delayed_ref_node, 3902d7df2c79SJosef Bacik rb_node); 3903d7df2c79SJosef Bacik ref->in_tree = 0; 3904d7df2c79SJosef Bacik rb_erase(&ref->rb_node, &head->ref_root); 3905d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 3906d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 3907d7df2c79SJosef Bacik } 390854067ae9SJosef Bacik if (head->must_insert_reserved) 3909e78417d1SJosef Bacik pin_bytes = true; 391078a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 3911acce952bSliubo delayed_refs->num_heads--; 3912d7df2c79SJosef Bacik if (head->processing == 0) 3913acce952bSliubo delayed_refs->num_heads_ready--; 3914d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 3915d7df2c79SJosef Bacik head->node.in_tree = 0; 3916c46effa6SLiu Bo rb_erase(&head->href_node, &delayed_refs->href_root); 3917d7df2c79SJosef Bacik spin_unlock(&head->lock); 3918acce952bSliubo spin_unlock(&delayed_refs->lock); 3919e78417d1SJosef Bacik mutex_unlock(&head->mutex); 3920acce952bSliubo 3921d7df2c79SJosef Bacik if (pin_bytes) 3922d7df2c79SJosef Bacik btrfs_pin_extent(root, head->node.bytenr, 3923d7df2c79SJosef Bacik head->node.num_bytes, 1); 3924d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 3925acce952bSliubo cond_resched(); 3926acce952bSliubo spin_lock(&delayed_refs->lock); 3927acce952bSliubo } 3928acce952bSliubo 3929acce952bSliubo spin_unlock(&delayed_refs->lock); 3930acce952bSliubo 3931acce952bSliubo return ret; 3932acce952bSliubo } 3933acce952bSliubo 3934143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 3935acce952bSliubo { 3936acce952bSliubo struct btrfs_inode *btrfs_inode; 3937acce952bSliubo struct list_head splice; 3938acce952bSliubo 3939acce952bSliubo INIT_LIST_HEAD(&splice); 3940acce952bSliubo 3941eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 3942eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 3943acce952bSliubo 3944acce952bSliubo while (!list_empty(&splice)) { 3945eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 3946acce952bSliubo delalloc_inodes); 3947acce952bSliubo 3948acce952bSliubo list_del_init(&btrfs_inode->delalloc_inodes); 3949df0af1a5SMiao Xie clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, 3950df0af1a5SMiao Xie &btrfs_inode->runtime_flags); 3951eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 3952acce952bSliubo 3953acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 3954b216cbfbSMiao Xie 3955eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 3956acce952bSliubo } 3957acce952bSliubo 3958eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 3959eb73c1b7SMiao Xie } 3960eb73c1b7SMiao Xie 3961eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 3962eb73c1b7SMiao Xie { 3963eb73c1b7SMiao Xie struct btrfs_root *root; 3964eb73c1b7SMiao Xie struct list_head splice; 3965eb73c1b7SMiao Xie 3966eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 3967eb73c1b7SMiao Xie 3968eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 3969eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 3970eb73c1b7SMiao Xie while (!list_empty(&splice)) { 3971eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 3972eb73c1b7SMiao Xie delalloc_root); 3973eb73c1b7SMiao Xie list_del_init(&root->delalloc_root); 3974eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 3975eb73c1b7SMiao Xie BUG_ON(!root); 3976eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 3977eb73c1b7SMiao Xie 3978eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 3979eb73c1b7SMiao Xie btrfs_put_fs_root(root); 3980eb73c1b7SMiao Xie 3981eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 3982eb73c1b7SMiao Xie } 3983eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 3984acce952bSliubo } 3985acce952bSliubo 3986acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 3987acce952bSliubo struct extent_io_tree *dirty_pages, 3988acce952bSliubo int mark) 3989acce952bSliubo { 3990acce952bSliubo int ret; 3991acce952bSliubo struct extent_buffer *eb; 3992acce952bSliubo u64 start = 0; 3993acce952bSliubo u64 end; 3994acce952bSliubo 3995acce952bSliubo while (1) { 3996acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 3997e6138876SJosef Bacik mark, NULL); 3998acce952bSliubo if (ret) 3999acce952bSliubo break; 4000acce952bSliubo 4001acce952bSliubo clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); 4002acce952bSliubo while (start <= end) { 4003fd8b2b61SJosef Bacik eb = btrfs_find_tree_block(root, start, 4004fd8b2b61SJosef Bacik root->leafsize); 400569a85bd8SJosef Bacik start += root->leafsize; 4006fd8b2b61SJosef Bacik if (!eb) 4007acce952bSliubo continue; 4008fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 4009acce952bSliubo 4010fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 4011fd8b2b61SJosef Bacik &eb->bflags)) 4012fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 4013fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 4014acce952bSliubo } 4015acce952bSliubo } 4016acce952bSliubo 4017acce952bSliubo return ret; 4018acce952bSliubo } 4019acce952bSliubo 4020acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 4021acce952bSliubo struct extent_io_tree *pinned_extents) 4022acce952bSliubo { 4023acce952bSliubo struct extent_io_tree *unpin; 4024acce952bSliubo u64 start; 4025acce952bSliubo u64 end; 4026acce952bSliubo int ret; 4027ed0eaa14SLiu Bo bool loop = true; 4028acce952bSliubo 4029acce952bSliubo unpin = pinned_extents; 4030ed0eaa14SLiu Bo again: 4031acce952bSliubo while (1) { 4032acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4033e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4034acce952bSliubo if (ret) 4035acce952bSliubo break; 4036acce952bSliubo 4037acce952bSliubo /* opt_discard */ 40385378e607SLi Dongyang if (btrfs_test_opt(root, DISCARD)) 40395378e607SLi Dongyang ret = btrfs_error_discard_extent(root, start, 40405378e607SLi Dongyang end + 1 - start, 40415378e607SLi Dongyang NULL); 4042acce952bSliubo 4043acce952bSliubo clear_extent_dirty(unpin, start, end, GFP_NOFS); 4044acce952bSliubo btrfs_error_unpin_extent_range(root, start, end); 4045acce952bSliubo cond_resched(); 4046acce952bSliubo } 4047acce952bSliubo 4048ed0eaa14SLiu Bo if (loop) { 4049ed0eaa14SLiu Bo if (unpin == &root->fs_info->freed_extents[0]) 4050ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[1]; 4051ed0eaa14SLiu Bo else 4052ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[0]; 4053ed0eaa14SLiu Bo loop = false; 4054ed0eaa14SLiu Bo goto again; 4055ed0eaa14SLiu Bo } 4056ed0eaa14SLiu Bo 4057acce952bSliubo return 0; 4058acce952bSliubo } 4059acce952bSliubo 406049b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 406149b25e05SJeff Mahoney struct btrfs_root *root) 406249b25e05SJeff Mahoney { 4063724e2315SJosef Bacik btrfs_destroy_ordered_operations(cur_trans, root); 4064724e2315SJosef Bacik 406549b25e05SJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, root); 406649b25e05SJeff Mahoney 40674a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 406849b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_blocked_wait); 406949b25e05SJeff Mahoney 40704a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 407149b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_wait); 407249b25e05SJeff Mahoney 407367cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 407467cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 407567cde344SMiao Xie 407649b25e05SJeff Mahoney btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, 407749b25e05SJeff Mahoney EXTENT_DIRTY); 40786e841e32SLiu Bo btrfs_destroy_pinned_extent(root, 40796e841e32SLiu Bo root->fs_info->pinned_extents); 408049b25e05SJeff Mahoney 40814a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 40824a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 40834a9d8bdeSMiao Xie 408449b25e05SJeff Mahoney /* 408549b25e05SJeff Mahoney memset(cur_trans, 0, sizeof(*cur_trans)); 408649b25e05SJeff Mahoney kmem_cache_free(btrfs_transaction_cachep, cur_trans); 408749b25e05SJeff Mahoney */ 408849b25e05SJeff Mahoney } 408949b25e05SJeff Mahoney 409048a3b636SEric Sandeen static int btrfs_cleanup_transaction(struct btrfs_root *root) 4091acce952bSliubo { 4092acce952bSliubo struct btrfs_transaction *t; 4093acce952bSliubo 4094acce952bSliubo mutex_lock(&root->fs_info->transaction_kthread_mutex); 4095acce952bSliubo 4096a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4097724e2315SJosef Bacik while (!list_empty(&root->fs_info->trans_list)) { 4098724e2315SJosef Bacik t = list_first_entry(&root->fs_info->trans_list, 4099724e2315SJosef Bacik struct btrfs_transaction, list); 4100724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 4101724e2315SJosef Bacik atomic_inc(&t->use_count); 4102724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4103724e2315SJosef Bacik btrfs_wait_for_commit(root, t->transid); 4104724e2315SJosef Bacik btrfs_put_transaction(t); 4105724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4106724e2315SJosef Bacik continue; 4107724e2315SJosef Bacik } 4108724e2315SJosef Bacik if (t == root->fs_info->running_transaction) { 4109724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 4110724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4111724e2315SJosef Bacik /* 4112724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4113724e2315SJosef Bacik * handle open currently for this transaction. 4114724e2315SJosef Bacik */ 4115724e2315SJosef Bacik wait_event(t->writer_wait, 4116724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4117724e2315SJosef Bacik } else { 4118724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4119724e2315SJosef Bacik } 4120724e2315SJosef Bacik btrfs_cleanup_one_transaction(t, root); 4121724e2315SJosef Bacik 4122724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4123724e2315SJosef Bacik if (t == root->fs_info->running_transaction) 4124ac673879SMiao Xie root->fs_info->running_transaction = NULL; 4125724e2315SJosef Bacik list_del_init(&t->list); 4126a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4127a4abeea4SJosef Bacik 4128724e2315SJosef Bacik btrfs_put_transaction(t); 4129724e2315SJosef Bacik trace_btrfs_transaction_commit(root); 4130724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4131724e2315SJosef Bacik } 4132724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4133199c2a9cSMiao Xie btrfs_destroy_all_ordered_extents(root->fs_info); 413467cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 413567cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 4136724e2315SJosef Bacik btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); 4137eb73c1b7SMiao Xie btrfs_destroy_all_delalloc_inodes(root->fs_info); 4138acce952bSliubo mutex_unlock(&root->fs_info->transaction_kthread_mutex); 4139acce952bSliubo 4140acce952bSliubo return 0; 4141acce952bSliubo } 4142acce952bSliubo 4143d1310b2eSChris Mason static struct extent_io_ops btree_extent_io_ops = { 4144ce9adaa5SChris Mason .readpage_end_io_hook = btree_readpage_end_io_hook, 41454bb31e92SArne Jansen .readpage_io_failed_hook = btree_io_failed_hook, 41460b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 4147239b14b3SChris Mason /* note we're sharing with inode.c for the merge bio hook */ 4148239b14b3SChris Mason .merge_bio_hook = btrfs_merge_bio_hook, 41490da5468fSChris Mason }; 4150