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; 58d458b054SQu Wenruo 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; 89d458b054SQu Wenruo 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; 111d458b054SQu Wenruo 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 763d458b054SQu Wenruo 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 776d458b054SQu Wenruo 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 803d458b054SQu Wenruo 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 { 857ce3ed71aSChris Mason struct bio_vec *bvec = bio->bi_io_vec; 858ce3ed71aSChris Mason int bio_index = 0; 859ce3ed71aSChris Mason struct btrfs_root *root; 86079787eaaSJeff Mahoney int ret = 0; 861ce3ed71aSChris Mason 862ce3ed71aSChris Mason WARN_ON(bio->bi_vcnt <= 0); 863ce3ed71aSChris Mason while (bio_index < bio->bi_vcnt) { 864ce3ed71aSChris Mason root = BTRFS_I(bvec->bv_page->mapping->host)->root; 86579787eaaSJeff Mahoney ret = csum_dirty_buffer(root, bvec->bv_page); 86679787eaaSJeff Mahoney if (ret) 86779787eaaSJeff Mahoney break; 868ce3ed71aSChris Mason bio_index++; 869ce3ed71aSChris Mason bvec++; 870ce3ed71aSChris Mason } 87179787eaaSJeff Mahoney return ret; 872ce3ed71aSChris Mason } 873ce3ed71aSChris Mason 8744a69a410SChris Mason static int __btree_submit_bio_start(struct inode *inode, int rw, 8754a69a410SChris Mason struct bio *bio, int mirror_num, 876eaf25d93SChris Mason unsigned long bio_flags, 877eaf25d93SChris Mason u64 bio_offset) 87822c59948SChris Mason { 8798b712842SChris Mason /* 8808b712842SChris Mason * when we're called for a write, we're already in the async 8815443be45SChris Mason * submission context. Just jump into btrfs_map_bio 8828b712842SChris Mason */ 88379787eaaSJeff Mahoney return btree_csum_one_bio(bio); 88422c59948SChris Mason } 88522c59948SChris Mason 8864a69a410SChris Mason static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, 887eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 888eaf25d93SChris Mason u64 bio_offset) 8894a69a410SChris Mason { 89061891923SStefan Behrens int ret; 89161891923SStefan Behrens 8928b712842SChris Mason /* 8934a69a410SChris Mason * when we're called for a write, we're already in the async 8944a69a410SChris Mason * submission context. Just jump into btrfs_map_bio 8958b712842SChris Mason */ 89661891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); 89761891923SStefan Behrens if (ret) 89861891923SStefan Behrens bio_endio(bio, ret); 89961891923SStefan Behrens return ret; 9000b86a832SChris Mason } 9010b86a832SChris Mason 902de0022b9SJosef Bacik static int check_async_write(struct inode *inode, unsigned long bio_flags) 903de0022b9SJosef Bacik { 904de0022b9SJosef Bacik if (bio_flags & EXTENT_BIO_TREE_LOG) 905de0022b9SJosef Bacik return 0; 906de0022b9SJosef Bacik #ifdef CONFIG_X86 907de0022b9SJosef Bacik if (cpu_has_xmm4_2) 908de0022b9SJosef Bacik return 0; 909de0022b9SJosef Bacik #endif 910de0022b9SJosef Bacik return 1; 911de0022b9SJosef Bacik } 912de0022b9SJosef Bacik 91344b8bd7eSChris Mason static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, 914eaf25d93SChris Mason int mirror_num, unsigned long bio_flags, 915eaf25d93SChris Mason u64 bio_offset) 91644b8bd7eSChris Mason { 917de0022b9SJosef Bacik int async = check_async_write(inode, bio_flags); 9184a69a410SChris Mason int ret; 919cad321adSChris Mason 9207b6d91daSChristoph Hellwig if (!(rw & REQ_WRITE)) { 921cad321adSChris Mason /* 922cad321adSChris Mason * called for a read, do the setup so that checksum validation 923cad321adSChris Mason * can happen in the async kernel threads 924cad321adSChris Mason */ 925f3f266abSChris Mason ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, 926f3f266abSChris Mason bio, 1); 9271d4284bdSChris Mason if (ret) 92861891923SStefan Behrens goto out_w_error; 92961891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 9306f3577bdSChris Mason mirror_num, 0); 931de0022b9SJosef Bacik } else if (!async) { 932de0022b9SJosef Bacik ret = btree_csum_one_bio(bio); 933de0022b9SJosef Bacik if (ret) 93461891923SStefan Behrens goto out_w_error; 93561891923SStefan Behrens ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, 936de0022b9SJosef Bacik mirror_num, 0); 93761891923SStefan Behrens } else { 938cad321adSChris Mason /* 93961891923SStefan Behrens * kthread helpers are used to submit writes so that 94061891923SStefan Behrens * checksumming can happen in parallel across all CPUs 941cad321adSChris Mason */ 94261891923SStefan Behrens ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, 943c8b97818SChris Mason inode, rw, bio, mirror_num, 0, 944eaf25d93SChris Mason bio_offset, 9454a69a410SChris Mason __btree_submit_bio_start, 9464a69a410SChris Mason __btree_submit_bio_done); 94744b8bd7eSChris Mason } 94844b8bd7eSChris Mason 94961891923SStefan Behrens if (ret) { 95061891923SStefan Behrens out_w_error: 95161891923SStefan Behrens bio_endio(bio, ret); 95261891923SStefan Behrens } 95361891923SStefan Behrens return ret; 95461891923SStefan Behrens } 95561891923SStefan Behrens 9563dd1462eSJan Beulich #ifdef CONFIG_MIGRATION 957784b4e29SChris Mason static int btree_migratepage(struct address_space *mapping, 958a6bc32b8SMel Gorman struct page *newpage, struct page *page, 959a6bc32b8SMel Gorman enum migrate_mode mode) 960784b4e29SChris Mason { 961784b4e29SChris Mason /* 962784b4e29SChris Mason * we can't safely write a btree page from here, 963784b4e29SChris Mason * we haven't done the locking hook 964784b4e29SChris Mason */ 965784b4e29SChris Mason if (PageDirty(page)) 966784b4e29SChris Mason return -EAGAIN; 967784b4e29SChris Mason /* 968784b4e29SChris Mason * Buffers may be managed in a filesystem specific way. 969784b4e29SChris Mason * We must have no buffers or drop them. 970784b4e29SChris Mason */ 971784b4e29SChris Mason if (page_has_private(page) && 972784b4e29SChris Mason !try_to_release_page(page, GFP_KERNEL)) 973784b4e29SChris Mason return -EAGAIN; 974a6bc32b8SMel Gorman return migrate_page(mapping, newpage, page, mode); 975784b4e29SChris Mason } 9763dd1462eSJan Beulich #endif 977784b4e29SChris Mason 9780da5468fSChris Mason 9790da5468fSChris Mason static int btree_writepages(struct address_space *mapping, 9800da5468fSChris Mason struct writeback_control *wbc) 9810da5468fSChris Mason { 982e2d84521SMiao Xie struct btrfs_fs_info *fs_info; 983e2d84521SMiao Xie int ret; 984e2d84521SMiao Xie 985d8d5f3e1SChris Mason if (wbc->sync_mode == WB_SYNC_NONE) { 986448d640bSChris Mason 987448d640bSChris Mason if (wbc->for_kupdate) 988448d640bSChris Mason return 0; 989448d640bSChris Mason 990e2d84521SMiao Xie fs_info = BTRFS_I(mapping->host)->root->fs_info; 991b9473439SChris Mason /* this is a bit racy, but that's ok */ 992e2d84521SMiao Xie ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes, 993e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 994e2d84521SMiao Xie if (ret < 0) 995793955bcSChris Mason return 0; 996793955bcSChris Mason } 9970b32f4bbSJosef Bacik return btree_write_cache_pages(mapping, wbc); 9980da5468fSChris Mason } 9990da5468fSChris Mason 1000b2950863SChristoph Hellwig static int btree_readpage(struct file *file, struct page *page) 10015f39d397SChris Mason { 1002d1310b2eSChris Mason struct extent_io_tree *tree; 1003d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10048ddc7d9cSJan Schmidt return extent_read_full_page(tree, page, btree_get_extent, 0); 10055f39d397SChris Mason } 10065f39d397SChris Mason 100770dec807SChris Mason static int btree_releasepage(struct page *page, gfp_t gfp_flags) 10085f39d397SChris Mason { 100998509cfcSChris Mason if (PageWriteback(page) || PageDirty(page)) 101098509cfcSChris Mason return 0; 10110c4e538bSDavid Sterba 1012f7a52a40SDavid Sterba return try_release_extent_buffer(page); 1013d98237b3SChris Mason } 1014d98237b3SChris Mason 1015d47992f8SLukas Czerner static void btree_invalidatepage(struct page *page, unsigned int offset, 1016d47992f8SLukas Czerner unsigned int length) 1017d98237b3SChris Mason { 1018d1310b2eSChris Mason struct extent_io_tree *tree; 1019d1310b2eSChris Mason tree = &BTRFS_I(page->mapping->host)->io_tree; 10205f39d397SChris Mason extent_invalidatepage(tree, page, offset); 10215f39d397SChris Mason btree_releasepage(page, GFP_NOFS); 10229ad6b7bcSChris Mason if (PagePrivate(page)) { 1023efe120a0SFrank Holton btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info, 1024efe120a0SFrank Holton "page private not zero on page %llu", 1025efe120a0SFrank Holton (unsigned long long)page_offset(page)); 10269ad6b7bcSChris Mason ClearPagePrivate(page); 10279ad6b7bcSChris Mason set_page_private(page, 0); 10289ad6b7bcSChris Mason page_cache_release(page); 10299ad6b7bcSChris Mason } 1030d98237b3SChris Mason } 1031d98237b3SChris Mason 10320b32f4bbSJosef Bacik static int btree_set_page_dirty(struct page *page) 10330b32f4bbSJosef Bacik { 1034bb146eb2SJosef Bacik #ifdef DEBUG 10350b32f4bbSJosef Bacik struct extent_buffer *eb; 10360b32f4bbSJosef Bacik 10370b32f4bbSJosef Bacik BUG_ON(!PagePrivate(page)); 10380b32f4bbSJosef Bacik eb = (struct extent_buffer *)page->private; 10390b32f4bbSJosef Bacik BUG_ON(!eb); 10400b32f4bbSJosef Bacik BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); 10410b32f4bbSJosef Bacik BUG_ON(!atomic_read(&eb->refs)); 10420b32f4bbSJosef Bacik btrfs_assert_tree_locked(eb); 1043bb146eb2SJosef Bacik #endif 10440b32f4bbSJosef Bacik return __set_page_dirty_nobuffers(page); 10450b32f4bbSJosef Bacik } 10460b32f4bbSJosef Bacik 10477f09410bSAlexey Dobriyan static const struct address_space_operations btree_aops = { 1048d98237b3SChris Mason .readpage = btree_readpage, 10490da5468fSChris Mason .writepages = btree_writepages, 10505f39d397SChris Mason .releasepage = btree_releasepage, 10515f39d397SChris Mason .invalidatepage = btree_invalidatepage, 10525a92bc88SChris Mason #ifdef CONFIG_MIGRATION 1053784b4e29SChris Mason .migratepage = btree_migratepage, 10545a92bc88SChris Mason #endif 10550b32f4bbSJosef Bacik .set_page_dirty = btree_set_page_dirty, 1056d98237b3SChris Mason }; 1057123abc88SChris Mason 1058ca7a79adSChris Mason int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1059ca7a79adSChris Mason u64 parent_transid) 1060090d1875SChris Mason { 10615f39d397SChris Mason struct extent_buffer *buf = NULL; 10625f39d397SChris Mason struct inode *btree_inode = root->fs_info->btree_inode; 1063de428b63SChris Mason int ret = 0; 1064090d1875SChris Mason 1065db94535dSChris Mason buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 10665f39d397SChris Mason if (!buf) 1067090d1875SChris Mason return 0; 1068d1310b2eSChris Mason read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 1069bb82ab88SArne Jansen buf, 0, WAIT_NONE, btree_get_extent, 0); 10705f39d397SChris Mason free_extent_buffer(buf); 1071de428b63SChris Mason return ret; 1072090d1875SChris Mason } 1073090d1875SChris Mason 1074ab0fff03SArne Jansen int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize, 1075ab0fff03SArne Jansen int mirror_num, struct extent_buffer **eb) 1076ab0fff03SArne Jansen { 1077ab0fff03SArne Jansen struct extent_buffer *buf = NULL; 1078ab0fff03SArne Jansen struct inode *btree_inode = root->fs_info->btree_inode; 1079ab0fff03SArne Jansen struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1080ab0fff03SArne Jansen int ret; 1081ab0fff03SArne Jansen 1082ab0fff03SArne Jansen buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 1083ab0fff03SArne Jansen if (!buf) 1084ab0fff03SArne Jansen return 0; 1085ab0fff03SArne Jansen 1086ab0fff03SArne Jansen set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags); 1087ab0fff03SArne Jansen 1088ab0fff03SArne Jansen ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK, 1089ab0fff03SArne Jansen btree_get_extent, mirror_num); 1090ab0fff03SArne Jansen if (ret) { 1091ab0fff03SArne Jansen free_extent_buffer(buf); 1092ab0fff03SArne Jansen return ret; 1093ab0fff03SArne Jansen } 1094ab0fff03SArne Jansen 1095ab0fff03SArne Jansen if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) { 1096ab0fff03SArne Jansen free_extent_buffer(buf); 1097ab0fff03SArne Jansen return -EIO; 10980b32f4bbSJosef Bacik } else if (extent_buffer_uptodate(buf)) { 1099ab0fff03SArne Jansen *eb = buf; 1100ab0fff03SArne Jansen } else { 1101ab0fff03SArne Jansen free_extent_buffer(buf); 1102ab0fff03SArne Jansen } 1103ab0fff03SArne Jansen return 0; 1104ab0fff03SArne Jansen } 1105ab0fff03SArne Jansen 11060999df54SChris Mason struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, 11070999df54SChris Mason u64 bytenr, u32 blocksize) 11080999df54SChris Mason { 1109f28491e0SJosef Bacik return find_extent_buffer(root->fs_info, bytenr); 11100999df54SChris Mason } 11110999df54SChris Mason 11120999df54SChris Mason struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, 11130999df54SChris Mason u64 bytenr, u32 blocksize) 11140999df54SChris Mason { 1115f28491e0SJosef Bacik return alloc_extent_buffer(root->fs_info, bytenr, blocksize); 11160999df54SChris Mason } 11170999df54SChris Mason 11180999df54SChris Mason 1119e02119d5SChris Mason int btrfs_write_tree_block(struct extent_buffer *buf) 1120e02119d5SChris Mason { 1121727011e0SChris Mason return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start, 11228aa38c31SChristoph Hellwig buf->start + buf->len - 1); 1123e02119d5SChris Mason } 1124e02119d5SChris Mason 1125e02119d5SChris Mason int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) 1126e02119d5SChris Mason { 1127727011e0SChris Mason return filemap_fdatawait_range(buf->pages[0]->mapping, 11288aa38c31SChristoph Hellwig buf->start, buf->start + buf->len - 1); 1129e02119d5SChris Mason } 1130e02119d5SChris Mason 1131db94535dSChris Mason struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, 1132ca7a79adSChris Mason u32 blocksize, u64 parent_transid) 1133e20d96d6SChris Mason { 11345f39d397SChris Mason struct extent_buffer *buf = NULL; 113519c00ddcSChris Mason int ret; 113619c00ddcSChris Mason 1137db94535dSChris Mason buf = btrfs_find_create_tree_block(root, bytenr, blocksize); 11385f39d397SChris Mason if (!buf) 1139d98237b3SChris Mason return NULL; 1140e4204dedSChris Mason 1141ca7a79adSChris Mason ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 11420f0fe8f7SFilipe David Borba Manana if (ret) { 11430f0fe8f7SFilipe David Borba Manana free_extent_buffer(buf); 11440f0fe8f7SFilipe David Borba Manana return NULL; 11450f0fe8f7SFilipe David Borba Manana } 11465f39d397SChris Mason return buf; 1147ce9adaa5SChris Mason 1148eb60ceacSChris Mason } 1149eb60ceacSChris Mason 1150d5c13f92SJeff Mahoney void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, 11515f39d397SChris Mason struct extent_buffer *buf) 1152ed2ff2cbSChris Mason { 1153e2d84521SMiao Xie struct btrfs_fs_info *fs_info = root->fs_info; 1154e2d84521SMiao Xie 115555c69072SChris Mason if (btrfs_header_generation(buf) == 1156e2d84521SMiao Xie fs_info->running_transaction->transid) { 1157b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 1158b4ce94deSChris Mason 1159b9473439SChris Mason if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { 1160e2d84521SMiao Xie __percpu_counter_add(&fs_info->dirty_metadata_bytes, 1161e2d84521SMiao Xie -buf->len, 1162e2d84521SMiao Xie fs_info->dirty_metadata_batch); 1163b9473439SChris Mason /* ugh, clear_extent_buffer_dirty needs to lock the page */ 1164b9473439SChris Mason btrfs_set_lock_blocking(buf); 11650b32f4bbSJosef Bacik clear_extent_buffer_dirty(buf); 1166925baeddSChris Mason } 11675f39d397SChris Mason } 1168ed7b63ebSJosef Bacik } 11695f39d397SChris Mason 11708257b2dcSMiao Xie static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void) 11718257b2dcSMiao Xie { 11728257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 11738257b2dcSMiao Xie int ret; 11748257b2dcSMiao Xie 11758257b2dcSMiao Xie writers = kmalloc(sizeof(*writers), GFP_NOFS); 11768257b2dcSMiao Xie if (!writers) 11778257b2dcSMiao Xie return ERR_PTR(-ENOMEM); 11788257b2dcSMiao Xie 11798257b2dcSMiao Xie ret = percpu_counter_init(&writers->counter, 0); 11808257b2dcSMiao Xie if (ret < 0) { 11818257b2dcSMiao Xie kfree(writers); 11828257b2dcSMiao Xie return ERR_PTR(ret); 11838257b2dcSMiao Xie } 11848257b2dcSMiao Xie 11858257b2dcSMiao Xie init_waitqueue_head(&writers->wait); 11868257b2dcSMiao Xie return writers; 11878257b2dcSMiao Xie } 11888257b2dcSMiao Xie 11898257b2dcSMiao Xie static void 11908257b2dcSMiao Xie btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers) 11918257b2dcSMiao Xie { 11928257b2dcSMiao Xie percpu_counter_destroy(&writers->counter); 11938257b2dcSMiao Xie kfree(writers); 11948257b2dcSMiao Xie } 11958257b2dcSMiao Xie 1196143bede5SJeff Mahoney static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, 119787ee04ebSChris Mason u32 stripesize, struct btrfs_root *root, 11989f5fae2fSChris Mason struct btrfs_fs_info *fs_info, 1199e20d96d6SChris Mason u64 objectid) 1200d97e63b6SChris Mason { 1201cfaa7295SChris Mason root->node = NULL; 1202a28ec197SChris Mason root->commit_root = NULL; 1203db94535dSChris Mason root->sectorsize = sectorsize; 1204db94535dSChris Mason root->nodesize = nodesize; 1205db94535dSChris Mason root->leafsize = leafsize; 120687ee04ebSChris Mason root->stripesize = stripesize; 1207123abc88SChris Mason root->ref_cows = 0; 12080b86a832SChris Mason root->track_dirty = 0; 1209c71bf099SYan, Zheng root->in_radix = 0; 1210d68fc57bSYan, Zheng root->orphan_item_inserted = 0; 1211d68fc57bSYan, Zheng root->orphan_cleanup_state = 0; 12120b86a832SChris Mason 12130f7d52f4SChris Mason root->objectid = objectid; 12140f7d52f4SChris Mason root->last_trans = 0; 121513a8a7c8SYan, Zheng root->highest_objectid = 0; 1216eb73c1b7SMiao Xie root->nr_delalloc_inodes = 0; 1217199c2a9cSMiao Xie root->nr_ordered_extents = 0; 121858176a96SJosef Bacik root->name = NULL; 12196bef4d31SEric Paris root->inode_tree = RB_ROOT; 122016cdcec7SMiao Xie INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); 1221f0486c68SYan, Zheng root->block_rsv = NULL; 1222d68fc57bSYan, Zheng root->orphan_block_rsv = NULL; 12230b86a832SChris Mason 12240b86a832SChris Mason INIT_LIST_HEAD(&root->dirty_list); 12255d4f98a2SYan Zheng INIT_LIST_HEAD(&root->root_list); 1226eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_inodes); 1227eb73c1b7SMiao Xie INIT_LIST_HEAD(&root->delalloc_root); 1228199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_extents); 1229199c2a9cSMiao Xie INIT_LIST_HEAD(&root->ordered_root); 12302ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[0]); 12312ab28f32SJosef Bacik INIT_LIST_HEAD(&root->logged_list[1]); 1232d68fc57bSYan, Zheng spin_lock_init(&root->orphan_lock); 12335d4f98a2SYan Zheng spin_lock_init(&root->inode_lock); 1234eb73c1b7SMiao Xie spin_lock_init(&root->delalloc_lock); 1235199c2a9cSMiao Xie spin_lock_init(&root->ordered_extent_lock); 1236f0486c68SYan, Zheng spin_lock_init(&root->accounting_lock); 12372ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[0]); 12382ab28f32SJosef Bacik spin_lock_init(&root->log_extents_lock[1]); 1239a2135011SChris Mason mutex_init(&root->objectid_mutex); 1240e02119d5SChris Mason mutex_init(&root->log_mutex); 124131f3d255SMiao Xie mutex_init(&root->ordered_extent_mutex); 1242573bfb72SMiao Xie mutex_init(&root->delalloc_mutex); 12437237f183SYan Zheng init_waitqueue_head(&root->log_writer_wait); 12447237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[0]); 12457237f183SYan Zheng init_waitqueue_head(&root->log_commit_wait[1]); 12468b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[0]); 12478b050d35SMiao Xie INIT_LIST_HEAD(&root->log_ctxs[1]); 12487237f183SYan Zheng atomic_set(&root->log_commit[0], 0); 12497237f183SYan Zheng atomic_set(&root->log_commit[1], 0); 12507237f183SYan Zheng atomic_set(&root->log_writers, 0); 12512ecb7923SMiao Xie atomic_set(&root->log_batch, 0); 12528a35d95fSJosef Bacik atomic_set(&root->orphan_inodes, 0); 1253b0feb9d9SMiao Xie atomic_set(&root->refs, 1); 12548257b2dcSMiao Xie atomic_set(&root->will_be_snapshoted, 0); 12557237f183SYan Zheng root->log_transid = 0; 1256d1433debSMiao Xie root->log_transid_committed = -1; 1257257c62e1SChris Mason root->last_log_commit = 0; 125806ea65a3SJosef Bacik if (fs_info) 1259d0c803c4SChris Mason extent_io_tree_init(&root->dirty_log_pages, 1260f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 1261017e5369SChris Mason 12623768f368SChris Mason memset(&root->root_key, 0, sizeof(root->root_key)); 12633768f368SChris Mason memset(&root->root_item, 0, sizeof(root->root_item)); 12646702ed49SChris Mason memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); 126558176a96SJosef Bacik memset(&root->root_kobj, 0, sizeof(root->root_kobj)); 126606ea65a3SJosef Bacik if (fs_info) 12673f157a2fSChris Mason root->defrag_trans_start = fs_info->generation; 126806ea65a3SJosef Bacik else 126906ea65a3SJosef Bacik root->defrag_trans_start = 0; 127058176a96SJosef Bacik init_completion(&root->kobj_unregister); 12716702ed49SChris Mason root->defrag_running = 0; 12724d775673SChris Mason root->root_key.objectid = objectid; 12730ee5dc67SAl Viro root->anon_dev = 0; 12748ea05e3aSAlexander Block 12755f3ab90aSAnand Jain spin_lock_init(&root->root_item_lock); 12763768f368SChris Mason } 12773768f368SChris Mason 1278f84a8bd6SAl Viro static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info) 12796f07e42eSAl Viro { 12806f07e42eSAl Viro struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS); 12816f07e42eSAl Viro if (root) 12826f07e42eSAl Viro root->fs_info = fs_info; 12836f07e42eSAl Viro return root; 12846f07e42eSAl Viro } 12856f07e42eSAl Viro 128606ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 128706ea65a3SJosef Bacik /* Should only be used by the testing infrastructure */ 128806ea65a3SJosef Bacik struct btrfs_root *btrfs_alloc_dummy_root(void) 128906ea65a3SJosef Bacik { 129006ea65a3SJosef Bacik struct btrfs_root *root; 129106ea65a3SJosef Bacik 129206ea65a3SJosef Bacik root = btrfs_alloc_root(NULL); 129306ea65a3SJosef Bacik if (!root) 129406ea65a3SJosef Bacik return ERR_PTR(-ENOMEM); 129506ea65a3SJosef Bacik __setup_root(4096, 4096, 4096, 4096, root, NULL, 1); 129606ea65a3SJosef Bacik root->dummy_root = 1; 129706ea65a3SJosef Bacik 129806ea65a3SJosef Bacik return root; 129906ea65a3SJosef Bacik } 130006ea65a3SJosef Bacik #endif 130106ea65a3SJosef Bacik 130220897f5cSArne Jansen struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, 130320897f5cSArne Jansen struct btrfs_fs_info *fs_info, 130420897f5cSArne Jansen u64 objectid) 130520897f5cSArne Jansen { 130620897f5cSArne Jansen struct extent_buffer *leaf; 130720897f5cSArne Jansen struct btrfs_root *tree_root = fs_info->tree_root; 130820897f5cSArne Jansen struct btrfs_root *root; 130920897f5cSArne Jansen struct btrfs_key key; 131020897f5cSArne Jansen int ret = 0; 13116463fe58SStefan Behrens uuid_le uuid; 131220897f5cSArne Jansen 131320897f5cSArne Jansen root = btrfs_alloc_root(fs_info); 131420897f5cSArne Jansen if (!root) 131520897f5cSArne Jansen return ERR_PTR(-ENOMEM); 131620897f5cSArne Jansen 131720897f5cSArne Jansen __setup_root(tree_root->nodesize, tree_root->leafsize, 131820897f5cSArne Jansen tree_root->sectorsize, tree_root->stripesize, 131920897f5cSArne Jansen root, fs_info, objectid); 132020897f5cSArne Jansen root->root_key.objectid = objectid; 132120897f5cSArne Jansen root->root_key.type = BTRFS_ROOT_ITEM_KEY; 132220897f5cSArne Jansen root->root_key.offset = 0; 132320897f5cSArne Jansen 132420897f5cSArne Jansen leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 132520897f5cSArne Jansen 0, objectid, NULL, 0, 0, 0); 132620897f5cSArne Jansen if (IS_ERR(leaf)) { 132720897f5cSArne Jansen ret = PTR_ERR(leaf); 13281dd05682STsutomu Itoh leaf = NULL; 132920897f5cSArne Jansen goto fail; 133020897f5cSArne Jansen } 133120897f5cSArne Jansen 133220897f5cSArne Jansen memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 133320897f5cSArne Jansen btrfs_set_header_bytenr(leaf, leaf->start); 133420897f5cSArne Jansen btrfs_set_header_generation(leaf, trans->transid); 133520897f5cSArne Jansen btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 133620897f5cSArne Jansen btrfs_set_header_owner(leaf, objectid); 133720897f5cSArne Jansen root->node = leaf; 133820897f5cSArne Jansen 13390a4e5586SRoss Kirk write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), 134020897f5cSArne Jansen BTRFS_FSID_SIZE); 134120897f5cSArne Jansen write_extent_buffer(leaf, fs_info->chunk_tree_uuid, 1342b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(leaf), 134320897f5cSArne Jansen BTRFS_UUID_SIZE); 134420897f5cSArne Jansen btrfs_mark_buffer_dirty(leaf); 134520897f5cSArne Jansen 134620897f5cSArne Jansen root->commit_root = btrfs_root_node(root); 134720897f5cSArne Jansen root->track_dirty = 1; 134820897f5cSArne Jansen 134920897f5cSArne Jansen 135020897f5cSArne Jansen root->root_item.flags = 0; 135120897f5cSArne Jansen root->root_item.byte_limit = 0; 135220897f5cSArne Jansen btrfs_set_root_bytenr(&root->root_item, leaf->start); 135320897f5cSArne Jansen btrfs_set_root_generation(&root->root_item, trans->transid); 135420897f5cSArne Jansen btrfs_set_root_level(&root->root_item, 0); 135520897f5cSArne Jansen btrfs_set_root_refs(&root->root_item, 1); 135620897f5cSArne Jansen btrfs_set_root_used(&root->root_item, leaf->len); 135720897f5cSArne Jansen btrfs_set_root_last_snapshot(&root->root_item, 0); 135820897f5cSArne Jansen btrfs_set_root_dirid(&root->root_item, 0); 13596463fe58SStefan Behrens uuid_le_gen(&uuid); 13606463fe58SStefan Behrens memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE); 136120897f5cSArne Jansen root->root_item.drop_level = 0; 136220897f5cSArne Jansen 136320897f5cSArne Jansen key.objectid = objectid; 136420897f5cSArne Jansen key.type = BTRFS_ROOT_ITEM_KEY; 136520897f5cSArne Jansen key.offset = 0; 136620897f5cSArne Jansen ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); 136720897f5cSArne Jansen if (ret) 136820897f5cSArne Jansen goto fail; 136920897f5cSArne Jansen 137020897f5cSArne Jansen btrfs_tree_unlock(leaf); 137120897f5cSArne Jansen 137220897f5cSArne Jansen return root; 13731dd05682STsutomu Itoh 13741dd05682STsutomu Itoh fail: 13751dd05682STsutomu Itoh if (leaf) { 13761dd05682STsutomu Itoh btrfs_tree_unlock(leaf); 13771dd05682STsutomu Itoh free_extent_buffer(leaf); 13781dd05682STsutomu Itoh } 13791dd05682STsutomu Itoh kfree(root); 13801dd05682STsutomu Itoh 13811dd05682STsutomu Itoh return ERR_PTR(ret); 138220897f5cSArne Jansen } 138320897f5cSArne Jansen 13847237f183SYan Zheng static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, 1385e02119d5SChris Mason struct btrfs_fs_info *fs_info) 13860f7d52f4SChris Mason { 13870f7d52f4SChris Mason struct btrfs_root *root; 13880f7d52f4SChris Mason struct btrfs_root *tree_root = fs_info->tree_root; 13897237f183SYan Zheng struct extent_buffer *leaf; 1390e02119d5SChris Mason 13916f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1392e02119d5SChris Mason if (!root) 13937237f183SYan Zheng return ERR_PTR(-ENOMEM); 1394e02119d5SChris Mason 1395e02119d5SChris Mason __setup_root(tree_root->nodesize, tree_root->leafsize, 1396e02119d5SChris Mason tree_root->sectorsize, tree_root->stripesize, 1397e02119d5SChris Mason root, fs_info, BTRFS_TREE_LOG_OBJECTID); 1398e02119d5SChris Mason 1399e02119d5SChris Mason root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1400e02119d5SChris Mason root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1401e02119d5SChris Mason root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; 14027237f183SYan Zheng /* 14037237f183SYan Zheng * log trees do not get reference counted because they go away 14047237f183SYan Zheng * before a real commit is actually done. They do store pointers 14057237f183SYan Zheng * to file data extents, and those reference counts still get 14067237f183SYan Zheng * updated (along with back refs to the log tree). 14077237f183SYan Zheng */ 1408e02119d5SChris Mason root->ref_cows = 0; 1409e02119d5SChris Mason 14105d4f98a2SYan Zheng leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 141166d7e7f0SArne Jansen BTRFS_TREE_LOG_OBJECTID, NULL, 14125581a51aSJan Schmidt 0, 0, 0); 14137237f183SYan Zheng if (IS_ERR(leaf)) { 14147237f183SYan Zheng kfree(root); 14157237f183SYan Zheng return ERR_CAST(leaf); 14167237f183SYan Zheng } 1417e02119d5SChris Mason 14185d4f98a2SYan Zheng memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 14195d4f98a2SYan Zheng btrfs_set_header_bytenr(leaf, leaf->start); 14205d4f98a2SYan Zheng btrfs_set_header_generation(leaf, trans->transid); 14215d4f98a2SYan Zheng btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 14225d4f98a2SYan Zheng btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 14237237f183SYan Zheng root->node = leaf; 1424e02119d5SChris Mason 1425e02119d5SChris Mason write_extent_buffer(root->node, root->fs_info->fsid, 14260a4e5586SRoss Kirk btrfs_header_fsid(), BTRFS_FSID_SIZE); 1427e02119d5SChris Mason btrfs_mark_buffer_dirty(root->node); 1428e02119d5SChris Mason btrfs_tree_unlock(root->node); 14297237f183SYan Zheng return root; 14307237f183SYan Zheng } 14317237f183SYan Zheng 14327237f183SYan Zheng int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, 14337237f183SYan Zheng struct btrfs_fs_info *fs_info) 14347237f183SYan Zheng { 14357237f183SYan Zheng struct btrfs_root *log_root; 14367237f183SYan Zheng 14377237f183SYan Zheng log_root = alloc_log_tree(trans, fs_info); 14387237f183SYan Zheng if (IS_ERR(log_root)) 14397237f183SYan Zheng return PTR_ERR(log_root); 14407237f183SYan Zheng WARN_ON(fs_info->log_root_tree); 14417237f183SYan Zheng fs_info->log_root_tree = log_root; 14427237f183SYan Zheng return 0; 14437237f183SYan Zheng } 14447237f183SYan Zheng 14457237f183SYan Zheng int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 14467237f183SYan Zheng struct btrfs_root *root) 14477237f183SYan Zheng { 14487237f183SYan Zheng struct btrfs_root *log_root; 14497237f183SYan Zheng struct btrfs_inode_item *inode_item; 14507237f183SYan Zheng 14517237f183SYan Zheng log_root = alloc_log_tree(trans, root->fs_info); 14527237f183SYan Zheng if (IS_ERR(log_root)) 14537237f183SYan Zheng return PTR_ERR(log_root); 14547237f183SYan Zheng 14557237f183SYan Zheng log_root->last_trans = trans->transid; 14567237f183SYan Zheng log_root->root_key.offset = root->root_key.objectid; 14577237f183SYan Zheng 14587237f183SYan Zheng inode_item = &log_root->root_item.inode; 14593cae210fSQu Wenruo btrfs_set_stack_inode_generation(inode_item, 1); 14603cae210fSQu Wenruo btrfs_set_stack_inode_size(inode_item, 3); 14613cae210fSQu Wenruo btrfs_set_stack_inode_nlink(inode_item, 1); 14623cae210fSQu Wenruo btrfs_set_stack_inode_nbytes(inode_item, root->leafsize); 14633cae210fSQu Wenruo btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 14647237f183SYan Zheng 14655d4f98a2SYan Zheng btrfs_set_root_node(&log_root->root_item, log_root->node); 14667237f183SYan Zheng 14677237f183SYan Zheng WARN_ON(root->log_root); 14687237f183SYan Zheng root->log_root = log_root; 14697237f183SYan Zheng root->log_transid = 0; 1470d1433debSMiao Xie root->log_transid_committed = -1; 1471257c62e1SChris Mason root->last_log_commit = 0; 1472e02119d5SChris Mason return 0; 1473e02119d5SChris Mason } 1474e02119d5SChris Mason 147535a3621bSStefan Behrens static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, 1476cb517eabSMiao Xie struct btrfs_key *key) 1477e02119d5SChris Mason { 1478e02119d5SChris Mason struct btrfs_root *root; 1479e02119d5SChris Mason struct btrfs_fs_info *fs_info = tree_root->fs_info; 14800f7d52f4SChris Mason struct btrfs_path *path; 148184234f3aSYan Zheng u64 generation; 1482db94535dSChris Mason u32 blocksize; 1483cb517eabSMiao Xie int ret; 1484cb517eabSMiao Xie 1485cb517eabSMiao Xie path = btrfs_alloc_path(); 1486cb517eabSMiao Xie if (!path) 1487cb517eabSMiao Xie return ERR_PTR(-ENOMEM); 14880f7d52f4SChris Mason 14896f07e42eSAl Viro root = btrfs_alloc_root(fs_info); 1490cb517eabSMiao Xie if (!root) { 1491cb517eabSMiao Xie ret = -ENOMEM; 1492cb517eabSMiao Xie goto alloc_fail; 14930f7d52f4SChris Mason } 14940f7d52f4SChris Mason 1495db94535dSChris Mason __setup_root(tree_root->nodesize, tree_root->leafsize, 149687ee04ebSChris Mason tree_root->sectorsize, tree_root->stripesize, 1497cb517eabSMiao Xie root, fs_info, key->objectid); 14980f7d52f4SChris Mason 1499cb517eabSMiao Xie ret = btrfs_find_root(tree_root, key, path, 1500cb517eabSMiao Xie &root->root_item, &root->root_key); 15010f7d52f4SChris Mason if (ret) { 150213a8a7c8SYan, Zheng if (ret > 0) 150313a8a7c8SYan, Zheng ret = -ENOENT; 1504cb517eabSMiao Xie goto find_fail; 15050f7d52f4SChris Mason } 150613a8a7c8SYan, Zheng 150784234f3aSYan Zheng generation = btrfs_root_generation(&root->root_item); 1508db94535dSChris Mason blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); 1509db94535dSChris Mason root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 151084234f3aSYan Zheng blocksize, generation); 1511cb517eabSMiao Xie if (!root->node) { 1512cb517eabSMiao Xie ret = -ENOMEM; 1513cb517eabSMiao Xie goto find_fail; 1514cb517eabSMiao Xie } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) { 1515cb517eabSMiao Xie ret = -EIO; 1516cb517eabSMiao Xie goto read_fail; 1517416bc658SJosef Bacik } 15185d4f98a2SYan Zheng root->commit_root = btrfs_root_node(root); 151913a8a7c8SYan, Zheng out: 1520cb517eabSMiao Xie btrfs_free_path(path); 1521cb517eabSMiao Xie return root; 1522cb517eabSMiao Xie 1523cb517eabSMiao Xie read_fail: 1524cb517eabSMiao Xie free_extent_buffer(root->node); 1525cb517eabSMiao Xie find_fail: 1526cb517eabSMiao Xie kfree(root); 1527cb517eabSMiao Xie alloc_fail: 1528cb517eabSMiao Xie root = ERR_PTR(ret); 1529cb517eabSMiao Xie goto out; 1530cb517eabSMiao Xie } 1531cb517eabSMiao Xie 1532cb517eabSMiao Xie struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root, 1533cb517eabSMiao Xie struct btrfs_key *location) 1534cb517eabSMiao Xie { 1535cb517eabSMiao Xie struct btrfs_root *root; 1536cb517eabSMiao Xie 1537cb517eabSMiao Xie root = btrfs_read_tree_root(tree_root, location); 1538cb517eabSMiao Xie if (IS_ERR(root)) 1539cb517eabSMiao Xie return root; 1540cb517eabSMiao Xie 1541cb517eabSMiao Xie if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) { 15420f7d52f4SChris Mason root->ref_cows = 1; 154308fe4db1SLi Zefan btrfs_check_and_init_root_item(&root->root_item); 154408fe4db1SLi Zefan } 154513a8a7c8SYan, Zheng 15465eda7b5eSChris Mason return root; 15475eda7b5eSChris Mason } 15485eda7b5eSChris Mason 1549cb517eabSMiao Xie int btrfs_init_fs_root(struct btrfs_root *root) 1550cb517eabSMiao Xie { 1551cb517eabSMiao Xie int ret; 15528257b2dcSMiao Xie struct btrfs_subvolume_writers *writers; 1553cb517eabSMiao Xie 1554cb517eabSMiao Xie root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS); 1555cb517eabSMiao Xie root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned), 1556cb517eabSMiao Xie GFP_NOFS); 1557cb517eabSMiao Xie if (!root->free_ino_pinned || !root->free_ino_ctl) { 1558cb517eabSMiao Xie ret = -ENOMEM; 1559cb517eabSMiao Xie goto fail; 1560cb517eabSMiao Xie } 1561cb517eabSMiao Xie 15628257b2dcSMiao Xie writers = btrfs_alloc_subvolume_writers(); 15638257b2dcSMiao Xie if (IS_ERR(writers)) { 15648257b2dcSMiao Xie ret = PTR_ERR(writers); 15658257b2dcSMiao Xie goto fail; 15668257b2dcSMiao Xie } 15678257b2dcSMiao Xie root->subv_writers = writers; 15688257b2dcSMiao Xie 1569cb517eabSMiao Xie btrfs_init_free_ino_ctl(root); 1570cb517eabSMiao Xie spin_lock_init(&root->cache_lock); 1571cb517eabSMiao Xie init_waitqueue_head(&root->cache_wait); 1572cb517eabSMiao Xie 1573cb517eabSMiao Xie ret = get_anon_bdev(&root->anon_dev); 1574cb517eabSMiao Xie if (ret) 15758257b2dcSMiao Xie goto free_writers; 1576cb517eabSMiao Xie return 0; 15778257b2dcSMiao Xie 15788257b2dcSMiao Xie free_writers: 15798257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 1580cb517eabSMiao Xie fail: 1581cb517eabSMiao Xie kfree(root->free_ino_ctl); 1582cb517eabSMiao Xie kfree(root->free_ino_pinned); 1583cb517eabSMiao Xie return ret; 1584cb517eabSMiao Xie } 1585cb517eabSMiao Xie 1586171170c1SSergei Trofimovich static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, 1587cb517eabSMiao Xie u64 root_id) 1588cb517eabSMiao Xie { 1589cb517eabSMiao Xie struct btrfs_root *root; 1590cb517eabSMiao Xie 1591cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1592cb517eabSMiao Xie root = radix_tree_lookup(&fs_info->fs_roots_radix, 1593cb517eabSMiao Xie (unsigned long)root_id); 1594cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1595cb517eabSMiao Xie return root; 1596cb517eabSMiao Xie } 1597cb517eabSMiao Xie 1598cb517eabSMiao Xie int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, 1599cb517eabSMiao Xie struct btrfs_root *root) 1600cb517eabSMiao Xie { 1601cb517eabSMiao Xie int ret; 1602cb517eabSMiao Xie 1603cb517eabSMiao Xie ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM); 1604cb517eabSMiao Xie if (ret) 1605cb517eabSMiao Xie return ret; 1606cb517eabSMiao Xie 1607cb517eabSMiao Xie spin_lock(&fs_info->fs_roots_radix_lock); 1608cb517eabSMiao Xie ret = radix_tree_insert(&fs_info->fs_roots_radix, 1609cb517eabSMiao Xie (unsigned long)root->root_key.objectid, 1610cb517eabSMiao Xie root); 1611cb517eabSMiao Xie if (ret == 0) 1612cb517eabSMiao Xie root->in_radix = 1; 1613cb517eabSMiao Xie spin_unlock(&fs_info->fs_roots_radix_lock); 1614cb517eabSMiao Xie radix_tree_preload_end(); 1615cb517eabSMiao Xie 1616cb517eabSMiao Xie return ret; 1617cb517eabSMiao Xie } 1618cb517eabSMiao Xie 1619c00869f1SMiao Xie struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, 1620c00869f1SMiao Xie struct btrfs_key *location, 1621c00869f1SMiao Xie bool check_ref) 16225eda7b5eSChris Mason { 16235eda7b5eSChris Mason struct btrfs_root *root; 16245eda7b5eSChris Mason int ret; 16255eda7b5eSChris Mason 1626edbd8d4eSChris Mason if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) 1627edbd8d4eSChris Mason return fs_info->tree_root; 1628edbd8d4eSChris Mason if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) 1629edbd8d4eSChris Mason return fs_info->extent_root; 16308f18cf13SChris Mason if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) 16318f18cf13SChris Mason return fs_info->chunk_root; 16328f18cf13SChris Mason if (location->objectid == BTRFS_DEV_TREE_OBJECTID) 16338f18cf13SChris Mason return fs_info->dev_root; 16340403e47eSYan Zheng if (location->objectid == BTRFS_CSUM_TREE_OBJECTID) 16350403e47eSYan Zheng return fs_info->csum_root; 1636bcef60f2SArne Jansen if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID) 1637bcef60f2SArne Jansen return fs_info->quota_root ? fs_info->quota_root : 1638bcef60f2SArne Jansen ERR_PTR(-ENOENT); 1639f7a81ea4SStefan Behrens if (location->objectid == BTRFS_UUID_TREE_OBJECTID) 1640f7a81ea4SStefan Behrens return fs_info->uuid_root ? fs_info->uuid_root : 1641f7a81ea4SStefan Behrens ERR_PTR(-ENOENT); 16424df27c4dSYan, Zheng again: 1643cb517eabSMiao Xie root = btrfs_lookup_fs_root(fs_info, location->objectid); 164448475471SStefan Behrens if (root) { 1645c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) 164648475471SStefan Behrens return ERR_PTR(-ENOENT); 16475eda7b5eSChris Mason return root; 164848475471SStefan Behrens } 16495eda7b5eSChris Mason 1650cb517eabSMiao Xie root = btrfs_read_fs_root(fs_info->tree_root, location); 16515eda7b5eSChris Mason if (IS_ERR(root)) 16525eda7b5eSChris Mason return root; 16533394e160SChris Mason 1654c00869f1SMiao Xie if (check_ref && btrfs_root_refs(&root->root_item) == 0) { 1655d68fc57bSYan, Zheng ret = -ENOENT; 1656d68fc57bSYan, Zheng goto fail; 1657d68fc57bSYan, Zheng } 1658d68fc57bSYan, Zheng 1659cb517eabSMiao Xie ret = btrfs_init_fs_root(root); 1660cb517eabSMiao Xie if (ret) 1661cb517eabSMiao Xie goto fail; 1662cb517eabSMiao Xie 16633f870c28SKelley Nielsen ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID, 16643f870c28SKelley Nielsen location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL); 1665d68fc57bSYan, Zheng if (ret < 0) 1666d68fc57bSYan, Zheng goto fail; 1667d68fc57bSYan, Zheng if (ret == 0) 1668d68fc57bSYan, Zheng root->orphan_item_inserted = 1; 1669d68fc57bSYan, Zheng 1670cb517eabSMiao Xie ret = btrfs_insert_fs_root(fs_info, root); 16710f7d52f4SChris Mason if (ret) { 16724df27c4dSYan, Zheng if (ret == -EEXIST) { 16734df27c4dSYan, Zheng free_fs_root(root); 16744df27c4dSYan, Zheng goto again; 16750f7d52f4SChris Mason } 16764df27c4dSYan, Zheng goto fail; 16774df27c4dSYan, Zheng } 1678edbd8d4eSChris Mason return root; 16794df27c4dSYan, Zheng fail: 16804df27c4dSYan, Zheng free_fs_root(root); 16814df27c4dSYan, Zheng return ERR_PTR(ret); 1682edbd8d4eSChris Mason } 1683edbd8d4eSChris Mason 168404160088SChris Mason static int btrfs_congested_fn(void *congested_data, int bdi_bits) 168504160088SChris Mason { 168604160088SChris Mason struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; 168704160088SChris Mason int ret = 0; 168804160088SChris Mason struct btrfs_device *device; 168904160088SChris Mason struct backing_dev_info *bdi; 1690b7967db7SChris Mason 16911f78160cSXiao Guangrong rcu_read_lock(); 16921f78160cSXiao Guangrong list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1693dfe25020SChris Mason if (!device->bdev) 1694dfe25020SChris Mason continue; 169504160088SChris Mason bdi = blk_get_backing_dev_info(device->bdev); 169604160088SChris Mason if (bdi && bdi_congested(bdi, bdi_bits)) { 169704160088SChris Mason ret = 1; 169804160088SChris Mason break; 169904160088SChris Mason } 170004160088SChris Mason } 17011f78160cSXiao Guangrong rcu_read_unlock(); 170204160088SChris Mason return ret; 170304160088SChris Mason } 170404160088SChris Mason 170538b66988SChris Mason /* 1706ad081f14SJens Axboe * If this fails, caller must call bdi_destroy() to get rid of the 1707ad081f14SJens Axboe * bdi again. 1708ad081f14SJens Axboe */ 170904160088SChris Mason static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) 171004160088SChris Mason { 1711ad081f14SJens Axboe int err; 1712ad081f14SJens Axboe 1713ad081f14SJens Axboe bdi->capabilities = BDI_CAP_MAP_COPY; 1714e6d086d8SJens Axboe err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY); 1715ad081f14SJens Axboe if (err) 1716ad081f14SJens Axboe return err; 1717ad081f14SJens Axboe 17184575c9ccSChris Mason bdi->ra_pages = default_backing_dev_info.ra_pages; 171904160088SChris Mason bdi->congested_fn = btrfs_congested_fn; 172004160088SChris Mason bdi->congested_data = info; 172104160088SChris Mason return 0; 172204160088SChris Mason } 172304160088SChris Mason 17248b712842SChris Mason /* 17258b712842SChris Mason * called by the kthread helper functions to finally call the bio end_io 17268b712842SChris Mason * functions. This is where read checksum verification actually happens 17278b712842SChris Mason */ 1728d458b054SQu Wenruo static void end_workqueue_fn(struct btrfs_work *work) 1729ce9adaa5SChris Mason { 1730ce9adaa5SChris Mason struct bio *bio; 17318b712842SChris Mason struct end_io_wq *end_io_wq; 1732ce9adaa5SChris Mason int error; 1733ce9adaa5SChris Mason 17348b712842SChris Mason end_io_wq = container_of(work, struct end_io_wq, work); 1735ce9adaa5SChris Mason bio = end_io_wq->bio; 17368b712842SChris Mason 1737ce9adaa5SChris Mason error = end_io_wq->error; 1738ce9adaa5SChris Mason bio->bi_private = end_io_wq->private; 1739ce9adaa5SChris Mason bio->bi_end_io = end_io_wq->end_io; 1740ce9adaa5SChris Mason kfree(end_io_wq); 1741ce9adaa5SChris Mason bio_endio(bio, error); 1742ce9adaa5SChris Mason } 174344b8bd7eSChris Mason 1744a74a4b97SChris Mason static int cleaner_kthread(void *arg) 1745a74a4b97SChris Mason { 1746a74a4b97SChris Mason struct btrfs_root *root = arg; 1747d0278245SMiao Xie int again; 1748a74a4b97SChris Mason 1749a74a4b97SChris Mason do { 1750d0278245SMiao Xie again = 0; 17519d1a2a3aSDavid Sterba 1752d0278245SMiao Xie /* Make the cleaner go to sleep early. */ 1753babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) 1754d0278245SMiao Xie goto sleep; 1755d0278245SMiao Xie 1756d0278245SMiao Xie if (!mutex_trylock(&root->fs_info->cleaner_mutex)) 1757d0278245SMiao Xie goto sleep; 1758d0278245SMiao Xie 1759dc7f370cSMiao Xie /* 1760dc7f370cSMiao Xie * Avoid the problem that we change the status of the fs 1761dc7f370cSMiao Xie * during the above check and trylock. 1762dc7f370cSMiao Xie */ 1763babbf170SMiao Xie if (btrfs_need_cleaner_sleep(root)) { 1764dc7f370cSMiao Xie mutex_unlock(&root->fs_info->cleaner_mutex); 1765dc7f370cSMiao Xie goto sleep; 1766dc7f370cSMiao Xie } 1767dc7f370cSMiao Xie 176824bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 17699d1a2a3aSDavid Sterba again = btrfs_clean_one_deleted_snapshot(root); 1770a74a4b97SChris Mason mutex_unlock(&root->fs_info->cleaner_mutex); 1771a74a4b97SChris Mason 1772d0278245SMiao Xie /* 177305323cd1SMiao Xie * The defragger has dealt with the R/O remount and umount, 177405323cd1SMiao Xie * needn't do anything special here. 1775d0278245SMiao Xie */ 1776d0278245SMiao Xie btrfs_run_defrag_inodes(root->fs_info); 1777d0278245SMiao Xie sleep: 17789d1a2a3aSDavid Sterba if (!try_to_freeze() && !again) { 1779a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 17808929ecfaSYan, Zheng if (!kthread_should_stop()) 1781a74a4b97SChris Mason schedule(); 1782a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1783a74a4b97SChris Mason } 1784a74a4b97SChris Mason } while (!kthread_should_stop()); 1785a74a4b97SChris Mason return 0; 1786a74a4b97SChris Mason } 1787a74a4b97SChris Mason 1788a74a4b97SChris Mason static int transaction_kthread(void *arg) 1789a74a4b97SChris Mason { 1790a74a4b97SChris Mason struct btrfs_root *root = arg; 1791a74a4b97SChris Mason struct btrfs_trans_handle *trans; 1792a74a4b97SChris Mason struct btrfs_transaction *cur; 17938929ecfaSYan, Zheng u64 transid; 1794a74a4b97SChris Mason unsigned long now; 1795a74a4b97SChris Mason unsigned long delay; 1796914b2007SJan Kara bool cannot_commit; 1797a74a4b97SChris Mason 1798a74a4b97SChris Mason do { 1799914b2007SJan Kara cannot_commit = false; 18008b87dc17SDavid Sterba delay = HZ * root->fs_info->commit_interval; 1801a74a4b97SChris Mason mutex_lock(&root->fs_info->transaction_kthread_mutex); 1802a74a4b97SChris Mason 1803a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 1804a74a4b97SChris Mason cur = root->fs_info->running_transaction; 1805a74a4b97SChris Mason if (!cur) { 1806a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1807a74a4b97SChris Mason goto sleep; 1808a74a4b97SChris Mason } 180931153d81SYan Zheng 1810a74a4b97SChris Mason now = get_seconds(); 18114a9d8bdeSMiao Xie if (cur->state < TRANS_STATE_BLOCKED && 18128b87dc17SDavid Sterba (now < cur->start_time || 18138b87dc17SDavid Sterba now - cur->start_time < root->fs_info->commit_interval)) { 1814a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 1815a74a4b97SChris Mason delay = HZ * 5; 1816a74a4b97SChris Mason goto sleep; 1817a74a4b97SChris Mason } 18188929ecfaSYan, Zheng transid = cur->transid; 1819a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 182056bec294SChris Mason 182179787eaaSJeff Mahoney /* If the file system is aborted, this will always fail. */ 1822354aa0fbSMiao Xie trans = btrfs_attach_transaction(root); 1823914b2007SJan Kara if (IS_ERR(trans)) { 1824354aa0fbSMiao Xie if (PTR_ERR(trans) != -ENOENT) 1825914b2007SJan Kara cannot_commit = true; 182679787eaaSJeff Mahoney goto sleep; 1827914b2007SJan Kara } 18288929ecfaSYan, Zheng if (transid == trans->transid) { 182979787eaaSJeff Mahoney btrfs_commit_transaction(trans, root); 18308929ecfaSYan, Zheng } else { 18318929ecfaSYan, Zheng btrfs_end_transaction(trans, root); 18328929ecfaSYan, Zheng } 1833a74a4b97SChris Mason sleep: 1834a74a4b97SChris Mason wake_up_process(root->fs_info->cleaner_kthread); 1835a74a4b97SChris Mason mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1836a74a4b97SChris Mason 18374e121c06SJosef Bacik if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 18384e121c06SJosef Bacik &root->fs_info->fs_state))) 18394e121c06SJosef Bacik btrfs_cleanup_transaction(root); 1840a0acae0eSTejun Heo if (!try_to_freeze()) { 1841a74a4b97SChris Mason set_current_state(TASK_INTERRUPTIBLE); 18428929ecfaSYan, Zheng if (!kthread_should_stop() && 1843914b2007SJan Kara (!btrfs_transaction_blocked(root->fs_info) || 1844914b2007SJan Kara cannot_commit)) 1845a74a4b97SChris Mason schedule_timeout(delay); 1846a74a4b97SChris Mason __set_current_state(TASK_RUNNING); 1847a74a4b97SChris Mason } 1848a74a4b97SChris Mason } while (!kthread_should_stop()); 1849a74a4b97SChris Mason return 0; 1850a74a4b97SChris Mason } 1851a74a4b97SChris Mason 1852af31f5e5SChris Mason /* 1853af31f5e5SChris Mason * this will find the highest generation in the array of 1854af31f5e5SChris Mason * root backups. The index of the highest array is returned, 1855af31f5e5SChris Mason * or -1 if we can't find anything. 1856af31f5e5SChris Mason * 1857af31f5e5SChris Mason * We check to make sure the array is valid by comparing the 1858af31f5e5SChris Mason * generation of the latest root in the array with the generation 1859af31f5e5SChris Mason * in the super block. If they don't match we pitch it. 1860af31f5e5SChris Mason */ 1861af31f5e5SChris Mason static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen) 1862af31f5e5SChris Mason { 1863af31f5e5SChris Mason u64 cur; 1864af31f5e5SChris Mason int newest_index = -1; 1865af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1866af31f5e5SChris Mason int i; 1867af31f5e5SChris Mason 1868af31f5e5SChris Mason for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { 1869af31f5e5SChris Mason root_backup = info->super_copy->super_roots + i; 1870af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1871af31f5e5SChris Mason if (cur == newest_gen) 1872af31f5e5SChris Mason newest_index = i; 1873af31f5e5SChris Mason } 1874af31f5e5SChris Mason 1875af31f5e5SChris Mason /* check to see if we actually wrapped around */ 1876af31f5e5SChris Mason if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) { 1877af31f5e5SChris Mason root_backup = info->super_copy->super_roots; 1878af31f5e5SChris Mason cur = btrfs_backup_tree_root_gen(root_backup); 1879af31f5e5SChris Mason if (cur == newest_gen) 1880af31f5e5SChris Mason newest_index = 0; 1881af31f5e5SChris Mason } 1882af31f5e5SChris Mason return newest_index; 1883af31f5e5SChris Mason } 1884af31f5e5SChris Mason 1885af31f5e5SChris Mason 1886af31f5e5SChris Mason /* 1887af31f5e5SChris Mason * find the oldest backup so we know where to store new entries 1888af31f5e5SChris Mason * in the backup array. This will set the backup_root_index 1889af31f5e5SChris Mason * field in the fs_info struct 1890af31f5e5SChris Mason */ 1891af31f5e5SChris Mason static void find_oldest_super_backup(struct btrfs_fs_info *info, 1892af31f5e5SChris Mason u64 newest_gen) 1893af31f5e5SChris Mason { 1894af31f5e5SChris Mason int newest_index = -1; 1895af31f5e5SChris Mason 1896af31f5e5SChris Mason newest_index = find_newest_super_backup(info, newest_gen); 1897af31f5e5SChris Mason /* if there was garbage in there, just move along */ 1898af31f5e5SChris Mason if (newest_index == -1) { 1899af31f5e5SChris Mason info->backup_root_index = 0; 1900af31f5e5SChris Mason } else { 1901af31f5e5SChris Mason info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS; 1902af31f5e5SChris Mason } 1903af31f5e5SChris Mason } 1904af31f5e5SChris Mason 1905af31f5e5SChris Mason /* 1906af31f5e5SChris Mason * copy all the root pointers into the super backup array. 1907af31f5e5SChris Mason * this will bump the backup pointer by one when it is 1908af31f5e5SChris Mason * done 1909af31f5e5SChris Mason */ 1910af31f5e5SChris Mason static void backup_super_roots(struct btrfs_fs_info *info) 1911af31f5e5SChris Mason { 1912af31f5e5SChris Mason int next_backup; 1913af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 1914af31f5e5SChris Mason int last_backup; 1915af31f5e5SChris Mason 1916af31f5e5SChris Mason next_backup = info->backup_root_index; 1917af31f5e5SChris Mason last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) % 1918af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 1919af31f5e5SChris Mason 1920af31f5e5SChris Mason /* 1921af31f5e5SChris Mason * just overwrite the last backup if we're at the same generation 1922af31f5e5SChris Mason * this happens only at umount 1923af31f5e5SChris Mason */ 1924af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + last_backup; 1925af31f5e5SChris Mason if (btrfs_backup_tree_root_gen(root_backup) == 1926af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)) 1927af31f5e5SChris Mason next_backup = last_backup; 1928af31f5e5SChris Mason 1929af31f5e5SChris Mason root_backup = info->super_for_commit->super_roots + next_backup; 1930af31f5e5SChris Mason 1931af31f5e5SChris Mason /* 1932af31f5e5SChris Mason * make sure all of our padding and empty slots get zero filled 1933af31f5e5SChris Mason * regardless of which ones we use today 1934af31f5e5SChris Mason */ 1935af31f5e5SChris Mason memset(root_backup, 0, sizeof(*root_backup)); 1936af31f5e5SChris Mason 1937af31f5e5SChris Mason info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; 1938af31f5e5SChris Mason 1939af31f5e5SChris Mason btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); 1940af31f5e5SChris Mason btrfs_set_backup_tree_root_gen(root_backup, 1941af31f5e5SChris Mason btrfs_header_generation(info->tree_root->node)); 1942af31f5e5SChris Mason 1943af31f5e5SChris Mason btrfs_set_backup_tree_root_level(root_backup, 1944af31f5e5SChris Mason btrfs_header_level(info->tree_root->node)); 1945af31f5e5SChris Mason 1946af31f5e5SChris Mason btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); 1947af31f5e5SChris Mason btrfs_set_backup_chunk_root_gen(root_backup, 1948af31f5e5SChris Mason btrfs_header_generation(info->chunk_root->node)); 1949af31f5e5SChris Mason btrfs_set_backup_chunk_root_level(root_backup, 1950af31f5e5SChris Mason btrfs_header_level(info->chunk_root->node)); 1951af31f5e5SChris Mason 1952af31f5e5SChris Mason btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start); 1953af31f5e5SChris Mason btrfs_set_backup_extent_root_gen(root_backup, 1954af31f5e5SChris Mason btrfs_header_generation(info->extent_root->node)); 1955af31f5e5SChris Mason btrfs_set_backup_extent_root_level(root_backup, 1956af31f5e5SChris Mason btrfs_header_level(info->extent_root->node)); 1957af31f5e5SChris Mason 19587c7e82a7SChris Mason /* 19597c7e82a7SChris Mason * we might commit during log recovery, which happens before we set 19607c7e82a7SChris Mason * the fs_root. Make sure it is valid before we fill it in. 19617c7e82a7SChris Mason */ 19627c7e82a7SChris Mason if (info->fs_root && info->fs_root->node) { 19637c7e82a7SChris Mason btrfs_set_backup_fs_root(root_backup, 19647c7e82a7SChris Mason info->fs_root->node->start); 1965af31f5e5SChris Mason btrfs_set_backup_fs_root_gen(root_backup, 1966af31f5e5SChris Mason btrfs_header_generation(info->fs_root->node)); 1967af31f5e5SChris Mason btrfs_set_backup_fs_root_level(root_backup, 1968af31f5e5SChris Mason btrfs_header_level(info->fs_root->node)); 19697c7e82a7SChris Mason } 1970af31f5e5SChris Mason 1971af31f5e5SChris Mason btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); 1972af31f5e5SChris Mason btrfs_set_backup_dev_root_gen(root_backup, 1973af31f5e5SChris Mason btrfs_header_generation(info->dev_root->node)); 1974af31f5e5SChris Mason btrfs_set_backup_dev_root_level(root_backup, 1975af31f5e5SChris Mason btrfs_header_level(info->dev_root->node)); 1976af31f5e5SChris Mason 1977af31f5e5SChris Mason btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start); 1978af31f5e5SChris Mason btrfs_set_backup_csum_root_gen(root_backup, 1979af31f5e5SChris Mason btrfs_header_generation(info->csum_root->node)); 1980af31f5e5SChris Mason btrfs_set_backup_csum_root_level(root_backup, 1981af31f5e5SChris Mason btrfs_header_level(info->csum_root->node)); 1982af31f5e5SChris Mason 1983af31f5e5SChris Mason btrfs_set_backup_total_bytes(root_backup, 1984af31f5e5SChris Mason btrfs_super_total_bytes(info->super_copy)); 1985af31f5e5SChris Mason btrfs_set_backup_bytes_used(root_backup, 1986af31f5e5SChris Mason btrfs_super_bytes_used(info->super_copy)); 1987af31f5e5SChris Mason btrfs_set_backup_num_devices(root_backup, 1988af31f5e5SChris Mason btrfs_super_num_devices(info->super_copy)); 1989af31f5e5SChris Mason 1990af31f5e5SChris Mason /* 1991af31f5e5SChris Mason * if we don't copy this out to the super_copy, it won't get remembered 1992af31f5e5SChris Mason * for the next commit 1993af31f5e5SChris Mason */ 1994af31f5e5SChris Mason memcpy(&info->super_copy->super_roots, 1995af31f5e5SChris Mason &info->super_for_commit->super_roots, 1996af31f5e5SChris Mason sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); 1997af31f5e5SChris Mason } 1998af31f5e5SChris Mason 1999af31f5e5SChris Mason /* 2000af31f5e5SChris Mason * this copies info out of the root backup array and back into 2001af31f5e5SChris Mason * the in-memory super block. It is meant to help iterate through 2002af31f5e5SChris Mason * the array, so you send it the number of backups you've already 2003af31f5e5SChris Mason * tried and the last backup index you used. 2004af31f5e5SChris Mason * 2005af31f5e5SChris Mason * this returns -1 when it has tried all the backups 2006af31f5e5SChris Mason */ 2007af31f5e5SChris Mason static noinline int next_root_backup(struct btrfs_fs_info *info, 2008af31f5e5SChris Mason struct btrfs_super_block *super, 2009af31f5e5SChris Mason int *num_backups_tried, int *backup_index) 2010af31f5e5SChris Mason { 2011af31f5e5SChris Mason struct btrfs_root_backup *root_backup; 2012af31f5e5SChris Mason int newest = *backup_index; 2013af31f5e5SChris Mason 2014af31f5e5SChris Mason if (*num_backups_tried == 0) { 2015af31f5e5SChris Mason u64 gen = btrfs_super_generation(super); 2016af31f5e5SChris Mason 2017af31f5e5SChris Mason newest = find_newest_super_backup(info, gen); 2018af31f5e5SChris Mason if (newest == -1) 2019af31f5e5SChris Mason return -1; 2020af31f5e5SChris Mason 2021af31f5e5SChris Mason *backup_index = newest; 2022af31f5e5SChris Mason *num_backups_tried = 1; 2023af31f5e5SChris Mason } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) { 2024af31f5e5SChris Mason /* we've tried all the backups, all done */ 2025af31f5e5SChris Mason return -1; 2026af31f5e5SChris Mason } else { 2027af31f5e5SChris Mason /* jump to the next oldest backup */ 2028af31f5e5SChris Mason newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) % 2029af31f5e5SChris Mason BTRFS_NUM_BACKUP_ROOTS; 2030af31f5e5SChris Mason *backup_index = newest; 2031af31f5e5SChris Mason *num_backups_tried += 1; 2032af31f5e5SChris Mason } 2033af31f5e5SChris Mason root_backup = super->super_roots + newest; 2034af31f5e5SChris Mason 2035af31f5e5SChris Mason btrfs_set_super_generation(super, 2036af31f5e5SChris Mason btrfs_backup_tree_root_gen(root_backup)); 2037af31f5e5SChris Mason btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); 2038af31f5e5SChris Mason btrfs_set_super_root_level(super, 2039af31f5e5SChris Mason btrfs_backup_tree_root_level(root_backup)); 2040af31f5e5SChris Mason btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); 2041af31f5e5SChris Mason 2042af31f5e5SChris Mason /* 2043af31f5e5SChris Mason * fixme: the total bytes and num_devices need to match or we should 2044af31f5e5SChris Mason * need a fsck 2045af31f5e5SChris Mason */ 2046af31f5e5SChris Mason btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); 2047af31f5e5SChris Mason btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); 2048af31f5e5SChris Mason return 0; 2049af31f5e5SChris Mason } 2050af31f5e5SChris Mason 20517abadb64SLiu Bo /* helper to cleanup workers */ 20527abadb64SLiu Bo static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) 20537abadb64SLiu Bo { 2054dc6e3209SQu Wenruo btrfs_destroy_workqueue(fs_info->fixup_workers); 2055afe3d242SQu Wenruo btrfs_destroy_workqueue(fs_info->delalloc_workers); 20565cdc7ad3SQu Wenruo btrfs_destroy_workqueue(fs_info->workers); 2057fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_workers); 2058fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_workers); 2059fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 2060d05a33acSQu Wenruo btrfs_destroy_workqueue(fs_info->rmw_workers); 2061fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_meta_write_workers); 2062fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_write_workers); 2063fccb5d86SQu Wenruo btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2064a8c93d4eSQu Wenruo btrfs_destroy_workqueue(fs_info->submit_workers); 20655b3bc44eSQu Wenruo btrfs_destroy_workqueue(fs_info->delayed_workers); 2066e66f0bb1SQu Wenruo btrfs_destroy_workqueue(fs_info->caching_workers); 2067736cfa15SQu Wenruo btrfs_destroy_workqueue(fs_info->readahead_workers); 2068a44903abSQu Wenruo btrfs_destroy_workqueue(fs_info->flush_workers); 2069fc97fab0SQu Wenruo btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 20707abadb64SLiu Bo } 20717abadb64SLiu Bo 20722e9f5954SRashika static void free_root_extent_buffers(struct btrfs_root *root) 20732e9f5954SRashika { 20742e9f5954SRashika if (root) { 20752e9f5954SRashika free_extent_buffer(root->node); 20762e9f5954SRashika free_extent_buffer(root->commit_root); 20772e9f5954SRashika root->node = NULL; 20782e9f5954SRashika root->commit_root = NULL; 20792e9f5954SRashika } 20802e9f5954SRashika } 20812e9f5954SRashika 2082af31f5e5SChris Mason /* helper to cleanup tree roots */ 2083af31f5e5SChris Mason static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) 2084af31f5e5SChris Mason { 20852e9f5954SRashika free_root_extent_buffers(info->tree_root); 2086655b09feSJosef Bacik 20872e9f5954SRashika free_root_extent_buffers(info->dev_root); 20882e9f5954SRashika free_root_extent_buffers(info->extent_root); 20892e9f5954SRashika free_root_extent_buffers(info->csum_root); 20902e9f5954SRashika free_root_extent_buffers(info->quota_root); 20912e9f5954SRashika free_root_extent_buffers(info->uuid_root); 20922e9f5954SRashika if (chunk_root) 20932e9f5954SRashika free_root_extent_buffers(info->chunk_root); 2094af31f5e5SChris Mason } 2095af31f5e5SChris Mason 2096171f6537SJosef Bacik static void del_fs_roots(struct btrfs_fs_info *fs_info) 2097171f6537SJosef Bacik { 2098171f6537SJosef Bacik int ret; 2099171f6537SJosef Bacik struct btrfs_root *gang[8]; 2100171f6537SJosef Bacik int i; 2101171f6537SJosef Bacik 2102171f6537SJosef Bacik while (!list_empty(&fs_info->dead_roots)) { 2103171f6537SJosef Bacik gang[0] = list_entry(fs_info->dead_roots.next, 2104171f6537SJosef Bacik struct btrfs_root, root_list); 2105171f6537SJosef Bacik list_del(&gang[0]->root_list); 2106171f6537SJosef Bacik 2107171f6537SJosef Bacik if (gang[0]->in_radix) { 2108cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[0]); 2109171f6537SJosef Bacik } else { 2110171f6537SJosef Bacik free_extent_buffer(gang[0]->node); 2111171f6537SJosef Bacik free_extent_buffer(gang[0]->commit_root); 2112b0feb9d9SMiao Xie btrfs_put_fs_root(gang[0]); 2113171f6537SJosef Bacik } 2114171f6537SJosef Bacik } 2115171f6537SJosef Bacik 2116171f6537SJosef Bacik while (1) { 2117171f6537SJosef Bacik ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 2118171f6537SJosef Bacik (void **)gang, 0, 2119171f6537SJosef Bacik ARRAY_SIZE(gang)); 2120171f6537SJosef Bacik if (!ret) 2121171f6537SJosef Bacik break; 2122171f6537SJosef Bacik for (i = 0; i < ret; i++) 2123cb517eabSMiao Xie btrfs_drop_and_free_fs_root(fs_info, gang[i]); 2124171f6537SJosef Bacik } 21251a4319ccSLiu Bo 21261a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 21271a4319ccSLiu Bo btrfs_free_log_root_tree(NULL, fs_info); 21281a4319ccSLiu Bo btrfs_destroy_pinned_extent(fs_info->tree_root, 21291a4319ccSLiu Bo fs_info->pinned_extents); 21301a4319ccSLiu Bo } 2131171f6537SJosef Bacik } 2132af31f5e5SChris Mason 2133ad2b2c80SAl Viro int open_ctree(struct super_block *sb, 2134dfe25020SChris Mason struct btrfs_fs_devices *fs_devices, 2135dfe25020SChris Mason char *options) 2136eb60ceacSChris Mason { 2137db94535dSChris Mason u32 sectorsize; 2138db94535dSChris Mason u32 nodesize; 2139db94535dSChris Mason u32 leafsize; 2140db94535dSChris Mason u32 blocksize; 214187ee04ebSChris Mason u32 stripesize; 214284234f3aSYan Zheng u64 generation; 2143f2b636e8SJosef Bacik u64 features; 21443de4586cSChris Mason struct btrfs_key location; 2145a061fc8dSChris Mason struct buffer_head *bh; 21464d34b278SIlya Dryomov struct btrfs_super_block *disk_super; 2147815745cfSAl Viro struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2148f84a8bd6SAl Viro struct btrfs_root *tree_root; 21494d34b278SIlya Dryomov struct btrfs_root *extent_root; 21504d34b278SIlya Dryomov struct btrfs_root *csum_root; 21514d34b278SIlya Dryomov struct btrfs_root *chunk_root; 21524d34b278SIlya Dryomov struct btrfs_root *dev_root; 2153bcef60f2SArne Jansen struct btrfs_root *quota_root; 2154f7a81ea4SStefan Behrens struct btrfs_root *uuid_root; 2155e02119d5SChris Mason struct btrfs_root *log_tree_root; 2156eb60ceacSChris Mason int ret; 2157e58ca020SYan int err = -EINVAL; 2158af31f5e5SChris Mason int num_backups_tried = 0; 2159af31f5e5SChris Mason int backup_index = 0; 21605cdc7ad3SQu Wenruo int max_active; 21615cdc7ad3SQu Wenruo int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; 216270f80175SStefan Behrens bool create_uuid_tree; 216370f80175SStefan Behrens bool check_uuid_tree; 21644543df7eSChris Mason 2165f84a8bd6SAl Viro tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info); 21666f07e42eSAl Viro chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info); 2167cb517eabSMiao Xie if (!tree_root || !chunk_root) { 216839279cc3SChris Mason err = -ENOMEM; 216939279cc3SChris Mason goto fail; 217039279cc3SChris Mason } 217176dda93cSYan, Zheng 217276dda93cSYan, Zheng ret = init_srcu_struct(&fs_info->subvol_srcu); 217376dda93cSYan, Zheng if (ret) { 217476dda93cSYan, Zheng err = ret; 217576dda93cSYan, Zheng goto fail; 217676dda93cSYan, Zheng } 217776dda93cSYan, Zheng 217876dda93cSYan, Zheng ret = setup_bdi(fs_info, &fs_info->bdi); 217976dda93cSYan, Zheng if (ret) { 218076dda93cSYan, Zheng err = ret; 218176dda93cSYan, Zheng goto fail_srcu; 218276dda93cSYan, Zheng } 218376dda93cSYan, Zheng 2184e2d84521SMiao Xie ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0); 2185e2d84521SMiao Xie if (ret) { 2186e2d84521SMiao Xie err = ret; 2187e2d84521SMiao Xie goto fail_bdi; 2188e2d84521SMiao Xie } 2189e2d84521SMiao Xie fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE * 2190e2d84521SMiao Xie (1 + ilog2(nr_cpu_ids)); 2191e2d84521SMiao Xie 2192963d678bSMiao Xie ret = percpu_counter_init(&fs_info->delalloc_bytes, 0); 2193963d678bSMiao Xie if (ret) { 2194963d678bSMiao Xie err = ret; 2195963d678bSMiao Xie goto fail_dirty_metadata_bytes; 2196963d678bSMiao Xie } 2197963d678bSMiao Xie 2198c404e0dcSMiao Xie ret = percpu_counter_init(&fs_info->bio_counter, 0); 2199c404e0dcSMiao Xie if (ret) { 2200c404e0dcSMiao Xie err = ret; 2201c404e0dcSMiao Xie goto fail_delalloc_bytes; 2202c404e0dcSMiao Xie } 2203c404e0dcSMiao Xie 220476dda93cSYan, Zheng fs_info->btree_inode = new_inode(sb); 220576dda93cSYan, Zheng if (!fs_info->btree_inode) { 220676dda93cSYan, Zheng err = -ENOMEM; 2207c404e0dcSMiao Xie goto fail_bio_counter; 220876dda93cSYan, Zheng } 220976dda93cSYan, Zheng 2210a6591715SChris Mason mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); 22111561dedaSMiao Xie 221276dda93cSYan, Zheng INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); 2213f28491e0SJosef Bacik INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); 22148fd17795SChris Mason INIT_LIST_HEAD(&fs_info->trans_list); 2215facda1e7SChris Mason INIT_LIST_HEAD(&fs_info->dead_roots); 221624bbcf04SYan, Zheng INIT_LIST_HEAD(&fs_info->delayed_iputs); 2217eb73c1b7SMiao Xie INIT_LIST_HEAD(&fs_info->delalloc_roots); 221811833d66SYan Zheng INIT_LIST_HEAD(&fs_info->caching_block_groups); 2219eb73c1b7SMiao Xie spin_lock_init(&fs_info->delalloc_root_lock); 2220a4abeea4SJosef Bacik spin_lock_init(&fs_info->trans_lock); 222176dda93cSYan, Zheng spin_lock_init(&fs_info->fs_roots_radix_lock); 222224bbcf04SYan, Zheng spin_lock_init(&fs_info->delayed_iput_lock); 22234cb5300bSChris Mason spin_lock_init(&fs_info->defrag_inodes_lock); 22242bf64758SJosef Bacik spin_lock_init(&fs_info->free_chunk_lock); 2225f29021b2SJan Schmidt spin_lock_init(&fs_info->tree_mod_seq_lock); 2226ceda0864SMiao Xie spin_lock_init(&fs_info->super_lock); 2227f28491e0SJosef Bacik spin_lock_init(&fs_info->buffer_lock); 2228f29021b2SJan Schmidt rwlock_init(&fs_info->tree_mod_log_lock); 22297585717fSChris Mason mutex_init(&fs_info->reloc_mutex); 2230573bfb72SMiao Xie mutex_init(&fs_info->delalloc_root_mutex); 2231de98ced9SMiao Xie seqlock_init(&fs_info->profiles_lock); 223219c00ddcSChris Mason 223358176a96SJosef Bacik init_completion(&fs_info->kobj_unregister); 22340b86a832SChris Mason INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); 22356324fbf3SChris Mason INIT_LIST_HEAD(&fs_info->space_info); 2236f29021b2SJan Schmidt INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); 22370b86a832SChris Mason btrfs_mapping_init(&fs_info->mapping_tree); 223866d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->global_block_rsv, 223966d8f3ddSMiao Xie BTRFS_BLOCK_RSV_GLOBAL); 224066d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delalloc_block_rsv, 224166d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELALLOC); 224266d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); 224366d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); 224466d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); 224566d8f3ddSMiao Xie btrfs_init_block_rsv(&fs_info->delayed_block_rsv, 224666d8f3ddSMiao Xie BTRFS_BLOCK_RSV_DELOPS); 2247cb03c743SChris Mason atomic_set(&fs_info->nr_async_submits, 0); 2248771ed689SChris Mason atomic_set(&fs_info->async_delalloc_pages, 0); 22498c8bee1dSChris Mason atomic_set(&fs_info->async_submit_draining, 0); 22500986fe9eSChris Mason atomic_set(&fs_info->nr_async_bios, 0); 22514cb5300bSChris Mason atomic_set(&fs_info->defrag_running, 0); 2252fc36ed7eSJan Schmidt atomic64_set(&fs_info->tree_mod_seq, 0); 2253e20d96d6SChris Mason fs_info->sb = sb; 22546f568d35SChris Mason fs_info->max_inline = 8192 * 1024; 22559ed74f2dSJosef Bacik fs_info->metadata_ratio = 0; 22564cb5300bSChris Mason fs_info->defrag_inodes = RB_ROOT; 22572bf64758SJosef Bacik fs_info->free_chunk_space = 0; 2258f29021b2SJan Schmidt fs_info->tree_mod_log = RB_ROOT; 22598b87dc17SDavid Sterba fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 22600a2b2a84SJosef Bacik fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64); 226190519d66SArne Jansen /* readahead state */ 226290519d66SArne Jansen INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT); 226390519d66SArne Jansen spin_lock_init(&fs_info->reada_lock); 2264c8b97818SChris Mason 2265b34b086cSChris Mason fs_info->thread_pool_size = min_t(unsigned long, 2266b34b086cSChris Mason num_online_cpus() + 2, 8); 22670afbaf8cSChris Mason 2268199c2a9cSMiao Xie INIT_LIST_HEAD(&fs_info->ordered_roots); 2269199c2a9cSMiao Xie spin_lock_init(&fs_info->ordered_root_lock); 227016cdcec7SMiao Xie fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), 227116cdcec7SMiao Xie GFP_NOFS); 227216cdcec7SMiao Xie if (!fs_info->delayed_root) { 227316cdcec7SMiao Xie err = -ENOMEM; 227416cdcec7SMiao Xie goto fail_iput; 227516cdcec7SMiao Xie } 227616cdcec7SMiao Xie btrfs_init_delayed_root(fs_info->delayed_root); 22773eaa2885SChris Mason 2278a2de733cSArne Jansen mutex_init(&fs_info->scrub_lock); 2279a2de733cSArne Jansen atomic_set(&fs_info->scrubs_running, 0); 2280a2de733cSArne Jansen atomic_set(&fs_info->scrub_pause_req, 0); 2281a2de733cSArne Jansen atomic_set(&fs_info->scrubs_paused, 0); 2282a2de733cSArne Jansen atomic_set(&fs_info->scrub_cancel_req, 0); 2283c404e0dcSMiao Xie init_waitqueue_head(&fs_info->replace_wait); 2284a2de733cSArne Jansen init_waitqueue_head(&fs_info->scrub_pause_wait); 2285a2de733cSArne Jansen fs_info->scrub_workers_refcnt = 0; 228621adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 228721adbd5cSStefan Behrens fs_info->check_integrity_print_mask = 0; 228821adbd5cSStefan Behrens #endif 2289a2de733cSArne Jansen 2290c9e9f97bSIlya Dryomov spin_lock_init(&fs_info->balance_lock); 2291c9e9f97bSIlya Dryomov mutex_init(&fs_info->balance_mutex); 2292837d5b6eSIlya Dryomov atomic_set(&fs_info->balance_running, 0); 2293837d5b6eSIlya Dryomov atomic_set(&fs_info->balance_pause_req, 0); 2294a7e99c69SIlya Dryomov atomic_set(&fs_info->balance_cancel_req, 0); 2295c9e9f97bSIlya Dryomov fs_info->balance_ctl = NULL; 2296837d5b6eSIlya Dryomov init_waitqueue_head(&fs_info->balance_wait_q); 2297a061fc8dSChris Mason 22980afbaf8cSChris Mason sb->s_blocksize = 4096; 22990afbaf8cSChris Mason sb->s_blocksize_bits = blksize_bits(4096); 230032a88aa1SJens Axboe sb->s_bdi = &fs_info->bdi; 23010afbaf8cSChris Mason 230276dda93cSYan, Zheng fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 2303bfe86848SMiklos Szeredi set_nlink(fs_info->btree_inode, 1); 23040afbaf8cSChris Mason /* 23050afbaf8cSChris Mason * we set the i_size on the btree inode to the max possible int. 23060afbaf8cSChris Mason * the real end of the address space is determined by all of 23070afbaf8cSChris Mason * the devices in the system 23080afbaf8cSChris Mason */ 23090afbaf8cSChris Mason fs_info->btree_inode->i_size = OFFSET_MAX; 2310d98237b3SChris Mason fs_info->btree_inode->i_mapping->a_ops = &btree_aops; 231104160088SChris Mason fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; 231204160088SChris Mason 23135d4f98a2SYan Zheng RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); 2314d1310b2eSChris Mason extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, 2315f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 23160b32f4bbSJosef Bacik BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; 2317a8067e02SDavid Sterba extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree); 23180da5468fSChris Mason 2319d1310b2eSChris Mason BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; 2320d1310b2eSChris Mason 23210f7d52f4SChris Mason BTRFS_I(fs_info->btree_inode)->root = tree_root; 23220f7d52f4SChris Mason memset(&BTRFS_I(fs_info->btree_inode)->location, 0, 23230f7d52f4SChris Mason sizeof(struct btrfs_key)); 232472ac3c0dSJosef Bacik set_bit(BTRFS_INODE_DUMMY, 232572ac3c0dSJosef Bacik &BTRFS_I(fs_info->btree_inode)->runtime_flags); 2326778ba82bSFilipe David Borba Manana btrfs_insert_inode_hash(fs_info->btree_inode); 232739279cc3SChris Mason 2328e02119d5SChris Mason spin_lock_init(&fs_info->block_group_cache_lock); 23296bef4d31SEric Paris fs_info->block_group_cache_tree = RB_ROOT; 2330a1897fddSLiu Bo fs_info->first_logical_byte = (u64)-1; 2331925baeddSChris Mason 233211833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[0], 2333f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 233411833d66SYan Zheng extent_io_tree_init(&fs_info->freed_extents[1], 2335f993c883SDavid Sterba fs_info->btree_inode->i_mapping); 233611833d66SYan Zheng fs_info->pinned_extents = &fs_info->freed_extents[0]; 2337e6dcd2dcSChris Mason fs_info->do_barriers = 1; 2338f9295749SChris Mason 2339d98237b3SChris Mason 23405a3f23d5SChris Mason mutex_init(&fs_info->ordered_operations_mutex); 23419ffba8cdSJosef Bacik mutex_init(&fs_info->ordered_extent_flush_mutex); 234230ae8467SChris Mason mutex_init(&fs_info->tree_log_mutex); 2343925baeddSChris Mason mutex_init(&fs_info->chunk_mutex); 2344925baeddSChris Mason mutex_init(&fs_info->transaction_kthread_mutex); 2345a74a4b97SChris Mason mutex_init(&fs_info->cleaner_mutex); 2346a74a4b97SChris Mason mutex_init(&fs_info->volume_mutex); 23479e351cc8SJosef Bacik init_rwsem(&fs_info->commit_root_sem); 2348c71bf099SYan, Zheng init_rwsem(&fs_info->cleanup_work_sem); 234976dda93cSYan, Zheng init_rwsem(&fs_info->subvol_sem); 2350803b2f54SStefan Behrens sema_init(&fs_info->uuid_tree_rescan_sem, 1); 2351e922e087SStefan Behrens fs_info->dev_replace.lock_owner = 0; 2352e922e087SStefan Behrens atomic_set(&fs_info->dev_replace.nesting_level, 0); 2353e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); 2354e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock_management_lock); 2355e922e087SStefan Behrens mutex_init(&fs_info->dev_replace.lock); 2356fa9c0d79SChris Mason 2357416ac51dSArne Jansen spin_lock_init(&fs_info->qgroup_lock); 2358f2f6ed3dSWang Shilong mutex_init(&fs_info->qgroup_ioctl_lock); 2359416ac51dSArne Jansen fs_info->qgroup_tree = RB_ROOT; 2360416ac51dSArne Jansen INIT_LIST_HEAD(&fs_info->dirty_qgroups); 2361416ac51dSArne Jansen fs_info->qgroup_seq = 1; 2362416ac51dSArne Jansen fs_info->quota_enabled = 0; 2363416ac51dSArne Jansen fs_info->pending_quota_state = 0; 23641e8f9158SWang Shilong fs_info->qgroup_ulist = NULL; 23652f232036SJan Schmidt mutex_init(&fs_info->qgroup_rescan_lock); 2366416ac51dSArne Jansen 2367fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); 2368fa9c0d79SChris Mason btrfs_init_free_cluster(&fs_info->data_alloc_cluster); 2369fa9c0d79SChris Mason 23707d9eb12cSChris Mason init_waitqueue_head(&fs_info->transaction_throttle); 23713768f368SChris Mason init_waitqueue_head(&fs_info->transaction_wait); 2372bb9c12c9SSage Weil init_waitqueue_head(&fs_info->transaction_blocked_wait); 23734854ddd0SChris Mason init_waitqueue_head(&fs_info->async_submit_wait); 23749a8dd150SChris Mason 237553b381b3SDavid Woodhouse ret = btrfs_alloc_stripe_hash_table(fs_info); 237653b381b3SDavid Woodhouse if (ret) { 237783c8266aSDavid Sterba err = ret; 237853b381b3SDavid Woodhouse goto fail_alloc; 237953b381b3SDavid Woodhouse } 238053b381b3SDavid Woodhouse 23810b86a832SChris Mason __setup_root(4096, 4096, 4096, 4096, tree_root, 23822c90e5d6SChris Mason fs_info, BTRFS_ROOT_TREE_OBJECTID); 23837eccb903SChris Mason 23843c4bb26bSChris Mason invalidate_bdev(fs_devices->latest_bdev); 23851104a885SDavid Sterba 23861104a885SDavid Sterba /* 23871104a885SDavid Sterba * Read super block and check the signature bytes only 23881104a885SDavid Sterba */ 2389a512bbf8SYan Zheng bh = btrfs_read_dev_super(fs_devices->latest_bdev); 239020b45077SDave Young if (!bh) { 239120b45077SDave Young err = -EINVAL; 239216cdcec7SMiao Xie goto fail_alloc; 239320b45077SDave Young } 239439279cc3SChris Mason 23951104a885SDavid Sterba /* 23961104a885SDavid Sterba * We want to check superblock checksum, the type is stored inside. 23971104a885SDavid Sterba * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). 23981104a885SDavid Sterba */ 23991104a885SDavid Sterba if (btrfs_check_super_csum(bh->b_data)) { 2400efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock checksum mismatch\n"); 24011104a885SDavid Sterba err = -EINVAL; 24021104a885SDavid Sterba goto fail_alloc; 24031104a885SDavid Sterba } 24041104a885SDavid Sterba 24051104a885SDavid Sterba /* 24061104a885SDavid Sterba * super_copy is zeroed at allocation time and we never touch the 24071104a885SDavid Sterba * following bytes up to INFO_SIZE, the checksum is calculated from 24081104a885SDavid Sterba * the whole block of INFO_SIZE 24091104a885SDavid Sterba */ 24106c41761fSDavid Sterba memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy)); 24116c41761fSDavid Sterba memcpy(fs_info->super_for_commit, fs_info->super_copy, 24126c41761fSDavid Sterba sizeof(*fs_info->super_for_commit)); 2413a061fc8dSChris Mason brelse(bh); 24145f39d397SChris Mason 24156c41761fSDavid Sterba memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 24160b86a832SChris Mason 24171104a885SDavid Sterba ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); 24181104a885SDavid Sterba if (ret) { 2419efe120a0SFrank Holton printk(KERN_ERR "BTRFS: superblock contains fatal errors\n"); 24201104a885SDavid Sterba err = -EINVAL; 24211104a885SDavid Sterba goto fail_alloc; 24221104a885SDavid Sterba } 24231104a885SDavid Sterba 24246c41761fSDavid Sterba disk_super = fs_info->super_copy; 24250f7d52f4SChris Mason if (!btrfs_super_root(disk_super)) 242616cdcec7SMiao Xie goto fail_alloc; 24270f7d52f4SChris Mason 2428acce952bSliubo /* check FS state, whether FS is broken. */ 242987533c47SMiao Xie if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) 243087533c47SMiao Xie set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 2431acce952bSliubo 243275e7cb7fSLiu Bo /* 2433af31f5e5SChris Mason * run through our array of backup supers and setup 2434af31f5e5SChris Mason * our ring pointer to the oldest one 2435af31f5e5SChris Mason */ 2436af31f5e5SChris Mason generation = btrfs_super_generation(disk_super); 2437af31f5e5SChris Mason find_oldest_super_backup(fs_info, generation); 2438af31f5e5SChris Mason 2439af31f5e5SChris Mason /* 244075e7cb7fSLiu Bo * In the long term, we'll store the compression type in the super 244175e7cb7fSLiu Bo * block, and it'll be used for per file compression control. 244275e7cb7fSLiu Bo */ 244375e7cb7fSLiu Bo fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 244475e7cb7fSLiu Bo 24452b82032cSYan Zheng ret = btrfs_parse_options(tree_root, options); 24462b82032cSYan Zheng if (ret) { 24472b82032cSYan Zheng err = ret; 244816cdcec7SMiao Xie goto fail_alloc; 24492b82032cSYan Zheng } 2450dfe25020SChris Mason 2451f2b636e8SJosef Bacik features = btrfs_super_incompat_flags(disk_super) & 2452f2b636e8SJosef Bacik ~BTRFS_FEATURE_INCOMPAT_SUPP; 2453f2b636e8SJosef Bacik if (features) { 2454f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount because of " 2455f2b636e8SJosef Bacik "unsupported optional features (%Lx).\n", 2456c1c9ff7cSGeert Uytterhoeven features); 2457f2b636e8SJosef Bacik err = -EINVAL; 245816cdcec7SMiao Xie goto fail_alloc; 2459f2b636e8SJosef Bacik } 2460f2b636e8SJosef Bacik 2461727011e0SChris Mason if (btrfs_super_leafsize(disk_super) != 2462727011e0SChris Mason btrfs_super_nodesize(disk_super)) { 2463727011e0SChris Mason printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2464727011e0SChris Mason "blocksizes don't match. node %d leaf %d\n", 2465727011e0SChris Mason btrfs_super_nodesize(disk_super), 2466727011e0SChris Mason btrfs_super_leafsize(disk_super)); 2467727011e0SChris Mason err = -EINVAL; 2468727011e0SChris Mason goto fail_alloc; 2469727011e0SChris Mason } 2470727011e0SChris Mason if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) { 2471727011e0SChris Mason printk(KERN_ERR "BTRFS: couldn't mount because metadata " 2472727011e0SChris Mason "blocksize (%d) was too large\n", 2473727011e0SChris Mason btrfs_super_leafsize(disk_super)); 2474727011e0SChris Mason err = -EINVAL; 2475727011e0SChris Mason goto fail_alloc; 2476727011e0SChris Mason } 2477727011e0SChris Mason 24785d4f98a2SYan Zheng features = btrfs_super_incompat_flags(disk_super); 24795d4f98a2SYan Zheng features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 248069e380d1SLi Zefan if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 2481a6fa6faeSLi Zefan features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 2482727011e0SChris Mason 24833173a18fSJosef Bacik if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 2484efe120a0SFrank Holton printk(KERN_ERR "BTRFS: has skinny extents\n"); 24853173a18fSJosef Bacik 2486727011e0SChris Mason /* 2487727011e0SChris Mason * flag our filesystem as having big metadata blocks if 2488727011e0SChris Mason * they are bigger than the page size 2489727011e0SChris Mason */ 2490727011e0SChris Mason if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) { 2491727011e0SChris Mason if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA)) 2492efe120a0SFrank Holton printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n"); 2493727011e0SChris Mason features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; 2494727011e0SChris Mason } 2495727011e0SChris Mason 2496bc3f116fSChris Mason nodesize = btrfs_super_nodesize(disk_super); 2497bc3f116fSChris Mason leafsize = btrfs_super_leafsize(disk_super); 2498bc3f116fSChris Mason sectorsize = btrfs_super_sectorsize(disk_super); 2499bc3f116fSChris Mason stripesize = btrfs_super_stripesize(disk_super); 2500e2d84521SMiao Xie fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids)); 2501963d678bSMiao Xie fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2502bc3f116fSChris Mason 2503bc3f116fSChris Mason /* 2504bc3f116fSChris Mason * mixed block groups end up with duplicate but slightly offset 2505bc3f116fSChris Mason * extent buffers for the same range. It leads to corruptions 2506bc3f116fSChris Mason */ 2507bc3f116fSChris Mason if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && 2508bc3f116fSChris Mason (sectorsize != leafsize)) { 2509efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes " 2510bc3f116fSChris Mason "are not allowed for mixed block groups on %s\n", 2511bc3f116fSChris Mason sb->s_id); 2512bc3f116fSChris Mason goto fail_alloc; 2513bc3f116fSChris Mason } 2514bc3f116fSChris Mason 2515ceda0864SMiao Xie /* 2516ceda0864SMiao Xie * Needn't use the lock because there is no other task which will 2517ceda0864SMiao Xie * update the flag. 2518ceda0864SMiao Xie */ 25195d4f98a2SYan Zheng btrfs_set_super_incompat_flags(disk_super, features); 25205d4f98a2SYan Zheng 2521f2b636e8SJosef Bacik features = btrfs_super_compat_ro_flags(disk_super) & 2522f2b636e8SJosef Bacik ~BTRFS_FEATURE_COMPAT_RO_SUPP; 2523f2b636e8SJosef Bacik if (!(sb->s_flags & MS_RDONLY) && features) { 2524f2b636e8SJosef Bacik printk(KERN_ERR "BTRFS: couldn't mount RDWR because of " 2525f2b636e8SJosef Bacik "unsupported option features (%Lx).\n", 2526c1c9ff7cSGeert Uytterhoeven features); 2527f2b636e8SJosef Bacik err = -EINVAL; 252816cdcec7SMiao Xie goto fail_alloc; 2529f2b636e8SJosef Bacik } 253061d92c32SChris Mason 25315cdc7ad3SQu Wenruo max_active = fs_info->thread_pool_size; 253261d92c32SChris Mason 25335cdc7ad3SQu Wenruo fs_info->workers = 25345cdc7ad3SQu Wenruo btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI, 25355cdc7ad3SQu Wenruo max_active, 16); 2536c8b97818SChris Mason 2537afe3d242SQu Wenruo fs_info->delalloc_workers = 2538afe3d242SQu Wenruo btrfs_alloc_workqueue("delalloc", flags, max_active, 2); 2539771ed689SChris Mason 2540a44903abSQu Wenruo fs_info->flush_workers = 2541a44903abSQu Wenruo btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0); 254261b49440SChris Mason 2543e66f0bb1SQu Wenruo fs_info->caching_workers = 2544e66f0bb1SQu Wenruo btrfs_alloc_workqueue("cache", flags, max_active, 0); 2545bab39bf9SJosef Bacik 2546a8c93d4eSQu Wenruo /* 2547a8c93d4eSQu Wenruo * a higher idle thresh on the submit workers makes it much more 254861b49440SChris Mason * likely that bios will be send down in a sane order to the 254961b49440SChris Mason * devices 255061b49440SChris Mason */ 2551a8c93d4eSQu Wenruo fs_info->submit_workers = 2552a8c93d4eSQu Wenruo btrfs_alloc_workqueue("submit", flags, 2553a8c93d4eSQu Wenruo min_t(u64, fs_devices->num_devices, 2554a8c93d4eSQu Wenruo max_active), 64); 255553863232SChris Mason 2556dc6e3209SQu Wenruo fs_info->fixup_workers = 2557dc6e3209SQu Wenruo btrfs_alloc_workqueue("fixup", flags, 1, 0); 2558fccb5d86SQu Wenruo 2559fccb5d86SQu Wenruo /* 2560fccb5d86SQu Wenruo * endios are largely parallel and should have a very 2561fccb5d86SQu Wenruo * low idle thresh 2562fccb5d86SQu Wenruo */ 2563fccb5d86SQu Wenruo fs_info->endio_workers = 2564fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio", flags, max_active, 4); 2565fccb5d86SQu Wenruo fs_info->endio_meta_workers = 2566fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-meta", flags, max_active, 4); 2567fccb5d86SQu Wenruo fs_info->endio_meta_write_workers = 2568fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2); 2569fccb5d86SQu Wenruo fs_info->endio_raid56_workers = 2570fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4); 2571d05a33acSQu Wenruo fs_info->rmw_workers = 2572d05a33acSQu Wenruo btrfs_alloc_workqueue("rmw", flags, max_active, 2); 2573fccb5d86SQu Wenruo fs_info->endio_write_workers = 2574fccb5d86SQu Wenruo btrfs_alloc_workqueue("endio-write", flags, max_active, 2); 2575fccb5d86SQu Wenruo fs_info->endio_freespace_worker = 2576fccb5d86SQu Wenruo btrfs_alloc_workqueue("freespace-write", flags, max_active, 0); 25775b3bc44eSQu Wenruo fs_info->delayed_workers = 25785b3bc44eSQu Wenruo btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0); 2579736cfa15SQu Wenruo fs_info->readahead_workers = 2580736cfa15SQu Wenruo btrfs_alloc_workqueue("readahead", flags, max_active, 2); 2581fc97fab0SQu Wenruo fs_info->qgroup_rescan_workers = 2582fc97fab0SQu Wenruo btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0); 258361b49440SChris Mason 2584a8c93d4eSQu Wenruo if (!(fs_info->workers && fs_info->delalloc_workers && 2585fccb5d86SQu Wenruo fs_info->submit_workers && fs_info->flush_workers && 2586fccb5d86SQu Wenruo fs_info->endio_workers && fs_info->endio_meta_workers && 2587fccb5d86SQu Wenruo fs_info->endio_meta_write_workers && 2588fccb5d86SQu Wenruo fs_info->endio_write_workers && fs_info->endio_raid56_workers && 2589e66f0bb1SQu Wenruo fs_info->endio_freespace_worker && fs_info->rmw_workers && 2590dc6e3209SQu Wenruo fs_info->caching_workers && fs_info->readahead_workers && 2591fc97fab0SQu Wenruo fs_info->fixup_workers && fs_info->delayed_workers && 2592fc97fab0SQu Wenruo fs_info->qgroup_rescan_workers)) { 25935cdc7ad3SQu Wenruo err = -ENOMEM; 25945cdc7ad3SQu Wenruo goto fail_sb_buffer; 25955cdc7ad3SQu Wenruo } 25964543df7eSChris Mason 25974575c9ccSChris Mason fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); 2598c8b97818SChris Mason fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 2599c8b97818SChris Mason 4 * 1024 * 1024 / PAGE_CACHE_SIZE); 26004575c9ccSChris Mason 2601db94535dSChris Mason tree_root->nodesize = nodesize; 2602db94535dSChris Mason tree_root->leafsize = leafsize; 2603db94535dSChris Mason tree_root->sectorsize = sectorsize; 260487ee04ebSChris Mason tree_root->stripesize = stripesize; 2605a061fc8dSChris Mason 2606a061fc8dSChris Mason sb->s_blocksize = sectorsize; 2607a061fc8dSChris Mason sb->s_blocksize_bits = blksize_bits(sectorsize); 2608db94535dSChris Mason 26093cae210fSQu Wenruo if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) { 2610efe120a0SFrank Holton printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id); 261139279cc3SChris Mason goto fail_sb_buffer; 261239279cc3SChris Mason } 261319c00ddcSChris Mason 26148d082fb7SLiu Bo if (sectorsize != PAGE_SIZE) { 2615efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) " 26168d082fb7SLiu Bo "found on %s\n", (unsigned long)sectorsize, sb->s_id); 2617941b2ddfSKeith Mannthey goto fail_sb_buffer; 2618941b2ddfSKeith Mannthey } 2619941b2ddfSKeith Mannthey 2620925baeddSChris Mason mutex_lock(&fs_info->chunk_mutex); 2621e4404d6eSYan Zheng ret = btrfs_read_sys_array(tree_root); 2622925baeddSChris Mason mutex_unlock(&fs_info->chunk_mutex); 262384eed90fSChris Mason if (ret) { 2624efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read the system " 2625d397712bSChris Mason "array on %s\n", sb->s_id); 26265d4f98a2SYan Zheng goto fail_sb_buffer; 262784eed90fSChris Mason } 26280b86a832SChris Mason 26290b86a832SChris Mason blocksize = btrfs_level_size(tree_root, 26300b86a832SChris Mason btrfs_super_chunk_root_level(disk_super)); 263184234f3aSYan Zheng generation = btrfs_super_chunk_root_generation(disk_super); 26320b86a832SChris Mason 26330b86a832SChris Mason __setup_root(nodesize, leafsize, sectorsize, stripesize, 26340b86a832SChris Mason chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); 26350b86a832SChris Mason 26360b86a832SChris Mason chunk_root->node = read_tree_block(chunk_root, 26370b86a832SChris Mason btrfs_super_chunk_root(disk_super), 263884234f3aSYan Zheng blocksize, generation); 2639416bc658SJosef Bacik if (!chunk_root->node || 2640416bc658SJosef Bacik !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) { 2641efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n", 264283121942SDavid Woodhouse sb->s_id); 2643af31f5e5SChris Mason goto fail_tree_roots; 264483121942SDavid Woodhouse } 26455d4f98a2SYan Zheng btrfs_set_root_node(&chunk_root->root_item, chunk_root->node); 26465d4f98a2SYan Zheng chunk_root->commit_root = btrfs_root_node(chunk_root); 26470b86a832SChris Mason 2648e17cade2SChris Mason read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2649b308bc2fSGeert Uytterhoeven btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2650e17cade2SChris Mason 26510b86a832SChris Mason ret = btrfs_read_chunk_tree(chunk_root); 26522b82032cSYan Zheng if (ret) { 2653efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n", 2654d397712bSChris Mason sb->s_id); 2655af31f5e5SChris Mason goto fail_tree_roots; 26562b82032cSYan Zheng } 26570b86a832SChris Mason 26588dabb742SStefan Behrens /* 26598dabb742SStefan Behrens * keep the device that is marked to be the target device for the 26608dabb742SStefan Behrens * dev_replace procedure 26618dabb742SStefan Behrens */ 26628dabb742SStefan Behrens btrfs_close_extra_devices(fs_info, fs_devices, 0); 2663dfe25020SChris Mason 2664a6b0d5c8SChris Mason if (!fs_devices->latest_bdev) { 2665efe120a0SFrank Holton printk(KERN_CRIT "BTRFS: failed to read devices on %s\n", 2666a6b0d5c8SChris Mason sb->s_id); 2667a6b0d5c8SChris Mason goto fail_tree_roots; 2668a6b0d5c8SChris Mason } 2669a6b0d5c8SChris Mason 2670af31f5e5SChris Mason retry_root_backup: 2671db94535dSChris Mason blocksize = btrfs_level_size(tree_root, 2672db94535dSChris Mason btrfs_super_root_level(disk_super)); 267384234f3aSYan Zheng generation = btrfs_super_generation(disk_super); 26740b86a832SChris Mason 2675e20d96d6SChris Mason tree_root->node = read_tree_block(tree_root, 2676db94535dSChris Mason btrfs_super_root(disk_super), 267784234f3aSYan Zheng blocksize, generation); 2678af31f5e5SChris Mason if (!tree_root->node || 2679af31f5e5SChris Mason !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) { 2680efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n", 268183121942SDavid Woodhouse sb->s_id); 2682af31f5e5SChris Mason 2683af31f5e5SChris Mason goto recovery_tree_root; 268483121942SDavid Woodhouse } 2685af31f5e5SChris Mason 26865d4f98a2SYan Zheng btrfs_set_root_node(&tree_root->root_item, tree_root->node); 26875d4f98a2SYan Zheng tree_root->commit_root = btrfs_root_node(tree_root); 268869e9c6c6SStefan Behrens btrfs_set_root_refs(&tree_root->root_item, 1); 2689db94535dSChris Mason 2690cb517eabSMiao Xie location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 2691cb517eabSMiao Xie location.type = BTRFS_ROOT_ITEM_KEY; 2692cb517eabSMiao Xie location.offset = 0; 2693cb517eabSMiao Xie 2694cb517eabSMiao Xie extent_root = btrfs_read_tree_root(tree_root, &location); 2695cb517eabSMiao Xie if (IS_ERR(extent_root)) { 2696cb517eabSMiao Xie ret = PTR_ERR(extent_root); 2697af31f5e5SChris Mason goto recovery_tree_root; 2698cb517eabSMiao Xie } 26990b86a832SChris Mason extent_root->track_dirty = 1; 2700cb517eabSMiao Xie fs_info->extent_root = extent_root; 27010b86a832SChris Mason 2702cb517eabSMiao Xie location.objectid = BTRFS_DEV_TREE_OBJECTID; 2703cb517eabSMiao Xie dev_root = btrfs_read_tree_root(tree_root, &location); 2704cb517eabSMiao Xie if (IS_ERR(dev_root)) { 2705cb517eabSMiao Xie ret = PTR_ERR(dev_root); 2706af31f5e5SChris Mason goto recovery_tree_root; 2707cb517eabSMiao Xie } 27085d4f98a2SYan Zheng dev_root->track_dirty = 1; 2709cb517eabSMiao Xie fs_info->dev_root = dev_root; 2710cb517eabSMiao Xie btrfs_init_devices_late(fs_info); 27113768f368SChris Mason 2712cb517eabSMiao Xie location.objectid = BTRFS_CSUM_TREE_OBJECTID; 2713cb517eabSMiao Xie csum_root = btrfs_read_tree_root(tree_root, &location); 2714cb517eabSMiao Xie if (IS_ERR(csum_root)) { 2715cb517eabSMiao Xie ret = PTR_ERR(csum_root); 2716af31f5e5SChris Mason goto recovery_tree_root; 2717cb517eabSMiao Xie } 2718d20f7043SChris Mason csum_root->track_dirty = 1; 2719cb517eabSMiao Xie fs_info->csum_root = csum_root; 2720d20f7043SChris Mason 2721cb517eabSMiao Xie location.objectid = BTRFS_QUOTA_TREE_OBJECTID; 2722cb517eabSMiao Xie quota_root = btrfs_read_tree_root(tree_root, &location); 2723cb517eabSMiao Xie if (!IS_ERR(quota_root)) { 2724bcef60f2SArne Jansen quota_root->track_dirty = 1; 2725bcef60f2SArne Jansen fs_info->quota_enabled = 1; 2726bcef60f2SArne Jansen fs_info->pending_quota_state = 1; 2727cb517eabSMiao Xie fs_info->quota_root = quota_root; 2728bcef60f2SArne Jansen } 2729bcef60f2SArne Jansen 2730f7a81ea4SStefan Behrens location.objectid = BTRFS_UUID_TREE_OBJECTID; 2731f7a81ea4SStefan Behrens uuid_root = btrfs_read_tree_root(tree_root, &location); 2732f7a81ea4SStefan Behrens if (IS_ERR(uuid_root)) { 2733f7a81ea4SStefan Behrens ret = PTR_ERR(uuid_root); 2734f7a81ea4SStefan Behrens if (ret != -ENOENT) 2735f7a81ea4SStefan Behrens goto recovery_tree_root; 2736f7a81ea4SStefan Behrens create_uuid_tree = true; 273770f80175SStefan Behrens check_uuid_tree = false; 2738f7a81ea4SStefan Behrens } else { 2739f7a81ea4SStefan Behrens uuid_root->track_dirty = 1; 2740f7a81ea4SStefan Behrens fs_info->uuid_root = uuid_root; 274170f80175SStefan Behrens create_uuid_tree = false; 274270f80175SStefan Behrens check_uuid_tree = 274370f80175SStefan Behrens generation != btrfs_super_uuid_tree_generation(disk_super); 2744f7a81ea4SStefan Behrens } 2745f7a81ea4SStefan Behrens 27468929ecfaSYan, Zheng fs_info->generation = generation; 27478929ecfaSYan, Zheng fs_info->last_trans_committed = generation; 27488929ecfaSYan, Zheng 274968310a5eSIlya Dryomov ret = btrfs_recover_balance(fs_info); 275068310a5eSIlya Dryomov if (ret) { 2751efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to recover balance\n"); 275268310a5eSIlya Dryomov goto fail_block_groups; 275368310a5eSIlya Dryomov } 275468310a5eSIlya Dryomov 2755733f4fbbSStefan Behrens ret = btrfs_init_dev_stats(fs_info); 2756733f4fbbSStefan Behrens if (ret) { 2757efe120a0SFrank Holton printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n", 2758733f4fbbSStefan Behrens ret); 2759733f4fbbSStefan Behrens goto fail_block_groups; 2760733f4fbbSStefan Behrens } 2761733f4fbbSStefan Behrens 27628dabb742SStefan Behrens ret = btrfs_init_dev_replace(fs_info); 27638dabb742SStefan Behrens if (ret) { 2764efe120a0SFrank Holton pr_err("BTRFS: failed to init dev_replace: %d\n", ret); 27658dabb742SStefan Behrens goto fail_block_groups; 27668dabb742SStefan Behrens } 27678dabb742SStefan Behrens 27688dabb742SStefan Behrens btrfs_close_extra_devices(fs_info, fs_devices, 1); 27698dabb742SStefan Behrens 27705ac1d209SJeff Mahoney ret = btrfs_sysfs_add_one(fs_info); 27715ac1d209SJeff Mahoney if (ret) { 2772efe120a0SFrank Holton pr_err("BTRFS: failed to init sysfs interface: %d\n", ret); 27735ac1d209SJeff Mahoney goto fail_block_groups; 27745ac1d209SJeff Mahoney } 27755ac1d209SJeff Mahoney 2776c59021f8Sliubo ret = btrfs_init_space_info(fs_info); 2777c59021f8Sliubo if (ret) { 2778efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret); 27792365dd3cSAnand Jain goto fail_sysfs; 2780c59021f8Sliubo } 2781c59021f8Sliubo 27821b1d1f66SJosef Bacik ret = btrfs_read_block_groups(extent_root); 27831b1d1f66SJosef Bacik if (ret) { 2784efe120a0SFrank Holton printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret); 27852365dd3cSAnand Jain goto fail_sysfs; 27861b1d1f66SJosef Bacik } 27875af3e8ccSStefan Behrens fs_info->num_tolerated_disk_barrier_failures = 27885af3e8ccSStefan Behrens btrfs_calc_num_tolerated_disk_barrier_failures(fs_info); 2789292fd7fcSStefan Behrens if (fs_info->fs_devices->missing_devices > 2790292fd7fcSStefan Behrens fs_info->num_tolerated_disk_barrier_failures && 2791292fd7fcSStefan Behrens !(sb->s_flags & MS_RDONLY)) { 2792efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: " 2793efe120a0SFrank Holton "too many missing devices, writeable mount is not allowed\n"); 27942365dd3cSAnand Jain goto fail_sysfs; 2795292fd7fcSStefan Behrens } 27969078a3e1SChris Mason 2797a74a4b97SChris Mason fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 2798a74a4b97SChris Mason "btrfs-cleaner"); 279957506d50SQinghuang Feng if (IS_ERR(fs_info->cleaner_kthread)) 28002365dd3cSAnand Jain goto fail_sysfs; 2801a74a4b97SChris Mason 2802a74a4b97SChris Mason fs_info->transaction_kthread = kthread_run(transaction_kthread, 2803a74a4b97SChris Mason tree_root, 2804a74a4b97SChris Mason "btrfs-transaction"); 280557506d50SQinghuang Feng if (IS_ERR(fs_info->transaction_kthread)) 28063f157a2fSChris Mason goto fail_cleaner; 2807a74a4b97SChris Mason 2808c289811cSChris Mason if (!btrfs_test_opt(tree_root, SSD) && 2809c289811cSChris Mason !btrfs_test_opt(tree_root, NOSSD) && 2810c289811cSChris Mason !fs_info->fs_devices->rotating) { 2811efe120a0SFrank Holton printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD " 2812c289811cSChris Mason "mode\n"); 2813c289811cSChris Mason btrfs_set_opt(fs_info->mount_opt, SSD); 2814c289811cSChris Mason } 2815c289811cSChris Mason 28163818aea2SQu Wenruo /* Set the real inode map cache flag */ 28173818aea2SQu Wenruo if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE)) 28183818aea2SQu Wenruo btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE); 28193818aea2SQu Wenruo 282021adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 282121adbd5cSStefan Behrens if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) { 282221adbd5cSStefan Behrens ret = btrfsic_mount(tree_root, fs_devices, 282321adbd5cSStefan Behrens btrfs_test_opt(tree_root, 282421adbd5cSStefan Behrens CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 282521adbd5cSStefan Behrens 1 : 0, 282621adbd5cSStefan Behrens fs_info->check_integrity_print_mask); 282721adbd5cSStefan Behrens if (ret) 2828efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to initialize" 282921adbd5cSStefan Behrens " integrity check module %s\n", sb->s_id); 283021adbd5cSStefan Behrens } 283121adbd5cSStefan Behrens #endif 2832bcef60f2SArne Jansen ret = btrfs_read_qgroup_config(fs_info); 2833bcef60f2SArne Jansen if (ret) 2834bcef60f2SArne Jansen goto fail_trans_kthread; 283521adbd5cSStefan Behrens 2836acce952bSliubo /* do not make disk changes in broken FS */ 283768ce9682SStefan Behrens if (btrfs_super_log_root(disk_super) != 0) { 2838e02119d5SChris Mason u64 bytenr = btrfs_super_log_root(disk_super); 2839d18a2c44SChris Mason 28407c2ca468SChris Mason if (fs_devices->rw_devices == 0) { 2841efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: log replay required " 2842d397712bSChris Mason "on RO media\n"); 28437c2ca468SChris Mason err = -EIO; 2844bcef60f2SArne Jansen goto fail_qgroup; 28457c2ca468SChris Mason } 2846e02119d5SChris Mason blocksize = 2847e02119d5SChris Mason btrfs_level_size(tree_root, 2848e02119d5SChris Mason btrfs_super_log_root_level(disk_super)); 2849e02119d5SChris Mason 28506f07e42eSAl Viro log_tree_root = btrfs_alloc_root(fs_info); 2851676e4c86SDan Carpenter if (!log_tree_root) { 2852676e4c86SDan Carpenter err = -ENOMEM; 2853bcef60f2SArne Jansen goto fail_qgroup; 2854676e4c86SDan Carpenter } 2855e02119d5SChris Mason 2856e02119d5SChris Mason __setup_root(nodesize, leafsize, sectorsize, stripesize, 2857e02119d5SChris Mason log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); 2858e02119d5SChris Mason 2859e02119d5SChris Mason log_tree_root->node = read_tree_block(tree_root, bytenr, 286084234f3aSYan Zheng blocksize, 286184234f3aSYan Zheng generation + 1); 2862416bc658SJosef Bacik if (!log_tree_root->node || 2863416bc658SJosef Bacik !extent_buffer_uptodate(log_tree_root->node)) { 2864efe120a0SFrank Holton printk(KERN_ERR "BTRFS: failed to read log tree\n"); 2865416bc658SJosef Bacik free_extent_buffer(log_tree_root->node); 2866416bc658SJosef Bacik kfree(log_tree_root); 2867416bc658SJosef Bacik goto fail_trans_kthread; 2868416bc658SJosef Bacik } 286979787eaaSJeff Mahoney /* returns with log_tree_root freed on success */ 2870e02119d5SChris Mason ret = btrfs_recover_log_trees(log_tree_root); 287179787eaaSJeff Mahoney if (ret) { 287279787eaaSJeff Mahoney btrfs_error(tree_root->fs_info, ret, 287379787eaaSJeff Mahoney "Failed to recover log tree"); 287479787eaaSJeff Mahoney free_extent_buffer(log_tree_root->node); 287579787eaaSJeff Mahoney kfree(log_tree_root); 287679787eaaSJeff Mahoney goto fail_trans_kthread; 287779787eaaSJeff Mahoney } 2878e556ce2cSYan Zheng 2879e556ce2cSYan Zheng if (sb->s_flags & MS_RDONLY) { 2880e556ce2cSYan Zheng ret = btrfs_commit_super(tree_root); 288179787eaaSJeff Mahoney if (ret) 288279787eaaSJeff Mahoney goto fail_trans_kthread; 2883e556ce2cSYan Zheng } 2884e02119d5SChris Mason } 28851a40e23bSZheng Yan 288676dda93cSYan, Zheng ret = btrfs_find_orphan_roots(tree_root); 288779787eaaSJeff Mahoney if (ret) 288879787eaaSJeff Mahoney goto fail_trans_kthread; 288976dda93cSYan, Zheng 28907c2ca468SChris Mason if (!(sb->s_flags & MS_RDONLY)) { 2891d68fc57bSYan, Zheng ret = btrfs_cleanup_fs_roots(fs_info); 289244c44af2SIlya Dryomov if (ret) 289344c44af2SIlya Dryomov goto fail_trans_kthread; 2894d68fc57bSYan, Zheng 28955d4f98a2SYan Zheng ret = btrfs_recover_relocation(tree_root); 2896d7ce5843SMiao Xie if (ret < 0) { 2897d7ce5843SMiao Xie printk(KERN_WARNING 2898efe120a0SFrank Holton "BTRFS: failed to recover relocation\n"); 2899d7ce5843SMiao Xie err = -EINVAL; 2900bcef60f2SArne Jansen goto fail_qgroup; 2901d7ce5843SMiao Xie } 29027c2ca468SChris Mason } 29031a40e23bSZheng Yan 29043de4586cSChris Mason location.objectid = BTRFS_FS_TREE_OBJECTID; 29053de4586cSChris Mason location.type = BTRFS_ROOT_ITEM_KEY; 2906cb517eabSMiao Xie location.offset = 0; 29073de4586cSChris Mason 29083de4586cSChris Mason fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); 29093140c9a3SDan Carpenter if (IS_ERR(fs_info->fs_root)) { 29103140c9a3SDan Carpenter err = PTR_ERR(fs_info->fs_root); 2911bcef60f2SArne Jansen goto fail_qgroup; 29123140c9a3SDan Carpenter } 2913c289811cSChris Mason 29142b6ba629SIlya Dryomov if (sb->s_flags & MS_RDONLY) 29152b6ba629SIlya Dryomov return 0; 29162b6ba629SIlya Dryomov 2917e3acc2a6SJosef Bacik down_read(&fs_info->cleanup_work_sem); 29182b6ba629SIlya Dryomov if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 29192b6ba629SIlya Dryomov (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 29202b6ba629SIlya Dryomov up_read(&fs_info->cleanup_work_sem); 29212b6ba629SIlya Dryomov close_ctree(tree_root); 29222b6ba629SIlya Dryomov return ret; 29232b6ba629SIlya Dryomov } 2924e3acc2a6SJosef Bacik up_read(&fs_info->cleanup_work_sem); 292559641015SIlya Dryomov 29262b6ba629SIlya Dryomov ret = btrfs_resume_balance_async(fs_info); 29272b6ba629SIlya Dryomov if (ret) { 2928efe120a0SFrank Holton printk(KERN_WARNING "BTRFS: failed to resume balance\n"); 292966b4ffd1SJosef Bacik close_ctree(tree_root); 29302b6ba629SIlya Dryomov return ret; 2931e3acc2a6SJosef Bacik } 2932e3acc2a6SJosef Bacik 29338dabb742SStefan Behrens ret = btrfs_resume_dev_replace_async(fs_info); 29348dabb742SStefan Behrens if (ret) { 2935efe120a0SFrank Holton pr_warn("BTRFS: failed to resume dev_replace\n"); 29368dabb742SStefan Behrens close_ctree(tree_root); 29378dabb742SStefan Behrens return ret; 29388dabb742SStefan Behrens } 29398dabb742SStefan Behrens 2940b382a324SJan Schmidt btrfs_qgroup_rescan_resume(fs_info); 2941b382a324SJan Schmidt 2942f7a81ea4SStefan Behrens if (create_uuid_tree) { 2943efe120a0SFrank Holton pr_info("BTRFS: creating UUID tree\n"); 2944f7a81ea4SStefan Behrens ret = btrfs_create_uuid_tree(fs_info); 2945f7a81ea4SStefan Behrens if (ret) { 2946efe120a0SFrank Holton pr_warn("BTRFS: failed to create the UUID tree %d\n", 2947f7a81ea4SStefan Behrens ret); 2948f7a81ea4SStefan Behrens close_ctree(tree_root); 2949f7a81ea4SStefan Behrens return ret; 2950f7a81ea4SStefan Behrens } 2951f420ee1eSStefan Behrens } else if (check_uuid_tree || 2952f420ee1eSStefan Behrens btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) { 2953efe120a0SFrank Holton pr_info("BTRFS: checking UUID tree\n"); 295470f80175SStefan Behrens ret = btrfs_check_uuid_tree(fs_info); 295570f80175SStefan Behrens if (ret) { 2956efe120a0SFrank Holton pr_warn("BTRFS: failed to check the UUID tree %d\n", 295770f80175SStefan Behrens ret); 295870f80175SStefan Behrens close_ctree(tree_root); 295970f80175SStefan Behrens return ret; 296070f80175SStefan Behrens } 296170f80175SStefan Behrens } else { 296270f80175SStefan Behrens fs_info->update_uuid_tree_gen = 1; 2963f7a81ea4SStefan Behrens } 2964f7a81ea4SStefan Behrens 2965ad2b2c80SAl Viro return 0; 296639279cc3SChris Mason 2967bcef60f2SArne Jansen fail_qgroup: 2968bcef60f2SArne Jansen btrfs_free_qgroup_config(fs_info); 29697c2ca468SChris Mason fail_trans_kthread: 29707c2ca468SChris Mason kthread_stop(fs_info->transaction_kthread); 297154067ae9SJosef Bacik btrfs_cleanup_transaction(fs_info->tree_root); 29727b5ff90eSJosef Bacik del_fs_roots(fs_info); 29733f157a2fSChris Mason fail_cleaner: 2974a74a4b97SChris Mason kthread_stop(fs_info->cleaner_kthread); 29757c2ca468SChris Mason 29767c2ca468SChris Mason /* 29777c2ca468SChris Mason * make sure we're done with the btree inode before we stop our 29787c2ca468SChris Mason * kthreads 29797c2ca468SChris Mason */ 29807c2ca468SChris Mason filemap_write_and_wait(fs_info->btree_inode->i_mapping); 29817c2ca468SChris Mason 29822365dd3cSAnand Jain fail_sysfs: 29832365dd3cSAnand Jain btrfs_sysfs_remove_one(fs_info); 29842365dd3cSAnand Jain 29851b1d1f66SJosef Bacik fail_block_groups: 298654067ae9SJosef Bacik btrfs_put_block_group_cache(fs_info); 29871b1d1f66SJosef Bacik btrfs_free_block_groups(fs_info); 2988af31f5e5SChris Mason 2989af31f5e5SChris Mason fail_tree_roots: 2990af31f5e5SChris Mason free_root_pointers(fs_info, 1); 29912b8195bbSMiao Xie invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 2992af31f5e5SChris Mason 299339279cc3SChris Mason fail_sb_buffer: 29947abadb64SLiu Bo btrfs_stop_all_workers(fs_info); 299516cdcec7SMiao Xie fail_alloc: 29964543df7eSChris Mason fail_iput: 2997586e46e2SIlya Dryomov btrfs_mapping_tree_free(&fs_info->mapping_tree); 2998586e46e2SIlya Dryomov 29994543df7eSChris Mason iput(fs_info->btree_inode); 3000c404e0dcSMiao Xie fail_bio_counter: 3001c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 3002963d678bSMiao Xie fail_delalloc_bytes: 3003963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3004e2d84521SMiao Xie fail_dirty_metadata_bytes: 3005e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3006ad081f14SJens Axboe fail_bdi: 30077e662854SQinghuang Feng bdi_destroy(&fs_info->bdi); 300876dda93cSYan, Zheng fail_srcu: 300976dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 30107e662854SQinghuang Feng fail: 301153b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 3012586e46e2SIlya Dryomov btrfs_close_devices(fs_info->fs_devices); 3013ad2b2c80SAl Viro return err; 3014af31f5e5SChris Mason 3015af31f5e5SChris Mason recovery_tree_root: 3016af31f5e5SChris Mason if (!btrfs_test_opt(tree_root, RECOVERY)) 3017af31f5e5SChris Mason goto fail_tree_roots; 3018af31f5e5SChris Mason 3019af31f5e5SChris Mason free_root_pointers(fs_info, 0); 3020af31f5e5SChris Mason 3021af31f5e5SChris Mason /* don't use the log in recovery mode, it won't be valid */ 3022af31f5e5SChris Mason btrfs_set_super_log_root(disk_super, 0); 3023af31f5e5SChris Mason 3024af31f5e5SChris Mason /* we can't trust the free space cache either */ 3025af31f5e5SChris Mason btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); 3026af31f5e5SChris Mason 3027af31f5e5SChris Mason ret = next_root_backup(fs_info, fs_info->super_copy, 3028af31f5e5SChris Mason &num_backups_tried, &backup_index); 3029af31f5e5SChris Mason if (ret == -1) 3030af31f5e5SChris Mason goto fail_block_groups; 3031af31f5e5SChris Mason goto retry_root_backup; 3032eb60ceacSChris Mason } 3033eb60ceacSChris Mason 3034f2984462SChris Mason static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) 3035f2984462SChris Mason { 3036f2984462SChris Mason if (uptodate) { 3037f2984462SChris Mason set_buffer_uptodate(bh); 3038f2984462SChris Mason } else { 3039442a4f63SStefan Behrens struct btrfs_device *device = (struct btrfs_device *) 3040442a4f63SStefan Behrens bh->b_private; 3041442a4f63SStefan Behrens 3042efe120a0SFrank Holton printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to " 3043606686eeSJosef Bacik "I/O error on %s\n", 3044606686eeSJosef Bacik rcu_str_deref(device->name)); 30451259ab75SChris Mason /* note, we dont' set_buffer_write_io_error because we have 30461259ab75SChris Mason * our own ways of dealing with the IO errors 30471259ab75SChris Mason */ 3048f2984462SChris Mason clear_buffer_uptodate(bh); 3049442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS); 3050f2984462SChris Mason } 3051f2984462SChris Mason unlock_buffer(bh); 3052f2984462SChris Mason put_bh(bh); 3053f2984462SChris Mason } 3054f2984462SChris Mason 3055a512bbf8SYan Zheng struct buffer_head *btrfs_read_dev_super(struct block_device *bdev) 3056a512bbf8SYan Zheng { 3057a512bbf8SYan Zheng struct buffer_head *bh; 3058a512bbf8SYan Zheng struct buffer_head *latest = NULL; 3059a512bbf8SYan Zheng struct btrfs_super_block *super; 3060a512bbf8SYan Zheng int i; 3061a512bbf8SYan Zheng u64 transid = 0; 3062a512bbf8SYan Zheng u64 bytenr; 3063a512bbf8SYan Zheng 3064a512bbf8SYan Zheng /* we would like to check all the supers, but that would make 3065a512bbf8SYan Zheng * a btrfs mount succeed after a mkfs from a different FS. 3066a512bbf8SYan Zheng * So, we need to add a special mount option to scan for 3067a512bbf8SYan Zheng * later supers, using BTRFS_SUPER_MIRROR_MAX instead 3068a512bbf8SYan Zheng */ 3069a512bbf8SYan Zheng for (i = 0; i < 1; i++) { 3070a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 30718068a47eSAnand Jain if (bytenr + BTRFS_SUPER_INFO_SIZE >= 30728068a47eSAnand Jain i_size_read(bdev->bd_inode)) 3073a512bbf8SYan Zheng break; 30748068a47eSAnand Jain bh = __bread(bdev, bytenr / 4096, 30758068a47eSAnand Jain BTRFS_SUPER_INFO_SIZE); 3076a512bbf8SYan Zheng if (!bh) 3077a512bbf8SYan Zheng continue; 3078a512bbf8SYan Zheng 3079a512bbf8SYan Zheng super = (struct btrfs_super_block *)bh->b_data; 3080a512bbf8SYan Zheng if (btrfs_super_bytenr(super) != bytenr || 30813cae210fSQu Wenruo btrfs_super_magic(super) != BTRFS_MAGIC) { 3082a512bbf8SYan Zheng brelse(bh); 3083a512bbf8SYan Zheng continue; 3084a512bbf8SYan Zheng } 3085a512bbf8SYan Zheng 3086a512bbf8SYan Zheng if (!latest || btrfs_super_generation(super) > transid) { 3087a512bbf8SYan Zheng brelse(latest); 3088a512bbf8SYan Zheng latest = bh; 3089a512bbf8SYan Zheng transid = btrfs_super_generation(super); 3090a512bbf8SYan Zheng } else { 3091a512bbf8SYan Zheng brelse(bh); 3092a512bbf8SYan Zheng } 3093a512bbf8SYan Zheng } 3094a512bbf8SYan Zheng return latest; 3095a512bbf8SYan Zheng } 3096a512bbf8SYan Zheng 30974eedeb75SHisashi Hifumi /* 30984eedeb75SHisashi Hifumi * this should be called twice, once with wait == 0 and 30994eedeb75SHisashi Hifumi * once with wait == 1. When wait == 0 is done, all the buffer heads 31004eedeb75SHisashi Hifumi * we write are pinned. 31014eedeb75SHisashi Hifumi * 31024eedeb75SHisashi Hifumi * They are released when wait == 1 is done. 31034eedeb75SHisashi Hifumi * max_mirrors must be the same for both runs, and it indicates how 31044eedeb75SHisashi Hifumi * many supers on this one device should be written. 31054eedeb75SHisashi Hifumi * 31064eedeb75SHisashi Hifumi * max_mirrors == 0 means to write them all. 31074eedeb75SHisashi Hifumi */ 3108a512bbf8SYan Zheng static int write_dev_supers(struct btrfs_device *device, 3109a512bbf8SYan Zheng struct btrfs_super_block *sb, 3110a512bbf8SYan Zheng int do_barriers, int wait, int max_mirrors) 3111a512bbf8SYan Zheng { 3112a512bbf8SYan Zheng struct buffer_head *bh; 3113a512bbf8SYan Zheng int i; 3114a512bbf8SYan Zheng int ret; 3115a512bbf8SYan Zheng int errors = 0; 3116a512bbf8SYan Zheng u32 crc; 3117a512bbf8SYan Zheng u64 bytenr; 3118a512bbf8SYan Zheng 3119a512bbf8SYan Zheng if (max_mirrors == 0) 3120a512bbf8SYan Zheng max_mirrors = BTRFS_SUPER_MIRROR_MAX; 3121a512bbf8SYan Zheng 3122a512bbf8SYan Zheng for (i = 0; i < max_mirrors; i++) { 3123a512bbf8SYan Zheng bytenr = btrfs_sb_offset(i); 3124a512bbf8SYan Zheng if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes) 3125a512bbf8SYan Zheng break; 3126a512bbf8SYan Zheng 3127a512bbf8SYan Zheng if (wait) { 3128a512bbf8SYan Zheng bh = __find_get_block(device->bdev, bytenr / 4096, 3129a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3130634554dcSJosef Bacik if (!bh) { 3131634554dcSJosef Bacik errors++; 3132634554dcSJosef Bacik continue; 3133634554dcSJosef Bacik } 3134a512bbf8SYan Zheng wait_on_buffer(bh); 31354eedeb75SHisashi Hifumi if (!buffer_uptodate(bh)) 31364eedeb75SHisashi Hifumi errors++; 31374eedeb75SHisashi Hifumi 31384eedeb75SHisashi Hifumi /* drop our reference */ 31394eedeb75SHisashi Hifumi brelse(bh); 31404eedeb75SHisashi Hifumi 31414eedeb75SHisashi Hifumi /* drop the reference from the wait == 0 run */ 3142a512bbf8SYan Zheng brelse(bh); 3143a512bbf8SYan Zheng continue; 3144a512bbf8SYan Zheng } else { 3145a512bbf8SYan Zheng btrfs_set_super_bytenr(sb, bytenr); 3146a512bbf8SYan Zheng 3147a512bbf8SYan Zheng crc = ~(u32)0; 3148b0496686SLiu Bo crc = btrfs_csum_data((char *)sb + 3149a512bbf8SYan Zheng BTRFS_CSUM_SIZE, crc, 3150a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE - 3151a512bbf8SYan Zheng BTRFS_CSUM_SIZE); 3152a512bbf8SYan Zheng btrfs_csum_final(crc, sb->csum); 3153a512bbf8SYan Zheng 31544eedeb75SHisashi Hifumi /* 31554eedeb75SHisashi Hifumi * one reference for us, and we leave it for the 31564eedeb75SHisashi Hifumi * caller 31574eedeb75SHisashi Hifumi */ 3158a512bbf8SYan Zheng bh = __getblk(device->bdev, bytenr / 4096, 3159a512bbf8SYan Zheng BTRFS_SUPER_INFO_SIZE); 3160634554dcSJosef Bacik if (!bh) { 3161efe120a0SFrank Holton printk(KERN_ERR "BTRFS: couldn't get super " 3162634554dcSJosef Bacik "buffer head for bytenr %Lu\n", bytenr); 3163634554dcSJosef Bacik errors++; 3164634554dcSJosef Bacik continue; 3165634554dcSJosef Bacik } 3166634554dcSJosef Bacik 3167a512bbf8SYan Zheng memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); 3168a512bbf8SYan Zheng 31694eedeb75SHisashi Hifumi /* one reference for submit_bh */ 3170a512bbf8SYan Zheng get_bh(bh); 31714eedeb75SHisashi Hifumi 31724eedeb75SHisashi Hifumi set_buffer_uptodate(bh); 3173a512bbf8SYan Zheng lock_buffer(bh); 3174a512bbf8SYan Zheng bh->b_end_io = btrfs_end_buffer_write_sync; 3175442a4f63SStefan Behrens bh->b_private = device; 3176a512bbf8SYan Zheng } 3177a512bbf8SYan Zheng 3178387125fcSChris Mason /* 3179387125fcSChris Mason * we fua the first super. The others we allow 3180387125fcSChris Mason * to go down lazy. 3181387125fcSChris Mason */ 3182e8117c26SWang Shilong if (i == 0) 318321adbd5cSStefan Behrens ret = btrfsic_submit_bh(WRITE_FUA, bh); 3184e8117c26SWang Shilong else 3185e8117c26SWang Shilong ret = btrfsic_submit_bh(WRITE_SYNC, bh); 31864eedeb75SHisashi Hifumi if (ret) 3187a512bbf8SYan Zheng errors++; 3188a512bbf8SYan Zheng } 3189a512bbf8SYan Zheng return errors < i ? 0 : -1; 3190a512bbf8SYan Zheng } 3191a512bbf8SYan Zheng 3192387125fcSChris Mason /* 3193387125fcSChris Mason * endio for the write_dev_flush, this will wake anyone waiting 3194387125fcSChris Mason * for the barrier when it is done 3195387125fcSChris Mason */ 3196387125fcSChris Mason static void btrfs_end_empty_barrier(struct bio *bio, int err) 3197387125fcSChris Mason { 3198387125fcSChris Mason if (err) { 3199387125fcSChris Mason if (err == -EOPNOTSUPP) 3200387125fcSChris Mason set_bit(BIO_EOPNOTSUPP, &bio->bi_flags); 3201387125fcSChris Mason clear_bit(BIO_UPTODATE, &bio->bi_flags); 3202387125fcSChris Mason } 3203387125fcSChris Mason if (bio->bi_private) 3204387125fcSChris Mason complete(bio->bi_private); 3205387125fcSChris Mason bio_put(bio); 3206387125fcSChris Mason } 3207387125fcSChris Mason 3208387125fcSChris Mason /* 3209387125fcSChris Mason * trigger flushes for one the devices. If you pass wait == 0, the flushes are 3210387125fcSChris Mason * sent down. With wait == 1, it waits for the previous flush. 3211387125fcSChris Mason * 3212387125fcSChris Mason * any device where the flush fails with eopnotsupp are flagged as not-barrier 3213387125fcSChris Mason * capable 3214387125fcSChris Mason */ 3215387125fcSChris Mason static int write_dev_flush(struct btrfs_device *device, int wait) 3216387125fcSChris Mason { 3217387125fcSChris Mason struct bio *bio; 3218387125fcSChris Mason int ret = 0; 3219387125fcSChris Mason 3220387125fcSChris Mason if (device->nobarriers) 3221387125fcSChris Mason return 0; 3222387125fcSChris Mason 3223387125fcSChris Mason if (wait) { 3224387125fcSChris Mason bio = device->flush_bio; 3225387125fcSChris Mason if (!bio) 3226387125fcSChris Mason return 0; 3227387125fcSChris Mason 3228387125fcSChris Mason wait_for_completion(&device->flush_wait); 3229387125fcSChris Mason 3230387125fcSChris Mason if (bio_flagged(bio, BIO_EOPNOTSUPP)) { 3231efe120a0SFrank Holton printk_in_rcu("BTRFS: disabling barriers on dev %s\n", 3232606686eeSJosef Bacik rcu_str_deref(device->name)); 3233387125fcSChris Mason device->nobarriers = 1; 32345af3e8ccSStefan Behrens } else if (!bio_flagged(bio, BIO_UPTODATE)) { 3235387125fcSChris Mason ret = -EIO; 3236442a4f63SStefan Behrens btrfs_dev_stat_inc_and_print(device, 3237442a4f63SStefan Behrens BTRFS_DEV_STAT_FLUSH_ERRS); 3238387125fcSChris Mason } 3239387125fcSChris Mason 3240387125fcSChris Mason /* drop the reference from the wait == 0 run */ 3241387125fcSChris Mason bio_put(bio); 3242387125fcSChris Mason device->flush_bio = NULL; 3243387125fcSChris Mason 3244387125fcSChris Mason return ret; 3245387125fcSChris Mason } 3246387125fcSChris Mason 3247387125fcSChris Mason /* 3248387125fcSChris Mason * one reference for us, and we leave it for the 3249387125fcSChris Mason * caller 3250387125fcSChris Mason */ 32519c017abcSJesper Juhl device->flush_bio = NULL; 32529be3395bSChris Mason bio = btrfs_io_bio_alloc(GFP_NOFS, 0); 3253387125fcSChris Mason if (!bio) 3254387125fcSChris Mason return -ENOMEM; 3255387125fcSChris Mason 3256387125fcSChris Mason bio->bi_end_io = btrfs_end_empty_barrier; 3257387125fcSChris Mason bio->bi_bdev = device->bdev; 3258387125fcSChris Mason init_completion(&device->flush_wait); 3259387125fcSChris Mason bio->bi_private = &device->flush_wait; 3260387125fcSChris Mason device->flush_bio = bio; 3261387125fcSChris Mason 3262387125fcSChris Mason bio_get(bio); 326321adbd5cSStefan Behrens btrfsic_submit_bio(WRITE_FLUSH, bio); 3264387125fcSChris Mason 3265387125fcSChris Mason return 0; 3266387125fcSChris Mason } 3267387125fcSChris Mason 3268387125fcSChris Mason /* 3269387125fcSChris Mason * send an empty flush down to each device in parallel, 3270387125fcSChris Mason * then wait for them 3271387125fcSChris Mason */ 3272387125fcSChris Mason static int barrier_all_devices(struct btrfs_fs_info *info) 3273387125fcSChris Mason { 3274387125fcSChris Mason struct list_head *head; 3275387125fcSChris Mason struct btrfs_device *dev; 32765af3e8ccSStefan Behrens int errors_send = 0; 32775af3e8ccSStefan Behrens int errors_wait = 0; 3278387125fcSChris Mason int ret; 3279387125fcSChris Mason 3280387125fcSChris Mason /* send down all the barriers */ 3281387125fcSChris Mason head = &info->fs_devices->devices; 3282387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3283f88ba6a2SHidetoshi Seto if (dev->missing) 3284f88ba6a2SHidetoshi Seto continue; 3285387125fcSChris Mason if (!dev->bdev) { 32865af3e8ccSStefan Behrens errors_send++; 3287387125fcSChris Mason continue; 3288387125fcSChris Mason } 3289387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3290387125fcSChris Mason continue; 3291387125fcSChris Mason 3292387125fcSChris Mason ret = write_dev_flush(dev, 0); 3293387125fcSChris Mason if (ret) 32945af3e8ccSStefan Behrens errors_send++; 3295387125fcSChris Mason } 3296387125fcSChris Mason 3297387125fcSChris Mason /* wait for all the barriers */ 3298387125fcSChris Mason list_for_each_entry_rcu(dev, head, dev_list) { 3299f88ba6a2SHidetoshi Seto if (dev->missing) 3300f88ba6a2SHidetoshi Seto continue; 3301387125fcSChris Mason if (!dev->bdev) { 33025af3e8ccSStefan Behrens errors_wait++; 3303387125fcSChris Mason continue; 3304387125fcSChris Mason } 3305387125fcSChris Mason if (!dev->in_fs_metadata || !dev->writeable) 3306387125fcSChris Mason continue; 3307387125fcSChris Mason 3308387125fcSChris Mason ret = write_dev_flush(dev, 1); 3309387125fcSChris Mason if (ret) 33105af3e8ccSStefan Behrens errors_wait++; 3311387125fcSChris Mason } 33125af3e8ccSStefan Behrens if (errors_send > info->num_tolerated_disk_barrier_failures || 33135af3e8ccSStefan Behrens errors_wait > info->num_tolerated_disk_barrier_failures) 3314387125fcSChris Mason return -EIO; 3315387125fcSChris Mason return 0; 3316387125fcSChris Mason } 3317387125fcSChris Mason 33185af3e8ccSStefan Behrens int btrfs_calc_num_tolerated_disk_barrier_failures( 33195af3e8ccSStefan Behrens struct btrfs_fs_info *fs_info) 33205af3e8ccSStefan Behrens { 33215af3e8ccSStefan Behrens struct btrfs_ioctl_space_info space; 33225af3e8ccSStefan Behrens struct btrfs_space_info *sinfo; 33235af3e8ccSStefan Behrens u64 types[] = {BTRFS_BLOCK_GROUP_DATA, 33245af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_SYSTEM, 33255af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_METADATA, 33265af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA}; 33275af3e8ccSStefan Behrens int num_types = 4; 33285af3e8ccSStefan Behrens int i; 33295af3e8ccSStefan Behrens int c; 33305af3e8ccSStefan Behrens int num_tolerated_disk_barrier_failures = 33315af3e8ccSStefan Behrens (int)fs_info->fs_devices->num_devices; 33325af3e8ccSStefan Behrens 33335af3e8ccSStefan Behrens for (i = 0; i < num_types; i++) { 33345af3e8ccSStefan Behrens struct btrfs_space_info *tmp; 33355af3e8ccSStefan Behrens 33365af3e8ccSStefan Behrens sinfo = NULL; 33375af3e8ccSStefan Behrens rcu_read_lock(); 33385af3e8ccSStefan Behrens list_for_each_entry_rcu(tmp, &fs_info->space_info, list) { 33395af3e8ccSStefan Behrens if (tmp->flags == types[i]) { 33405af3e8ccSStefan Behrens sinfo = tmp; 33415af3e8ccSStefan Behrens break; 33425af3e8ccSStefan Behrens } 33435af3e8ccSStefan Behrens } 33445af3e8ccSStefan Behrens rcu_read_unlock(); 33455af3e8ccSStefan Behrens 33465af3e8ccSStefan Behrens if (!sinfo) 33475af3e8ccSStefan Behrens continue; 33485af3e8ccSStefan Behrens 33495af3e8ccSStefan Behrens down_read(&sinfo->groups_sem); 33505af3e8ccSStefan Behrens for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) { 33515af3e8ccSStefan Behrens if (!list_empty(&sinfo->block_groups[c])) { 33525af3e8ccSStefan Behrens u64 flags; 33535af3e8ccSStefan Behrens 33545af3e8ccSStefan Behrens btrfs_get_block_group_info( 33555af3e8ccSStefan Behrens &sinfo->block_groups[c], &space); 33565af3e8ccSStefan Behrens if (space.total_bytes == 0 || 33575af3e8ccSStefan Behrens space.used_bytes == 0) 33585af3e8ccSStefan Behrens continue; 33595af3e8ccSStefan Behrens flags = space.flags; 33605af3e8ccSStefan Behrens /* 33615af3e8ccSStefan Behrens * return 33625af3e8ccSStefan Behrens * 0: if dup, single or RAID0 is configured for 33635af3e8ccSStefan Behrens * any of metadata, system or data, else 33645af3e8ccSStefan Behrens * 1: if RAID5 is configured, or if RAID1 or 33655af3e8ccSStefan Behrens * RAID10 is configured and only two mirrors 33665af3e8ccSStefan Behrens * are used, else 33675af3e8ccSStefan Behrens * 2: if RAID6 is configured, else 33685af3e8ccSStefan Behrens * num_mirrors - 1: if RAID1 or RAID10 is 33695af3e8ccSStefan Behrens * configured and more than 33705af3e8ccSStefan Behrens * 2 mirrors are used. 33715af3e8ccSStefan Behrens */ 33725af3e8ccSStefan Behrens if (num_tolerated_disk_barrier_failures > 0 && 33735af3e8ccSStefan Behrens ((flags & (BTRFS_BLOCK_GROUP_DUP | 33745af3e8ccSStefan Behrens BTRFS_BLOCK_GROUP_RAID0)) || 33755af3e8ccSStefan Behrens ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) 33765af3e8ccSStefan Behrens == 0))) 33775af3e8ccSStefan Behrens num_tolerated_disk_barrier_failures = 0; 337853b381b3SDavid Woodhouse else if (num_tolerated_disk_barrier_failures > 1) { 337953b381b3SDavid Woodhouse if (flags & (BTRFS_BLOCK_GROUP_RAID1 | 338053b381b3SDavid Woodhouse BTRFS_BLOCK_GROUP_RAID5 | 338153b381b3SDavid Woodhouse BTRFS_BLOCK_GROUP_RAID10)) { 33825af3e8ccSStefan Behrens num_tolerated_disk_barrier_failures = 1; 338353b381b3SDavid Woodhouse } else if (flags & 338415b0a89dSHenrik Nordvik BTRFS_BLOCK_GROUP_RAID6) { 338553b381b3SDavid Woodhouse num_tolerated_disk_barrier_failures = 2; 338653b381b3SDavid Woodhouse } 338753b381b3SDavid Woodhouse } 33885af3e8ccSStefan Behrens } 33895af3e8ccSStefan Behrens } 33905af3e8ccSStefan Behrens up_read(&sinfo->groups_sem); 33915af3e8ccSStefan Behrens } 33925af3e8ccSStefan Behrens 33935af3e8ccSStefan Behrens return num_tolerated_disk_barrier_failures; 33945af3e8ccSStefan Behrens } 33955af3e8ccSStefan Behrens 339648a3b636SEric Sandeen static int write_all_supers(struct btrfs_root *root, int max_mirrors) 3397f2984462SChris Mason { 3398e5e9a520SChris Mason struct list_head *head; 3399f2984462SChris Mason struct btrfs_device *dev; 3400a061fc8dSChris Mason struct btrfs_super_block *sb; 3401f2984462SChris Mason struct btrfs_dev_item *dev_item; 3402f2984462SChris Mason int ret; 3403f2984462SChris Mason int do_barriers; 3404a236aed1SChris Mason int max_errors; 3405a236aed1SChris Mason int total_errors = 0; 3406a061fc8dSChris Mason u64 flags; 3407f2984462SChris Mason 3408f2984462SChris Mason do_barriers = !btrfs_test_opt(root, NOBARRIER); 3409af31f5e5SChris Mason backup_super_roots(root->fs_info); 3410f2984462SChris Mason 34116c41761fSDavid Sterba sb = root->fs_info->super_for_commit; 3412a061fc8dSChris Mason dev_item = &sb->dev_item; 3413e5e9a520SChris Mason 3414174ba509SChris Mason mutex_lock(&root->fs_info->fs_devices->device_list_mutex); 3415e5e9a520SChris Mason head = &root->fs_info->fs_devices->devices; 3416d7306801SFilipe David Borba Manana max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; 3417387125fcSChris Mason 34185af3e8ccSStefan Behrens if (do_barriers) { 34195af3e8ccSStefan Behrens ret = barrier_all_devices(root->fs_info); 34205af3e8ccSStefan Behrens if (ret) { 34215af3e8ccSStefan Behrens mutex_unlock( 34225af3e8ccSStefan Behrens &root->fs_info->fs_devices->device_list_mutex); 34235af3e8ccSStefan Behrens btrfs_error(root->fs_info, ret, 34245af3e8ccSStefan Behrens "errors while submitting device barriers."); 34255af3e8ccSStefan Behrens return ret; 34265af3e8ccSStefan Behrens } 34275af3e8ccSStefan Behrens } 3428387125fcSChris Mason 34291f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3430dfe25020SChris Mason if (!dev->bdev) { 3431dfe25020SChris Mason total_errors++; 3432dfe25020SChris Mason continue; 3433dfe25020SChris Mason } 34342b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3435dfe25020SChris Mason continue; 3436dfe25020SChris Mason 34372b82032cSYan Zheng btrfs_set_stack_device_generation(dev_item, 0); 3438a061fc8dSChris Mason btrfs_set_stack_device_type(dev_item, dev->type); 3439a061fc8dSChris Mason btrfs_set_stack_device_id(dev_item, dev->devid); 3440a061fc8dSChris Mason btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); 3441a061fc8dSChris Mason btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); 3442a061fc8dSChris Mason btrfs_set_stack_device_io_align(dev_item, dev->io_align); 3443a061fc8dSChris Mason btrfs_set_stack_device_io_width(dev_item, dev->io_width); 3444a061fc8dSChris Mason btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); 3445a061fc8dSChris Mason memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); 34462b82032cSYan Zheng memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); 3447a512bbf8SYan Zheng 3448a061fc8dSChris Mason flags = btrfs_super_flags(sb); 3449a061fc8dSChris Mason btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3450f2984462SChris Mason 3451a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); 3452a236aed1SChris Mason if (ret) 3453a236aed1SChris Mason total_errors++; 3454f2984462SChris Mason } 3455a236aed1SChris Mason if (total_errors > max_errors) { 3456efe120a0SFrank Holton btrfs_err(root->fs_info, "%d errors while writing supers", 3457d397712bSChris Mason total_errors); 3458a724b436SStefan Behrens mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 345979787eaaSJeff Mahoney 34609d565ba4SStefan Behrens /* FUA is masked off if unsupported and can't be the reason */ 34619d565ba4SStefan Behrens btrfs_error(root->fs_info, -EIO, 34629d565ba4SStefan Behrens "%d errors while writing supers", total_errors); 34639d565ba4SStefan Behrens return -EIO; 3464a236aed1SChris Mason } 3465f2984462SChris Mason 3466a512bbf8SYan Zheng total_errors = 0; 34671f78160cSXiao Guangrong list_for_each_entry_rcu(dev, head, dev_list) { 3468dfe25020SChris Mason if (!dev->bdev) 3469dfe25020SChris Mason continue; 34702b82032cSYan Zheng if (!dev->in_fs_metadata || !dev->writeable) 3471dfe25020SChris Mason continue; 3472dfe25020SChris Mason 3473a512bbf8SYan Zheng ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); 3474a512bbf8SYan Zheng if (ret) 34751259ab75SChris Mason total_errors++; 3476f2984462SChris Mason } 3477174ba509SChris Mason mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3478a236aed1SChris Mason if (total_errors > max_errors) { 347979787eaaSJeff Mahoney btrfs_error(root->fs_info, -EIO, 348079787eaaSJeff Mahoney "%d errors while writing supers", total_errors); 348179787eaaSJeff Mahoney return -EIO; 3482a236aed1SChris Mason } 3483f2984462SChris Mason return 0; 3484f2984462SChris Mason } 3485f2984462SChris Mason 3486a512bbf8SYan Zheng int write_ctree_super(struct btrfs_trans_handle *trans, 3487a512bbf8SYan Zheng struct btrfs_root *root, int max_mirrors) 3488cfaa7295SChris Mason { 3489f570e757SRashika return write_all_supers(root, max_mirrors); 3490cfaa7295SChris Mason } 3491cfaa7295SChris Mason 3492cb517eabSMiao Xie /* Drop a fs root from the radix tree and free it. */ 3493cb517eabSMiao Xie void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3494cb517eabSMiao Xie struct btrfs_root *root) 34952619ba1fSChris Mason { 34964df27c4dSYan, Zheng spin_lock(&fs_info->fs_roots_radix_lock); 34972619ba1fSChris Mason radix_tree_delete(&fs_info->fs_roots_radix, 34982619ba1fSChris Mason (unsigned long)root->root_key.objectid); 34994df27c4dSYan, Zheng spin_unlock(&fs_info->fs_roots_radix_lock); 350076dda93cSYan, Zheng 350176dda93cSYan, Zheng if (btrfs_root_refs(&root->root_item) == 0) 350276dda93cSYan, Zheng synchronize_srcu(&fs_info->subvol_srcu); 350376dda93cSYan, Zheng 35041a4319ccSLiu Bo if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 35053321719eSLiu Bo btrfs_free_log(NULL, root); 35063321719eSLiu Bo 3507581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_pinned); 3508581bb050SLi Zefan __btrfs_remove_free_space_cache(root->free_ino_ctl); 35094df27c4dSYan, Zheng free_fs_root(root); 35104df27c4dSYan, Zheng } 35114df27c4dSYan, Zheng 35124df27c4dSYan, Zheng static void free_fs_root(struct btrfs_root *root) 35134df27c4dSYan, Zheng { 351482d5902dSLi Zefan iput(root->cache_inode); 35154df27c4dSYan, Zheng WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 35161cb048f5SFilipe David Borba Manana btrfs_free_block_rsv(root, root->orphan_block_rsv); 35171cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 35180ee5dc67SAl Viro if (root->anon_dev) 35190ee5dc67SAl Viro free_anon_bdev(root->anon_dev); 35208257b2dcSMiao Xie if (root->subv_writers) 35218257b2dcSMiao Xie btrfs_free_subvolume_writers(root->subv_writers); 35225f39d397SChris Mason free_extent_buffer(root->node); 35235f39d397SChris Mason free_extent_buffer(root->commit_root); 3524581bb050SLi Zefan kfree(root->free_ino_ctl); 3525581bb050SLi Zefan kfree(root->free_ino_pinned); 352658176a96SJosef Bacik kfree(root->name); 3527b0feb9d9SMiao Xie btrfs_put_fs_root(root); 35282619ba1fSChris Mason } 35292619ba1fSChris Mason 3530cb517eabSMiao Xie void btrfs_free_fs_root(struct btrfs_root *root) 3531cb517eabSMiao Xie { 3532cb517eabSMiao Xie free_fs_root(root); 3533eb60ceacSChris Mason } 3534eb60ceacSChris Mason 3535c146afadSYan Zheng int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) 3536c146afadSYan Zheng { 3537c146afadSYan Zheng u64 root_objectid = 0; 3538c146afadSYan Zheng struct btrfs_root *gang[8]; 3539c146afadSYan Zheng int i; 3540c146afadSYan Zheng int ret; 3541c146afadSYan Zheng 3542c146afadSYan Zheng while (1) { 3543c146afadSYan Zheng ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, 3544c146afadSYan Zheng (void **)gang, root_objectid, 3545c146afadSYan Zheng ARRAY_SIZE(gang)); 3546c146afadSYan Zheng if (!ret) 3547c146afadSYan Zheng break; 35485d4f98a2SYan Zheng 35495d4f98a2SYan Zheng root_objectid = gang[ret - 1]->root_key.objectid + 1; 3550c146afadSYan Zheng for (i = 0; i < ret; i++) { 355166b4ffd1SJosef Bacik int err; 355266b4ffd1SJosef Bacik 3553c146afadSYan Zheng root_objectid = gang[i]->root_key.objectid; 355466b4ffd1SJosef Bacik err = btrfs_orphan_cleanup(gang[i]); 355566b4ffd1SJosef Bacik if (err) 355666b4ffd1SJosef Bacik return err; 3557c146afadSYan Zheng } 3558c146afadSYan Zheng root_objectid++; 3559c146afadSYan Zheng } 3560c146afadSYan Zheng return 0; 3561c146afadSYan Zheng } 3562c146afadSYan Zheng 3563c146afadSYan Zheng int btrfs_commit_super(struct btrfs_root *root) 3564c146afadSYan Zheng { 3565c146afadSYan Zheng struct btrfs_trans_handle *trans; 3566c146afadSYan Zheng 3567c146afadSYan Zheng mutex_lock(&root->fs_info->cleaner_mutex); 356824bbcf04SYan, Zheng btrfs_run_delayed_iputs(root); 3569c146afadSYan Zheng mutex_unlock(&root->fs_info->cleaner_mutex); 35709d1a2a3aSDavid Sterba wake_up_process(root->fs_info->cleaner_kthread); 3571c71bf099SYan, Zheng 3572c71bf099SYan, Zheng /* wait until ongoing cleanup work done */ 3573c71bf099SYan, Zheng down_write(&root->fs_info->cleanup_work_sem); 3574c71bf099SYan, Zheng up_write(&root->fs_info->cleanup_work_sem); 3575c71bf099SYan, Zheng 35767a7eaa40SJosef Bacik trans = btrfs_join_transaction(root); 35773612b495STsutomu Itoh if (IS_ERR(trans)) 35783612b495STsutomu Itoh return PTR_ERR(trans); 3579d52c1bccSLiu Bo return btrfs_commit_transaction(trans, root); 3580c146afadSYan Zheng } 3581c146afadSYan Zheng 3582e20d96d6SChris Mason int close_ctree(struct btrfs_root *root) 3583eb60ceacSChris Mason { 35840f7d52f4SChris Mason struct btrfs_fs_info *fs_info = root->fs_info; 3585c146afadSYan Zheng int ret; 3586e089f05cSChris Mason 3587facda1e7SChris Mason fs_info->closing = 1; 3588a2135011SChris Mason smp_mb(); 3589a2135011SChris Mason 3590803b2f54SStefan Behrens /* wait for the uuid_scan task to finish */ 3591803b2f54SStefan Behrens down(&fs_info->uuid_tree_rescan_sem); 3592803b2f54SStefan Behrens /* avoid complains from lockdep et al., set sem back to initial state */ 3593803b2f54SStefan Behrens up(&fs_info->uuid_tree_rescan_sem); 3594803b2f54SStefan Behrens 3595837d5b6eSIlya Dryomov /* pause restriper - we want to resume on mount */ 3596aa1b8cd4SStefan Behrens btrfs_pause_balance(fs_info); 3597837d5b6eSIlya Dryomov 35988dabb742SStefan Behrens btrfs_dev_replace_suspend_for_unmount(fs_info); 35998dabb742SStefan Behrens 3600aa1b8cd4SStefan Behrens btrfs_scrub_cancel(fs_info); 36014cb5300bSChris Mason 36024cb5300bSChris Mason /* wait for any defraggers to finish */ 36034cb5300bSChris Mason wait_event(fs_info->transaction_wait, 36044cb5300bSChris Mason (atomic_read(&fs_info->defrag_running) == 0)); 36054cb5300bSChris Mason 36064cb5300bSChris Mason /* clear out the rbtree of defraggable inodes */ 360726176e7cSMiao Xie btrfs_cleanup_defrag_inodes(fs_info); 36084cb5300bSChris Mason 3609c146afadSYan Zheng if (!(fs_info->sb->s_flags & MS_RDONLY)) { 3610c146afadSYan Zheng ret = btrfs_commit_super(root); 3611d397712bSChris Mason if (ret) 3612efe120a0SFrank Holton btrfs_err(root->fs_info, "commit super ret %d", ret); 3613c146afadSYan Zheng } 3614ed2ff2cbSChris Mason 361587533c47SMiao Xie if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 361668ce9682SStefan Behrens btrfs_error_commit_super(root); 3617acce952bSliubo 3618e3029d9fSAl Viro kthread_stop(fs_info->transaction_kthread); 3619e3029d9fSAl Viro kthread_stop(fs_info->cleaner_kthread); 36208929ecfaSYan, Zheng 3621f25784b3SYan Zheng fs_info->closing = 2; 3622f25784b3SYan Zheng smp_mb(); 3623f25784b3SYan Zheng 3624bcef60f2SArne Jansen btrfs_free_qgroup_config(root->fs_info); 3625bcef60f2SArne Jansen 3626963d678bSMiao Xie if (percpu_counter_sum(&fs_info->delalloc_bytes)) { 3627efe120a0SFrank Holton btrfs_info(root->fs_info, "at unmount delalloc count %lld", 3628963d678bSMiao Xie percpu_counter_sum(&fs_info->delalloc_bytes)); 3629b0c68f8bSChris Mason } 363031153d81SYan Zheng 36315ac1d209SJeff Mahoney btrfs_sysfs_remove_one(fs_info); 36325ac1d209SJeff Mahoney 36330f7d52f4SChris Mason del_fs_roots(fs_info); 3634d10c5f31SChris Mason 36351a4319ccSLiu Bo btrfs_put_block_group_cache(fs_info); 36361a4319ccSLiu Bo 36372b1360daSJosef Bacik btrfs_free_block_groups(fs_info); 36382b1360daSJosef Bacik 363996192499SJosef Bacik btrfs_stop_all_workers(fs_info); 364096192499SJosef Bacik 364113e6c37bSJosef Bacik free_root_pointers(fs_info, 1); 36429ad6b7bcSChris Mason 364313e6c37bSJosef Bacik iput(fs_info->btree_inode); 3644d6bfde87SChris Mason 364521adbd5cSStefan Behrens #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 364621adbd5cSStefan Behrens if (btrfs_test_opt(root, CHECK_INTEGRITY)) 364721adbd5cSStefan Behrens btrfsic_unmount(root, fs_info->fs_devices); 364821adbd5cSStefan Behrens #endif 364921adbd5cSStefan Behrens 3650dfe25020SChris Mason btrfs_close_devices(fs_info->fs_devices); 36510b86a832SChris Mason btrfs_mapping_tree_free(&fs_info->mapping_tree); 3652b248a415SChris Mason 3653e2d84521SMiao Xie percpu_counter_destroy(&fs_info->dirty_metadata_bytes); 3654963d678bSMiao Xie percpu_counter_destroy(&fs_info->delalloc_bytes); 3655c404e0dcSMiao Xie percpu_counter_destroy(&fs_info->bio_counter); 365604160088SChris Mason bdi_destroy(&fs_info->bdi); 365776dda93cSYan, Zheng cleanup_srcu_struct(&fs_info->subvol_srcu); 36580b86a832SChris Mason 365953b381b3SDavid Woodhouse btrfs_free_stripe_hash_table(fs_info); 366053b381b3SDavid Woodhouse 36611cb048f5SFilipe David Borba Manana btrfs_free_block_rsv(root, root->orphan_block_rsv); 36621cb048f5SFilipe David Borba Manana root->orphan_block_rsv = NULL; 36631cb048f5SFilipe David Borba Manana 3664eb60ceacSChris Mason return 0; 3665eb60ceacSChris Mason } 3666eb60ceacSChris Mason 3667b9fab919SChris Mason int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 3668b9fab919SChris Mason int atomic) 3669ccd467d6SChris Mason { 36701259ab75SChris Mason int ret; 3671727011e0SChris Mason struct inode *btree_inode = buf->pages[0]->mapping->host; 36721259ab75SChris Mason 36730b32f4bbSJosef Bacik ret = extent_buffer_uptodate(buf); 36741259ab75SChris Mason if (!ret) 36751259ab75SChris Mason return ret; 36761259ab75SChris Mason 36771259ab75SChris Mason ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, 3678b9fab919SChris Mason parent_transid, atomic); 3679b9fab919SChris Mason if (ret == -EAGAIN) 3680b9fab919SChris Mason return ret; 36811259ab75SChris Mason return !ret; 36825f39d397SChris Mason } 36836702ed49SChris Mason 36845f39d397SChris Mason int btrfs_set_buffer_uptodate(struct extent_buffer *buf) 36855f39d397SChris Mason { 36860b32f4bbSJosef Bacik return set_extent_buffer_uptodate(buf); 36875f39d397SChris Mason } 36885f39d397SChris Mason 36895f39d397SChris Mason void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 36905f39d397SChris Mason { 369106ea65a3SJosef Bacik struct btrfs_root *root; 36925f39d397SChris Mason u64 transid = btrfs_header_generation(buf); 3693b9473439SChris Mason int was_dirty; 3694b4ce94deSChris Mason 369506ea65a3SJosef Bacik #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 369606ea65a3SJosef Bacik /* 369706ea65a3SJosef Bacik * This is a fast path so only do this check if we have sanity tests 369806ea65a3SJosef Bacik * enabled. Normal people shouldn't be marking dummy buffers as dirty 369906ea65a3SJosef Bacik * outside of the sanity tests. 370006ea65a3SJosef Bacik */ 370106ea65a3SJosef Bacik if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags))) 370206ea65a3SJosef Bacik return; 370306ea65a3SJosef Bacik #endif 370406ea65a3SJosef Bacik root = BTRFS_I(buf->pages[0]->mapping->host)->root; 3705b9447ef8SChris Mason btrfs_assert_tree_locked(buf); 370631b1a2bdSJulia Lawall if (transid != root->fs_info->generation) 370731b1a2bdSJulia Lawall WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, " 3708d397712bSChris Mason "found %llu running %llu\n", 3709c1c9ff7cSGeert Uytterhoeven buf->start, transid, root->fs_info->generation); 37100b32f4bbSJosef Bacik was_dirty = set_extent_buffer_dirty(buf); 3711e2d84521SMiao Xie if (!was_dirty) 3712e2d84521SMiao Xie __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, 3713e2d84521SMiao Xie buf->len, 3714e2d84521SMiao Xie root->fs_info->dirty_metadata_batch); 3715eb60ceacSChris Mason } 3716eb60ceacSChris Mason 3717b53d3f5dSLiu Bo static void __btrfs_btree_balance_dirty(struct btrfs_root *root, 3718b53d3f5dSLiu Bo int flush_delayed) 371935b7e476SChris Mason { 3720188de649SChris Mason /* 3721188de649SChris Mason * looks as though older kernels can get into trouble with 3722188de649SChris Mason * this code, they end up stuck in balance_dirty_pages forever 3723188de649SChris Mason */ 3724e2d84521SMiao Xie int ret; 3725d6bfde87SChris Mason 37266933c02eSJens Axboe if (current->flags & PF_MEMALLOC) 3727d6bfde87SChris Mason return; 3728d6bfde87SChris Mason 3729b53d3f5dSLiu Bo if (flush_delayed) 373016cdcec7SMiao Xie btrfs_balance_delayed_items(root); 373116cdcec7SMiao Xie 3732e2d84521SMiao Xie ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, 3733e2d84521SMiao Xie BTRFS_DIRTY_METADATA_THRESH); 3734e2d84521SMiao Xie if (ret > 0) { 3735d0e1d66bSNamjae Jeon balance_dirty_pages_ratelimited( 3736d0e1d66bSNamjae Jeon root->fs_info->btree_inode->i_mapping); 373716cdcec7SMiao Xie } 373816cdcec7SMiao Xie return; 373916cdcec7SMiao Xie } 374016cdcec7SMiao Xie 3741b53d3f5dSLiu Bo void btrfs_btree_balance_dirty(struct btrfs_root *root) 374216cdcec7SMiao Xie { 3743b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 1); 374435b7e476SChris Mason } 3745b53d3f5dSLiu Bo 3746b53d3f5dSLiu Bo void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) 3747b53d3f5dSLiu Bo { 3748b53d3f5dSLiu Bo __btrfs_btree_balance_dirty(root, 0); 3749d6bfde87SChris Mason } 37506b80053dSChris Mason 3751ca7a79adSChris Mason int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 37526b80053dSChris Mason { 3753727011e0SChris Mason struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 37540b32f4bbSJosef Bacik return btree_read_extent_buffer_pages(root, buf, 0, parent_transid); 37556b80053dSChris Mason } 37560da5468fSChris Mason 3757fcd1f065SDavid Sterba static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 3758acce952bSliubo int read_only) 3759acce952bSliubo { 37601104a885SDavid Sterba /* 37611104a885SDavid Sterba * Placeholder for checks 37621104a885SDavid Sterba */ 3763fcd1f065SDavid Sterba return 0; 3764fcd1f065SDavid Sterba } 3765fcd1f065SDavid Sterba 376648a3b636SEric Sandeen static void btrfs_error_commit_super(struct btrfs_root *root) 3767acce952bSliubo { 3768acce952bSliubo mutex_lock(&root->fs_info->cleaner_mutex); 3769acce952bSliubo btrfs_run_delayed_iputs(root); 3770acce952bSliubo mutex_unlock(&root->fs_info->cleaner_mutex); 3771acce952bSliubo 3772acce952bSliubo down_write(&root->fs_info->cleanup_work_sem); 3773acce952bSliubo up_write(&root->fs_info->cleanup_work_sem); 3774acce952bSliubo 3775acce952bSliubo /* cleanup FS via transaction */ 3776acce952bSliubo btrfs_cleanup_transaction(root); 3777acce952bSliubo } 3778acce952bSliubo 3779569e0f35SJosef Bacik static void btrfs_destroy_ordered_operations(struct btrfs_transaction *t, 3780569e0f35SJosef Bacik struct btrfs_root *root) 3781acce952bSliubo { 3782acce952bSliubo struct btrfs_inode *btrfs_inode; 3783acce952bSliubo struct list_head splice; 3784acce952bSliubo 3785acce952bSliubo INIT_LIST_HEAD(&splice); 3786acce952bSliubo 3787acce952bSliubo mutex_lock(&root->fs_info->ordered_operations_mutex); 3788199c2a9cSMiao Xie spin_lock(&root->fs_info->ordered_root_lock); 3789acce952bSliubo 3790569e0f35SJosef Bacik list_splice_init(&t->ordered_operations, &splice); 3791acce952bSliubo while (!list_empty(&splice)) { 3792acce952bSliubo btrfs_inode = list_entry(splice.next, struct btrfs_inode, 3793acce952bSliubo ordered_operations); 3794acce952bSliubo 3795acce952bSliubo list_del_init(&btrfs_inode->ordered_operations); 3796199c2a9cSMiao Xie spin_unlock(&root->fs_info->ordered_root_lock); 3797acce952bSliubo 3798acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 3799b216cbfbSMiao Xie 3800199c2a9cSMiao Xie spin_lock(&root->fs_info->ordered_root_lock); 3801acce952bSliubo } 3802acce952bSliubo 3803199c2a9cSMiao Xie spin_unlock(&root->fs_info->ordered_root_lock); 3804acce952bSliubo mutex_unlock(&root->fs_info->ordered_operations_mutex); 3805acce952bSliubo } 3806acce952bSliubo 3807143bede5SJeff Mahoney static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 3808acce952bSliubo { 3809acce952bSliubo struct btrfs_ordered_extent *ordered; 3810acce952bSliubo 3811199c2a9cSMiao Xie spin_lock(&root->ordered_extent_lock); 3812779880efSJosef Bacik /* 3813779880efSJosef Bacik * This will just short circuit the ordered completion stuff which will 3814779880efSJosef Bacik * make sure the ordered extent gets properly cleaned up. 3815779880efSJosef Bacik */ 3816199c2a9cSMiao Xie list_for_each_entry(ordered, &root->ordered_extents, 3817779880efSJosef Bacik root_extent_list) 3818779880efSJosef Bacik set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); 3819199c2a9cSMiao Xie spin_unlock(&root->ordered_extent_lock); 3820199c2a9cSMiao Xie } 3821199c2a9cSMiao Xie 3822199c2a9cSMiao Xie static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) 3823199c2a9cSMiao Xie { 3824199c2a9cSMiao Xie struct btrfs_root *root; 3825199c2a9cSMiao Xie struct list_head splice; 3826199c2a9cSMiao Xie 3827199c2a9cSMiao Xie INIT_LIST_HEAD(&splice); 3828199c2a9cSMiao Xie 3829199c2a9cSMiao Xie spin_lock(&fs_info->ordered_root_lock); 3830199c2a9cSMiao Xie list_splice_init(&fs_info->ordered_roots, &splice); 3831199c2a9cSMiao Xie while (!list_empty(&splice)) { 3832199c2a9cSMiao Xie root = list_first_entry(&splice, struct btrfs_root, 3833199c2a9cSMiao Xie ordered_root); 38341de2cfdeSJosef Bacik list_move_tail(&root->ordered_root, 38351de2cfdeSJosef Bacik &fs_info->ordered_roots); 3836199c2a9cSMiao Xie 38372a85d9caSLiu Bo spin_unlock(&fs_info->ordered_root_lock); 3838199c2a9cSMiao Xie btrfs_destroy_ordered_extents(root); 3839199c2a9cSMiao Xie 38402a85d9caSLiu Bo cond_resched(); 38412a85d9caSLiu Bo spin_lock(&fs_info->ordered_root_lock); 3842199c2a9cSMiao Xie } 3843199c2a9cSMiao Xie spin_unlock(&fs_info->ordered_root_lock); 3844acce952bSliubo } 3845acce952bSliubo 384635a3621bSStefan Behrens static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 3847acce952bSliubo struct btrfs_root *root) 3848acce952bSliubo { 3849acce952bSliubo struct rb_node *node; 3850acce952bSliubo struct btrfs_delayed_ref_root *delayed_refs; 3851acce952bSliubo struct btrfs_delayed_ref_node *ref; 3852acce952bSliubo int ret = 0; 3853acce952bSliubo 3854acce952bSliubo delayed_refs = &trans->delayed_refs; 3855acce952bSliubo 3856acce952bSliubo spin_lock(&delayed_refs->lock); 3857d7df2c79SJosef Bacik if (atomic_read(&delayed_refs->num_entries) == 0) { 3858cfece4dbSDavid Sterba spin_unlock(&delayed_refs->lock); 3859efe120a0SFrank Holton btrfs_info(root->fs_info, "delayed_refs has NO entry"); 3860acce952bSliubo return ret; 3861acce952bSliubo } 3862acce952bSliubo 3863d7df2c79SJosef Bacik while ((node = rb_first(&delayed_refs->href_root)) != NULL) { 3864d7df2c79SJosef Bacik struct btrfs_delayed_ref_head *head; 3865e78417d1SJosef Bacik bool pin_bytes = false; 3866acce952bSliubo 3867d7df2c79SJosef Bacik head = rb_entry(node, struct btrfs_delayed_ref_head, 3868d7df2c79SJosef Bacik href_node); 3869b939d1abSJosef Bacik if (!mutex_trylock(&head->mutex)) { 3870d7df2c79SJosef Bacik atomic_inc(&head->node.refs); 387179787eaaSJeff Mahoney spin_unlock(&delayed_refs->lock); 3872b939d1abSJosef Bacik 3873acce952bSliubo mutex_lock(&head->mutex); 3874b939d1abSJosef Bacik mutex_unlock(&head->mutex); 3875d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 3876e18fca73SJosef Bacik spin_lock(&delayed_refs->lock); 3877b939d1abSJosef Bacik continue; 3878b939d1abSJosef Bacik } 3879d7df2c79SJosef Bacik spin_lock(&head->lock); 3880d7df2c79SJosef Bacik while ((node = rb_first(&head->ref_root)) != NULL) { 3881d7df2c79SJosef Bacik ref = rb_entry(node, struct btrfs_delayed_ref_node, 3882d7df2c79SJosef Bacik rb_node); 3883d7df2c79SJosef Bacik ref->in_tree = 0; 3884d7df2c79SJosef Bacik rb_erase(&ref->rb_node, &head->ref_root); 3885d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 3886d7df2c79SJosef Bacik btrfs_put_delayed_ref(ref); 3887d7df2c79SJosef Bacik } 388854067ae9SJosef Bacik if (head->must_insert_reserved) 3889e78417d1SJosef Bacik pin_bytes = true; 389078a6184aSMiao Xie btrfs_free_delayed_extent_op(head->extent_op); 3891acce952bSliubo delayed_refs->num_heads--; 3892d7df2c79SJosef Bacik if (head->processing == 0) 3893acce952bSliubo delayed_refs->num_heads_ready--; 3894d7df2c79SJosef Bacik atomic_dec(&delayed_refs->num_entries); 3895d7df2c79SJosef Bacik head->node.in_tree = 0; 3896c46effa6SLiu Bo rb_erase(&head->href_node, &delayed_refs->href_root); 3897d7df2c79SJosef Bacik spin_unlock(&head->lock); 3898acce952bSliubo spin_unlock(&delayed_refs->lock); 3899e78417d1SJosef Bacik mutex_unlock(&head->mutex); 3900acce952bSliubo 3901d7df2c79SJosef Bacik if (pin_bytes) 3902d7df2c79SJosef Bacik btrfs_pin_extent(root, head->node.bytenr, 3903d7df2c79SJosef Bacik head->node.num_bytes, 1); 3904d7df2c79SJosef Bacik btrfs_put_delayed_ref(&head->node); 3905acce952bSliubo cond_resched(); 3906acce952bSliubo spin_lock(&delayed_refs->lock); 3907acce952bSliubo } 3908acce952bSliubo 3909acce952bSliubo spin_unlock(&delayed_refs->lock); 3910acce952bSliubo 3911acce952bSliubo return ret; 3912acce952bSliubo } 3913acce952bSliubo 3914143bede5SJeff Mahoney static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) 3915acce952bSliubo { 3916acce952bSliubo struct btrfs_inode *btrfs_inode; 3917acce952bSliubo struct list_head splice; 3918acce952bSliubo 3919acce952bSliubo INIT_LIST_HEAD(&splice); 3920acce952bSliubo 3921eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 3922eb73c1b7SMiao Xie list_splice_init(&root->delalloc_inodes, &splice); 3923acce952bSliubo 3924acce952bSliubo while (!list_empty(&splice)) { 3925eb73c1b7SMiao Xie btrfs_inode = list_first_entry(&splice, struct btrfs_inode, 3926acce952bSliubo delalloc_inodes); 3927acce952bSliubo 3928acce952bSliubo list_del_init(&btrfs_inode->delalloc_inodes); 3929df0af1a5SMiao Xie clear_bit(BTRFS_INODE_IN_DELALLOC_LIST, 3930df0af1a5SMiao Xie &btrfs_inode->runtime_flags); 3931eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 3932acce952bSliubo 3933acce952bSliubo btrfs_invalidate_inodes(btrfs_inode->root); 3934b216cbfbSMiao Xie 3935eb73c1b7SMiao Xie spin_lock(&root->delalloc_lock); 3936acce952bSliubo } 3937acce952bSliubo 3938eb73c1b7SMiao Xie spin_unlock(&root->delalloc_lock); 3939eb73c1b7SMiao Xie } 3940eb73c1b7SMiao Xie 3941eb73c1b7SMiao Xie static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) 3942eb73c1b7SMiao Xie { 3943eb73c1b7SMiao Xie struct btrfs_root *root; 3944eb73c1b7SMiao Xie struct list_head splice; 3945eb73c1b7SMiao Xie 3946eb73c1b7SMiao Xie INIT_LIST_HEAD(&splice); 3947eb73c1b7SMiao Xie 3948eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 3949eb73c1b7SMiao Xie list_splice_init(&fs_info->delalloc_roots, &splice); 3950eb73c1b7SMiao Xie while (!list_empty(&splice)) { 3951eb73c1b7SMiao Xie root = list_first_entry(&splice, struct btrfs_root, 3952eb73c1b7SMiao Xie delalloc_root); 3953eb73c1b7SMiao Xie list_del_init(&root->delalloc_root); 3954eb73c1b7SMiao Xie root = btrfs_grab_fs_root(root); 3955eb73c1b7SMiao Xie BUG_ON(!root); 3956eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 3957eb73c1b7SMiao Xie 3958eb73c1b7SMiao Xie btrfs_destroy_delalloc_inodes(root); 3959eb73c1b7SMiao Xie btrfs_put_fs_root(root); 3960eb73c1b7SMiao Xie 3961eb73c1b7SMiao Xie spin_lock(&fs_info->delalloc_root_lock); 3962eb73c1b7SMiao Xie } 3963eb73c1b7SMiao Xie spin_unlock(&fs_info->delalloc_root_lock); 3964acce952bSliubo } 3965acce952bSliubo 3966acce952bSliubo static int btrfs_destroy_marked_extents(struct btrfs_root *root, 3967acce952bSliubo struct extent_io_tree *dirty_pages, 3968acce952bSliubo int mark) 3969acce952bSliubo { 3970acce952bSliubo int ret; 3971acce952bSliubo struct extent_buffer *eb; 3972acce952bSliubo u64 start = 0; 3973acce952bSliubo u64 end; 3974acce952bSliubo 3975acce952bSliubo while (1) { 3976acce952bSliubo ret = find_first_extent_bit(dirty_pages, start, &start, &end, 3977e6138876SJosef Bacik mark, NULL); 3978acce952bSliubo if (ret) 3979acce952bSliubo break; 3980acce952bSliubo 3981acce952bSliubo clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS); 3982acce952bSliubo while (start <= end) { 3983fd8b2b61SJosef Bacik eb = btrfs_find_tree_block(root, start, 3984fd8b2b61SJosef Bacik root->leafsize); 398569a85bd8SJosef Bacik start += root->leafsize; 3986fd8b2b61SJosef Bacik if (!eb) 3987acce952bSliubo continue; 3988fd8b2b61SJosef Bacik wait_on_extent_buffer_writeback(eb); 3989acce952bSliubo 3990fd8b2b61SJosef Bacik if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, 3991fd8b2b61SJosef Bacik &eb->bflags)) 3992fd8b2b61SJosef Bacik clear_extent_buffer_dirty(eb); 3993fd8b2b61SJosef Bacik free_extent_buffer_stale(eb); 3994acce952bSliubo } 3995acce952bSliubo } 3996acce952bSliubo 3997acce952bSliubo return ret; 3998acce952bSliubo } 3999acce952bSliubo 4000acce952bSliubo static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 4001acce952bSliubo struct extent_io_tree *pinned_extents) 4002acce952bSliubo { 4003acce952bSliubo struct extent_io_tree *unpin; 4004acce952bSliubo u64 start; 4005acce952bSliubo u64 end; 4006acce952bSliubo int ret; 4007ed0eaa14SLiu Bo bool loop = true; 4008acce952bSliubo 4009acce952bSliubo unpin = pinned_extents; 4010ed0eaa14SLiu Bo again: 4011acce952bSliubo while (1) { 4012acce952bSliubo ret = find_first_extent_bit(unpin, 0, &start, &end, 4013e6138876SJosef Bacik EXTENT_DIRTY, NULL); 4014acce952bSliubo if (ret) 4015acce952bSliubo break; 4016acce952bSliubo 4017acce952bSliubo /* opt_discard */ 40185378e607SLi Dongyang if (btrfs_test_opt(root, DISCARD)) 40195378e607SLi Dongyang ret = btrfs_error_discard_extent(root, start, 40205378e607SLi Dongyang end + 1 - start, 40215378e607SLi Dongyang NULL); 4022acce952bSliubo 4023acce952bSliubo clear_extent_dirty(unpin, start, end, GFP_NOFS); 4024acce952bSliubo btrfs_error_unpin_extent_range(root, start, end); 4025acce952bSliubo cond_resched(); 4026acce952bSliubo } 4027acce952bSliubo 4028ed0eaa14SLiu Bo if (loop) { 4029ed0eaa14SLiu Bo if (unpin == &root->fs_info->freed_extents[0]) 4030ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[1]; 4031ed0eaa14SLiu Bo else 4032ed0eaa14SLiu Bo unpin = &root->fs_info->freed_extents[0]; 4033ed0eaa14SLiu Bo loop = false; 4034ed0eaa14SLiu Bo goto again; 4035ed0eaa14SLiu Bo } 4036ed0eaa14SLiu Bo 4037acce952bSliubo return 0; 4038acce952bSliubo } 4039acce952bSliubo 404049b25e05SJeff Mahoney void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 404149b25e05SJeff Mahoney struct btrfs_root *root) 404249b25e05SJeff Mahoney { 4043724e2315SJosef Bacik btrfs_destroy_ordered_operations(cur_trans, root); 4044724e2315SJosef Bacik 404549b25e05SJeff Mahoney btrfs_destroy_delayed_refs(cur_trans, root); 404649b25e05SJeff Mahoney 40474a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_COMMIT_START; 404849b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_blocked_wait); 404949b25e05SJeff Mahoney 40504a9d8bdeSMiao Xie cur_trans->state = TRANS_STATE_UNBLOCKED; 405149b25e05SJeff Mahoney wake_up(&root->fs_info->transaction_wait); 405249b25e05SJeff Mahoney 405367cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 405467cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 405567cde344SMiao Xie 405649b25e05SJeff Mahoney btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, 405749b25e05SJeff Mahoney EXTENT_DIRTY); 40586e841e32SLiu Bo btrfs_destroy_pinned_extent(root, 40596e841e32SLiu Bo root->fs_info->pinned_extents); 406049b25e05SJeff Mahoney 40614a9d8bdeSMiao Xie cur_trans->state =TRANS_STATE_COMPLETED; 40624a9d8bdeSMiao Xie wake_up(&cur_trans->commit_wait); 40634a9d8bdeSMiao Xie 406449b25e05SJeff Mahoney /* 406549b25e05SJeff Mahoney memset(cur_trans, 0, sizeof(*cur_trans)); 406649b25e05SJeff Mahoney kmem_cache_free(btrfs_transaction_cachep, cur_trans); 406749b25e05SJeff Mahoney */ 406849b25e05SJeff Mahoney } 406949b25e05SJeff Mahoney 407048a3b636SEric Sandeen static int btrfs_cleanup_transaction(struct btrfs_root *root) 4071acce952bSliubo { 4072acce952bSliubo struct btrfs_transaction *t; 4073acce952bSliubo 4074acce952bSliubo mutex_lock(&root->fs_info->transaction_kthread_mutex); 4075acce952bSliubo 4076a4abeea4SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4077724e2315SJosef Bacik while (!list_empty(&root->fs_info->trans_list)) { 4078724e2315SJosef Bacik t = list_first_entry(&root->fs_info->trans_list, 4079724e2315SJosef Bacik struct btrfs_transaction, list); 4080724e2315SJosef Bacik if (t->state >= TRANS_STATE_COMMIT_START) { 4081724e2315SJosef Bacik atomic_inc(&t->use_count); 4082724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4083724e2315SJosef Bacik btrfs_wait_for_commit(root, t->transid); 4084724e2315SJosef Bacik btrfs_put_transaction(t); 4085724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4086724e2315SJosef Bacik continue; 4087724e2315SJosef Bacik } 4088724e2315SJosef Bacik if (t == root->fs_info->running_transaction) { 4089724e2315SJosef Bacik t->state = TRANS_STATE_COMMIT_DOING; 4090724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4091724e2315SJosef Bacik /* 4092724e2315SJosef Bacik * We wait for 0 num_writers since we don't hold a trans 4093724e2315SJosef Bacik * handle open currently for this transaction. 4094724e2315SJosef Bacik */ 4095724e2315SJosef Bacik wait_event(t->writer_wait, 4096724e2315SJosef Bacik atomic_read(&t->num_writers) == 0); 4097724e2315SJosef Bacik } else { 4098724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4099724e2315SJosef Bacik } 4100724e2315SJosef Bacik btrfs_cleanup_one_transaction(t, root); 4101724e2315SJosef Bacik 4102724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4103724e2315SJosef Bacik if (t == root->fs_info->running_transaction) 4104ac673879SMiao Xie root->fs_info->running_transaction = NULL; 4105724e2315SJosef Bacik list_del_init(&t->list); 4106a4abeea4SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4107a4abeea4SJosef Bacik 4108724e2315SJosef Bacik btrfs_put_transaction(t); 4109724e2315SJosef Bacik trace_btrfs_transaction_commit(root); 4110724e2315SJosef Bacik spin_lock(&root->fs_info->trans_lock); 4111724e2315SJosef Bacik } 4112724e2315SJosef Bacik spin_unlock(&root->fs_info->trans_lock); 4113199c2a9cSMiao Xie btrfs_destroy_all_ordered_extents(root->fs_info); 411467cde344SMiao Xie btrfs_destroy_delayed_inodes(root); 411567cde344SMiao Xie btrfs_assert_delayed_root_empty(root); 4116724e2315SJosef Bacik btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); 4117eb73c1b7SMiao Xie btrfs_destroy_all_delalloc_inodes(root->fs_info); 4118acce952bSliubo mutex_unlock(&root->fs_info->transaction_kthread_mutex); 4119acce952bSliubo 4120acce952bSliubo return 0; 4121acce952bSliubo } 4122acce952bSliubo 4123d1310b2eSChris Mason static struct extent_io_ops btree_extent_io_ops = { 4124ce9adaa5SChris Mason .readpage_end_io_hook = btree_readpage_end_io_hook, 41254bb31e92SArne Jansen .readpage_io_failed_hook = btree_io_failed_hook, 41260b86a832SChris Mason .submit_bio_hook = btree_submit_bio_hook, 4127239b14b3SChris Mason /* note we're sharing with inode.c for the merge bio hook */ 4128239b14b3SChris Mason .merge_bio_hook = btrfs_merge_bio_hook, 41290da5468fSChris Mason }; 4130