aboutsummaryrefslogtreecommitdiffstats
path: root/fs/btrfs/disk-io.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/btrfs/disk-io.c')
-rw-r--r--fs/btrfs/disk-io.c656
1 files changed, 331 insertions, 325 deletions
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3a57f99d96aa..08b74daf35d0 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -64,19 +64,18 @@
64static const struct extent_io_ops btree_extent_io_ops; 64static const struct extent_io_ops btree_extent_io_ops;
65static void end_workqueue_fn(struct btrfs_work *work); 65static void end_workqueue_fn(struct btrfs_work *work);
66static void free_fs_root(struct btrfs_root *root); 66static void free_fs_root(struct btrfs_root *root);
67static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 67static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info);
68 int read_only);
69static void btrfs_destroy_ordered_extents(struct btrfs_root *root); 68static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
70static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 69static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
71 struct btrfs_root *root); 70 struct btrfs_fs_info *fs_info);
72static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); 71static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
73static int btrfs_destroy_marked_extents(struct btrfs_root *root, 72static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info,
74 struct extent_io_tree *dirty_pages, 73 struct extent_io_tree *dirty_pages,
75 int mark); 74 int mark);
76static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 75static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info,
77 struct extent_io_tree *pinned_extents); 76 struct extent_io_tree *pinned_extents);
78static int btrfs_cleanup_transaction(struct btrfs_root *root); 77static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info);
79static void btrfs_error_commit_super(struct btrfs_root *root); 78static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info);
80 79
81/* 80/*
82 * btrfs_end_io_wq structs are used to do processing in task context when an IO 81 * btrfs_end_io_wq structs are used to do processing in task context when an IO
@@ -220,19 +219,19 @@ void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
220 * extents on the btree inode are pretty simple, there's one extent 219 * extents on the btree inode are pretty simple, there's one extent
221 * that covers the entire device 220 * that covers the entire device
222 */ 221 */
223static struct extent_map *btree_get_extent(struct inode *inode, 222static struct extent_map *btree_get_extent(struct btrfs_inode *inode,
224 struct page *page, size_t pg_offset, u64 start, u64 len, 223 struct page *page, size_t pg_offset, u64 start, u64 len,
225 int create) 224 int create)
226{ 225{
227 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; 226 struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
227 struct extent_map_tree *em_tree = &inode->extent_tree;
228 struct extent_map *em; 228 struct extent_map *em;
229 int ret; 229 int ret;
230 230
231 read_lock(&em_tree->lock); 231 read_lock(&em_tree->lock);
232 em = lookup_extent_mapping(em_tree, start, len); 232 em = lookup_extent_mapping(em_tree, start, len);
233 if (em) { 233 if (em) {
234 em->bdev = 234 em->bdev = fs_info->fs_devices->latest_bdev;
235 BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
236 read_unlock(&em_tree->lock); 235 read_unlock(&em_tree->lock);
237 goto out; 236 goto out;
238 } 237 }
@@ -247,7 +246,7 @@ static struct extent_map *btree_get_extent(struct inode *inode,
247 em->len = (u64)-1; 246 em->len = (u64)-1;
248 em->block_len = (u64)-1; 247 em->block_len = (u64)-1;
249 em->block_start = 0; 248 em->block_start = 0;
250 em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 249 em->bdev = fs_info->fs_devices->latest_bdev;
251 250
252 write_lock(&em_tree->lock); 251 write_lock(&em_tree->lock);
253 ret = add_extent_mapping(em_tree, em, 0); 252 ret = add_extent_mapping(em_tree, em, 0);
@@ -266,12 +265,12 @@ out:
266 return em; 265 return em;
267} 266}
268 267
269u32 btrfs_csum_data(char *data, u32 seed, size_t len) 268u32 btrfs_csum_data(const char *data, u32 seed, size_t len)
270{ 269{
271 return btrfs_crc32c(seed, data, len); 270 return btrfs_crc32c(seed, data, len);
272} 271}
273 272
274void btrfs_csum_final(u32 crc, char *result) 273void btrfs_csum_final(u32 crc, u8 *result)
275{ 274{
276 put_unaligned_le32(~crc, result); 275 put_unaligned_le32(~crc, result);
277} 276}
@@ -440,7 +439,7 @@ static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info,
440 * helper to read a given tree block, doing retries as required when 439 * helper to read a given tree block, doing retries as required when
441 * the checksums don't match and we have alternate mirrors to try. 440 * the checksums don't match and we have alternate mirrors to try.
442 */ 441 */
443static int btree_read_extent_buffer_pages(struct btrfs_root *root, 442static int btree_read_extent_buffer_pages(struct btrfs_fs_info *fs_info,
444 struct extent_buffer *eb, 443 struct extent_buffer *eb,
445 u64 parent_transid) 444 u64 parent_transid)
446{ 445{
@@ -452,7 +451,7 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root,
452 int failed_mirror = 0; 451 int failed_mirror = 0;
453 452
454 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); 453 clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
455 io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; 454 io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
456 while (1) { 455 while (1) {
457 ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE, 456 ret = read_extent_buffer_pages(io_tree, eb, WAIT_COMPLETE,
458 btree_get_extent, mirror_num); 457 btree_get_extent, mirror_num);
@@ -472,7 +471,7 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root,
472 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags)) 471 if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
473 break; 472 break;
474 473
475 num_copies = btrfs_num_copies(root->fs_info, 474 num_copies = btrfs_num_copies(fs_info,
476 eb->start, eb->len); 475 eb->start, eb->len);
477 if (num_copies == 1) 476 if (num_copies == 1)
478 break; 477 break;
@@ -491,7 +490,7 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root,
491 } 490 }
492 491
493 if (failed && !ret && failed_mirror) 492 if (failed && !ret && failed_mirror)
494 repair_eb_io_failure(root, eb, failed_mirror); 493 repair_eb_io_failure(fs_info, eb, failed_mirror);
495 494
496 return ret; 495 return ret;
497} 496}
@@ -545,47 +544,63 @@ static int check_tree_block_fsid(struct btrfs_fs_info *fs_info,
545 return ret; 544 return ret;
546} 545}
547 546
548#define CORRUPT(reason, eb, root, slot) \ 547#define CORRUPT(reason, eb, root, slot) \
549 btrfs_crit(root->fs_info, "corrupt %s, %s: block=%llu," \ 548 btrfs_crit(root->fs_info, \
550 " root=%llu, slot=%d", \ 549 "corrupt %s, %s: block=%llu, root=%llu, slot=%d", \
551 btrfs_header_level(eb) == 0 ? "leaf" : "node",\ 550 btrfs_header_level(eb) == 0 ? "leaf" : "node", \
552 reason, btrfs_header_bytenr(eb), root->objectid, slot) 551 reason, btrfs_header_bytenr(eb), root->objectid, slot)
553 552
554static noinline int check_leaf(struct btrfs_root *root, 553static noinline int check_leaf(struct btrfs_root *root,
555 struct extent_buffer *leaf) 554 struct extent_buffer *leaf)
556{ 555{
556 struct btrfs_fs_info *fs_info = root->fs_info;
557 struct btrfs_key key; 557 struct btrfs_key key;
558 struct btrfs_key leaf_key; 558 struct btrfs_key leaf_key;
559 u32 nritems = btrfs_header_nritems(leaf); 559 u32 nritems = btrfs_header_nritems(leaf);
560 int slot; 560 int slot;
561 561
562 if (nritems == 0) { 562 /*
563 * Extent buffers from a relocation tree have a owner field that
564 * corresponds to the subvolume tree they are based on. So just from an
565 * extent buffer alone we can not find out what is the id of the
566 * corresponding subvolume tree, so we can not figure out if the extent
567 * buffer corresponds to the root of the relocation tree or not. So skip
568 * this check for relocation trees.
569 */
570 if (nritems == 0 && !btrfs_header_flag(leaf, BTRFS_HEADER_FLAG_RELOC)) {
563 struct btrfs_root *check_root; 571 struct btrfs_root *check_root;
564 572
565 key.objectid = btrfs_header_owner(leaf); 573 key.objectid = btrfs_header_owner(leaf);
566 key.type = BTRFS_ROOT_ITEM_KEY; 574 key.type = BTRFS_ROOT_ITEM_KEY;
567 key.offset = (u64)-1; 575 key.offset = (u64)-1;
568 576
569 check_root = btrfs_get_fs_root(root->fs_info, &key, false); 577 check_root = btrfs_get_fs_root(fs_info, &key, false);
570 /* 578 /*
571 * The only reason we also check NULL here is that during 579 * The only reason we also check NULL here is that during
572 * open_ctree() some roots has not yet been set up. 580 * open_ctree() some roots has not yet been set up.
573 */ 581 */
574 if (!IS_ERR_OR_NULL(check_root)) { 582 if (!IS_ERR_OR_NULL(check_root)) {
583 struct extent_buffer *eb;
584
585 eb = btrfs_root_node(check_root);
575 /* if leaf is the root, then it's fine */ 586 /* if leaf is the root, then it's fine */
576 if (leaf->start != 587 if (leaf != eb) {
577 btrfs_root_bytenr(&check_root->root_item)) {
578 CORRUPT("non-root leaf's nritems is 0", 588 CORRUPT("non-root leaf's nritems is 0",
579 leaf, root, 0); 589 leaf, check_root, 0);
590 free_extent_buffer(eb);
580 return -EIO; 591 return -EIO;
581 } 592 }
593 free_extent_buffer(eb);
582 } 594 }
583 return 0; 595 return 0;
584 } 596 }
585 597
598 if (nritems == 0)
599 return 0;
600
586 /* Check the 0 item */ 601 /* Check the 0 item */
587 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) != 602 if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
588 BTRFS_LEAF_DATA_SIZE(root)) { 603 BTRFS_LEAF_DATA_SIZE(fs_info)) {
589 CORRUPT("invalid item offset size pair", leaf, root, 0); 604 CORRUPT("invalid item offset size pair", leaf, root, 0);
590 return -EIO; 605 return -EIO;
591 } 606 }
@@ -624,7 +639,7 @@ static noinline int check_leaf(struct btrfs_root *root,
624 * all point outside of the leaf. 639 * all point outside of the leaf.
625 */ 640 */
626 if (btrfs_item_end_nr(leaf, slot) > 641 if (btrfs_item_end_nr(leaf, slot) >
627 BTRFS_LEAF_DATA_SIZE(root)) { 642 BTRFS_LEAF_DATA_SIZE(fs_info)) {
628 CORRUPT("slot end outside of leaf", leaf, root, slot); 643 CORRUPT("slot end outside of leaf", leaf, root, slot);
629 return -EIO; 644 return -EIO;
630 } 645 }
@@ -641,7 +656,7 @@ static int check_node(struct btrfs_root *root, struct extent_buffer *node)
641 u64 bytenr; 656 u64 bytenr;
642 int ret = 0; 657 int ret = 0;
643 658
644 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root)) { 659 if (nr == 0 || nr > BTRFS_NODEPTRS_PER_BLOCK(root->fs_info)) {
645 btrfs_crit(root->fs_info, 660 btrfs_crit(root->fs_info,
646 "corrupt node: block %llu root %llu nritems %lu", 661 "corrupt node: block %llu root %llu nritems %lu",
647 node->start, root->objectid, nr); 662 node->start, root->objectid, nr);
@@ -747,7 +762,7 @@ static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
747err: 762err:
748 if (reads_done && 763 if (reads_done &&
749 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 764 test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
750 btree_readahead_hook(fs_info, eb, eb->start, ret); 765 btree_readahead_hook(fs_info, eb, ret);
751 766
752 if (ret) { 767 if (ret) {
753 /* 768 /*
@@ -772,7 +787,7 @@ static int btree_io_failed_hook(struct page *page, int failed_mirror)
772 eb->read_mirror = failed_mirror; 787 eb->read_mirror = failed_mirror;
773 atomic_dec(&eb->io_pages); 788 atomic_dec(&eb->io_pages);
774 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) 789 if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
775 btree_readahead_hook(eb->fs_info, eb, eb->start, -EIO); 790 btree_readahead_hook(eb->fs_info, eb, -EIO);
776 return -EIO; /* we fixed nothing */ 791 return -EIO; /* we fixed nothing */
777} 792}
778 793
@@ -930,7 +945,7 @@ int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
930 945
931 atomic_inc(&fs_info->nr_async_submits); 946 atomic_inc(&fs_info->nr_async_submits);
932 947
933 if (bio->bi_opf & REQ_SYNC) 948 if (op_is_sync(bio->bi_opf))
934 btrfs_set_work_high_priority(&async->work); 949 btrfs_set_work_high_priority(&async->work);
935 950
936 btrfs_queue_work(fs_info->workers, &async->work); 951 btrfs_queue_work(fs_info->workers, &async->work);
@@ -981,7 +996,7 @@ static int __btree_submit_bio_done(struct inode *inode, struct bio *bio,
981 * when we're called for a write, we're already in the async 996 * when we're called for a write, we're already in the async
982 * submission context. Just jump into btrfs_map_bio 997 * submission context. Just jump into btrfs_map_bio
983 */ 998 */
984 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 1); 999 ret = btrfs_map_bio(btrfs_sb(inode->i_sb), bio, mirror_num, 1);
985 if (ret) { 1000 if (ret) {
986 bio->bi_error = ret; 1001 bio->bi_error = ret;
987 bio_endio(bio); 1002 bio_endio(bio);
@@ -989,7 +1004,7 @@ static int __btree_submit_bio_done(struct inode *inode, struct bio *bio,
989 return ret; 1004 return ret;
990} 1005}
991 1006
992static int check_async_write(struct inode *inode, unsigned long bio_flags) 1007static int check_async_write(unsigned long bio_flags)
993{ 1008{
994 if (bio_flags & EXTENT_BIO_TREE_LOG) 1009 if (bio_flags & EXTENT_BIO_TREE_LOG)
995 return 0; 1010 return 0;
@@ -1004,7 +1019,8 @@ static int btree_submit_bio_hook(struct inode *inode, struct bio *bio,
1004 int mirror_num, unsigned long bio_flags, 1019 int mirror_num, unsigned long bio_flags,
1005 u64 bio_offset) 1020 u64 bio_offset)
1006{ 1021{
1007 int async = check_async_write(inode, bio_flags); 1022 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1023 int async = check_async_write(bio_flags);
1008 int ret; 1024 int ret;
1009 1025
1010 if (bio_op(bio) != REQ_OP_WRITE) { 1026 if (bio_op(bio) != REQ_OP_WRITE) {
@@ -1012,23 +1028,22 @@ static int btree_submit_bio_hook(struct inode *inode, struct bio *bio,
1012 * called for a read, do the setup so that checksum validation 1028 * called for a read, do the setup so that checksum validation
1013 * can happen in the async kernel threads 1029 * can happen in the async kernel threads
1014 */ 1030 */
1015 ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, 1031 ret = btrfs_bio_wq_end_io(fs_info, bio,
1016 bio, BTRFS_WQ_ENDIO_METADATA); 1032 BTRFS_WQ_ENDIO_METADATA);
1017 if (ret) 1033 if (ret)
1018 goto out_w_error; 1034 goto out_w_error;
1019 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0); 1035 ret = btrfs_map_bio(fs_info, bio, mirror_num, 0);
1020 } else if (!async) { 1036 } else if (!async) {
1021 ret = btree_csum_one_bio(bio); 1037 ret = btree_csum_one_bio(bio);
1022 if (ret) 1038 if (ret)
1023 goto out_w_error; 1039 goto out_w_error;
1024 ret = btrfs_map_bio(BTRFS_I(inode)->root, bio, mirror_num, 0); 1040 ret = btrfs_map_bio(fs_info, bio, mirror_num, 0);
1025 } else { 1041 } else {
1026 /* 1042 /*
1027 * kthread helpers are used to submit writes so that 1043 * kthread helpers are used to submit writes so that
1028 * checksumming can happen in parallel across all CPUs 1044 * checksumming can happen in parallel across all CPUs
1029 */ 1045 */
1030 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, 1046 ret = btrfs_wq_submit_bio(fs_info, inode, bio, mirror_num, 0,
1031 inode, bio, mirror_num, 0,
1032 bio_offset, 1047 bio_offset,
1033 __btree_submit_bio_start, 1048 __btree_submit_bio_start,
1034 __btree_submit_bio_done); 1049 __btree_submit_bio_done);
@@ -1146,12 +1161,12 @@ static const struct address_space_operations btree_aops = {
1146 .set_page_dirty = btree_set_page_dirty, 1161 .set_page_dirty = btree_set_page_dirty,
1147}; 1162};
1148 1163
1149void readahead_tree_block(struct btrfs_root *root, u64 bytenr) 1164void readahead_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr)
1150{ 1165{
1151 struct extent_buffer *buf = NULL; 1166 struct extent_buffer *buf = NULL;
1152 struct inode *btree_inode = root->fs_info->btree_inode; 1167 struct inode *btree_inode = fs_info->btree_inode;
1153 1168
1154 buf = btrfs_find_create_tree_block(root, bytenr); 1169 buf = btrfs_find_create_tree_block(fs_info, bytenr);
1155 if (IS_ERR(buf)) 1170 if (IS_ERR(buf))
1156 return; 1171 return;
1157 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, 1172 read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
@@ -1159,15 +1174,15 @@ void readahead_tree_block(struct btrfs_root *root, u64 bytenr)
1159 free_extent_buffer(buf); 1174 free_extent_buffer(buf);
1160} 1175}
1161 1176
1162int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, 1177int reada_tree_block_flagged(struct btrfs_fs_info *fs_info, u64 bytenr,
1163 int mirror_num, struct extent_buffer **eb) 1178 int mirror_num, struct extent_buffer **eb)
1164{ 1179{
1165 struct extent_buffer *buf = NULL; 1180 struct extent_buffer *buf = NULL;
1166 struct inode *btree_inode = root->fs_info->btree_inode; 1181 struct inode *btree_inode = fs_info->btree_inode;
1167 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree; 1182 struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1168 int ret; 1183 int ret;
1169 1184
1170 buf = btrfs_find_create_tree_block(root, bytenr); 1185 buf = btrfs_find_create_tree_block(fs_info, bytenr);
1171 if (IS_ERR(buf)) 1186 if (IS_ERR(buf))
1172 return 0; 1187 return 0;
1173 1188
@@ -1191,19 +1206,13 @@ int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr,
1191 return 0; 1206 return 0;
1192} 1207}
1193 1208
1194struct extent_buffer *btrfs_find_tree_block(struct btrfs_fs_info *fs_info, 1209struct extent_buffer *btrfs_find_create_tree_block(
1195 u64 bytenr) 1210 struct btrfs_fs_info *fs_info,
1211 u64 bytenr)
1196{ 1212{
1197 return find_extent_buffer(fs_info, bytenr); 1213 if (btrfs_is_testing(fs_info))
1198} 1214 return alloc_test_extent_buffer(fs_info, bytenr);
1199 1215 return alloc_extent_buffer(fs_info, bytenr);
1200struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1201 u64 bytenr)
1202{
1203 if (btrfs_is_testing(root->fs_info))
1204 return alloc_test_extent_buffer(root->fs_info, bytenr,
1205 root->nodesize);
1206 return alloc_extent_buffer(root->fs_info, bytenr);
1207} 1216}
1208 1217
1209 1218
@@ -1219,17 +1228,17 @@ int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1219 buf->start, buf->start + buf->len - 1); 1228 buf->start, buf->start + buf->len - 1);
1220} 1229}
1221 1230
1222struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, 1231struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr,
1223 u64 parent_transid) 1232 u64 parent_transid)
1224{ 1233{
1225 struct extent_buffer *buf = NULL; 1234 struct extent_buffer *buf = NULL;
1226 int ret; 1235 int ret;
1227 1236
1228 buf = btrfs_find_create_tree_block(root, bytenr); 1237 buf = btrfs_find_create_tree_block(fs_info, bytenr);
1229 if (IS_ERR(buf)) 1238 if (IS_ERR(buf))
1230 return buf; 1239 return buf;
1231 1240
1232 ret = btree_read_extent_buffer_pages(root, buf, parent_transid); 1241 ret = btree_read_extent_buffer_pages(fs_info, buf, parent_transid);
1233 if (ret) { 1242 if (ret) {
1234 free_extent_buffer(buf); 1243 free_extent_buffer(buf);
1235 return ERR_PTR(ret); 1244 return ERR_PTR(ret);
@@ -1238,8 +1247,7 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
1238 1247
1239} 1248}
1240 1249
1241void clean_tree_block(struct btrfs_trans_handle *trans, 1250void clean_tree_block(struct btrfs_fs_info *fs_info,
1242 struct btrfs_fs_info *fs_info,
1243 struct extent_buffer *buf) 1251 struct extent_buffer *buf)
1244{ 1252{
1245 if (btrfs_header_generation(buf) == 1253 if (btrfs_header_generation(buf) ==
@@ -1283,16 +1291,12 @@ btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1283 kfree(writers); 1291 kfree(writers);
1284} 1292}
1285 1293
1286static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize, 1294static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info,
1287 struct btrfs_root *root, struct btrfs_fs_info *fs_info,
1288 u64 objectid) 1295 u64 objectid)
1289{ 1296{
1290 bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); 1297 bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state);
1291 root->node = NULL; 1298 root->node = NULL;
1292 root->commit_root = NULL; 1299 root->commit_root = NULL;
1293 root->sectorsize = sectorsize;
1294 root->nodesize = nodesize;
1295 root->stripesize = stripesize;
1296 root->state = 0; 1300 root->state = 0;
1297 root->orphan_cleanup_state = 0; 1301 root->orphan_cleanup_state = 0;
1298 1302
@@ -1370,8 +1374,7 @@ static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info,
1370 1374
1371#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 1375#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1372/* Should only be used by the testing infrastructure */ 1376/* Should only be used by the testing infrastructure */
1373struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info, 1377struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info)
1374 u32 sectorsize, u32 nodesize)
1375{ 1378{
1376 struct btrfs_root *root; 1379 struct btrfs_root *root;
1377 1380
@@ -1381,9 +1384,9 @@ struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info,
1381 root = btrfs_alloc_root(fs_info, GFP_KERNEL); 1384 root = btrfs_alloc_root(fs_info, GFP_KERNEL);
1382 if (!root) 1385 if (!root)
1383 return ERR_PTR(-ENOMEM); 1386 return ERR_PTR(-ENOMEM);
1387
1384 /* We don't use the stripesize in selftest, set it as sectorsize */ 1388 /* We don't use the stripesize in selftest, set it as sectorsize */
1385 __setup_root(nodesize, sectorsize, sectorsize, root, fs_info, 1389 __setup_root(root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
1386 BTRFS_ROOT_TREE_OBJECTID);
1387 root->alloc_bytenr = 0; 1390 root->alloc_bytenr = 0;
1388 1391
1389 return root; 1392 return root;
@@ -1405,8 +1408,7 @@ struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1405 if (!root) 1408 if (!root)
1406 return ERR_PTR(-ENOMEM); 1409 return ERR_PTR(-ENOMEM);
1407 1410
1408 __setup_root(tree_root->nodesize, tree_root->sectorsize, 1411 __setup_root(root, fs_info, objectid);
1409 tree_root->stripesize, root, fs_info, objectid);
1410 root->root_key.objectid = objectid; 1412 root->root_key.objectid = objectid;
1411 root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1413 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1412 root->root_key.offset = 0; 1414 root->root_key.offset = 0;
@@ -1418,18 +1420,15 @@ struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1418 goto fail; 1420 goto fail;
1419 } 1421 }
1420 1422
1421 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 1423 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
1422 btrfs_set_header_bytenr(leaf, leaf->start); 1424 btrfs_set_header_bytenr(leaf, leaf->start);
1423 btrfs_set_header_generation(leaf, trans->transid); 1425 btrfs_set_header_generation(leaf, trans->transid);
1424 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 1426 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1425 btrfs_set_header_owner(leaf, objectid); 1427 btrfs_set_header_owner(leaf, objectid);
1426 root->node = leaf; 1428 root->node = leaf;
1427 1429
1428 write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(), 1430 write_extent_buffer_fsid(leaf, fs_info->fsid);
1429 BTRFS_FSID_SIZE); 1431 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
1430 write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
1431 btrfs_header_chunk_tree_uuid(leaf),
1432 BTRFS_UUID_SIZE);
1433 btrfs_mark_buffer_dirty(leaf); 1432 btrfs_mark_buffer_dirty(leaf);
1434 1433
1435 root->commit_root = btrfs_root_node(root); 1434 root->commit_root = btrfs_root_node(root);
@@ -1474,16 +1473,13 @@ static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1474 struct btrfs_fs_info *fs_info) 1473 struct btrfs_fs_info *fs_info)
1475{ 1474{
1476 struct btrfs_root *root; 1475 struct btrfs_root *root;
1477 struct btrfs_root *tree_root = fs_info->tree_root;
1478 struct extent_buffer *leaf; 1476 struct extent_buffer *leaf;
1479 1477
1480 root = btrfs_alloc_root(fs_info, GFP_NOFS); 1478 root = btrfs_alloc_root(fs_info, GFP_NOFS);
1481 if (!root) 1479 if (!root)
1482 return ERR_PTR(-ENOMEM); 1480 return ERR_PTR(-ENOMEM);
1483 1481
1484 __setup_root(tree_root->nodesize, tree_root->sectorsize, 1482 __setup_root(root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1485 tree_root->stripesize, root, fs_info,
1486 BTRFS_TREE_LOG_OBJECTID);
1487 1483
1488 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; 1484 root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1489 root->root_key.type = BTRFS_ROOT_ITEM_KEY; 1485 root->root_key.type = BTRFS_ROOT_ITEM_KEY;
@@ -1505,15 +1501,14 @@ static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1505 return ERR_CAST(leaf); 1501 return ERR_CAST(leaf);
1506 } 1502 }
1507 1503
1508 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header)); 1504 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
1509 btrfs_set_header_bytenr(leaf, leaf->start); 1505 btrfs_set_header_bytenr(leaf, leaf->start);
1510 btrfs_set_header_generation(leaf, trans->transid); 1506 btrfs_set_header_generation(leaf, trans->transid);
1511 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV); 1507 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1512 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID); 1508 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
1513 root->node = leaf; 1509 root->node = leaf;
1514 1510
1515 write_extent_buffer(root->node, root->fs_info->fsid, 1511 write_extent_buffer_fsid(root->node, fs_info->fsid);
1516 btrfs_header_fsid(), BTRFS_FSID_SIZE);
1517 btrfs_mark_buffer_dirty(root->node); 1512 btrfs_mark_buffer_dirty(root->node);
1518 btrfs_tree_unlock(root->node); 1513 btrfs_tree_unlock(root->node);
1519 return root; 1514 return root;
@@ -1535,10 +1530,11 @@ int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1535int btrfs_add_log_tree(struct btrfs_trans_handle *trans, 1530int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1536 struct btrfs_root *root) 1531 struct btrfs_root *root)
1537{ 1532{
1533 struct btrfs_fs_info *fs_info = root->fs_info;
1538 struct btrfs_root *log_root; 1534 struct btrfs_root *log_root;
1539 struct btrfs_inode_item *inode_item; 1535 struct btrfs_inode_item *inode_item;
1540 1536
1541 log_root = alloc_log_tree(trans, root->fs_info); 1537 log_root = alloc_log_tree(trans, fs_info);
1542 if (IS_ERR(log_root)) 1538 if (IS_ERR(log_root))
1543 return PTR_ERR(log_root); 1539 return PTR_ERR(log_root);
1544 1540
@@ -1549,7 +1545,8 @@ int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1549 btrfs_set_stack_inode_generation(inode_item, 1); 1545 btrfs_set_stack_inode_generation(inode_item, 1);
1550 btrfs_set_stack_inode_size(inode_item, 3); 1546 btrfs_set_stack_inode_size(inode_item, 3);
1551 btrfs_set_stack_inode_nlink(inode_item, 1); 1547 btrfs_set_stack_inode_nlink(inode_item, 1);
1552 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize); 1548 btrfs_set_stack_inode_nbytes(inode_item,
1549 fs_info->nodesize);
1553 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); 1550 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1554 1551
1555 btrfs_set_root_node(&log_root->root_item, log_root->node); 1552 btrfs_set_root_node(&log_root->root_item, log_root->node);
@@ -1581,8 +1578,7 @@ static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1581 goto alloc_fail; 1578 goto alloc_fail;
1582 } 1579 }
1583 1580
1584 __setup_root(tree_root->nodesize, tree_root->sectorsize, 1581 __setup_root(root, fs_info, key->objectid);
1585 tree_root->stripesize, root, fs_info, key->objectid);
1586 1582
1587 ret = btrfs_find_root(tree_root, key, path, 1583 ret = btrfs_find_root(tree_root, key, path,
1588 &root->root_item, &root->root_key); 1584 &root->root_item, &root->root_key);
@@ -1593,7 +1589,8 @@ static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1593 } 1589 }
1594 1590
1595 generation = btrfs_root_generation(&root->root_item); 1591 generation = btrfs_root_generation(&root->root_item);
1596 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), 1592 root->node = read_tree_block(fs_info,
1593 btrfs_root_bytenr(&root->root_item),
1597 generation); 1594 generation);
1598 if (IS_ERR(root->node)) { 1595 if (IS_ERR(root->node)) {
1599 ret = PTR_ERR(root->node); 1596 ret = PTR_ERR(root->node);
@@ -1801,7 +1798,7 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1801 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) { 1798 list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
1802 if (!device->bdev) 1799 if (!device->bdev)
1803 continue; 1800 continue;
1804 bdi = blk_get_backing_dev_info(device->bdev); 1801 bdi = device->bdev->bd_bdi;
1805 if (bdi_congested(bdi, bdi_bits)) { 1802 if (bdi_congested(bdi, bdi_bits)) {
1806 ret = 1; 1803 ret = 1;
1807 break; 1804 break;
@@ -1848,6 +1845,7 @@ static void end_workqueue_fn(struct btrfs_work *work)
1848static int cleaner_kthread(void *arg) 1845static int cleaner_kthread(void *arg)
1849{ 1846{
1850 struct btrfs_root *root = arg; 1847 struct btrfs_root *root = arg;
1848 struct btrfs_fs_info *fs_info = root->fs_info;
1851 int again; 1849 int again;
1852 struct btrfs_trans_handle *trans; 1850 struct btrfs_trans_handle *trans;
1853 1851
@@ -1855,40 +1853,40 @@ static int cleaner_kthread(void *arg)
1855 again = 0; 1853 again = 0;
1856 1854
1857 /* Make the cleaner go to sleep early. */ 1855 /* Make the cleaner go to sleep early. */
1858 if (btrfs_need_cleaner_sleep(root)) 1856 if (btrfs_need_cleaner_sleep(fs_info))
1859 goto sleep; 1857 goto sleep;
1860 1858
1861 /* 1859 /*
1862 * Do not do anything if we might cause open_ctree() to block 1860 * Do not do anything if we might cause open_ctree() to block
1863 * before we have finished mounting the filesystem. 1861 * before we have finished mounting the filesystem.
1864 */ 1862 */
1865 if (!test_bit(BTRFS_FS_OPEN, &root->fs_info->flags)) 1863 if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))
1866 goto sleep; 1864 goto sleep;
1867 1865
1868 if (!mutex_trylock(&root->fs_info->cleaner_mutex)) 1866 if (!mutex_trylock(&fs_info->cleaner_mutex))
1869 goto sleep; 1867 goto sleep;
1870 1868
1871 /* 1869 /*
1872 * Avoid the problem that we change the status of the fs 1870 * Avoid the problem that we change the status of the fs
1873 * during the above check and trylock. 1871 * during the above check and trylock.
1874 */ 1872 */
1875 if (btrfs_need_cleaner_sleep(root)) { 1873 if (btrfs_need_cleaner_sleep(fs_info)) {
1876 mutex_unlock(&root->fs_info->cleaner_mutex); 1874 mutex_unlock(&fs_info->cleaner_mutex);
1877 goto sleep; 1875 goto sleep;
1878 } 1876 }
1879 1877
1880 mutex_lock(&root->fs_info->cleaner_delayed_iput_mutex); 1878 mutex_lock(&fs_info->cleaner_delayed_iput_mutex);
1881 btrfs_run_delayed_iputs(root); 1879 btrfs_run_delayed_iputs(fs_info);
1882 mutex_unlock(&root->fs_info->cleaner_delayed_iput_mutex); 1880 mutex_unlock(&fs_info->cleaner_delayed_iput_mutex);
1883 1881
1884 again = btrfs_clean_one_deleted_snapshot(root); 1882 again = btrfs_clean_one_deleted_snapshot(root);
1885 mutex_unlock(&root->fs_info->cleaner_mutex); 1883 mutex_unlock(&fs_info->cleaner_mutex);
1886 1884
1887 /* 1885 /*
1888 * The defragger has dealt with the R/O remount and umount, 1886 * The defragger has dealt with the R/O remount and umount,
1889 * needn't do anything special here. 1887 * needn't do anything special here.
1890 */ 1888 */
1891 btrfs_run_defrag_inodes(root->fs_info); 1889 btrfs_run_defrag_inodes(fs_info);
1892 1890
1893 /* 1891 /*
1894 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing 1892 * Acquires fs_info->delete_unused_bgs_mutex to avoid racing
@@ -1898,7 +1896,7 @@ static int cleaner_kthread(void *arg)
1898 * can't hold, nor need to, fs_info->cleaner_mutex when deleting 1896 * can't hold, nor need to, fs_info->cleaner_mutex when deleting
1899 * unused block groups. 1897 * unused block groups.
1900 */ 1898 */
1901 btrfs_delete_unused_bgs(root->fs_info); 1899 btrfs_delete_unused_bgs(fs_info);
1902sleep: 1900sleep:
1903 if (!again) { 1901 if (!again) {
1904 set_current_state(TASK_INTERRUPTIBLE); 1902 set_current_state(TASK_INTERRUPTIBLE);
@@ -1922,15 +1920,15 @@ sleep:
1922 trans = btrfs_attach_transaction(root); 1920 trans = btrfs_attach_transaction(root);
1923 if (IS_ERR(trans)) { 1921 if (IS_ERR(trans)) {
1924 if (PTR_ERR(trans) != -ENOENT) 1922 if (PTR_ERR(trans) != -ENOENT)
1925 btrfs_err(root->fs_info, 1923 btrfs_err(fs_info,
1926 "cleaner transaction attach returned %ld", 1924 "cleaner transaction attach returned %ld",
1927 PTR_ERR(trans)); 1925 PTR_ERR(trans));
1928 } else { 1926 } else {
1929 int ret; 1927 int ret;
1930 1928
1931 ret = btrfs_commit_transaction(trans, root); 1929 ret = btrfs_commit_transaction(trans);
1932 if (ret) 1930 if (ret)
1933 btrfs_err(root->fs_info, 1931 btrfs_err(fs_info,
1934 "cleaner open transaction commit returned %d", 1932 "cleaner open transaction commit returned %d",
1935 ret); 1933 ret);
1936 } 1934 }
@@ -1941,6 +1939,7 @@ sleep:
1941static int transaction_kthread(void *arg) 1939static int transaction_kthread(void *arg)
1942{ 1940{
1943 struct btrfs_root *root = arg; 1941 struct btrfs_root *root = arg;
1942 struct btrfs_fs_info *fs_info = root->fs_info;
1944 struct btrfs_trans_handle *trans; 1943 struct btrfs_trans_handle *trans;
1945 struct btrfs_transaction *cur; 1944 struct btrfs_transaction *cur;
1946 u64 transid; 1945 u64 transid;
@@ -1950,26 +1949,26 @@ static int transaction_kthread(void *arg)
1950 1949
1951 do { 1950 do {
1952 cannot_commit = false; 1951 cannot_commit = false;
1953 delay = HZ * root->fs_info->commit_interval; 1952 delay = HZ * fs_info->commit_interval;
1954 mutex_lock(&root->fs_info->transaction_kthread_mutex); 1953 mutex_lock(&fs_info->transaction_kthread_mutex);
1955 1954
1956 spin_lock(&root->fs_info->trans_lock); 1955 spin_lock(&fs_info->trans_lock);
1957 cur = root->fs_info->running_transaction; 1956 cur = fs_info->running_transaction;
1958 if (!cur) { 1957 if (!cur) {
1959 spin_unlock(&root->fs_info->trans_lock); 1958 spin_unlock(&fs_info->trans_lock);
1960 goto sleep; 1959 goto sleep;
1961 } 1960 }
1962 1961
1963 now = get_seconds(); 1962 now = get_seconds();
1964 if (cur->state < TRANS_STATE_BLOCKED && 1963 if (cur->state < TRANS_STATE_BLOCKED &&
1965 (now < cur->start_time || 1964 (now < cur->start_time ||
1966 now - cur->start_time < root->fs_info->commit_interval)) { 1965 now - cur->start_time < fs_info->commit_interval)) {
1967 spin_unlock(&root->fs_info->trans_lock); 1966 spin_unlock(&fs_info->trans_lock);
1968 delay = HZ * 5; 1967 delay = HZ * 5;
1969 goto sleep; 1968 goto sleep;
1970 } 1969 }
1971 transid = cur->transid; 1970 transid = cur->transid;
1972 spin_unlock(&root->fs_info->trans_lock); 1971 spin_unlock(&fs_info->trans_lock);
1973 1972
1974 /* If the file system is aborted, this will always fail. */ 1973 /* If the file system is aborted, this will always fail. */
1975 trans = btrfs_attach_transaction(root); 1974 trans = btrfs_attach_transaction(root);
@@ -1979,20 +1978,20 @@ static int transaction_kthread(void *arg)
1979 goto sleep; 1978 goto sleep;
1980 } 1979 }
1981 if (transid == trans->transid) { 1980 if (transid == trans->transid) {
1982 btrfs_commit_transaction(trans, root); 1981 btrfs_commit_transaction(trans);
1983 } else { 1982 } else {
1984 btrfs_end_transaction(trans, root); 1983 btrfs_end_transaction(trans);
1985 } 1984 }
1986sleep: 1985sleep:
1987 wake_up_process(root->fs_info->cleaner_kthread); 1986 wake_up_process(fs_info->cleaner_kthread);
1988 mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1987 mutex_unlock(&fs_info->transaction_kthread_mutex);
1989 1988
1990 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR, 1989 if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1991 &root->fs_info->fs_state))) 1990 &fs_info->fs_state)))
1992 btrfs_cleanup_transaction(root); 1991 btrfs_cleanup_transaction(fs_info);
1993 set_current_state(TASK_INTERRUPTIBLE); 1992 set_current_state(TASK_INTERRUPTIBLE);
1994 if (!kthread_should_stop() && 1993 if (!kthread_should_stop() &&
1995 (!btrfs_transaction_blocked(root->fs_info) || 1994 (!btrfs_transaction_blocked(fs_info) ||
1996 cannot_commit)) 1995 cannot_commit))
1997 schedule_timeout(delay); 1996 schedule_timeout(delay);
1998 __set_current_state(TASK_RUNNING); 1997 __set_current_state(TASK_RUNNING);
@@ -2206,11 +2205,9 @@ static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2206 btrfs_destroy_workqueue(fs_info->delalloc_workers); 2205 btrfs_destroy_workqueue(fs_info->delalloc_workers);
2207 btrfs_destroy_workqueue(fs_info->workers); 2206 btrfs_destroy_workqueue(fs_info->workers);
2208 btrfs_destroy_workqueue(fs_info->endio_workers); 2207 btrfs_destroy_workqueue(fs_info->endio_workers);
2209 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2210 btrfs_destroy_workqueue(fs_info->endio_raid56_workers); 2208 btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
2211 btrfs_destroy_workqueue(fs_info->endio_repair_workers); 2209 btrfs_destroy_workqueue(fs_info->endio_repair_workers);
2212 btrfs_destroy_workqueue(fs_info->rmw_workers); 2210 btrfs_destroy_workqueue(fs_info->rmw_workers);
2213 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2214 btrfs_destroy_workqueue(fs_info->endio_write_workers); 2211 btrfs_destroy_workqueue(fs_info->endio_write_workers);
2215 btrfs_destroy_workqueue(fs_info->endio_freespace_worker); 2212 btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
2216 btrfs_destroy_workqueue(fs_info->submit_workers); 2213 btrfs_destroy_workqueue(fs_info->submit_workers);
@@ -2220,6 +2217,13 @@ static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2220 btrfs_destroy_workqueue(fs_info->flush_workers); 2217 btrfs_destroy_workqueue(fs_info->flush_workers);
2221 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); 2218 btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
2222 btrfs_destroy_workqueue(fs_info->extent_workers); 2219 btrfs_destroy_workqueue(fs_info->extent_workers);
2220 /*
2221 * Now that all other work queues are destroyed, we can safely destroy
2222 * the queues used for metadata I/O, since tasks from those other work
2223 * queues can do metadata I/O operations.
2224 */
2225 btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2226 btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2223} 2227}
2224 2228
2225static void free_root_extent_buffers(struct btrfs_root *root) 2229static void free_root_extent_buffers(struct btrfs_root *root)
@@ -2279,8 +2283,7 @@ void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
2279 2283
2280 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 2284 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2281 btrfs_free_log_root_tree(NULL, fs_info); 2285 btrfs_free_log_root_tree(NULL, fs_info);
2282 btrfs_destroy_pinned_extent(fs_info->tree_root, 2286 btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents);
2283 fs_info->pinned_extents);
2284 } 2287 }
2285} 2288}
2286 2289
@@ -2306,33 +2309,31 @@ static void btrfs_init_balance(struct btrfs_fs_info *fs_info)
2306 init_waitqueue_head(&fs_info->balance_wait_q); 2309 init_waitqueue_head(&fs_info->balance_wait_q);
2307} 2310}
2308 2311
2309static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info, 2312static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info)
2310 struct btrfs_root *tree_root)
2311{ 2313{
2312 fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; 2314 struct inode *inode = fs_info->btree_inode;
2313 set_nlink(fs_info->btree_inode, 1); 2315
2316 inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2317 set_nlink(inode, 1);
2314 /* 2318 /*
2315 * we set the i_size on the btree inode to the max possible int. 2319 * we set the i_size on the btree inode to the max possible int.
2316 * the real end of the address space is determined by all of 2320 * the real end of the address space is determined by all of
2317 * the devices in the system 2321 * the devices in the system
2318 */ 2322 */
2319 fs_info->btree_inode->i_size = OFFSET_MAX; 2323 inode->i_size = OFFSET_MAX;
2320 fs_info->btree_inode->i_mapping->a_ops = &btree_aops; 2324 inode->i_mapping->a_ops = &btree_aops;
2321 2325
2322 RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node); 2326 RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
2323 extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, 2327 extent_io_tree_init(&BTRFS_I(inode)->io_tree, inode->i_mapping);
2324 fs_info->btree_inode->i_mapping); 2328 BTRFS_I(inode)->io_tree.track_uptodate = 0;
2325 BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0; 2329 extent_map_tree_init(&BTRFS_I(inode)->extent_tree);
2326 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2327 2330
2328 BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; 2331 BTRFS_I(inode)->io_tree.ops = &btree_extent_io_ops;
2329 2332
2330 BTRFS_I(fs_info->btree_inode)->root = tree_root; 2333 BTRFS_I(inode)->root = fs_info->tree_root;
2331 memset(&BTRFS_I(fs_info->btree_inode)->location, 0, 2334 memset(&BTRFS_I(inode)->location, 0, sizeof(struct btrfs_key));
2332 sizeof(struct btrfs_key)); 2335 set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags);
2333 set_bit(BTRFS_INODE_DUMMY, 2336 btrfs_insert_inode_hash(inode);
2334 &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2335 btrfs_insert_inode_hash(fs_info->btree_inode);
2336} 2337}
2337 2338
2338static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) 2339static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info)
@@ -2453,7 +2454,6 @@ static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2453 struct btrfs_fs_devices *fs_devices) 2454 struct btrfs_fs_devices *fs_devices)
2454{ 2455{
2455 int ret; 2456 int ret;
2456 struct btrfs_root *tree_root = fs_info->tree_root;
2457 struct btrfs_root *log_tree_root; 2457 struct btrfs_root *log_tree_root;
2458 struct btrfs_super_block *disk_super = fs_info->super_copy; 2458 struct btrfs_super_block *disk_super = fs_info->super_copy;
2459 u64 bytenr = btrfs_super_log_root(disk_super); 2459 u64 bytenr = btrfs_super_log_root(disk_super);
@@ -2467,12 +2467,10 @@ static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2467 if (!log_tree_root) 2467 if (!log_tree_root)
2468 return -ENOMEM; 2468 return -ENOMEM;
2469 2469
2470 __setup_root(tree_root->nodesize, tree_root->sectorsize, 2470 __setup_root(log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2471 tree_root->stripesize, log_tree_root, fs_info,
2472 BTRFS_TREE_LOG_OBJECTID);
2473 2471
2474 log_tree_root->node = read_tree_block(tree_root, bytenr, 2472 log_tree_root->node = read_tree_block(fs_info, bytenr,
2475 fs_info->generation + 1); 2473 fs_info->generation + 1);
2476 if (IS_ERR(log_tree_root->node)) { 2474 if (IS_ERR(log_tree_root->node)) {
2477 btrfs_warn(fs_info, "failed to read log tree"); 2475 btrfs_warn(fs_info, "failed to read log tree");
2478 ret = PTR_ERR(log_tree_root->node); 2476 ret = PTR_ERR(log_tree_root->node);
@@ -2487,15 +2485,15 @@ static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2487 /* returns with log_tree_root freed on success */ 2485 /* returns with log_tree_root freed on success */
2488 ret = btrfs_recover_log_trees(log_tree_root); 2486 ret = btrfs_recover_log_trees(log_tree_root);
2489 if (ret) { 2487 if (ret) {
2490 btrfs_handle_fs_error(tree_root->fs_info, ret, 2488 btrfs_handle_fs_error(fs_info, ret,
2491 "Failed to recover log tree"); 2489 "Failed to recover log tree");
2492 free_extent_buffer(log_tree_root->node); 2490 free_extent_buffer(log_tree_root->node);
2493 kfree(log_tree_root); 2491 kfree(log_tree_root);
2494 return ret; 2492 return ret;
2495 } 2493 }
2496 2494
2497 if (fs_info->sb->s_flags & MS_RDONLY) { 2495 if (fs_info->sb->s_flags & MS_RDONLY) {
2498 ret = btrfs_commit_super(tree_root); 2496 ret = btrfs_commit_super(fs_info);
2499 if (ret) 2497 if (ret)
2500 return ret; 2498 return ret;
2501 } 2499 }
@@ -2503,13 +2501,15 @@ static int btrfs_replay_log(struct btrfs_fs_info *fs_info,
2503 return 0; 2501 return 0;
2504} 2502}
2505 2503
2506static int btrfs_read_roots(struct btrfs_fs_info *fs_info, 2504static int btrfs_read_roots(struct btrfs_fs_info *fs_info)
2507 struct btrfs_root *tree_root)
2508{ 2505{
2506 struct btrfs_root *tree_root = fs_info->tree_root;
2509 struct btrfs_root *root; 2507 struct btrfs_root *root;
2510 struct btrfs_key location; 2508 struct btrfs_key location;
2511 int ret; 2509 int ret;
2512 2510
2511 BUG_ON(!fs_info->tree_root);
2512
2513 location.objectid = BTRFS_EXTENT_TREE_OBJECTID; 2513 location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2514 location.type = BTRFS_ROOT_ITEM_KEY; 2514 location.type = BTRFS_ROOT_ITEM_KEY;
2515 location.offset = 0; 2515 location.offset = 0;
@@ -2720,7 +2720,7 @@ int open_ctree(struct super_block *sb,
2720 sb->s_blocksize_bits = blksize_bits(4096); 2720 sb->s_blocksize_bits = blksize_bits(4096);
2721 sb->s_bdi = &fs_info->bdi; 2721 sb->s_bdi = &fs_info->bdi;
2722 2722
2723 btrfs_init_btree_inode(fs_info, tree_root); 2723 btrfs_init_btree_inode(fs_info);
2724 2724
2725 spin_lock_init(&fs_info->block_group_cache_lock); 2725 spin_lock_init(&fs_info->block_group_cache_lock);
2726 fs_info->block_group_cache_tree = RB_ROOT; 2726 fs_info->block_group_cache_tree = RB_ROOT;
@@ -2758,14 +2758,18 @@ int open_ctree(struct super_block *sb,
2758 2758
2759 INIT_LIST_HEAD(&fs_info->pinned_chunks); 2759 INIT_LIST_HEAD(&fs_info->pinned_chunks);
2760 2760
2761 /* Usable values until the real ones are cached from the superblock */
2762 fs_info->nodesize = 4096;
2763 fs_info->sectorsize = 4096;
2764 fs_info->stripesize = 4096;
2765
2761 ret = btrfs_alloc_stripe_hash_table(fs_info); 2766 ret = btrfs_alloc_stripe_hash_table(fs_info);
2762 if (ret) { 2767 if (ret) {
2763 err = ret; 2768 err = ret;
2764 goto fail_alloc; 2769 goto fail_alloc;
2765 } 2770 }
2766 2771
2767 __setup_root(4096, 4096, 4096, tree_root, 2772 __setup_root(tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
2768 fs_info, BTRFS_ROOT_TREE_OBJECTID);
2769 2773
2770 invalidate_bdev(fs_devices->latest_bdev); 2774 invalidate_bdev(fs_devices->latest_bdev);
2771 2775
@@ -2801,7 +2805,7 @@ int open_ctree(struct super_block *sb,
2801 2805
2802 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE); 2806 memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
2803 2807
2804 ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY); 2808 ret = btrfs_check_super_valid(fs_info);
2805 if (ret) { 2809 if (ret) {
2806 btrfs_err(fs_info, "superblock contains fatal errors"); 2810 btrfs_err(fs_info, "superblock contains fatal errors");
2807 err = -EINVAL; 2811 err = -EINVAL;
@@ -2829,7 +2833,7 @@ int open_ctree(struct super_block *sb,
2829 */ 2833 */
2830 fs_info->compress_type = BTRFS_COMPRESS_ZLIB; 2834 fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2831 2835
2832 ret = btrfs_parse_options(tree_root, options, sb->s_flags); 2836 ret = btrfs_parse_options(fs_info, options, sb->s_flags);
2833 if (ret) { 2837 if (ret) {
2834 err = ret; 2838 err = ret;
2835 goto fail_alloc; 2839 goto fail_alloc;
@@ -2847,7 +2851,7 @@ int open_ctree(struct super_block *sb,
2847 2851
2848 features = btrfs_super_incompat_flags(disk_super); 2852 features = btrfs_super_incompat_flags(disk_super);
2849 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; 2853 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
2850 if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO) 2854 if (fs_info->compress_type == BTRFS_COMPRESS_LZO)
2851 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; 2855 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
2852 2856
2853 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA) 2857 if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
@@ -2870,6 +2874,11 @@ int open_ctree(struct super_block *sb,
2870 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); 2874 fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids));
2871 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); 2875 fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
2872 2876
2877 /* Cache block sizes */
2878 fs_info->nodesize = nodesize;
2879 fs_info->sectorsize = sectorsize;
2880 fs_info->stripesize = stripesize;
2881
2873 /* 2882 /*
2874 * mixed block groups end up with duplicate but slightly offset 2883 * mixed block groups end up with duplicate but slightly offset
2875 * extent buffers for the same range. It leads to corruptions 2884 * extent buffers for the same range. It leads to corruptions
@@ -2910,15 +2919,11 @@ int open_ctree(struct super_block *sb,
2910 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, 2919 fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2911 SZ_4M / PAGE_SIZE); 2920 SZ_4M / PAGE_SIZE);
2912 2921
2913 tree_root->nodesize = nodesize;
2914 tree_root->sectorsize = sectorsize;
2915 tree_root->stripesize = stripesize;
2916
2917 sb->s_blocksize = sectorsize; 2922 sb->s_blocksize = sectorsize;
2918 sb->s_blocksize_bits = blksize_bits(sectorsize); 2923 sb->s_blocksize_bits = blksize_bits(sectorsize);
2919 2924
2920 mutex_lock(&fs_info->chunk_mutex); 2925 mutex_lock(&fs_info->chunk_mutex);
2921 ret = btrfs_read_sys_array(tree_root); 2926 ret = btrfs_read_sys_array(fs_info);
2922 mutex_unlock(&fs_info->chunk_mutex); 2927 mutex_unlock(&fs_info->chunk_mutex);
2923 if (ret) { 2928 if (ret) {
2924 btrfs_err(fs_info, "failed to read the system array: %d", ret); 2929 btrfs_err(fs_info, "failed to read the system array: %d", ret);
@@ -2927,10 +2932,9 @@ int open_ctree(struct super_block *sb,
2927 2932
2928 generation = btrfs_super_chunk_root_generation(disk_super); 2933 generation = btrfs_super_chunk_root_generation(disk_super);
2929 2934
2930 __setup_root(nodesize, sectorsize, stripesize, chunk_root, 2935 __setup_root(chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2931 fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2932 2936
2933 chunk_root->node = read_tree_block(chunk_root, 2937 chunk_root->node = read_tree_block(fs_info,
2934 btrfs_super_chunk_root(disk_super), 2938 btrfs_super_chunk_root(disk_super),
2935 generation); 2939 generation);
2936 if (IS_ERR(chunk_root->node) || 2940 if (IS_ERR(chunk_root->node) ||
@@ -2947,7 +2951,7 @@ int open_ctree(struct super_block *sb,
2947 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, 2951 read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
2948 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE); 2952 btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
2949 2953
2950 ret = btrfs_read_chunk_tree(chunk_root); 2954 ret = btrfs_read_chunk_tree(fs_info);
2951 if (ret) { 2955 if (ret) {
2952 btrfs_err(fs_info, "failed to read chunk tree: %d", ret); 2956 btrfs_err(fs_info, "failed to read chunk tree: %d", ret);
2953 goto fail_tree_roots; 2957 goto fail_tree_roots;
@@ -2967,7 +2971,7 @@ int open_ctree(struct super_block *sb,
2967retry_root_backup: 2971retry_root_backup:
2968 generation = btrfs_super_generation(disk_super); 2972 generation = btrfs_super_generation(disk_super);
2969 2973
2970 tree_root->node = read_tree_block(tree_root, 2974 tree_root->node = read_tree_block(fs_info,
2971 btrfs_super_root(disk_super), 2975 btrfs_super_root(disk_super),
2972 generation); 2976 generation);
2973 if (IS_ERR(tree_root->node) || 2977 if (IS_ERR(tree_root->node) ||
@@ -2995,7 +2999,7 @@ retry_root_backup:
2995 2999
2996 mutex_unlock(&tree_root->objectid_mutex); 3000 mutex_unlock(&tree_root->objectid_mutex);
2997 3001
2998 ret = btrfs_read_roots(fs_info, tree_root); 3002 ret = btrfs_read_roots(fs_info);
2999 if (ret) 3003 if (ret)
3000 goto recovery_tree_root; 3004 goto recovery_tree_root;
3001 3005
@@ -3048,7 +3052,7 @@ retry_root_backup:
3048 goto fail_sysfs; 3052 goto fail_sysfs;
3049 } 3053 }
3050 3054
3051 ret = btrfs_read_block_groups(fs_info->extent_root); 3055 ret = btrfs_read_block_groups(fs_info);
3052 if (ret) { 3056 if (ret) {
3053 btrfs_err(fs_info, "failed to read block groups: %d", ret); 3057 btrfs_err(fs_info, "failed to read block groups: %d", ret);
3054 goto fail_sysfs; 3058 goto fail_sysfs;
@@ -3076,8 +3080,8 @@ retry_root_backup:
3076 if (IS_ERR(fs_info->transaction_kthread)) 3080 if (IS_ERR(fs_info->transaction_kthread))
3077 goto fail_cleaner; 3081 goto fail_cleaner;
3078 3082
3079 if (!btrfs_test_opt(tree_root->fs_info, SSD) && 3083 if (!btrfs_test_opt(fs_info, SSD) &&
3080 !btrfs_test_opt(tree_root->fs_info, NOSSD) && 3084 !btrfs_test_opt(fs_info, NOSSD) &&
3081 !fs_info->fs_devices->rotating) { 3085 !fs_info->fs_devices->rotating) {
3082 btrfs_info(fs_info, "detected SSD devices, enabling SSD mode"); 3086 btrfs_info(fs_info, "detected SSD devices, enabling SSD mode");
3083 btrfs_set_opt(fs_info->mount_opt, SSD); 3087 btrfs_set_opt(fs_info->mount_opt, SSD);
@@ -3090,9 +3094,9 @@ retry_root_backup:
3090 btrfs_apply_pending_changes(fs_info); 3094 btrfs_apply_pending_changes(fs_info);
3091 3095
3092#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3096#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3093 if (btrfs_test_opt(tree_root->fs_info, CHECK_INTEGRITY)) { 3097 if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) {
3094 ret = btrfsic_mount(tree_root, fs_devices, 3098 ret = btrfsic_mount(fs_info, fs_devices,
3095 btrfs_test_opt(tree_root->fs_info, 3099 btrfs_test_opt(fs_info,
3096 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ? 3100 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
3097 1 : 0, 3101 1 : 0,
3098 fs_info->check_integrity_print_mask); 3102 fs_info->check_integrity_print_mask);
@@ -3108,7 +3112,7 @@ retry_root_backup:
3108 3112
3109 /* do not make disk changes in broken FS or nologreplay is given */ 3113 /* do not make disk changes in broken FS or nologreplay is given */
3110 if (btrfs_super_log_root(disk_super) != 0 && 3114 if (btrfs_super_log_root(disk_super) != 0 &&
3111 !btrfs_test_opt(tree_root->fs_info, NOLOGREPLAY)) { 3115 !btrfs_test_opt(fs_info, NOLOGREPLAY)) {
3112 ret = btrfs_replay_log(fs_info, fs_devices); 3116 ret = btrfs_replay_log(fs_info, fs_devices);
3113 if (ret) { 3117 if (ret) {
3114 err = ret; 3118 err = ret;
@@ -3116,7 +3120,7 @@ retry_root_backup:
3116 } 3120 }
3117 } 3121 }
3118 3122
3119 ret = btrfs_find_orphan_roots(tree_root); 3123 ret = btrfs_find_orphan_roots(fs_info);
3120 if (ret) 3124 if (ret)
3121 goto fail_qgroup; 3125 goto fail_qgroup;
3122 3126
@@ -3164,19 +3168,19 @@ retry_root_backup:
3164 if (ret) { 3168 if (ret) {
3165 btrfs_warn(fs_info, 3169 btrfs_warn(fs_info,
3166 "failed to clear free space tree: %d", ret); 3170 "failed to clear free space tree: %d", ret);
3167 close_ctree(tree_root); 3171 close_ctree(fs_info);
3168 return ret; 3172 return ret;
3169 } 3173 }
3170 } 3174 }
3171 3175
3172 if (btrfs_test_opt(tree_root->fs_info, FREE_SPACE_TREE) && 3176 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) &&
3173 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 3177 !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
3174 btrfs_info(fs_info, "creating free space tree"); 3178 btrfs_info(fs_info, "creating free space tree");
3175 ret = btrfs_create_free_space_tree(fs_info); 3179 ret = btrfs_create_free_space_tree(fs_info);
3176 if (ret) { 3180 if (ret) {
3177 btrfs_warn(fs_info, 3181 btrfs_warn(fs_info,
3178 "failed to create free space tree: %d", ret); 3182 "failed to create free space tree: %d", ret);
3179 close_ctree(tree_root); 3183 close_ctree(fs_info);
3180 return ret; 3184 return ret;
3181 } 3185 }
3182 } 3186 }
@@ -3185,7 +3189,7 @@ retry_root_backup:
3185 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || 3189 if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
3186 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { 3190 (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
3187 up_read(&fs_info->cleanup_work_sem); 3191 up_read(&fs_info->cleanup_work_sem);
3188 close_ctree(tree_root); 3192 close_ctree(fs_info);
3189 return ret; 3193 return ret;
3190 } 3194 }
3191 up_read(&fs_info->cleanup_work_sem); 3195 up_read(&fs_info->cleanup_work_sem);
@@ -3193,14 +3197,14 @@ retry_root_backup:
3193 ret = btrfs_resume_balance_async(fs_info); 3197 ret = btrfs_resume_balance_async(fs_info);
3194 if (ret) { 3198 if (ret) {
3195 btrfs_warn(fs_info, "failed to resume balance: %d", ret); 3199 btrfs_warn(fs_info, "failed to resume balance: %d", ret);
3196 close_ctree(tree_root); 3200 close_ctree(fs_info);
3197 return ret; 3201 return ret;
3198 } 3202 }
3199 3203
3200 ret = btrfs_resume_dev_replace_async(fs_info); 3204 ret = btrfs_resume_dev_replace_async(fs_info);
3201 if (ret) { 3205 if (ret) {
3202 btrfs_warn(fs_info, "failed to resume device replace: %d", ret); 3206 btrfs_warn(fs_info, "failed to resume device replace: %d", ret);
3203 close_ctree(tree_root); 3207 close_ctree(fs_info);
3204 return ret; 3208 return ret;
3205 } 3209 }
3206 3210
@@ -3212,10 +3216,10 @@ retry_root_backup:
3212 if (ret) { 3216 if (ret) {
3213 btrfs_warn(fs_info, 3217 btrfs_warn(fs_info,
3214 "failed to create the UUID tree: %d", ret); 3218 "failed to create the UUID tree: %d", ret);
3215 close_ctree(tree_root); 3219 close_ctree(fs_info);
3216 return ret; 3220 return ret;
3217 } 3221 }
3218 } else if (btrfs_test_opt(tree_root->fs_info, RESCAN_UUID_TREE) || 3222 } else if (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) ||
3219 fs_info->generation != 3223 fs_info->generation !=
3220 btrfs_super_uuid_tree_generation(disk_super)) { 3224 btrfs_super_uuid_tree_generation(disk_super)) {
3221 btrfs_info(fs_info, "checking UUID tree"); 3225 btrfs_info(fs_info, "checking UUID tree");
@@ -3223,7 +3227,7 @@ retry_root_backup:
3223 if (ret) { 3227 if (ret) {
3224 btrfs_warn(fs_info, 3228 btrfs_warn(fs_info,
3225 "failed to check the UUID tree: %d", ret); 3229 "failed to check the UUID tree: %d", ret);
3226 close_ctree(tree_root); 3230 close_ctree(fs_info);
3227 return ret; 3231 return ret;
3228 } 3232 }
3229 } else { 3233 } else {
@@ -3243,7 +3247,7 @@ fail_qgroup:
3243 btrfs_free_qgroup_config(fs_info); 3247 btrfs_free_qgroup_config(fs_info);
3244fail_trans_kthread: 3248fail_trans_kthread:
3245 kthread_stop(fs_info->transaction_kthread); 3249 kthread_stop(fs_info->transaction_kthread);
3246 btrfs_cleanup_transaction(fs_info->tree_root); 3250 btrfs_cleanup_transaction(fs_info);
3247 btrfs_free_fs_roots(fs_info); 3251 btrfs_free_fs_roots(fs_info);
3248fail_cleaner: 3252fail_cleaner:
3249 kthread_stop(fs_info->cleaner_kthread); 3253 kthread_stop(fs_info->cleaner_kthread);
@@ -3262,7 +3266,6 @@ fail_fsdev_sysfs:
3262 3266
3263fail_block_groups: 3267fail_block_groups:
3264 btrfs_put_block_group_cache(fs_info); 3268 btrfs_put_block_group_cache(fs_info);
3265 btrfs_free_block_groups(fs_info);
3266 3269
3267fail_tree_roots: 3270fail_tree_roots:
3268 free_root_pointers(fs_info, 1); 3271 free_root_pointers(fs_info, 1);
@@ -3270,6 +3273,7 @@ fail_tree_roots:
3270 3273
3271fail_sb_buffer: 3274fail_sb_buffer:
3272 btrfs_stop_all_workers(fs_info); 3275 btrfs_stop_all_workers(fs_info);
3276 btrfs_free_block_groups(fs_info);
3273fail_alloc: 3277fail_alloc:
3274fail_iput: 3278fail_iput:
3275 btrfs_mapping_tree_free(&fs_info->mapping_tree); 3279 btrfs_mapping_tree_free(&fs_info->mapping_tree);
@@ -3291,7 +3295,7 @@ fail:
3291 return err; 3295 return err;
3292 3296
3293recovery_tree_root: 3297recovery_tree_root:
3294 if (!btrfs_test_opt(tree_root->fs_info, USEBACKUPROOT)) 3298 if (!btrfs_test_opt(fs_info, USEBACKUPROOT))
3295 goto fail_tree_roots; 3299 goto fail_tree_roots;
3296 3300
3297 free_root_pointers(fs_info, 0); 3301 free_root_pointers(fs_info, 0);
@@ -3317,7 +3321,7 @@ static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3317 struct btrfs_device *device = (struct btrfs_device *) 3321 struct btrfs_device *device = (struct btrfs_device *)
3318 bh->b_private; 3322 bh->b_private;
3319 3323
3320 btrfs_warn_rl_in_rcu(device->dev_root->fs_info, 3324 btrfs_warn_rl_in_rcu(device->fs_info,
3321 "lost page write due to IO error on %s", 3325 "lost page write due to IO error on %s",
3322 rcu_str_deref(device->name)); 3326 rcu_str_deref(device->name));
3323 /* note, we don't set_buffer_write_io_error because we have 3327 /* note, we don't set_buffer_write_io_error because we have
@@ -3410,7 +3414,7 @@ struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3410 */ 3414 */
3411static int write_dev_supers(struct btrfs_device *device, 3415static int write_dev_supers(struct btrfs_device *device,
3412 struct btrfs_super_block *sb, 3416 struct btrfs_super_block *sb,
3413 int do_barriers, int wait, int max_mirrors) 3417 int wait, int max_mirrors)
3414{ 3418{
3415 struct buffer_head *bh; 3419 struct buffer_head *bh;
3416 int i; 3420 int i;
@@ -3449,7 +3453,7 @@ static int write_dev_supers(struct btrfs_device *device,
3449 btrfs_set_super_bytenr(sb, bytenr); 3453 btrfs_set_super_bytenr(sb, bytenr);
3450 3454
3451 crc = ~(u32)0; 3455 crc = ~(u32)0;
3452 crc = btrfs_csum_data((char *)sb + 3456 crc = btrfs_csum_data((const char *)sb +
3453 BTRFS_CSUM_SIZE, crc, 3457 BTRFS_CSUM_SIZE, crc,
3454 BTRFS_SUPER_INFO_SIZE - 3458 BTRFS_SUPER_INFO_SIZE -
3455 BTRFS_CSUM_SIZE); 3459 BTRFS_CSUM_SIZE);
@@ -3462,7 +3466,7 @@ static int write_dev_supers(struct btrfs_device *device,
3462 bh = __getblk(device->bdev, bytenr / 4096, 3466 bh = __getblk(device->bdev, bytenr / 4096,
3463 BTRFS_SUPER_INFO_SIZE); 3467 BTRFS_SUPER_INFO_SIZE);
3464 if (!bh) { 3468 if (!bh) {
3465 btrfs_err(device->dev_root->fs_info, 3469 btrfs_err(device->fs_info,
3466 "couldn't get super buffer head for bytenr %llu", 3470 "couldn't get super buffer head for bytenr %llu",
3467 bytenr); 3471 bytenr);
3468 errors++; 3472 errors++;
@@ -3485,9 +3489,9 @@ static int write_dev_supers(struct btrfs_device *device,
3485 * to go down lazy. 3489 * to go down lazy.
3486 */ 3490 */
3487 if (i == 0) 3491 if (i == 0)
3488 ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_FUA, bh); 3492 ret = btrfsic_submit_bh(REQ_OP_WRITE, REQ_FUA, bh);
3489 else 3493 else
3490 ret = btrfsic_submit_bh(REQ_OP_WRITE, WRITE_SYNC, bh); 3494 ret = btrfsic_submit_bh(REQ_OP_WRITE, REQ_SYNC, bh);
3491 if (ret) 3495 if (ret)
3492 errors++; 3496 errors++;
3493 } 3497 }
@@ -3551,7 +3555,7 @@ static int write_dev_flush(struct btrfs_device *device, int wait)
3551 3555
3552 bio->bi_end_io = btrfs_end_empty_barrier; 3556 bio->bi_end_io = btrfs_end_empty_barrier;
3553 bio->bi_bdev = device->bdev; 3557 bio->bi_bdev = device->bdev;
3554 bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_FLUSH); 3558 bio->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
3555 init_completion(&device->flush_wait); 3559 init_completion(&device->flush_wait);
3556 bio->bi_private = &device->flush_wait; 3560 bio->bi_private = &device->flush_wait;
3557 device->flush_bio = bio; 3561 device->flush_bio = bio;
@@ -3695,7 +3699,7 @@ int btrfs_calc_num_tolerated_disk_barrier_failures(
3695 return num_tolerated_disk_barrier_failures; 3699 return num_tolerated_disk_barrier_failures;
3696} 3700}
3697 3701
3698static int write_all_supers(struct btrfs_root *root, int max_mirrors) 3702int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors)
3699{ 3703{
3700 struct list_head *head; 3704 struct list_head *head;
3701 struct btrfs_device *dev; 3705 struct btrfs_device *dev;
@@ -3707,23 +3711,23 @@ static int write_all_supers(struct btrfs_root *root, int max_mirrors)
3707 int total_errors = 0; 3711 int total_errors = 0;
3708 u64 flags; 3712 u64 flags;
3709 3713
3710 do_barriers = !btrfs_test_opt(root->fs_info, NOBARRIER); 3714 do_barriers = !btrfs_test_opt(fs_info, NOBARRIER);
3711 backup_super_roots(root->fs_info); 3715 backup_super_roots(fs_info);
3712 3716
3713 sb = root->fs_info->super_for_commit; 3717 sb = fs_info->super_for_commit;
3714 dev_item = &sb->dev_item; 3718 dev_item = &sb->dev_item;
3715 3719
3716 mutex_lock(&root->fs_info->fs_devices->device_list_mutex); 3720 mutex_lock(&fs_info->fs_devices->device_list_mutex);
3717 head = &root->fs_info->fs_devices->devices; 3721 head = &fs_info->fs_devices->devices;
3718 max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1; 3722 max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1;
3719 3723
3720 if (do_barriers) { 3724 if (do_barriers) {
3721 ret = barrier_all_devices(root->fs_info); 3725 ret = barrier_all_devices(fs_info);
3722 if (ret) { 3726 if (ret) {
3723 mutex_unlock( 3727 mutex_unlock(
3724 &root->fs_info->fs_devices->device_list_mutex); 3728 &fs_info->fs_devices->device_list_mutex);
3725 btrfs_handle_fs_error(root->fs_info, ret, 3729 btrfs_handle_fs_error(fs_info, ret,
3726 "errors while submitting device barriers."); 3730 "errors while submitting device barriers.");
3727 return ret; 3731 return ret;
3728 } 3732 }
3729 } 3733 }
@@ -3752,18 +3756,19 @@ static int write_all_supers(struct btrfs_root *root, int max_mirrors)
3752 flags = btrfs_super_flags(sb); 3756 flags = btrfs_super_flags(sb);
3753 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); 3757 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3754 3758
3755 ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors); 3759 ret = write_dev_supers(dev, sb, 0, max_mirrors);
3756 if (ret) 3760 if (ret)
3757 total_errors++; 3761 total_errors++;
3758 } 3762 }
3759 if (total_errors > max_errors) { 3763 if (total_errors > max_errors) {
3760 btrfs_err(root->fs_info, "%d errors while writing supers", 3764 btrfs_err(fs_info, "%d errors while writing supers",
3761 total_errors); 3765 total_errors);
3762 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3766 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
3763 3767
3764 /* FUA is masked off if unsupported and can't be the reason */ 3768 /* FUA is masked off if unsupported and can't be the reason */
3765 btrfs_handle_fs_error(root->fs_info, -EIO, 3769 btrfs_handle_fs_error(fs_info, -EIO,
3766 "%d errors while writing supers", total_errors); 3770 "%d errors while writing supers",
3771 total_errors);
3767 return -EIO; 3772 return -EIO;
3768 } 3773 }
3769 3774
@@ -3774,25 +3779,20 @@ static int write_all_supers(struct btrfs_root *root, int max_mirrors)
3774 if (!dev->in_fs_metadata || !dev->writeable) 3779 if (!dev->in_fs_metadata || !dev->writeable)
3775 continue; 3780 continue;
3776 3781
3777 ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors); 3782 ret = write_dev_supers(dev, sb, 1, max_mirrors);
3778 if (ret) 3783 if (ret)
3779 total_errors++; 3784 total_errors++;
3780 } 3785 }
3781 mutex_unlock(&root->fs_info->fs_devices->device_list_mutex); 3786 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
3782 if (total_errors > max_errors) { 3787 if (total_errors > max_errors) {
3783 btrfs_handle_fs_error(root->fs_info, -EIO, 3788 btrfs_handle_fs_error(fs_info, -EIO,
3784 "%d errors while writing supers", total_errors); 3789 "%d errors while writing supers",
3790 total_errors);
3785 return -EIO; 3791 return -EIO;
3786 } 3792 }
3787 return 0; 3793 return 0;
3788} 3794}
3789 3795
3790int write_ctree_super(struct btrfs_trans_handle *trans,
3791 struct btrfs_root *root, int max_mirrors)
3792{
3793 return write_all_supers(root, max_mirrors);
3794}
3795
3796/* Drop a fs root from the radix tree and free it. */ 3796/* Drop a fs root from the radix tree and free it. */
3797void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, 3797void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3798 struct btrfs_root *root) 3798 struct btrfs_root *root)
@@ -3826,7 +3826,7 @@ static void free_fs_root(struct btrfs_root *root)
3826{ 3826{
3827 iput(root->ino_cache_inode); 3827 iput(root->ino_cache_inode);
3828 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); 3828 WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
3829 btrfs_free_block_rsv(root, root->orphan_block_rsv); 3829 btrfs_free_block_rsv(root->fs_info, root->orphan_block_rsv);
3830 root->orphan_block_rsv = NULL; 3830 root->orphan_block_rsv = NULL;
3831 if (root->anon_dev) 3831 if (root->anon_dev)
3832 free_anon_bdev(root->anon_dev); 3832 free_anon_bdev(root->anon_dev);
@@ -3896,28 +3896,29 @@ int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3896 return err; 3896 return err;
3897} 3897}
3898 3898
3899int btrfs_commit_super(struct btrfs_root *root) 3899int btrfs_commit_super(struct btrfs_fs_info *fs_info)
3900{ 3900{
3901 struct btrfs_root *root = fs_info->tree_root;
3901 struct btrfs_trans_handle *trans; 3902 struct btrfs_trans_handle *trans;
3902 3903
3903 mutex_lock(&root->fs_info->cleaner_mutex); 3904 mutex_lock(&fs_info->cleaner_mutex);
3904 btrfs_run_delayed_iputs(root); 3905 btrfs_run_delayed_iputs(fs_info);
3905 mutex_unlock(&root->fs_info->cleaner_mutex); 3906 mutex_unlock(&fs_info->cleaner_mutex);
3906 wake_up_process(root->fs_info->cleaner_kthread); 3907 wake_up_process(fs_info->cleaner_kthread);
3907 3908
3908 /* wait until ongoing cleanup work done */ 3909 /* wait until ongoing cleanup work done */
3909 down_write(&root->fs_info->cleanup_work_sem); 3910 down_write(&fs_info->cleanup_work_sem);
3910 up_write(&root->fs_info->cleanup_work_sem); 3911 up_write(&fs_info->cleanup_work_sem);
3911 3912
3912 trans = btrfs_join_transaction(root); 3913 trans = btrfs_join_transaction(root);
3913 if (IS_ERR(trans)) 3914 if (IS_ERR(trans))
3914 return PTR_ERR(trans); 3915 return PTR_ERR(trans);
3915 return btrfs_commit_transaction(trans, root); 3916 return btrfs_commit_transaction(trans);
3916} 3917}
3917 3918
3918void close_ctree(struct btrfs_root *root) 3919void close_ctree(struct btrfs_fs_info *fs_info)
3919{ 3920{
3920 struct btrfs_fs_info *fs_info = root->fs_info; 3921 struct btrfs_root *root = fs_info->tree_root;
3921 int ret; 3922 int ret;
3922 3923
3923 set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); 3924 set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags);
@@ -3952,15 +3953,15 @@ void close_ctree(struct btrfs_root *root)
3952 * block groups queued for removal, the deletion will be 3953 * block groups queued for removal, the deletion will be
3953 * skipped when we quit the cleaner thread. 3954 * skipped when we quit the cleaner thread.
3954 */ 3955 */
3955 btrfs_delete_unused_bgs(root->fs_info); 3956 btrfs_delete_unused_bgs(fs_info);
3956 3957
3957 ret = btrfs_commit_super(root); 3958 ret = btrfs_commit_super(fs_info);
3958 if (ret) 3959 if (ret)
3959 btrfs_err(fs_info, "commit super ret %d", ret); 3960 btrfs_err(fs_info, "commit super ret %d", ret);
3960 } 3961 }
3961 3962
3962 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) 3963 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3963 btrfs_error_commit_super(root); 3964 btrfs_error_commit_super(fs_info);
3964 3965
3965 kthread_stop(fs_info->transaction_kthread); 3966 kthread_stop(fs_info->transaction_kthread);
3966 kthread_stop(fs_info->cleaner_kthread); 3967 kthread_stop(fs_info->cleaner_kthread);
@@ -3981,8 +3982,6 @@ void close_ctree(struct btrfs_root *root)
3981 3982
3982 btrfs_put_block_group_cache(fs_info); 3983 btrfs_put_block_group_cache(fs_info);
3983 3984
3984 btrfs_free_block_groups(fs_info);
3985
3986 /* 3985 /*
3987 * we must make sure there is not any read request to 3986 * we must make sure there is not any read request to
3988 * submit after we stopping all workers. 3987 * submit after we stopping all workers.
@@ -3990,14 +3989,16 @@ void close_ctree(struct btrfs_root *root)
3990 invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 3989 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
3991 btrfs_stop_all_workers(fs_info); 3990 btrfs_stop_all_workers(fs_info);
3992 3991
3992 btrfs_free_block_groups(fs_info);
3993
3993 clear_bit(BTRFS_FS_OPEN, &fs_info->flags); 3994 clear_bit(BTRFS_FS_OPEN, &fs_info->flags);
3994 free_root_pointers(fs_info, 1); 3995 free_root_pointers(fs_info, 1);
3995 3996
3996 iput(fs_info->btree_inode); 3997 iput(fs_info->btree_inode);
3997 3998
3998#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 3999#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3999 if (btrfs_test_opt(root->fs_info, CHECK_INTEGRITY)) 4000 if (btrfs_test_opt(fs_info, CHECK_INTEGRITY))
4000 btrfsic_unmount(root, fs_info->fs_devices); 4001 btrfsic_unmount(fs_info->fs_devices);
4001#endif 4002#endif
4002 4003
4003 btrfs_close_devices(fs_info->fs_devices); 4004 btrfs_close_devices(fs_info->fs_devices);
@@ -4014,7 +4015,7 @@ void close_ctree(struct btrfs_root *root)
4014 __btrfs_free_block_rsv(root->orphan_block_rsv); 4015 __btrfs_free_block_rsv(root->orphan_block_rsv);
4015 root->orphan_block_rsv = NULL; 4016 root->orphan_block_rsv = NULL;
4016 4017
4017 lock_chunks(root); 4018 mutex_lock(&fs_info->chunk_mutex);
4018 while (!list_empty(&fs_info->pinned_chunks)) { 4019 while (!list_empty(&fs_info->pinned_chunks)) {
4019 struct extent_map *em; 4020 struct extent_map *em;
4020 4021
@@ -4023,7 +4024,7 @@ void close_ctree(struct btrfs_root *root)
4023 list_del_init(&em->list); 4024 list_del_init(&em->list);
4024 free_extent_map(em); 4025 free_extent_map(em);
4025 } 4026 }
4026 unlock_chunks(root); 4027 mutex_unlock(&fs_info->chunk_mutex);
4027} 4028}
4028 4029
4029int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, 4030int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
@@ -4045,6 +4046,7 @@ int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
4045 4046
4046void btrfs_mark_buffer_dirty(struct extent_buffer *buf) 4047void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
4047{ 4048{
4049 struct btrfs_fs_info *fs_info;
4048 struct btrfs_root *root; 4050 struct btrfs_root *root;
4049 u64 transid = btrfs_header_generation(buf); 4051 u64 transid = btrfs_header_generation(buf);
4050 int was_dirty; 4052 int was_dirty;
@@ -4059,24 +4061,25 @@ void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
4059 return; 4061 return;
4060#endif 4062#endif
4061 root = BTRFS_I(buf->pages[0]->mapping->host)->root; 4063 root = BTRFS_I(buf->pages[0]->mapping->host)->root;
4064 fs_info = root->fs_info;
4062 btrfs_assert_tree_locked(buf); 4065 btrfs_assert_tree_locked(buf);
4063 if (transid != root->fs_info->generation) 4066 if (transid != fs_info->generation)
4064 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", 4067 WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n",
4065 buf->start, transid, root->fs_info->generation); 4068 buf->start, transid, fs_info->generation);
4066 was_dirty = set_extent_buffer_dirty(buf); 4069 was_dirty = set_extent_buffer_dirty(buf);
4067 if (!was_dirty) 4070 if (!was_dirty)
4068 __percpu_counter_add(&root->fs_info->dirty_metadata_bytes, 4071 __percpu_counter_add(&fs_info->dirty_metadata_bytes,
4069 buf->len, 4072 buf->len,
4070 root->fs_info->dirty_metadata_batch); 4073 fs_info->dirty_metadata_batch);
4071#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 4074#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
4072 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) { 4075 if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
4073 btrfs_print_leaf(root, buf); 4076 btrfs_print_leaf(fs_info, buf);
4074 ASSERT(0); 4077 ASSERT(0);
4075 } 4078 }
4076#endif 4079#endif
4077} 4080}
4078 4081
4079static void __btrfs_btree_balance_dirty(struct btrfs_root *root, 4082static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info,
4080 int flush_delayed) 4083 int flush_delayed)
4081{ 4084{
4082 /* 4085 /*
@@ -4089,34 +4092,34 @@ static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
4089 return; 4092 return;
4090 4093
4091 if (flush_delayed) 4094 if (flush_delayed)
4092 btrfs_balance_delayed_items(root); 4095 btrfs_balance_delayed_items(fs_info);
4093 4096
4094 ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes, 4097 ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
4095 BTRFS_DIRTY_METADATA_THRESH); 4098 BTRFS_DIRTY_METADATA_THRESH);
4096 if (ret > 0) { 4099 if (ret > 0) {
4097 balance_dirty_pages_ratelimited( 4100 balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping);
4098 root->fs_info->btree_inode->i_mapping);
4099 } 4101 }
4100} 4102}
4101 4103
4102void btrfs_btree_balance_dirty(struct btrfs_root *root) 4104void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info)
4103{ 4105{
4104 __btrfs_btree_balance_dirty(root, 1); 4106 __btrfs_btree_balance_dirty(fs_info, 1);
4105} 4107}
4106 4108
4107void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root) 4109void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info)
4108{ 4110{
4109 __btrfs_btree_balance_dirty(root, 0); 4111 __btrfs_btree_balance_dirty(fs_info, 0);
4110} 4112}
4111 4113
4112int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) 4114int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
4113{ 4115{
4114 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root; 4116 struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
4115 return btree_read_extent_buffer_pages(root, buf, parent_transid); 4117 struct btrfs_fs_info *fs_info = root->fs_info;
4118
4119 return btree_read_extent_buffer_pages(fs_info, buf, parent_transid);
4116} 4120}
4117 4121
4118static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info, 4122static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info)
4119 int read_only)
4120{ 4123{
4121 struct btrfs_super_block *sb = fs_info->super_copy; 4124 struct btrfs_super_block *sb = fs_info->super_copy;
4122 u64 nodesize = btrfs_super_nodesize(sb); 4125 u64 nodesize = btrfs_super_nodesize(sb);
@@ -4263,17 +4266,17 @@ static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
4263 return ret; 4266 return ret;
4264} 4267}
4265 4268
4266static void btrfs_error_commit_super(struct btrfs_root *root) 4269static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info)
4267{ 4270{
4268 mutex_lock(&root->fs_info->cleaner_mutex); 4271 mutex_lock(&fs_info->cleaner_mutex);
4269 btrfs_run_delayed_iputs(root); 4272 btrfs_run_delayed_iputs(fs_info);
4270 mutex_unlock(&root->fs_info->cleaner_mutex); 4273 mutex_unlock(&fs_info->cleaner_mutex);
4271 4274
4272 down_write(&root->fs_info->cleanup_work_sem); 4275 down_write(&fs_info->cleanup_work_sem);
4273 up_write(&root->fs_info->cleanup_work_sem); 4276 up_write(&fs_info->cleanup_work_sem);
4274 4277
4275 /* cleanup FS via transaction */ 4278 /* cleanup FS via transaction */
4276 btrfs_cleanup_transaction(root); 4279 btrfs_cleanup_transaction(fs_info);
4277} 4280}
4278 4281
4279static void btrfs_destroy_ordered_extents(struct btrfs_root *root) 4282static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
@@ -4316,7 +4319,7 @@ static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
4316} 4319}
4317 4320
4318static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, 4321static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4319 struct btrfs_root *root) 4322 struct btrfs_fs_info *fs_info)
4320{ 4323{
4321 struct rb_node *node; 4324 struct rb_node *node;
4322 struct btrfs_delayed_ref_root *delayed_refs; 4325 struct btrfs_delayed_ref_root *delayed_refs;
@@ -4328,7 +4331,7 @@ static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4328 spin_lock(&delayed_refs->lock); 4331 spin_lock(&delayed_refs->lock);
4329 if (atomic_read(&delayed_refs->num_entries) == 0) { 4332 if (atomic_read(&delayed_refs->num_entries) == 0) {
4330 spin_unlock(&delayed_refs->lock); 4333 spin_unlock(&delayed_refs->lock);
4331 btrfs_info(root->fs_info, "delayed_refs has NO entry"); 4334 btrfs_info(fs_info, "delayed_refs has NO entry");
4332 return ret; 4335 return ret;
4333 } 4336 }
4334 4337
@@ -4354,6 +4357,8 @@ static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4354 list) { 4357 list) {
4355 ref->in_tree = 0; 4358 ref->in_tree = 0;
4356 list_del(&ref->list); 4359 list_del(&ref->list);
4360 if (!list_empty(&ref->add_list))
4361 list_del(&ref->add_list);
4357 atomic_dec(&delayed_refs->num_entries); 4362 atomic_dec(&delayed_refs->num_entries);
4358 btrfs_put_delayed_ref(ref); 4363 btrfs_put_delayed_ref(ref);
4359 } 4364 }
@@ -4371,7 +4376,7 @@ static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
4371 mutex_unlock(&head->mutex); 4376 mutex_unlock(&head->mutex);
4372 4377
4373 if (pin_bytes) 4378 if (pin_bytes)
4374 btrfs_pin_extent(root, head->node.bytenr, 4379 btrfs_pin_extent(fs_info, head->node.bytenr,
4375 head->node.num_bytes, 1); 4380 head->node.num_bytes, 1);
4376 btrfs_put_delayed_ref(&head->node); 4381 btrfs_put_delayed_ref(&head->node);
4377 cond_resched(); 4382 cond_resched();
@@ -4435,7 +4440,7 @@ static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
4435 spin_unlock(&fs_info->delalloc_root_lock); 4440 spin_unlock(&fs_info->delalloc_root_lock);
4436} 4441}
4437 4442
4438static int btrfs_destroy_marked_extents(struct btrfs_root *root, 4443static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info,
4439 struct extent_io_tree *dirty_pages, 4444 struct extent_io_tree *dirty_pages,
4440 int mark) 4445 int mark)
4441{ 4446{
@@ -4452,8 +4457,8 @@ static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4452 4457
4453 clear_extent_bits(dirty_pages, start, end, mark); 4458 clear_extent_bits(dirty_pages, start, end, mark);
4454 while (start <= end) { 4459 while (start <= end) {
4455 eb = btrfs_find_tree_block(root->fs_info, start); 4460 eb = find_extent_buffer(fs_info, start);
4456 start += root->nodesize; 4461 start += fs_info->nodesize;
4457 if (!eb) 4462 if (!eb)
4458 continue; 4463 continue;
4459 wait_on_extent_buffer_writeback(eb); 4464 wait_on_extent_buffer_writeback(eb);
@@ -4468,7 +4473,7 @@ static int btrfs_destroy_marked_extents(struct btrfs_root *root,
4468 return ret; 4473 return ret;
4469} 4474}
4470 4475
4471static int btrfs_destroy_pinned_extent(struct btrfs_root *root, 4476static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info,
4472 struct extent_io_tree *pinned_extents) 4477 struct extent_io_tree *pinned_extents)
4473{ 4478{
4474 struct extent_io_tree *unpin; 4479 struct extent_io_tree *unpin;
@@ -4486,15 +4491,15 @@ again:
4486 break; 4491 break;
4487 4492
4488 clear_extent_dirty(unpin, start, end); 4493 clear_extent_dirty(unpin, start, end);
4489 btrfs_error_unpin_extent_range(root, start, end); 4494 btrfs_error_unpin_extent_range(fs_info, start, end);
4490 cond_resched(); 4495 cond_resched();
4491 } 4496 }
4492 4497
4493 if (loop) { 4498 if (loop) {
4494 if (unpin == &root->fs_info->freed_extents[0]) 4499 if (unpin == &fs_info->freed_extents[0])
4495 unpin = &root->fs_info->freed_extents[1]; 4500 unpin = &fs_info->freed_extents[1];
4496 else 4501 else
4497 unpin = &root->fs_info->freed_extents[0]; 4502 unpin = &fs_info->freed_extents[0];
4498 loop = false; 4503 loop = false;
4499 goto again; 4504 goto again;
4500 } 4505 }
@@ -4517,7 +4522,7 @@ static void btrfs_cleanup_bg_io(struct btrfs_block_group_cache *cache)
4517} 4522}
4518 4523
4519void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, 4524void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans,
4520 struct btrfs_root *root) 4525 struct btrfs_fs_info *fs_info)
4521{ 4526{
4522 struct btrfs_block_group_cache *cache; 4527 struct btrfs_block_group_cache *cache;
4523 4528
@@ -4527,8 +4532,7 @@ void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans,
4527 struct btrfs_block_group_cache, 4532 struct btrfs_block_group_cache,
4528 dirty_list); 4533 dirty_list);
4529 if (!cache) { 4534 if (!cache) {
4530 btrfs_err(root->fs_info, 4535 btrfs_err(fs_info, "orphan block group dirty_bgs list");
4531 "orphan block group dirty_bgs list");
4532 spin_unlock(&cur_trans->dirty_bgs_lock); 4536 spin_unlock(&cur_trans->dirty_bgs_lock);
4533 return; 4537 return;
4534 } 4538 }
@@ -4556,8 +4560,7 @@ void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans,
4556 struct btrfs_block_group_cache, 4560 struct btrfs_block_group_cache,
4557 io_list); 4561 io_list);
4558 if (!cache) { 4562 if (!cache) {
4559 btrfs_err(root->fs_info, 4563 btrfs_err(fs_info, "orphan block group on io_bgs list");
4560 "orphan block group on io_bgs list");
4561 return; 4564 return;
4562 } 4565 }
4563 4566
@@ -4570,27 +4573,27 @@ void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans,
4570} 4573}
4571 4574
4572void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, 4575void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4573 struct btrfs_root *root) 4576 struct btrfs_fs_info *fs_info)
4574{ 4577{
4575 btrfs_cleanup_dirty_bgs(cur_trans, root); 4578 btrfs_cleanup_dirty_bgs(cur_trans, fs_info);
4576 ASSERT(list_empty(&cur_trans->dirty_bgs)); 4579 ASSERT(list_empty(&cur_trans->dirty_bgs));
4577 ASSERT(list_empty(&cur_trans->io_bgs)); 4580 ASSERT(list_empty(&cur_trans->io_bgs));
4578 4581
4579 btrfs_destroy_delayed_refs(cur_trans, root); 4582 btrfs_destroy_delayed_refs(cur_trans, fs_info);
4580 4583
4581 cur_trans->state = TRANS_STATE_COMMIT_START; 4584 cur_trans->state = TRANS_STATE_COMMIT_START;
4582 wake_up(&root->fs_info->transaction_blocked_wait); 4585 wake_up(&fs_info->transaction_blocked_wait);
4583 4586
4584 cur_trans->state = TRANS_STATE_UNBLOCKED; 4587 cur_trans->state = TRANS_STATE_UNBLOCKED;
4585 wake_up(&root->fs_info->transaction_wait); 4588 wake_up(&fs_info->transaction_wait);
4586 4589
4587 btrfs_destroy_delayed_inodes(root); 4590 btrfs_destroy_delayed_inodes(fs_info);
4588 btrfs_assert_delayed_root_empty(root); 4591 btrfs_assert_delayed_root_empty(fs_info);
4589 4592
4590 btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages, 4593 btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages,
4591 EXTENT_DIRTY); 4594 EXTENT_DIRTY);
4592 btrfs_destroy_pinned_extent(root, 4595 btrfs_destroy_pinned_extent(fs_info,
4593 root->fs_info->pinned_extents); 4596 fs_info->pinned_extents);
4594 4597
4595 cur_trans->state =TRANS_STATE_COMPLETED; 4598 cur_trans->state =TRANS_STATE_COMPLETED;
4596 wake_up(&cur_trans->commit_wait); 4599 wake_up(&cur_trans->commit_wait);
@@ -4601,27 +4604,27 @@ void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4601 */ 4604 */
4602} 4605}
4603 4606
4604static int btrfs_cleanup_transaction(struct btrfs_root *root) 4607static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info)
4605{ 4608{
4606 struct btrfs_transaction *t; 4609 struct btrfs_transaction *t;
4607 4610
4608 mutex_lock(&root->fs_info->transaction_kthread_mutex); 4611 mutex_lock(&fs_info->transaction_kthread_mutex);
4609 4612
4610 spin_lock(&root->fs_info->trans_lock); 4613 spin_lock(&fs_info->trans_lock);
4611 while (!list_empty(&root->fs_info->trans_list)) { 4614 while (!list_empty(&fs_info->trans_list)) {
4612 t = list_first_entry(&root->fs_info->trans_list, 4615 t = list_first_entry(&fs_info->trans_list,
4613 struct btrfs_transaction, list); 4616 struct btrfs_transaction, list);
4614 if (t->state >= TRANS_STATE_COMMIT_START) { 4617 if (t->state >= TRANS_STATE_COMMIT_START) {
4615 atomic_inc(&t->use_count); 4618 atomic_inc(&t->use_count);
4616 spin_unlock(&root->fs_info->trans_lock); 4619 spin_unlock(&fs_info->trans_lock);
4617 btrfs_wait_for_commit(root, t->transid); 4620 btrfs_wait_for_commit(fs_info, t->transid);
4618 btrfs_put_transaction(t); 4621 btrfs_put_transaction(t);
4619 spin_lock(&root->fs_info->trans_lock); 4622 spin_lock(&fs_info->trans_lock);
4620 continue; 4623 continue;
4621 } 4624 }
4622 if (t == root->fs_info->running_transaction) { 4625 if (t == fs_info->running_transaction) {
4623 t->state = TRANS_STATE_COMMIT_DOING; 4626 t->state = TRANS_STATE_COMMIT_DOING;
4624 spin_unlock(&root->fs_info->trans_lock); 4627 spin_unlock(&fs_info->trans_lock);
4625 /* 4628 /*
4626 * We wait for 0 num_writers since we don't hold a trans 4629 * We wait for 0 num_writers since we don't hold a trans
4627 * handle open currently for this transaction. 4630 * handle open currently for this transaction.
@@ -4629,35 +4632,38 @@ static int btrfs_cleanup_transaction(struct btrfs_root *root)
4629 wait_event(t->writer_wait, 4632 wait_event(t->writer_wait,
4630 atomic_read(&t->num_writers) == 0); 4633 atomic_read(&t->num_writers) == 0);
4631 } else { 4634 } else {
4632 spin_unlock(&root->fs_info->trans_lock); 4635 spin_unlock(&fs_info->trans_lock);
4633 } 4636 }
4634 btrfs_cleanup_one_transaction(t, root); 4637 btrfs_cleanup_one_transaction(t, fs_info);
4635 4638
4636 spin_lock(&root->fs_info->trans_lock); 4639 spin_lock(&fs_info->trans_lock);
4637 if (t == root->fs_info->running_transaction) 4640 if (t == fs_info->running_transaction)
4638 root->fs_info->running_transaction = NULL; 4641 fs_info->running_transaction = NULL;
4639 list_del_init(&t->list); 4642 list_del_init(&t->list);
4640 spin_unlock(&root->fs_info->trans_lock); 4643 spin_unlock(&fs_info->trans_lock);
4641 4644
4642 btrfs_put_transaction(t); 4645 btrfs_put_transaction(t);
4643 trace_btrfs_transaction_commit(root); 4646 trace_btrfs_transaction_commit(fs_info->tree_root);
4644 spin_lock(&root->fs_info->trans_lock); 4647 spin_lock(&fs_info->trans_lock);
4645 } 4648 }
4646 spin_unlock(&root->fs_info->trans_lock); 4649 spin_unlock(&fs_info->trans_lock);
4647 btrfs_destroy_all_ordered_extents(root->fs_info); 4650 btrfs_destroy_all_ordered_extents(fs_info);
4648 btrfs_destroy_delayed_inodes(root); 4651 btrfs_destroy_delayed_inodes(fs_info);
4649 btrfs_assert_delayed_root_empty(root); 4652 btrfs_assert_delayed_root_empty(fs_info);
4650 btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents); 4653 btrfs_destroy_pinned_extent(fs_info, fs_info->pinned_extents);
4651 btrfs_destroy_all_delalloc_inodes(root->fs_info); 4654 btrfs_destroy_all_delalloc_inodes(fs_info);
4652 mutex_unlock(&root->fs_info->transaction_kthread_mutex); 4655 mutex_unlock(&fs_info->transaction_kthread_mutex);
4653 4656
4654 return 0; 4657 return 0;
4655} 4658}
4656 4659
4657static const struct extent_io_ops btree_extent_io_ops = { 4660static const struct extent_io_ops btree_extent_io_ops = {
4658 .readpage_end_io_hook = btree_readpage_end_io_hook, 4661 /* mandatory callbacks */
4659 .readpage_io_failed_hook = btree_io_failed_hook,
4660 .submit_bio_hook = btree_submit_bio_hook, 4662 .submit_bio_hook = btree_submit_bio_hook,
4663 .readpage_end_io_hook = btree_readpage_end_io_hook,
4661 /* note we're sharing with inode.c for the merge bio hook */ 4664 /* note we're sharing with inode.c for the merge bio hook */
4662 .merge_bio_hook = btrfs_merge_bio_hook, 4665 .merge_bio_hook = btrfs_merge_bio_hook,
4666 .readpage_io_failed_hook = btree_io_failed_hook,
4667
4668 /* optional callbacks */
4663}; 4669};
generated by cgit v1.2.2 (git 2.25.0) at 2025-10-13 22:05:47 -0400