aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--fs/btrfs/ctree.c155
-rw-r--r--fs/btrfs/ctree.h26
-rw-r--r--fs/btrfs/disk-io.c9
-rw-r--r--fs/btrfs/extent-tree.c2034
-rw-r--r--fs/btrfs/root-tree.c5
-rw-r--r--fs/btrfs/transaction.c15
-rw-r--r--fs/btrfs/volumes.c9
7 files changed, 1828 insertions, 425 deletions
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c
index f9cd40967d04..50e81f43e6d4 100644
--- a/fs/btrfs/ctree.c
+++ b/fs/btrfs/ctree.c
@@ -179,7 +179,6 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans,
179 struct extent_buffer *cow; 179 struct extent_buffer *cow;
180 u32 nritems; 180 u32 nritems;
181 int ret = 0; 181 int ret = 0;
182 int different_trans = 0;
183 int level; 182 int level;
184 int unlock_orig = 0; 183 int unlock_orig = 0;
185 184
@@ -233,13 +232,33 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans,
233 WARN_ON(btrfs_header_generation(buf) > trans->transid); 232 WARN_ON(btrfs_header_generation(buf) > trans->transid);
234 if (btrfs_header_generation(buf) != trans->transid) { 233 if (btrfs_header_generation(buf) != trans->transid) {
235 u32 nr_extents; 234 u32 nr_extents;
236 different_trans = 1;
237 ret = btrfs_inc_ref(trans, root, buf, cow, &nr_extents); 235 ret = btrfs_inc_ref(trans, root, buf, cow, &nr_extents);
238 if (ret) 236 if (ret)
239 return ret; 237 return ret;
240 238
241 ret = btrfs_cache_ref(trans, root, buf, nr_extents); 239 ret = btrfs_cache_ref(trans, root, buf, nr_extents);
242 WARN_ON(ret); 240 WARN_ON(ret);
241 } else if (btrfs_header_owner(buf) == BTRFS_TREE_RELOC_OBJECTID) {
242 /*
243 * There are only two places that can drop reference to
244 * tree blocks owned by living reloc trees, one is here,
245 * the other place is btrfs_merge_path. In both places,
246 * we check reference count while tree block is locked.
247 * Furthermore, if reference count is one, it won't get
248 * increased by someone else.
249 */
250 u32 refs;
251 ret = btrfs_lookup_extent_ref(trans, root, buf->start,
252 buf->len, &refs);
253 BUG_ON(ret);
254 if (refs == 1) {
255 ret = btrfs_update_ref(trans, root, buf, cow,
256 0, nritems);
257 clean_tree_block(trans, root, buf);
258 } else {
259 ret = btrfs_inc_ref(trans, root, buf, cow, NULL);
260 }
261 BUG_ON(ret);
243 } else { 262 } else {
244 ret = btrfs_update_ref(trans, root, buf, cow, 0, nritems); 263 ret = btrfs_update_ref(trans, root, buf, cow, 0, nritems);
245 if (ret) 264 if (ret)
@@ -247,6 +266,14 @@ int noinline __btrfs_cow_block(struct btrfs_trans_handle *trans,
247 clean_tree_block(trans, root, buf); 266 clean_tree_block(trans, root, buf);
248 } 267 }
249 268
269 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
270 ret = btrfs_add_reloc_mapping(root, buf->start,
271 buf->len, cow->start);
272 BUG_ON(ret);
273 ret = btrfs_reloc_tree_cache_ref(trans, root, cow, buf->start);
274 WARN_ON(ret);
275 }
276
250 if (buf == root->node) { 277 if (buf == root->node) {
251 WARN_ON(parent && parent != buf); 278 WARN_ON(parent && parent != buf);
252 279
@@ -1466,6 +1493,130 @@ done:
1466 return ret; 1493 return ret;
1467} 1494}
1468 1495
1496int btrfs_merge_path(struct btrfs_trans_handle *trans,
1497 struct btrfs_root *root,
1498 struct btrfs_key *node_keys,
1499 u64 *nodes, int lowest_level)
1500{
1501 struct extent_buffer *eb;
1502 struct extent_buffer *parent;
1503 struct btrfs_key key;
1504 u64 bytenr;
1505 u64 generation;
1506 u32 blocksize;
1507 int level;
1508 int slot;
1509 int key_match;
1510 int ret;
1511
1512 eb = btrfs_lock_root_node(root);
1513 ret = btrfs_cow_block(trans, root, eb, NULL, 0, &eb, 0);
1514 BUG_ON(ret);
1515
1516 parent = eb;
1517 while (1) {
1518 level = btrfs_header_level(parent);
1519 if (level == 0 || level <= lowest_level)
1520 break;
1521
1522 ret = bin_search(parent, &node_keys[lowest_level], level,
1523 &slot);
1524 if (ret && slot > 0)
1525 slot--;
1526
1527 bytenr = btrfs_node_blockptr(parent, slot);
1528 if (nodes[level - 1] == bytenr)
1529 break;
1530
1531 blocksize = btrfs_level_size(root, level - 1);
1532 generation = btrfs_node_ptr_generation(parent, slot);
1533 btrfs_node_key_to_cpu(eb, &key, slot);
1534 key_match = !memcmp(&key, &node_keys[level - 1], sizeof(key));
1535
1536 /*
1537 * if node keys match and node pointer hasn't been modified
1538 * in the running transaction, we can merge the path. for
1539 * reloc trees, the node pointer check is skipped, this is
1540 * because the reloc trees are fully controlled by the space
1541 * balance code, no one else can modify them.
1542 */
1543 if (!nodes[level - 1] || !key_match ||
1544 (generation == trans->transid &&
1545 root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID)) {
1546next_level:
1547 if (level == 1 || level == lowest_level + 1)
1548 break;
1549
1550 eb = read_tree_block(root, bytenr, blocksize,
1551 generation);
1552 btrfs_tree_lock(eb);
1553
1554 ret = btrfs_cow_block(trans, root, eb, parent, slot,
1555 &eb, 0);
1556 BUG_ON(ret);
1557
1558 btrfs_tree_unlock(parent);
1559 free_extent_buffer(parent);
1560 parent = eb;
1561 continue;
1562 }
1563
1564 if (generation == trans->transid) {
1565 u32 refs;
1566 BUG_ON(btrfs_header_owner(eb) !=
1567 BTRFS_TREE_RELOC_OBJECTID);
1568 /*
1569 * lock the block to keep __btrfs_cow_block from
1570 * changing the reference count.
1571 */
1572 eb = read_tree_block(root, bytenr, blocksize,
1573 generation);
1574 btrfs_tree_lock(eb);
1575
1576 ret = btrfs_lookup_extent_ref(trans, root, bytenr,
1577 blocksize, &refs);
1578 BUG_ON(ret);
1579 /*
1580 * if replace block whose reference count is one,
1581 * we have to "drop the subtree". so skip it for
1582 * simplicity
1583 */
1584 if (refs == 1) {
1585 btrfs_tree_unlock(eb);
1586 free_extent_buffer(eb);
1587 goto next_level;
1588 }
1589 }
1590
1591 btrfs_set_node_blockptr(parent, slot, nodes[level - 1]);
1592 btrfs_set_node_ptr_generation(parent, slot, trans->transid);
1593 btrfs_mark_buffer_dirty(parent);
1594
1595 ret = btrfs_inc_extent_ref(trans, root,
1596 nodes[level - 1],
1597 blocksize, parent->start,
1598 btrfs_header_owner(parent),
1599 btrfs_header_generation(parent),
1600 level - 1, 0);
1601 BUG_ON(ret);
1602 ret = btrfs_free_extent(trans, root, bytenr,
1603 blocksize, parent->start,
1604 btrfs_header_owner(parent),
1605 btrfs_header_generation(parent),
1606 level - 1, 0, 1);
1607 BUG_ON(ret);
1608
1609 if (generation == trans->transid) {
1610 btrfs_tree_unlock(eb);
1611 free_extent_buffer(eb);
1612 }
1613 break;
1614 }
1615 btrfs_tree_unlock(parent);
1616 free_extent_buffer(parent);
1617 return 0;
1618}
1619
1469/* 1620/*
1470 * adjust the pointers going up the tree, starting at level 1621 * adjust the pointers going up the tree, starting at level
1471 * making sure the right key of each node is points to 'key'. 1622 * making sure the right key of each node is points to 'key'.
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 3e62a1b0a1f7..2775e270881e 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -604,6 +604,7 @@ struct btrfs_fs_info {
604 struct mutex chunk_mutex; 604 struct mutex chunk_mutex;
605 struct mutex drop_mutex; 605 struct mutex drop_mutex;
606 struct mutex volume_mutex; 606 struct mutex volume_mutex;
607 struct mutex tree_reloc_mutex;
607 struct list_head trans_list; 608 struct list_head trans_list;
608 struct list_head hashers; 609 struct list_head hashers;
609 struct list_head dead_roots; 610 struct list_head dead_roots;
@@ -647,6 +648,10 @@ struct btrfs_fs_info {
647 struct task_struct *cleaner_kthread; 648 struct task_struct *cleaner_kthread;
648 int thread_pool_size; 649 int thread_pool_size;
649 650
651 /* tree relocation relocated fields */
652 struct extent_io_tree reloc_mapping_tree;
653 struct list_head dead_reloc_roots;
654 struct btrfs_leaf_ref_tree reloc_ref_tree;
650 struct btrfs_leaf_ref_tree shared_ref_tree; 655 struct btrfs_leaf_ref_tree shared_ref_tree;
651 656
652 struct kobject super_kobj; 657 struct kobject super_kobj;
@@ -698,6 +703,7 @@ struct btrfs_root {
698 struct btrfs_leaf_ref_tree ref_tree_struct; 703 struct btrfs_leaf_ref_tree ref_tree_struct;
699 struct btrfs_dirty_root *dirty_root; 704 struct btrfs_dirty_root *dirty_root;
700 struct btrfs_root *log_root; 705 struct btrfs_root *log_root;
706 struct btrfs_root *reloc_root;
701 707
702 struct btrfs_root_item root_item; 708 struct btrfs_root_item root_item;
703 struct btrfs_key root_key; 709 struct btrfs_key root_key;
@@ -1517,7 +1523,6 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
1517struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans, 1523struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
1518 struct btrfs_root *root, 1524 struct btrfs_root *root,
1519 u64 bytenr, u32 blocksize); 1525 u64 bytenr, u32 blocksize);
1520int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size);
1521int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, 1526int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans,
1522 struct btrfs_root *root, 1527 struct btrfs_root *root,
1523 struct btrfs_path *path, 1528 struct btrfs_path *path,
@@ -1582,10 +1587,29 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
1582 struct btrfs_root *root, u64 bytes_used, 1587 struct btrfs_root *root, u64 bytes_used,
1583 u64 type, u64 chunk_objectid, u64 chunk_offset, 1588 u64 type, u64 chunk_objectid, u64 chunk_offset,
1584 u64 size); 1589 u64 size);
1590int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
1591 struct btrfs_root *root, u64 group_start);
1592int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start);
1593int btrfs_free_reloc_root(struct btrfs_root *root);
1594int btrfs_drop_dead_reloc_roots(struct btrfs_root *root);
1595int btrfs_add_reloc_mapping(struct btrfs_root *root, u64 orig_bytenr,
1596 u64 num_bytes, u64 new_bytenr);
1597int btrfs_get_reloc_mapping(struct btrfs_root *root, u64 orig_bytenr,
1598 u64 num_bytes, u64 *new_bytenr);
1599void btrfs_free_reloc_mappings(struct btrfs_root *root);
1600int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
1601 struct btrfs_root *root,
1602 struct extent_buffer *buf, u64 orig_start);
1603int btrfs_add_dead_reloc_root(struct btrfs_root *root);
1604int btrfs_cleanup_reloc_trees(struct btrfs_root *root);
1585/* ctree.c */ 1605/* ctree.c */
1586int btrfs_previous_item(struct btrfs_root *root, 1606int btrfs_previous_item(struct btrfs_root *root,
1587 struct btrfs_path *path, u64 min_objectid, 1607 struct btrfs_path *path, u64 min_objectid,
1588 int type); 1608 int type);
1609int btrfs_merge_path(struct btrfs_trans_handle *trans,
1610 struct btrfs_root *root,
1611 struct btrfs_key *node_keys,
1612 u64 *nodes, int lowest_level);
1589int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans, 1613int btrfs_set_item_key_safe(struct btrfs_trans_handle *trans,
1590 struct btrfs_root *root, struct btrfs_path *path, 1614 struct btrfs_root *root, struct btrfs_path *path,
1591 struct btrfs_key *new_key); 1615 struct btrfs_key *new_key);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 8969fee23318..45bc3132b054 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1406,6 +1406,10 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1406 fs_info->btree_inode->i_mapping, GFP_NOFS); 1406 fs_info->btree_inode->i_mapping, GFP_NOFS);
1407 fs_info->do_barriers = 1; 1407 fs_info->do_barriers = 1;
1408 1408
1409 extent_io_tree_init(&fs_info->reloc_mapping_tree,
1410 fs_info->btree_inode->i_mapping, GFP_NOFS);
1411 INIT_LIST_HEAD(&fs_info->dead_reloc_roots);
1412 btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree);
1409 btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree); 1413 btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree);
1410 1414
1411 BTRFS_I(fs_info->btree_inode)->root = tree_root; 1415 BTRFS_I(fs_info->btree_inode)->root = tree_root;
@@ -1421,6 +1425,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1421 mutex_init(&fs_info->transaction_kthread_mutex); 1425 mutex_init(&fs_info->transaction_kthread_mutex);
1422 mutex_init(&fs_info->cleaner_mutex); 1426 mutex_init(&fs_info->cleaner_mutex);
1423 mutex_init(&fs_info->volume_mutex); 1427 mutex_init(&fs_info->volume_mutex);
1428 mutex_init(&fs_info->tree_reloc_mutex);
1424 init_waitqueue_head(&fs_info->transaction_throttle); 1429 init_waitqueue_head(&fs_info->transaction_throttle);
1425 init_waitqueue_head(&fs_info->transaction_wait); 1430 init_waitqueue_head(&fs_info->transaction_wait);
1426 init_waitqueue_head(&fs_info->async_submit_wait); 1431 init_waitqueue_head(&fs_info->async_submit_wait);
@@ -1627,6 +1632,10 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1627 ret = btrfs_recover_log_trees(log_tree_root); 1632 ret = btrfs_recover_log_trees(log_tree_root);
1628 BUG_ON(ret); 1633 BUG_ON(ret);
1629 } 1634 }
1635
1636 ret = btrfs_cleanup_reloc_trees(tree_root);
1637 BUG_ON(ret);
1638
1630 fs_info->last_trans_committed = btrfs_super_generation(disk_super); 1639 fs_info->last_trans_committed = btrfs_super_generation(disk_super);
1631 return tree_root; 1640 return tree_root;
1632 1641
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 9ab099bc01a4..8043b9d584a9 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -1834,6 +1834,7 @@ static int pin_down_bytes(struct btrfs_trans_handle *trans,
1834 u64 header_owner = btrfs_header_owner(buf); 1834 u64 header_owner = btrfs_header_owner(buf);
1835 u64 header_transid = btrfs_header_generation(buf); 1835 u64 header_transid = btrfs_header_generation(buf);
1836 if (header_owner != BTRFS_TREE_LOG_OBJECTID && 1836 if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
1837 header_owner != BTRFS_TREE_RELOC_OBJECTID &&
1837 header_transid == trans->transid && 1838 header_transid == trans->transid &&
1838 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) { 1839 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
1839 clean_tree_block(NULL, root, buf); 1840 clean_tree_block(NULL, root, buf);
@@ -2487,6 +2488,7 @@ int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
2487 return -ENOSPC; 2488 return -ENOSPC;
2488 } 2489 }
2489 btrfs_add_free_space(cache, start, len); 2490 btrfs_add_free_space(cache, start, len);
2491 update_reserved_extents(root, start, len, 0);
2490 maybe_unlock_mutex(root); 2492 maybe_unlock_mutex(root);
2491 return 0; 2493 return 0;
2492} 2494}
@@ -2947,6 +2949,10 @@ static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
2947 */ 2949 */
2948 if (*level == 1) { 2950 if (*level == 1) {
2949 ref = btrfs_lookup_leaf_ref(root, bytenr); 2951 ref = btrfs_lookup_leaf_ref(root, bytenr);
2952 if (ref && ref->generation != ptr_gen) {
2953 btrfs_free_leaf_ref(root, ref);
2954 ref = NULL;
2955 }
2950 if (ref) { 2956 if (ref) {
2951 ret = cache_drop_leaf_ref(trans, root, ref); 2957 ret = cache_drop_leaf_ref(trans, root, ref);
2952 BUG_ON(ret); 2958 BUG_ON(ret);
@@ -3153,34 +3159,6 @@ out:
3153 return ret; 3159 return ret;
3154} 3160}
3155 3161
3156int btrfs_free_block_groups(struct btrfs_fs_info *info)
3157{
3158 struct btrfs_block_group_cache *block_group;
3159 struct rb_node *n;
3160
3161 mutex_lock(&info->alloc_mutex);
3162 spin_lock(&info->block_group_cache_lock);
3163 while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
3164 block_group = rb_entry(n, struct btrfs_block_group_cache,
3165 cache_node);
3166
3167 spin_unlock(&info->block_group_cache_lock);
3168 btrfs_remove_free_space_cache(block_group);
3169 spin_lock(&info->block_group_cache_lock);
3170
3171 rb_erase(&block_group->cache_node,
3172 &info->block_group_cache_tree);
3173
3174 spin_lock(&block_group->space_info->lock);
3175 list_del(&block_group->list);
3176 spin_unlock(&block_group->space_info->lock);
3177 kfree(block_group);
3178 }
3179 spin_unlock(&info->block_group_cache_lock);
3180 mutex_unlock(&info->alloc_mutex);
3181 return 0;
3182}
3183
3184static unsigned long calc_ra(unsigned long start, unsigned long last, 3162static unsigned long calc_ra(unsigned long start, unsigned long last,
3185 unsigned long nr) 3163 unsigned long nr)
3186{ 3164{
@@ -3192,37 +3170,43 @@ static int noinline relocate_inode_pages(struct inode *inode, u64 start,
3192{ 3170{
3193 u64 page_start; 3171 u64 page_start;
3194 u64 page_end; 3172 u64 page_end;
3173 unsigned long first_index;
3195 unsigned long last_index; 3174 unsigned long last_index;
3196 unsigned long i; 3175 unsigned long i;
3197 struct page *page; 3176 struct page *page;
3198 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 3177 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3199 struct file_ra_state *ra; 3178 struct file_ra_state *ra;
3200 unsigned long total_read = 0;
3201 unsigned long ra_pages;
3202 struct btrfs_ordered_extent *ordered; 3179 struct btrfs_ordered_extent *ordered;
3203 struct btrfs_trans_handle *trans; 3180 unsigned int total_read = 0;
3181 unsigned int total_dirty = 0;
3182 int ret = 0;
3204 3183
3205 ra = kzalloc(sizeof(*ra), GFP_NOFS); 3184 ra = kzalloc(sizeof(*ra), GFP_NOFS);
3206 3185
3207 mutex_lock(&inode->i_mutex); 3186 mutex_lock(&inode->i_mutex);
3208 i = start >> PAGE_CACHE_SHIFT; 3187 first_index = start >> PAGE_CACHE_SHIFT;
3209 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; 3188 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
3210 3189
3211 ra_pages = BTRFS_I(inode)->root->fs_info->bdi.ra_pages; 3190 /* make sure the dirty trick played by the caller work */
3191 ret = invalidate_inode_pages2_range(inode->i_mapping,
3192 first_index, last_index);
3193 if (ret)
3194 goto out_unlock;
3212 3195
3213 file_ra_state_init(ra, inode->i_mapping); 3196 file_ra_state_init(ra, inode->i_mapping);
3214 3197
3215 for (; i <= last_index; i++) { 3198 for (i = first_index ; i <= last_index; i++) {
3216 if (total_read % ra_pages == 0) { 3199 if (total_read % ra->ra_pages == 0) {
3217 btrfs_force_ra(inode->i_mapping, ra, NULL, i, 3200 btrfs_force_ra(inode->i_mapping, ra, NULL, i,
3218 calc_ra(i, last_index, ra_pages)); 3201 calc_ra(i, last_index, ra->ra_pages));
3219 } 3202 }
3220 total_read++; 3203 total_read++;
3221again: 3204again:
3222 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode)) 3205 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
3223 goto truncate_racing; 3206 BUG_ON(1);
3224 page = grab_cache_page(inode->i_mapping, i); 3207 page = grab_cache_page(inode->i_mapping, i);
3225 if (!page) { 3208 if (!page) {
3209 ret = -ENOMEM;
3226 goto out_unlock; 3210 goto out_unlock;
3227 } 3211 }
3228 if (!PageUptodate(page)) { 3212 if (!PageUptodate(page)) {
@@ -3231,6 +3215,7 @@ again:
3231 if (!PageUptodate(page)) { 3215 if (!PageUptodate(page)) {
3232 unlock_page(page); 3216 unlock_page(page);
3233 page_cache_release(page); 3217 page_cache_release(page);
3218 ret = -EIO;
3234 goto out_unlock; 3219 goto out_unlock;
3235 } 3220 }
3236 } 3221 }
@@ -3251,14 +3236,13 @@ again:
3251 } 3236 }
3252 set_page_extent_mapped(page); 3237 set_page_extent_mapped(page);
3253 3238
3254 /*
3255 * make sure page_mkwrite is called for this page if userland
3256 * wants to change it from mmap
3257 */
3258 clear_page_dirty_for_io(page);
3259
3260 btrfs_set_extent_delalloc(inode, page_start, page_end); 3239 btrfs_set_extent_delalloc(inode, page_start, page_end);
3240 if (i == first_index)
3241 set_extent_bits(io_tree, page_start, page_end,
3242 EXTENT_BOUNDARY, GFP_NOFS);
3243
3261 set_page_dirty(page); 3244 set_page_dirty(page);
3245 total_dirty++;
3262 3246
3263 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 3247 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3264 unlock_page(page); 3248 unlock_page(page);
@@ -3266,347 +3250,1457 @@ again:
3266 } 3250 }
3267 3251
3268out_unlock: 3252out_unlock:
3269 /* we have to start the IO in order to get the ordered extents
3270 * instantiated. This allows the relocation to code to wait
3271 * for all the ordered extents to hit the disk.
3272 *
3273 * Otherwise, it would constantly loop over the same extents
3274 * because the old ones don't get deleted until the IO is
3275 * started
3276 */
3277 btrfs_fdatawrite_range(inode->i_mapping, start, start + len - 1,
3278 WB_SYNC_NONE);
3279 kfree(ra); 3253 kfree(ra);
3280 trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1);
3281 if (trans) {
3282 btrfs_end_transaction(trans, BTRFS_I(inode)->root);
3283 mark_inode_dirty(inode);
3284 }
3285 mutex_unlock(&inode->i_mutex); 3254 mutex_unlock(&inode->i_mutex);
3286 return 0; 3255 balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
3256 return ret;
3257}
3258
3259static int noinline relocate_data_extent(struct inode *reloc_inode,
3260 struct btrfs_key *extent_key,
3261 u64 offset)
3262{
3263 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
3264 struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
3265 struct extent_map *em;
3266
3267 em = alloc_extent_map(GFP_NOFS);
3268 BUG_ON(!em || IS_ERR(em));
3269
3270 em->start = extent_key->objectid - offset;
3271 em->len = extent_key->offset;
3272 em->block_start = extent_key->objectid;
3273 em->bdev = root->fs_info->fs_devices->latest_bdev;
3274 set_bit(EXTENT_FLAG_PINNED, &em->flags);
3275
3276 /* setup extent map to cheat btrfs_readpage */
3277 mutex_lock(&BTRFS_I(reloc_inode)->extent_mutex);
3278 while (1) {
3279 int ret;
3280 spin_lock(&em_tree->lock);
3281 ret = add_extent_mapping(em_tree, em);
3282 spin_unlock(&em_tree->lock);
3283 if (ret != -EEXIST) {
3284 free_extent_map(em);
3285 break;
3286 }
3287 btrfs_drop_extent_cache(reloc_inode, em->start,
3288 em->start + em->len - 1, 0);
3289 }
3290 mutex_unlock(&BTRFS_I(reloc_inode)->extent_mutex);
3287 3291
3288truncate_racing: 3292 return relocate_inode_pages(reloc_inode, extent_key->objectid - offset,
3289 vmtruncate(inode, inode->i_size); 3293 extent_key->offset);
3290 balance_dirty_pages_ratelimited_nr(inode->i_mapping,
3291 total_read);
3292 goto out_unlock;
3293} 3294}
3294 3295
3295/* 3296struct btrfs_ref_path {
3296 * The back references tell us which tree holds a ref on a block, 3297 u64 extent_start;
3297 * but it is possible for the tree root field in the reference to 3298 u64 nodes[BTRFS_MAX_LEVEL];
3298 * reflect the original root before a snapshot was made. In this 3299 u64 root_objectid;
3299 * case we should search through all the children of a given root 3300 u64 root_generation;
3300 * to find potential holders of references on a block. 3301 u64 owner_objectid;
3301 * 3302 u64 owner_offset;
3302 * Instead, we do something a little less fancy and just search 3303 u32 num_refs;
3303 * all the roots for a given key/block combination. 3304 int lowest_level;
3304 */ 3305 int current_level;
3305static int find_root_for_ref(struct btrfs_root *root, 3306};
3306 struct btrfs_path *path,
3307 struct btrfs_key *key0,
3308 int level,
3309 int file_key,
3310 struct btrfs_root **found_root,
3311 u64 bytenr)
3312{
3313 struct btrfs_key root_location;
3314 struct btrfs_root *cur_root = *found_root;
3315 struct btrfs_file_extent_item *file_extent;
3316 u64 root_search_start = BTRFS_FS_TREE_OBJECTID;
3317 u64 found_bytenr;
3318 int ret;
3319 3307
3320 root_location.offset = (u64)-1; 3308struct disk_extent {
3321 root_location.type = BTRFS_ROOT_ITEM_KEY; 3309 u64 disk_bytenr;
3322 path->lowest_level = level; 3310 u64 disk_num_bytes;
3323 path->reada = 0; 3311 u64 offset;
3324 while(1) { 3312 u64 num_bytes;
3325 ret = btrfs_search_slot(NULL, cur_root, key0, path, 0, 0); 3313};
3326 found_bytenr = 0; 3314
3327 if (ret == 0 && file_key) { 3315static int is_cowonly_root(u64 root_objectid)
3328 struct extent_buffer *leaf = path->nodes[0]; 3316{
3329 file_extent = btrfs_item_ptr(leaf, path->slots[0], 3317 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
3330 struct btrfs_file_extent_item); 3318 root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
3331 if (btrfs_file_extent_type(leaf, file_extent) == 3319 root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
3332 BTRFS_FILE_EXTENT_REG) { 3320 root_objectid == BTRFS_DEV_TREE_OBJECTID ||
3333 found_bytenr = 3321 root_objectid == BTRFS_TREE_LOG_OBJECTID)
3334 btrfs_file_extent_disk_bytenr(leaf, 3322 return 1;
3335 file_extent); 3323 return 0;
3336 } 3324}
3337 } else if (!file_key) { 3325
3338 if (path->nodes[level]) 3326static int noinline __next_ref_path(struct btrfs_trans_handle *trans,
3339 found_bytenr = path->nodes[level]->start; 3327 struct btrfs_root *extent_root,
3340 } 3328 struct btrfs_ref_path *ref_path,
3341 3329 int first_time)
3342 btrfs_release_path(cur_root, path); 3330{
3343 3331 struct extent_buffer *leaf;
3344 if (found_bytenr == bytenr) { 3332 struct btrfs_path *path;
3345 *found_root = cur_root; 3333 struct btrfs_extent_ref *ref;
3334 struct btrfs_key key;
3335 struct btrfs_key found_key;
3336 u64 bytenr;
3337 u32 nritems;
3338 int level;
3339 int ret = 1;
3340
3341 path = btrfs_alloc_path();
3342 if (!path)
3343 return -ENOMEM;
3344
3345 mutex_lock(&extent_root->fs_info->alloc_mutex);
3346
3347 if (first_time) {
3348 ref_path->lowest_level = -1;
3349 ref_path->current_level = -1;
3350 goto walk_up;
3351 }
3352walk_down:
3353 level = ref_path->current_level - 1;
3354 while (level >= -1) {
3355 u64 parent;
3356 if (level < ref_path->lowest_level)
3357 break;
3358
3359 if (level >= 0) {
3360 bytenr = ref_path->nodes[level];
3361 } else {
3362 bytenr = ref_path->extent_start;
3363 }
3364 BUG_ON(bytenr == 0);
3365
3366 parent = ref_path->nodes[level + 1];
3367 ref_path->nodes[level + 1] = 0;
3368 ref_path->current_level = level;
3369 BUG_ON(parent == 0);
3370
3371 key.objectid = bytenr;
3372 key.offset = parent + 1;
3373 key.type = BTRFS_EXTENT_REF_KEY;
3374
3375 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
3376 if (ret < 0)
3377 goto out;
3378 BUG_ON(ret == 0);
3379
3380 leaf = path->nodes[0];
3381 nritems = btrfs_header_nritems(leaf);
3382 if (path->slots[0] >= nritems) {
3383 ret = btrfs_next_leaf(extent_root, path);
3384 if (ret < 0)
3385 goto out;
3386 if (ret > 0)
3387 goto next;
3388 leaf = path->nodes[0];
3389 }
3390
3391 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
3392 if (found_key.objectid == bytenr &&
3393 found_key.type == BTRFS_EXTENT_REF_KEY)
3394 goto found;
3395next:
3396 level--;
3397 btrfs_release_path(extent_root, path);
3398 if (need_resched()) {
3399 mutex_unlock(&extent_root->fs_info->alloc_mutex);
3400 cond_resched();
3401 mutex_lock(&extent_root->fs_info->alloc_mutex);
3402 }
3403 }
3404 /* reached lowest level */
3405 ret = 1;
3406 goto out;
3407walk_up:
3408 level = ref_path->current_level;
3409 while (level < BTRFS_MAX_LEVEL - 1) {
3410 u64 ref_objectid;
3411 if (level >= 0) {
3412 bytenr = ref_path->nodes[level];
3413 } else {
3414 bytenr = ref_path->extent_start;
3415 }
3416 BUG_ON(bytenr == 0);
3417
3418 key.objectid = bytenr;
3419 key.offset = 0;
3420 key.type = BTRFS_EXTENT_REF_KEY;
3421
3422 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
3423 if (ret < 0)
3424 goto out;
3425
3426 leaf = path->nodes[0];
3427 nritems = btrfs_header_nritems(leaf);
3428 if (path->slots[0] >= nritems) {
3429 ret = btrfs_next_leaf(extent_root, path);
3430 if (ret < 0)
3431 goto out;
3432 if (ret > 0) {
3433 /* the extent was freed by someone */
3434 if (ref_path->lowest_level == level)
3435 goto out;
3436 btrfs_release_path(extent_root, path);
3437 goto walk_down;
3438 }
3439 leaf = path->nodes[0];
3440 }
3441
3442 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
3443 if (found_key.objectid != bytenr ||
3444 found_key.type != BTRFS_EXTENT_REF_KEY) {
3445 /* the extent was freed by someone */
3446 if (ref_path->lowest_level == level) {
3447 ret = 1;
3448 goto out;
3449 }
3450 btrfs_release_path(extent_root, path);
3451 goto walk_down;
3452 }
3453found:
3454 ref = btrfs_item_ptr(leaf, path->slots[0],
3455 struct btrfs_extent_ref);
3456 ref_objectid = btrfs_ref_objectid(leaf, ref);
3457 if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
3458 if (first_time) {
3459 level = (int)ref_objectid;
3460 BUG_ON(level >= BTRFS_MAX_LEVEL);
3461 ref_path->lowest_level = level;
3462 ref_path->current_level = level;
3463 ref_path->nodes[level] = bytenr;
3464 } else {
3465 WARN_ON(ref_objectid != level);
3466 }
3467 } else {
3468 WARN_ON(level != -1);
3469 }
3470 first_time = 0;
3471
3472 if (ref_path->lowest_level == level) {
3473 ref_path->owner_objectid = ref_objectid;
3474 ref_path->owner_offset = btrfs_ref_offset(leaf, ref);
3475 ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
3476 }
3477
3478 /*
3479 * the block is tree root or the block isn't in reference
3480 * counted tree.
3481 */
3482 if (found_key.objectid == found_key.offset ||
3483 is_cowonly_root(btrfs_ref_root(leaf, ref))) {
3484 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
3485 ref_path->root_generation =
3486 btrfs_ref_generation(leaf, ref);
3487 if (level < 0) {
3488 /* special reference from the tree log */
3489 ref_path->nodes[0] = found_key.offset;
3490 ref_path->current_level = 0;
3491 }
3346 ret = 0; 3492 ret = 0;
3347 goto out; 3493 goto out;
3348 } 3494 }
3349 ret = btrfs_search_root(root->fs_info->tree_root,
3350 root_search_start, &root_search_start);
3351 if (ret)
3352 break;
3353 3495
3354 root_location.objectid = root_search_start; 3496 level++;
3355 cur_root = btrfs_read_fs_root_no_name(root->fs_info, 3497 BUG_ON(ref_path->nodes[level] != 0);
3356 &root_location); 3498 ref_path->nodes[level] = found_key.offset;
3357 if (!cur_root) { 3499 ref_path->current_level = level;
3358 ret = 1; 3500
3359 break; 3501 /*
3502 * the reference was created in the running transaction,
3503 * no need to continue walking up.
3504 */
3505 if (btrfs_ref_generation(leaf, ref) == trans->transid) {
3506 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
3507 ref_path->root_generation =
3508 btrfs_ref_generation(leaf, ref);
3509 ret = 0;
3510 goto out;
3511 }
3512
3513 btrfs_release_path(extent_root, path);
3514 if (need_resched()) {
3515 mutex_unlock(&extent_root->fs_info->alloc_mutex);
3516 cond_resched();
3517 mutex_lock(&extent_root->fs_info->alloc_mutex);
3360 } 3518 }
3361 } 3519 }
3520 /* reached max tree level, but no tree root found. */
3521 BUG();
3362out: 3522out:
3363 path->lowest_level = 0; 3523 mutex_unlock(&extent_root->fs_info->alloc_mutex);
3524 btrfs_free_path(path);
3364 return ret; 3525 return ret;
3365} 3526}
3366 3527
3367/* 3528static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
3368 * note, this releases the path 3529 struct btrfs_root *extent_root,
3369 */ 3530 struct btrfs_ref_path *ref_path,
3370static int noinline relocate_one_reference(struct btrfs_root *extent_root, 3531 u64 extent_start)
3371 struct btrfs_path *path, 3532{
3372 struct btrfs_key *extent_key, 3533 memset(ref_path, 0, sizeof(*ref_path));
3373 u64 *last_file_objectid, 3534 ref_path->extent_start = extent_start;
3374 u64 *last_file_offset, 3535
3375 u64 *last_file_root, 3536 return __next_ref_path(trans, extent_root, ref_path, 1);
3376 u64 last_extent) 3537}
3377{ 3538
3378 struct inode *inode; 3539static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
3379 struct btrfs_root *found_root; 3540 struct btrfs_root *extent_root,
3380 struct btrfs_key root_location; 3541 struct btrfs_ref_path *ref_path)
3542{
3543 return __next_ref_path(trans, extent_root, ref_path, 0);
3544}
3545
3546static int noinline get_new_locations(struct inode *reloc_inode,
3547 struct btrfs_key *extent_key,
3548 u64 offset, int no_fragment,
3549 struct disk_extent **extents,
3550 int *nr_extents)
3551{
3552 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
3553 struct btrfs_path *path;
3554 struct btrfs_file_extent_item *fi;
3555 struct extent_buffer *leaf;
3556 struct disk_extent *exts = *extents;
3381 struct btrfs_key found_key; 3557 struct btrfs_key found_key;
3382 struct btrfs_extent_ref *ref; 3558 u64 cur_pos;
3383 u64 ref_root; 3559 u64 last_byte;
3384 u64 ref_gen; 3560 u32 nritems;
3385 u64 ref_objectid; 3561 int nr = 0;
3386 u64 ref_offset; 3562 int max = *nr_extents;
3387 int ret; 3563 int ret;
3388 int level;
3389 3564
3390 WARN_ON(!mutex_is_locked(&extent_root->fs_info->alloc_mutex)); 3565 WARN_ON(!no_fragment && *extents);
3566 if (!exts) {
3567 max = 1;
3568 exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
3569 if (!exts)
3570 return -ENOMEM;
3571 }
3391 3572
3392 ref = btrfs_item_ptr(path->nodes[0], path->slots[0], 3573 path = btrfs_alloc_path();
3393 struct btrfs_extent_ref); 3574 BUG_ON(!path);
3394 ref_root = btrfs_ref_root(path->nodes[0], ref);
3395 ref_gen = btrfs_ref_generation(path->nodes[0], ref);
3396 ref_objectid = btrfs_ref_objectid(path->nodes[0], ref);
3397 ref_offset = btrfs_ref_offset(path->nodes[0], ref);
3398 btrfs_release_path(extent_root, path);
3399 3575
3400 root_location.objectid = ref_root; 3576 cur_pos = extent_key->objectid - offset;
3401 if (ref_gen == 0) 3577 last_byte = extent_key->objectid + extent_key->offset;
3402 root_location.offset = 0; 3578 ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
3403 else 3579 cur_pos, 0);
3404 root_location.offset = (u64)-1; 3580 if (ret < 0)
3405 root_location.type = BTRFS_ROOT_ITEM_KEY; 3581 goto out;
3582 if (ret > 0) {
3583 ret = -ENOENT;
3584 goto out;
3585 }
3406 3586
3407 found_root = btrfs_read_fs_root_no_name(extent_root->fs_info, 3587 while (1) {
3408 &root_location); 3588 leaf = path->nodes[0];
3409 BUG_ON(!found_root); 3589 nritems = btrfs_header_nritems(leaf);
3410 mutex_unlock(&extent_root->fs_info->alloc_mutex); 3590 if (path->slots[0] >= nritems) {
3591 ret = btrfs_next_leaf(root, path);
3592 if (ret < 0)
3593 goto out;
3594 if (ret > 0)
3595 break;
3596 leaf = path->nodes[0];
3597 }
3411 3598
3412 if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) { 3599 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
3413 found_key.objectid = ref_objectid; 3600 if (found_key.offset != cur_pos ||
3414 found_key.type = BTRFS_EXTENT_DATA_KEY; 3601 found_key.type != BTRFS_EXTENT_DATA_KEY ||
3415 found_key.offset = ref_offset; 3602 found_key.objectid != reloc_inode->i_ino)
3416 level = 0; 3603 break;
3417 3604
3418 if (last_extent == extent_key->objectid && 3605 fi = btrfs_item_ptr(leaf, path->slots[0],
3419 *last_file_objectid == ref_objectid && 3606 struct btrfs_file_extent_item);
3420 *last_file_offset == ref_offset && 3607 if (btrfs_file_extent_type(leaf, fi) !=
3421 *last_file_root == ref_root) 3608 BTRFS_FILE_EXTENT_REG ||
3422 goto out; 3609 btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
3610 break;
3423 3611
3424 ret = find_root_for_ref(extent_root, path, &found_key, 3612 if (nr == max) {
3425 level, 1, &found_root, 3613 struct disk_extent *old = exts;
3426 extent_key->objectid); 3614 max *= 2;
3615 exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
3616 memcpy(exts, old, sizeof(*exts) * nr);
3617 if (old != *extents)
3618 kfree(old);
3619 }
3427 3620
3428 if (ret) 3621 exts[nr].disk_bytenr =
3622 btrfs_file_extent_disk_bytenr(leaf, fi);
3623 exts[nr].disk_num_bytes =
3624 btrfs_file_extent_disk_num_bytes(leaf, fi);
3625 exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
3626 exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
3627 WARN_ON(exts[nr].offset > 0);
3628 WARN_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
3629
3630 cur_pos += exts[nr].num_bytes;
3631 nr++;
3632
3633 if (cur_pos + offset >= last_byte)
3634 break;
3635
3636 if (no_fragment) {
3637 ret = 1;
3429 goto out; 3638 goto out;
3639 }
3640 path->slots[0]++;
3641 }
3430 3642
3431 if (last_extent == extent_key->objectid && 3643 WARN_ON(cur_pos + offset > last_byte);
3432 *last_file_objectid == ref_objectid && 3644 if (cur_pos + offset < last_byte) {
3433 *last_file_offset == ref_offset && 3645 ret = -ENOENT;
3434 *last_file_root == ref_root) 3646 goto out;
3647 }
3648 ret = 0;
3649out:
3650 btrfs_free_path(path);
3651 if (ret) {
3652 if (exts != *extents)
3653 kfree(exts);
3654 } else {
3655 *extents = exts;
3656 *nr_extents = nr;
3657 }
3658 return ret;
3659}
3660
3661static int noinline replace_one_extent(struct btrfs_trans_handle *trans,
3662 struct btrfs_root *root,
3663 struct btrfs_path *path,
3664 struct btrfs_key *extent_key,
3665 struct btrfs_key *leaf_key,
3666 struct btrfs_ref_path *ref_path,
3667 struct disk_extent *new_extents,
3668 int nr_extents)
3669{
3670 struct extent_buffer *leaf;
3671 struct btrfs_file_extent_item *fi;
3672 struct inode *inode = NULL;
3673 struct btrfs_key key;
3674 u64 lock_start = 0;
3675 u64 lock_end = 0;
3676 u64 num_bytes;
3677 u64 ext_offset;
3678 u64 first_pos;
3679 u32 nritems;
3680 int extent_locked = 0;
3681 int ret;
3682
3683 first_pos = ref_path->owner_offset;
3684 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
3685 key.objectid = ref_path->owner_objectid;
3686 key.offset = ref_path->owner_offset;
3687 key.type = BTRFS_EXTENT_DATA_KEY;
3688 } else {
3689 memcpy(&key, leaf_key, sizeof(key));
3690 }
3691
3692 while (1) {
3693 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
3694 if (ret < 0)
3435 goto out; 3695 goto out;
3436 3696
3437 inode = btrfs_iget_locked(extent_root->fs_info->sb, 3697 leaf = path->nodes[0];
3438 ref_objectid, found_root); 3698 nritems = btrfs_header_nritems(leaf);
3439 if (inode->i_state & I_NEW) { 3699next:
3440 /* the inode and parent dir are two different roots */ 3700 if (extent_locked && ret > 0) {
3441 BTRFS_I(inode)->root = found_root; 3701 /*
3442 BTRFS_I(inode)->location.objectid = ref_objectid; 3702 * the file extent item was modified by someone
3443 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; 3703 * before the extent got locked.
3444 BTRFS_I(inode)->location.offset = 0; 3704 */
3445 btrfs_read_locked_inode(inode); 3705 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
3446 unlock_new_inode(inode); 3706 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
3707 lock_end, GFP_NOFS);
3708 extent_locked = 0;
3709 }
3710
3711 if (path->slots[0] >= nritems) {
3712 if (ref_path->owner_objectid ==
3713 BTRFS_MULTIPLE_OBJECTIDS)
3714 break;
3715
3716 BUG_ON(extent_locked);
3717 ret = btrfs_next_leaf(root, path);
3718 if (ret < 0)
3719 goto out;
3720 if (ret > 0)
3721 break;
3722 leaf = path->nodes[0];
3723 nritems = btrfs_header_nritems(leaf);
3724 }
3725
3726 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3447 3727
3728 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
3729 if ((key.objectid > ref_path->owner_objectid) ||
3730 (key.objectid == ref_path->owner_objectid &&
3731 key.type > BTRFS_EXTENT_DATA_KEY) ||
3732 (key.offset >= first_pos + extent_key->offset))
3733 break;
3448 } 3734 }
3449 /* this can happen if the reference is not against
3450 * the latest version of the tree root
3451 */
3452 if (is_bad_inode(inode))
3453 goto out;
3454 3735
3455 *last_file_objectid = inode->i_ino; 3736 if (inode && key.objectid != inode->i_ino) {
3456 *last_file_root = found_root->root_key.objectid; 3737 BUG_ON(extent_locked);
3457 *last_file_offset = ref_offset; 3738 btrfs_release_path(root, path);
3739 mutex_unlock(&inode->i_mutex);
3740 iput(inode);
3741 inode = NULL;
3742 continue;
3743 }
3458 3744
3459 relocate_inode_pages(inode, ref_offset, extent_key->offset); 3745 if (key.type != BTRFS_EXTENT_DATA_KEY) {
3460 iput(inode); 3746 path->slots[0]++;
3461 } else { 3747 ret = 1;
3462 struct btrfs_trans_handle *trans; 3748 goto next;
3463 struct extent_buffer *eb; 3749 }
3464 int needs_lock = 0; 3750 fi = btrfs_item_ptr(leaf, path->slots[0],
3751 struct btrfs_file_extent_item);
3752 if ((btrfs_file_extent_type(leaf, fi) !=
3753 BTRFS_FILE_EXTENT_REG) ||
3754 (btrfs_file_extent_disk_bytenr(leaf, fi) !=
3755 extent_key->objectid)) {
3756 path->slots[0]++;
3757 ret = 1;
3758 goto next;
3759 }
3465 3760
3466 eb = read_tree_block(found_root, extent_key->objectid, 3761 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
3467 extent_key->offset, 0); 3762 ext_offset = btrfs_file_extent_offset(leaf, fi);
3468 btrfs_tree_lock(eb);
3469 level = btrfs_header_level(eb);
3470 3763
3471 if (level == 0) 3764 if (first_pos > key.offset - ext_offset)
3472 btrfs_item_key_to_cpu(eb, &found_key, 0); 3765 first_pos = key.offset - ext_offset;
3473 else
3474 btrfs_node_key_to_cpu(eb, &found_key, 0);
3475 3766
3476 btrfs_tree_unlock(eb); 3767 if (!extent_locked) {
3477 free_extent_buffer(eb); 3768 lock_start = key.offset;
3769 lock_end = lock_start + num_bytes - 1;
3770 } else {
3771 BUG_ON(lock_start != key.offset);
3772 BUG_ON(lock_end - lock_start + 1 < num_bytes);
3773 }
3774
3775 if (!inode) {
3776 btrfs_release_path(root, path);
3777
3778 inode = btrfs_iget_locked(root->fs_info->sb,
3779 key.objectid, root);
3780 if (inode->i_state & I_NEW) {
3781 BTRFS_I(inode)->root = root;
3782 BTRFS_I(inode)->location.objectid =
3783 key.objectid;
3784 BTRFS_I(inode)->location.type =
3785 BTRFS_INODE_ITEM_KEY;
3786 BTRFS_I(inode)->location.offset = 0;
3787 btrfs_read_locked_inode(inode);
3788 unlock_new_inode(inode);
3789 }
3790 /*
3791 * some code call btrfs_commit_transaction while
3792 * holding the i_mutex, so we can't use mutex_lock
3793 * here.
3794 */
3795 if (is_bad_inode(inode) ||
3796 !mutex_trylock(&inode->i_mutex)) {
3797 iput(inode);
3798 inode = NULL;
3799 key.offset = (u64)-1;
3800 goto skip;
3801 }
3802 }
3803
3804 if (!extent_locked) {
3805 struct btrfs_ordered_extent *ordered;
3806
3807 btrfs_release_path(root, path);
3478 3808
3479 ret = find_root_for_ref(extent_root, path, &found_key, 3809 lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
3480 level, 0, &found_root, 3810 lock_end, GFP_NOFS);
3481 extent_key->objectid); 3811 ordered = btrfs_lookup_first_ordered_extent(inode,
3812 lock_end);
3813 if (ordered &&
3814 ordered->file_offset <= lock_end &&
3815 ordered->file_offset + ordered->len > lock_start) {
3816 unlock_extent(&BTRFS_I(inode)->io_tree,
3817 lock_start, lock_end, GFP_NOFS);
3818 btrfs_start_ordered_extent(inode, ordered, 1);
3819 btrfs_put_ordered_extent(ordered);
3820 key.offset += num_bytes;
3821 goto skip;
3822 }
3823 if (ordered)
3824 btrfs_put_ordered_extent(ordered);
3825
3826 mutex_lock(&BTRFS_I(inode)->extent_mutex);
3827 extent_locked = 1;
3828 continue;
3829 }
3482 3830
3831 if (nr_extents == 1) {
3832 /* update extent pointer in place */
3833 btrfs_set_file_extent_generation(leaf, fi,
3834 trans->transid);
3835 btrfs_set_file_extent_disk_bytenr(leaf, fi,
3836 new_extents[0].disk_bytenr);
3837 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
3838 new_extents[0].disk_num_bytes);
3839 ext_offset += new_extents[0].offset;
3840 btrfs_set_file_extent_offset(leaf, fi, ext_offset);
3841 btrfs_mark_buffer_dirty(leaf);
3842
3843 btrfs_drop_extent_cache(inode, key.offset,
3844 key.offset + num_bytes - 1, 0);
3845
3846 ret = btrfs_inc_extent_ref(trans, root,
3847 new_extents[0].disk_bytenr,
3848 new_extents[0].disk_num_bytes,
3849 leaf->start,
3850 root->root_key.objectid,
3851 trans->transid,
3852 key.objectid, key.offset);
3853 BUG_ON(ret);
3854
3855 ret = btrfs_free_extent(trans, root,
3856 extent_key->objectid,
3857 extent_key->offset,
3858 leaf->start,
3859 btrfs_header_owner(leaf),
3860 btrfs_header_generation(leaf),
3861 key.objectid, key.offset, 0);
3862 BUG_ON(ret);
3863
3864 btrfs_release_path(root, path);
3865 key.offset += num_bytes;
3866 } else {
3867 u64 alloc_hint;
3868 u64 extent_len;
3869 int i;
3870 /*
3871 * drop old extent pointer at first, then insert the
3872 * new pointers one bye one
3873 */
3874 btrfs_release_path(root, path);
3875 ret = btrfs_drop_extents(trans, root, inode, key.offset,
3876 key.offset + num_bytes,
3877 key.offset, &alloc_hint);
3878 BUG_ON(ret);
3879
3880 for (i = 0; i < nr_extents; i++) {
3881 if (ext_offset >= new_extents[i].num_bytes) {
3882 ext_offset -= new_extents[i].num_bytes;
3883 continue;
3884 }
3885 extent_len = min(new_extents[i].num_bytes -
3886 ext_offset, num_bytes);
3887
3888 ret = btrfs_insert_empty_item(trans, root,
3889 path, &key,
3890 sizeof(*fi));
3891 BUG_ON(ret);
3892
3893 leaf = path->nodes[0];
3894 fi = btrfs_item_ptr(leaf, path->slots[0],
3895 struct btrfs_file_extent_item);
3896 btrfs_set_file_extent_generation(leaf, fi,
3897 trans->transid);
3898 btrfs_set_file_extent_type(leaf, fi,
3899 BTRFS_FILE_EXTENT_REG);
3900 btrfs_set_file_extent_disk_bytenr(leaf, fi,
3901 new_extents[i].disk_bytenr);
3902 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
3903 new_extents[i].disk_num_bytes);
3904 btrfs_set_file_extent_num_bytes(leaf, fi,
3905 extent_len);
3906 ext_offset += new_extents[i].offset;
3907 btrfs_set_file_extent_offset(leaf, fi,
3908 ext_offset);
3909 btrfs_mark_buffer_dirty(leaf);
3910
3911 btrfs_drop_extent_cache(inode, key.offset,
3912 key.offset + extent_len - 1, 0);
3913
3914 ret = btrfs_inc_extent_ref(trans, root,
3915 new_extents[i].disk_bytenr,
3916 new_extents[i].disk_num_bytes,
3917 leaf->start,
3918 root->root_key.objectid,
3919 trans->transid,
3920 key.objectid, key.offset);
3921 BUG_ON(ret);
3922 btrfs_release_path(root, path);
3923
3924 inode->i_blocks += extent_len >> 9;
3925
3926 ext_offset = 0;
3927 num_bytes -= extent_len;
3928 key.offset += extent_len;
3929
3930 if (num_bytes == 0)
3931 break;
3932 }
3933 BUG_ON(i >= nr_extents);
3934 }
3935
3936 if (extent_locked) {
3937 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
3938 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
3939 lock_end, GFP_NOFS);
3940 extent_locked = 0;
3941 }
3942skip:
3943 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
3944 key.offset >= first_pos + extent_key->offset)
3945 break;
3946
3947 cond_resched();
3948 }
3949 ret = 0;
3950out:
3951 btrfs_release_path(root, path);
3952 if (inode) {
3953 mutex_unlock(&inode->i_mutex);
3954 if (extent_locked) {
3955 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
3956 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
3957 lock_end, GFP_NOFS);
3958 }
3959 iput(inode);
3960 }
3961 return ret;
3962}
3963
3964int btrfs_add_reloc_mapping(struct btrfs_root *root, u64 orig_bytenr,
3965 u64 num_bytes, u64 new_bytenr)
3966{
3967 set_extent_bits(&root->fs_info->reloc_mapping_tree,
3968 orig_bytenr, orig_bytenr + num_bytes - 1,
3969 EXTENT_LOCKED, GFP_NOFS);
3970 set_state_private(&root->fs_info->reloc_mapping_tree,
3971 orig_bytenr, new_bytenr);
3972 return 0;
3973}
3974
3975int btrfs_get_reloc_mapping(struct btrfs_root *root, u64 orig_bytenr,
3976 u64 num_bytes, u64 *new_bytenr)
3977{
3978 u64 bytenr;
3979 u64 cur_bytenr = orig_bytenr;
3980 u64 prev_bytenr = orig_bytenr;
3981 int ret;
3982
3983 while (1) {
3984 ret = get_state_private(&root->fs_info->reloc_mapping_tree,
3985 cur_bytenr, &bytenr);
3483 if (ret) 3986 if (ret)
3484 goto out; 3987 break;
3988 prev_bytenr = cur_bytenr;
3989 cur_bytenr = bytenr;
3990 }
3485 3991
3486 /* 3992 if (orig_bytenr == cur_bytenr)
3487 * right here almost anything could happen to our key, 3993 return -ENOENT;
3488 * but that's ok. The cow below will either relocate it
3489 * or someone else will have relocated it. Either way,
3490 * it is in a different spot than it was before and
3491 * we're happy.
3492 */
3493 3994
3494 trans = btrfs_start_transaction(found_root, 1); 3995 if (prev_bytenr != orig_bytenr) {
3996 set_state_private(&root->fs_info->reloc_mapping_tree,
3997 orig_bytenr, cur_bytenr);
3998 }
3999 *new_bytenr = cur_bytenr;
4000 return 0;
4001}
3495 4002
3496 if (found_root == extent_root->fs_info->extent_root || 4003void btrfs_free_reloc_mappings(struct btrfs_root *root)
3497 found_root == extent_root->fs_info->chunk_root || 4004{
3498 found_root == extent_root->fs_info->dev_root) { 4005 clear_extent_bits(&root->fs_info->reloc_mapping_tree,
3499 needs_lock = 1; 4006 0, (u64)-1, -1, GFP_NOFS);
3500 mutex_lock(&extent_root->fs_info->alloc_mutex); 4007}
4008
4009int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
4010 struct btrfs_root *root,
4011 struct extent_buffer *buf, u64 orig_start)
4012{
4013 int level;
4014 int ret;
4015
4016 BUG_ON(btrfs_header_generation(buf) != trans->transid);
4017 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
4018
4019 level = btrfs_header_level(buf);
4020 if (level == 0) {
4021 struct btrfs_leaf_ref *ref;
4022 struct btrfs_leaf_ref *orig_ref;
4023
4024 orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
4025 if (!orig_ref)
4026 return -ENOENT;
4027
4028 ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
4029 if (!ref) {
4030 btrfs_free_leaf_ref(root, orig_ref);
4031 return -ENOMEM;
3501 } 4032 }
3502 4033
3503 path->lowest_level = level; 4034 ref->nritems = orig_ref->nritems;
3504 path->reada = 2; 4035 memcpy(ref->extents, orig_ref->extents,
3505 ret = btrfs_search_slot(trans, found_root, &found_key, path, 4036 sizeof(ref->extents[0]) * ref->nritems);
3506 0, 1); 4037
4038 btrfs_free_leaf_ref(root, orig_ref);
4039
4040 ref->root_gen = trans->transid;
4041 ref->bytenr = buf->start;
4042 ref->owner = btrfs_header_owner(buf);
4043 ref->generation = btrfs_header_generation(buf);
4044 ret = btrfs_add_leaf_ref(root, ref, 0);
4045 WARN_ON(ret);
4046 btrfs_free_leaf_ref(root, ref);
4047 }
4048 return 0;
4049}
4050
4051static int noinline invalidate_extent_cache(struct btrfs_root *root,
4052 struct extent_buffer *leaf,
4053 struct btrfs_block_group_cache *group,
4054 struct btrfs_root *target_root)
4055{
4056 struct btrfs_key key;
4057 struct inode *inode = NULL;
4058 struct btrfs_file_extent_item *fi;
4059 u64 num_bytes;
4060 u64 skip_objectid = 0;
4061 u32 nritems;
4062 u32 i;
4063
4064 nritems = btrfs_header_nritems(leaf);
4065 for (i = 0; i < nritems; i++) {
4066 btrfs_item_key_to_cpu(leaf, &key, i);
4067 if (key.objectid == skip_objectid ||
4068 key.type != BTRFS_EXTENT_DATA_KEY)
4069 continue;
4070 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
4071 if (btrfs_file_extent_type(leaf, fi) ==
4072 BTRFS_FILE_EXTENT_INLINE)
4073 continue;
4074 if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
4075 continue;
4076 if (!inode || inode->i_ino != key.objectid) {
4077 iput(inode);
4078 inode = btrfs_ilookup(target_root->fs_info->sb,
4079 key.objectid, target_root, 1);
4080 }
4081 if (!inode) {
4082 skip_objectid = key.objectid;
4083 continue;
4084 }
4085 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
4086
4087 lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
4088 key.offset + num_bytes - 1, GFP_NOFS);
4089 mutex_lock(&BTRFS_I(inode)->extent_mutex);
4090 btrfs_drop_extent_cache(inode, key.offset,
4091 key.offset + num_bytes - 1, 1);
4092 mutex_unlock(&BTRFS_I(inode)->extent_mutex);
4093 unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
4094 key.offset + num_bytes - 1, GFP_NOFS);
4095 cond_resched();
4096 }
4097 iput(inode);
4098 return 0;
4099}
4100
4101static int noinline replace_extents_in_leaf(struct btrfs_trans_handle *trans,
4102 struct btrfs_root *root,
4103 struct extent_buffer *leaf,
4104 struct btrfs_block_group_cache *group,
4105 struct inode *reloc_inode)
4106{
4107 struct btrfs_key key;
4108 struct btrfs_key extent_key;
4109 struct btrfs_file_extent_item *fi;
4110 struct btrfs_leaf_ref *ref;
4111 struct disk_extent *new_extent;
4112 u64 bytenr;
4113 u64 num_bytes;
4114 u32 nritems;
4115 u32 i;
4116 int ext_index;
4117 int nr_extent;
4118 int ret;
4119
4120 new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
4121 BUG_ON(!new_extent);
4122
4123 ref = btrfs_lookup_leaf_ref(root, leaf->start);
4124 BUG_ON(!ref);
4125
4126 ext_index = -1;
4127 nritems = btrfs_header_nritems(leaf);
4128 for (i = 0; i < nritems; i++) {
4129 btrfs_item_key_to_cpu(leaf, &key, i);
4130 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
4131 continue;
4132 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
4133 if (btrfs_file_extent_type(leaf, fi) ==
4134 BTRFS_FILE_EXTENT_INLINE)
4135 continue;
4136 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
4137 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
4138 if (bytenr == 0)
4139 continue;
4140
4141 ext_index++;
4142 if (bytenr >= group->key.objectid + group->key.offset ||
4143 bytenr + num_bytes <= group->key.objectid)
4144 continue;
4145
4146 extent_key.objectid = bytenr;
4147 extent_key.offset = num_bytes;
4148 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
4149 nr_extent = 1;
4150 ret = get_new_locations(reloc_inode, &extent_key,
4151 group->key.objectid, 1,
4152 &new_extent, &nr_extent);
4153 if (ret > 0)
4154 continue;
4155 BUG_ON(ret < 0);
4156
4157 BUG_ON(ref->extents[ext_index].bytenr != bytenr);
4158 BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
4159 ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
4160 ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
4161
4162 btrfs_set_file_extent_generation(leaf, fi, trans->transid);
4163 btrfs_set_file_extent_disk_bytenr(leaf, fi,
4164 new_extent->disk_bytenr);
4165 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
4166 new_extent->disk_num_bytes);
4167 new_extent->offset += btrfs_file_extent_offset(leaf, fi);
4168 btrfs_set_file_extent_offset(leaf, fi, new_extent->offset);
4169 btrfs_mark_buffer_dirty(leaf);
4170
4171 ret = btrfs_inc_extent_ref(trans, root,
4172 new_extent->disk_bytenr,
4173 new_extent->disk_num_bytes,
4174 leaf->start,
4175 root->root_key.objectid,
4176 trans->transid,
4177 key.objectid, key.offset);
4178 BUG_ON(ret);
4179 ret = btrfs_free_extent(trans, root,
4180 bytenr, num_bytes, leaf->start,
4181 btrfs_header_owner(leaf),
4182 btrfs_header_generation(leaf),
4183 key.objectid, key.offset, 0);
4184 BUG_ON(ret);
4185 cond_resched();
4186 }
4187 kfree(new_extent);
4188 BUG_ON(ext_index + 1 != ref->nritems);
4189 btrfs_free_leaf_ref(root, ref);
4190 return 0;
4191}
4192
4193int btrfs_free_reloc_root(struct btrfs_root *root)
4194{
4195 struct btrfs_root *reloc_root;
4196
4197 if (root->reloc_root) {
4198 reloc_root = root->reloc_root;
4199 root->reloc_root = NULL;
4200 list_add(&reloc_root->dead_list,
4201 &root->fs_info->dead_reloc_roots);
4202 }
4203 return 0;
4204}
4205
4206int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
4207{
4208 struct btrfs_trans_handle *trans;
4209 struct btrfs_root *reloc_root;
4210 struct btrfs_root *prev_root = NULL;
4211 struct list_head dead_roots;
4212 int ret;
4213 unsigned long nr;
4214
4215 INIT_LIST_HEAD(&dead_roots);
4216 list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
4217
4218 while (!list_empty(&dead_roots)) {
4219 reloc_root = list_entry(dead_roots.prev,
4220 struct btrfs_root, dead_list);
4221 list_del_init(&reloc_root->dead_list);
4222
4223 BUG_ON(reloc_root->commit_root != NULL);
4224 while (1) {
4225 trans = btrfs_join_transaction(root, 1);
4226 BUG_ON(!trans);
4227
4228 mutex_lock(&root->fs_info->drop_mutex);
4229 ret = btrfs_drop_snapshot(trans, reloc_root);
4230 if (ret != -EAGAIN)
4231 break;
4232 mutex_unlock(&root->fs_info->drop_mutex);
4233
4234 nr = trans->blocks_used;
4235 ret = btrfs_end_transaction(trans, root);
4236 BUG_ON(ret);
4237 btrfs_btree_balance_dirty(root, nr);
4238 }
4239
4240 free_extent_buffer(reloc_root->node);
4241
4242 ret = btrfs_del_root(trans, root->fs_info->tree_root,
4243 &reloc_root->root_key);
4244 BUG_ON(ret);
4245 mutex_unlock(&root->fs_info->drop_mutex);
4246
4247 nr = trans->blocks_used;
4248 ret = btrfs_end_transaction(trans, root);
4249 BUG_ON(ret);
4250 btrfs_btree_balance_dirty(root, nr);
4251
4252 kfree(prev_root);
4253 prev_root = reloc_root;
4254 }
4255 if (prev_root) {
4256 btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
4257 kfree(prev_root);
4258 }
4259 return 0;
4260}
4261
4262int btrfs_add_dead_reloc_root(struct btrfs_root *root)
4263{
4264 list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
4265 return 0;
4266}
4267
4268int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
4269{
4270 struct btrfs_root *reloc_root;
4271 struct btrfs_trans_handle *trans;
4272 struct btrfs_key location;
4273 int found;
4274 int ret;
4275
4276 mutex_lock(&root->fs_info->tree_reloc_mutex);
4277 ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
4278 BUG_ON(ret);
4279 found = !list_empty(&root->fs_info->dead_reloc_roots);
4280 mutex_unlock(&root->fs_info->tree_reloc_mutex);
4281
4282 if (found) {
4283 trans = btrfs_start_transaction(root, 1);
4284 BUG_ON(!trans);
4285 ret = btrfs_commit_transaction(trans, root);
4286 BUG_ON(ret);
4287 }
4288
4289 location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
4290 location.offset = (u64)-1;
4291 location.type = BTRFS_ROOT_ITEM_KEY;
4292
4293 reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
4294 BUG_ON(!reloc_root);
4295 btrfs_orphan_cleanup(reloc_root);
4296 return 0;
4297}
4298
4299static int noinline init_reloc_tree(struct btrfs_trans_handle *trans,
4300 struct btrfs_root *root)
4301{
4302 struct btrfs_root *reloc_root;
4303 struct extent_buffer *eb;
4304 struct btrfs_root_item *root_item;
4305 struct btrfs_key root_key;
4306 int ret;
4307
4308 BUG_ON(!root->ref_cows);
4309 if (root->reloc_root)
4310 return 0;
4311
4312 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
4313 BUG_ON(!root_item);
4314
4315 ret = btrfs_copy_root(trans, root, root->commit_root,
4316 &eb, BTRFS_TREE_RELOC_OBJECTID);
4317 BUG_ON(ret);
4318
4319 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
4320 root_key.offset = root->root_key.objectid;
4321 root_key.type = BTRFS_ROOT_ITEM_KEY;
4322
4323 memcpy(root_item, &root->root_item, sizeof(root_item));
4324 btrfs_set_root_refs(root_item, 0);
4325 btrfs_set_root_bytenr(root_item, eb->start);
4326 btrfs_set_root_level(root_item, btrfs_header_level(eb));
4327 memset(&root_item->drop_progress, 0, sizeof(root_item->drop_progress));
4328 root_item->drop_level = 0;
4329
4330 btrfs_tree_unlock(eb);
4331 free_extent_buffer(eb);
4332
4333 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
4334 &root_key, root_item);
4335 BUG_ON(ret);
4336 kfree(root_item);
4337
4338 reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
4339 &root_key);
4340 BUG_ON(!reloc_root);
4341 reloc_root->last_trans = trans->transid;
4342 reloc_root->commit_root = NULL;
4343 reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
4344
4345 root->reloc_root = reloc_root;
4346 return 0;
4347}
4348
4349/*
4350 * Core function of space balance.
4351 *
4352 * The idea is using reloc trees to relocate tree blocks in reference
4353 * counted roots. There is one reloc tree for each subvol, all reloc
4354 * trees share same key objectid. Reloc trees are snapshots of the
4355 * latest committed roots (subvol root->commit_root). To relocate a tree
4356 * block referenced by a subvol, the code COW the block through the reloc
4357 * tree, then update pointer in the subvol to point to the new block.
4358 * Since all reloc trees share same key objectid, we can easily do special
4359 * handing to share tree blocks between reloc trees. Once a tree block has
4360 * been COWed in one reloc tree, we can use the result when the same block
4361 * is COWed again through other reloc trees.
4362 */
4363static int noinline relocate_one_path(struct btrfs_trans_handle *trans,
4364 struct btrfs_root *root,
4365 struct btrfs_path *path,
4366 struct btrfs_key *first_key,
4367 struct btrfs_ref_path *ref_path,
4368 struct btrfs_block_group_cache *group,
4369 struct inode *reloc_inode)
4370{
4371 struct btrfs_root *reloc_root;
4372 struct extent_buffer *eb = NULL;
4373 struct btrfs_key *keys;
4374 u64 *nodes;
4375 int level;
4376 int lowest_merge;
4377 int lowest_level = 0;
4378 int update_refs;
4379 int ret;
4380
4381 if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
4382 lowest_level = ref_path->owner_objectid;
4383
4384 if (is_cowonly_root(ref_path->root_objectid)) {
4385 path->lowest_level = lowest_level;
4386 ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
4387 BUG_ON(ret < 0);
3507 path->lowest_level = 0; 4388 path->lowest_level = 0;
3508 btrfs_release_path(found_root, path); 4389 btrfs_release_path(root, path);
4390 return 0;
4391 }
3509 4392
3510 if (found_root == found_root->fs_info->extent_root) 4393 keys = kzalloc(sizeof(*keys) * BTRFS_MAX_LEVEL, GFP_NOFS);
3511 btrfs_extent_post_op(trans, found_root); 4394 BUG_ON(!keys);
3512 if (needs_lock) 4395 nodes = kzalloc(sizeof(*nodes) * BTRFS_MAX_LEVEL, GFP_NOFS);
3513 mutex_unlock(&extent_root->fs_info->alloc_mutex); 4396 BUG_ON(!nodes);
4397
4398 mutex_lock(&root->fs_info->tree_reloc_mutex);
4399 ret = init_reloc_tree(trans, root);
4400 BUG_ON(ret);
4401 reloc_root = root->reloc_root;
4402
4403 path->lowest_level = lowest_level;
4404 ret = btrfs_search_slot(trans, reloc_root, first_key, path, 0, 0);
4405 BUG_ON(ret);
4406 /*
4407 * get relocation mapping for tree blocks in the path
4408 */
4409 lowest_merge = BTRFS_MAX_LEVEL;
4410 for (level = BTRFS_MAX_LEVEL - 1; level >= lowest_level; level--) {
4411 u64 new_bytenr;
4412 eb = path->nodes[level];
4413 if (!eb || eb == reloc_root->node)
4414 continue;
4415 ret = btrfs_get_reloc_mapping(reloc_root, eb->start, eb->len,
4416 &new_bytenr);
4417 if (ret)
4418 continue;
4419 if (level == 0)
4420 btrfs_item_key_to_cpu(eb, &keys[level], 0);
4421 else
4422 btrfs_node_key_to_cpu(eb, &keys[level], 0);
4423 nodes[level] = new_bytenr;
4424 lowest_merge = level;
4425 }
3514 4426
3515 btrfs_end_transaction(trans, found_root); 4427 update_refs = 0;
4428 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
4429 eb = path->nodes[0];
4430 if (btrfs_header_generation(eb) < trans->transid)
4431 update_refs = 1;
4432 }
3516 4433
4434 btrfs_release_path(reloc_root, path);
4435 /*
4436 * merge tree blocks that already relocated in other reloc trees
4437 */
4438 if (lowest_merge != BTRFS_MAX_LEVEL) {
4439 ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
4440 lowest_merge);
4441 BUG_ON(ret < 0);
3517 } 4442 }
3518out: 4443 /*
3519 mutex_lock(&extent_root->fs_info->alloc_mutex); 4444 * cow any tree blocks that still haven't been relocated
4445 */
4446 ret = btrfs_search_slot(trans, reloc_root, first_key, path, 0, 1);
4447 BUG_ON(ret);
4448 /*
4449 * if we are relocating data block group, update extent pointers
4450 * in the newly created tree leaf.
4451 */
4452 eb = path->nodes[0];
4453 if (update_refs && nodes[0] != eb->start) {
4454 ret = replace_extents_in_leaf(trans, reloc_root, eb, group,
4455 reloc_inode);
4456 BUG_ON(ret);
4457 }
4458
4459 memset(keys, 0, sizeof(*keys) * BTRFS_MAX_LEVEL);
4460 memset(nodes, 0, sizeof(*nodes) * BTRFS_MAX_LEVEL);
4461 for (level = BTRFS_MAX_LEVEL - 1; level >= lowest_level; level--) {
4462 eb = path->nodes[level];
4463 if (!eb || eb == reloc_root->node)
4464 continue;
4465 BUG_ON(btrfs_header_owner(eb) != BTRFS_TREE_RELOC_OBJECTID);
4466 nodes[level] = eb->start;
4467 if (level == 0)
4468 btrfs_item_key_to_cpu(eb, &keys[level], 0);
4469 else
4470 btrfs_node_key_to_cpu(eb, &keys[level], 0);
4471 }
4472
4473 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
4474 eb = path->nodes[0];
4475 extent_buffer_get(eb);
4476 }
4477 btrfs_release_path(reloc_root, path);
4478 /*
4479 * replace tree blocks in the fs tree with tree blocks in
4480 * the reloc tree.
4481 */
4482 ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
4483 BUG_ON(ret < 0);
4484
4485 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
4486 ret = invalidate_extent_cache(reloc_root, eb, group, root);
4487 BUG_ON(ret);
4488 free_extent_buffer(eb);
4489 }
4490 mutex_unlock(&root->fs_info->tree_reloc_mutex);
4491
4492 path->lowest_level = 0;
4493 kfree(nodes);
4494 kfree(keys);
4495 return 0;
4496}
4497
4498static int noinline relocate_tree_block(struct btrfs_trans_handle *trans,
4499 struct btrfs_root *root,
4500 struct btrfs_path *path,
4501 struct btrfs_key *first_key,
4502 struct btrfs_ref_path *ref_path)
4503{
4504 int ret;
4505 int needs_lock = 0;
4506
4507 if (root == root->fs_info->extent_root ||
4508 root == root->fs_info->chunk_root ||
4509 root == root->fs_info->dev_root) {
4510 needs_lock = 1;
4511 mutex_lock(&root->fs_info->alloc_mutex);
4512 }
4513
4514 ret = relocate_one_path(trans, root, path, first_key,
4515 ref_path, NULL, NULL);
4516 BUG_ON(ret);
4517
4518 if (root == root->fs_info->extent_root)
4519 btrfs_extent_post_op(trans, root);
4520 if (needs_lock)
4521 mutex_unlock(&root->fs_info->alloc_mutex);
4522
3520 return 0; 4523 return 0;
3521} 4524}
3522 4525
3523static int noinline del_extent_zero(struct btrfs_root *extent_root, 4526static int noinline del_extent_zero(struct btrfs_trans_handle *trans,
4527 struct btrfs_root *extent_root,
3524 struct btrfs_path *path, 4528 struct btrfs_path *path,
3525 struct btrfs_key *extent_key) 4529 struct btrfs_key *extent_key)
3526{ 4530{
3527 int ret; 4531 int ret;
3528 struct btrfs_trans_handle *trans;
3529 4532
3530 trans = btrfs_start_transaction(extent_root, 1); 4533 mutex_lock(&extent_root->fs_info->alloc_mutex);
3531 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1); 4534 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
3532 if (ret > 0) { 4535 if (ret)
3533 ret = -EIO;
3534 goto out;
3535 }
3536 if (ret < 0)
3537 goto out; 4536 goto out;
3538 ret = btrfs_del_item(trans, extent_root, path); 4537 ret = btrfs_del_item(trans, extent_root, path);
3539out: 4538out:
3540 btrfs_end_transaction(trans, extent_root); 4539 btrfs_release_path(extent_root, path);
4540 mutex_unlock(&extent_root->fs_info->alloc_mutex);
3541 return ret; 4541 return ret;
3542} 4542}
3543 4543
4544static struct btrfs_root noinline *read_ref_root(struct btrfs_fs_info *fs_info,
4545 struct btrfs_ref_path *ref_path)
4546{
4547 struct btrfs_key root_key;
4548
4549 root_key.objectid = ref_path->root_objectid;
4550 root_key.type = BTRFS_ROOT_ITEM_KEY;
4551 if (is_cowonly_root(ref_path->root_objectid))
4552 root_key.offset = 0;
4553 else
4554 root_key.offset = (u64)-1;
4555
4556 return btrfs_read_fs_root_no_name(fs_info, &root_key);
4557}
4558
3544static int noinline relocate_one_extent(struct btrfs_root *extent_root, 4559static int noinline relocate_one_extent(struct btrfs_root *extent_root,
3545 struct btrfs_path *path, 4560 struct btrfs_path *path,
3546 struct btrfs_key *extent_key) 4561 struct btrfs_key *extent_key,
4562 struct btrfs_block_group_cache *group,
4563 struct inode *reloc_inode, int pass)
3547{ 4564{
3548 struct btrfs_key key; 4565 struct btrfs_trans_handle *trans;
3549 struct btrfs_key found_key; 4566 struct btrfs_root *found_root;
3550 struct extent_buffer *leaf; 4567 struct btrfs_ref_path *ref_path = NULL;
3551 u64 last_file_objectid = 0; 4568 struct disk_extent *new_extents = NULL;
3552 u64 last_file_root = 0; 4569 int nr_extents = 0;
3553 u64 last_file_offset = (u64)-1; 4570 int loops;
3554 u64 last_extent = 0; 4571 int ret;
3555 u32 nritems; 4572 int level;
3556 u32 item_size; 4573 struct btrfs_key first_key;
3557 int ret = 0; 4574 u64 prev_block = 0;
4575
4576 mutex_unlock(&extent_root->fs_info->alloc_mutex);
4577
4578 trans = btrfs_start_transaction(extent_root, 1);
4579 BUG_ON(!trans);
3558 4580
3559 if (extent_key->objectid == 0) { 4581 if (extent_key->objectid == 0) {
3560 ret = del_extent_zero(extent_root, path, extent_key); 4582 ret = del_extent_zero(trans, extent_root, path, extent_key);
3561 goto out; 4583 goto out;
3562 } 4584 }
3563 key.objectid = extent_key->objectid;
3564 key.type = BTRFS_EXTENT_REF_KEY;
3565 key.offset = 0;
3566 4585
3567 while(1) { 4586 ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
3568 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); 4587 if (!ref_path) {
4588 ret = -ENOMEM;
4589 goto out;
4590 }
3569 4591
4592 for (loops = 0; ; loops++) {
4593 if (loops == 0) {
4594 ret = btrfs_first_ref_path(trans, extent_root, ref_path,
4595 extent_key->objectid);
4596 } else {
4597 ret = btrfs_next_ref_path(trans, extent_root, ref_path);
4598 }
3570 if (ret < 0) 4599 if (ret < 0)
3571 goto out; 4600 goto out;
4601 if (ret > 0)
4602 break;
3572 4603
3573 ret = 0; 4604 if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
3574 leaf = path->nodes[0]; 4605 ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
3575 nritems = btrfs_header_nritems(leaf); 4606 continue;
3576 if (path->slots[0] == nritems) { 4607
3577 ret = btrfs_next_leaf(extent_root, path); 4608 found_root = read_ref_root(extent_root->fs_info, ref_path);
3578 if (ret > 0) { 4609 BUG_ON(!found_root);
3579 ret = 0; 4610 /*
3580 goto out; 4611 * for reference counted tree, only process reference paths
4612 * rooted at the latest committed root.
4613 */
4614 if (found_root->ref_cows &&
4615 ref_path->root_generation != found_root->root_key.offset)
4616 continue;
4617
4618 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
4619 if (pass == 0) {
4620 /*
4621 * copy data extents to new locations
4622 */
4623 u64 group_start = group->key.objectid;
4624 ret = relocate_data_extent(reloc_inode,
4625 extent_key,
4626 group_start);
4627 if (ret < 0)
4628 goto out;
4629 break;
3581 } 4630 }
3582 if (ret < 0) 4631 level = 0;
3583 goto out; 4632 } else {
3584 leaf = path->nodes[0]; 4633 level = ref_path->owner_objectid;
3585 } 4634 }
3586 4635
3587 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 4636 if (prev_block != ref_path->nodes[level]) {
3588 if (found_key.objectid != extent_key->objectid) { 4637 struct extent_buffer *eb;
3589 break; 4638 u64 block_start = ref_path->nodes[level];
3590 } 4639 u64 block_size = btrfs_level_size(found_root, level);
3591 4640
3592 if (found_key.type != BTRFS_EXTENT_REF_KEY) { 4641 eb = read_tree_block(found_root, block_start,
3593 break; 4642 block_size, 0);
4643 btrfs_tree_lock(eb);
4644 BUG_ON(level != btrfs_header_level(eb));
4645
4646 if (level == 0)
4647 btrfs_item_key_to_cpu(eb, &first_key, 0);
4648 else
4649 btrfs_node_key_to_cpu(eb, &first_key, 0);
4650
4651 btrfs_tree_unlock(eb);
4652 free_extent_buffer(eb);
4653 prev_block = block_start;
3594 } 4654 }
3595 4655
3596 key.offset = found_key.offset + 1; 4656 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID &&
3597 item_size = btrfs_item_size_nr(leaf, path->slots[0]); 4657 pass >= 2) {
4658 /*
4659 * use fallback method to process the remaining
4660 * references.
4661 */
4662 if (!new_extents) {
4663 u64 group_start = group->key.objectid;
4664 ret = get_new_locations(reloc_inode,
4665 extent_key,
4666 group_start, 0,
4667 &new_extents,
4668 &nr_extents);
4669 if (ret < 0)
4670 goto out;
4671 }
4672 btrfs_record_root_in_trans(found_root);
4673 ret = replace_one_extent(trans, found_root,
4674 path, extent_key,
4675 &first_key, ref_path,
4676 new_extents, nr_extents);
4677 if (ret < 0)
4678 goto out;
4679 continue;
4680 }
3598 4681
3599 ret = relocate_one_reference(extent_root, path, extent_key, 4682 btrfs_record_root_in_trans(found_root);
3600 &last_file_objectid, 4683 if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
3601 &last_file_offset, 4684 ret = relocate_tree_block(trans, found_root, path,
3602 &last_file_root, last_extent); 4685 &first_key, ref_path);
3603 if (ret) 4686 } else {
4687 /*
4688 * try to update data extent references while
4689 * keeping metadata shared between snapshots.
4690 */
4691 ret = relocate_one_path(trans, found_root, path,
4692 &first_key, ref_path,
4693 group, reloc_inode);
4694 }
4695 if (ret < 0)
3604 goto out; 4696 goto out;
3605 last_extent = extent_key->objectid;
3606 } 4697 }
3607 ret = 0; 4698 ret = 0;
3608out: 4699out:
3609 btrfs_release_path(extent_root, path); 4700 btrfs_end_transaction(trans, extent_root);
4701 kfree(new_extents);
4702 kfree(ref_path);
4703 mutex_lock(&extent_root->fs_info->alloc_mutex);
3610 return ret; 4704 return ret;
3611} 4705}
3612 4706
@@ -3686,84 +4780,155 @@ int __alloc_chunk_for_shrink(struct btrfs_root *root,
3686 return 0; 4780 return 0;
3687} 4781}
3688 4782
3689int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) 4783static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
4784 struct btrfs_root *root,
4785 u64 objectid, u64 size)
4786{
4787 struct btrfs_path *path;
4788 struct btrfs_inode_item *item;
4789 struct extent_buffer *leaf;
4790 int ret;
4791
4792 path = btrfs_alloc_path();
4793 if (!path)
4794 return -ENOMEM;
4795
4796 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
4797 if (ret)
4798 goto out;
4799
4800 leaf = path->nodes[0];
4801 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
4802 memset_extent_buffer(leaf, 0, (unsigned long)item, sizeof(*item));
4803 btrfs_set_inode_generation(leaf, item, 1);
4804 btrfs_set_inode_size(leaf, item, size);
4805 btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
4806 btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NODATASUM);
4807 btrfs_mark_buffer_dirty(leaf);
4808 btrfs_release_path(root, path);
4809out:
4810 btrfs_free_path(path);
4811 return ret;
4812}
4813
4814static struct inode noinline *create_reloc_inode(struct btrfs_fs_info *fs_info,
4815 struct btrfs_block_group_cache *group)
4816{
4817 struct inode *inode = NULL;
4818 struct btrfs_trans_handle *trans;
4819 struct btrfs_root *root;
4820 struct btrfs_key root_key;
4821 u64 objectid = BTRFS_FIRST_FREE_OBJECTID;
4822 int err = 0;
4823
4824 root_key.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
4825 root_key.type = BTRFS_ROOT_ITEM_KEY;
4826 root_key.offset = (u64)-1;
4827 root = btrfs_read_fs_root_no_name(fs_info, &root_key);
4828 if (IS_ERR(root))
4829 return ERR_CAST(root);
4830
4831 trans = btrfs_start_transaction(root, 1);
4832 BUG_ON(!trans);
4833
4834 err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
4835 if (err)
4836 goto out;
4837
4838 err = __insert_orphan_inode(trans, root, objectid, group->key.offset);
4839 BUG_ON(err);
4840
4841 err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0,
4842 group->key.offset, 0);
4843 BUG_ON(err);
4844
4845 inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
4846 if (inode->i_state & I_NEW) {
4847 BTRFS_I(inode)->root = root;
4848 BTRFS_I(inode)->location.objectid = objectid;
4849 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
4850 BTRFS_I(inode)->location.offset = 0;
4851 btrfs_read_locked_inode(inode);
4852 unlock_new_inode(inode);
4853 BUG_ON(is_bad_inode(inode));
4854 } else {
4855 BUG_ON(1);
4856 }
4857
4858 err = btrfs_orphan_add(trans, inode);
4859out:
4860 btrfs_end_transaction(trans, root);
4861 if (err) {
4862 if (inode)
4863 iput(inode);
4864 inode = ERR_PTR(err);
4865 }
4866 return inode;
4867}
4868
4869int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start)
3690{ 4870{
3691 struct btrfs_trans_handle *trans; 4871 struct btrfs_trans_handle *trans;
3692 struct btrfs_root *tree_root = root->fs_info->tree_root;
3693 struct btrfs_path *path; 4872 struct btrfs_path *path;
4873 struct btrfs_fs_info *info = root->fs_info;
4874 struct extent_buffer *leaf;
4875 struct inode *reloc_inode;
4876 struct btrfs_block_group_cache *block_group;
4877 struct btrfs_key key;
3694 u64 cur_byte; 4878 u64 cur_byte;
3695 u64 total_found; 4879 u64 total_found;
3696 u64 shrink_last_byte;
3697 struct btrfs_block_group_cache *shrink_block_group;
3698 struct btrfs_key key;
3699 struct btrfs_key found_key;
3700 struct extent_buffer *leaf;
3701 u32 nritems; 4880 u32 nritems;
3702 int ret; 4881 int ret;
3703 int progress; 4882 int progress;
4883 int pass = 0;
3704 4884
3705 mutex_lock(&root->fs_info->alloc_mutex); 4885 root = root->fs_info->extent_root;
3706 shrink_block_group = btrfs_lookup_block_group(root->fs_info,
3707 shrink_start);
3708 BUG_ON(!shrink_block_group);
3709 4886
3710 shrink_last_byte = shrink_block_group->key.objectid + 4887 block_group = btrfs_lookup_block_group(info, group_start);
3711 shrink_block_group->key.offset; 4888 BUG_ON(!block_group);
4889
4890 printk("btrfs relocating block group %llu flags %llu\n",
4891 (unsigned long long)block_group->key.objectid,
4892 (unsigned long long)block_group->flags);
3712 4893
3713 shrink_block_group->space_info->total_bytes -=
3714 shrink_block_group->key.offset;
3715 path = btrfs_alloc_path(); 4894 path = btrfs_alloc_path();
3716 root = root->fs_info->extent_root; 4895 BUG_ON(!path);
3717 path->reada = 2;
3718 4896
3719 printk("btrfs relocating block group %llu flags %llu\n", 4897 reloc_inode = create_reloc_inode(info, block_group);
3720 (unsigned long long)shrink_start, 4898 BUG_ON(IS_ERR(reloc_inode));
3721 (unsigned long long)shrink_block_group->flags);
3722 4899
3723 __alloc_chunk_for_shrink(root, shrink_block_group, 1); 4900 mutex_lock(&root->fs_info->alloc_mutex);
3724 4901
3725again: 4902 __alloc_chunk_for_shrink(root, block_group, 1);
4903 block_group->ro = 1;
4904 block_group->space_info->total_bytes -= block_group->key.offset;
3726 4905
3727 shrink_block_group->ro = 1; 4906 mutex_unlock(&root->fs_info->alloc_mutex);
3728 4907
4908 btrfs_start_delalloc_inodes(info->tree_root);
4909 btrfs_wait_ordered_extents(info->tree_root, 0);
4910again:
3729 total_found = 0; 4911 total_found = 0;
3730 progress = 0; 4912 progress = 0;
3731 key.objectid = shrink_start; 4913 key.objectid = block_group->key.objectid;
3732 key.offset = 0; 4914 key.offset = 0;
3733 key.type = 0; 4915 key.type = 0;
3734 cur_byte = key.objectid; 4916 cur_byte = key.objectid;
3735 4917
3736 mutex_unlock(&root->fs_info->alloc_mutex); 4918 trans = btrfs_start_transaction(info->tree_root, 1);
4919 btrfs_commit_transaction(trans, info->tree_root);
3737 4920
3738 btrfs_start_delalloc_inodes(root); 4921 mutex_lock(&root->fs_info->cleaner_mutex);
3739 btrfs_wait_ordered_extents(tree_root, 0); 4922 btrfs_clean_old_snapshots(info->tree_root);
4923 btrfs_remove_leaf_refs(info->tree_root, (u64)-1, 1);
4924 mutex_unlock(&root->fs_info->cleaner_mutex);
3740 4925
3741 mutex_lock(&root->fs_info->alloc_mutex); 4926 mutex_lock(&root->fs_info->alloc_mutex);
3742 4927
3743 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3744 if (ret < 0)
3745 goto out;
3746
3747 ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
3748 if (ret < 0)
3749 goto out;
3750
3751 if (ret == 0) {
3752 leaf = path->nodes[0];
3753 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
3754 if (found_key.objectid + found_key.offset > shrink_start &&
3755 found_key.objectid < shrink_last_byte) {
3756 cur_byte = found_key.objectid;
3757 key.objectid = cur_byte;
3758 }
3759 }
3760 btrfs_release_path(root, path);
3761
3762 while(1) { 4928 while(1) {
3763 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 4929 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3764 if (ret < 0) 4930 if (ret < 0)
3765 goto out; 4931 goto out;
3766
3767next: 4932next:
3768 leaf = path->nodes[0]; 4933 leaf = path->nodes[0];
3769 nritems = btrfs_header_nritems(leaf); 4934 nritems = btrfs_header_nritems(leaf);
@@ -3779,109 +4944,76 @@ next:
3779 nritems = btrfs_header_nritems(leaf); 4944 nritems = btrfs_header_nritems(leaf);
3780 } 4945 }
3781 4946
3782 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 4947 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3783 4948
3784 if (found_key.objectid >= shrink_last_byte) 4949 if (key.objectid >= block_group->key.objectid +
4950 block_group->key.offset)
3785 break; 4951 break;
3786 4952
3787 if (progress && need_resched()) { 4953 if (progress && need_resched()) {
3788 memcpy(&key, &found_key, sizeof(key));
3789 cond_resched();
3790 btrfs_release_path(root, path); 4954 btrfs_release_path(root, path);
3791 btrfs_search_slot(NULL, root, &key, path, 0, 0); 4955 mutex_unlock(&root->fs_info->alloc_mutex);
4956 cond_resched();
4957 mutex_lock(&root->fs_info->alloc_mutex);
3792 progress = 0; 4958 progress = 0;
3793 goto next; 4959 continue;
3794 } 4960 }
3795 progress = 1; 4961 progress = 1;
3796 4962
3797 if (btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY || 4963 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY ||
3798 found_key.objectid + found_key.offset <= cur_byte) { 4964 key.objectid + key.offset <= cur_byte) {
3799 memcpy(&key, &found_key, sizeof(key));
3800 key.offset++;
3801 path->slots[0]++; 4965 path->slots[0]++;
3802 goto next; 4966 goto next;
3803 } 4967 }
3804 4968
3805 total_found++; 4969 total_found++;
3806 cur_byte = found_key.objectid + found_key.offset; 4970 cur_byte = key.objectid + key.offset;
3807 key.objectid = cur_byte;
3808 btrfs_release_path(root, path); 4971 btrfs_release_path(root, path);
3809 ret = relocate_one_extent(root, path, &found_key);
3810 __alloc_chunk_for_shrink(root, shrink_block_group, 0);
3811 }
3812
3813 btrfs_release_path(root, path);
3814
3815 if (total_found > 0) {
3816 printk("btrfs relocate found %llu last extent was %llu\n",
3817 (unsigned long long)total_found,
3818 (unsigned long long)found_key.objectid);
3819 mutex_unlock(&root->fs_info->alloc_mutex);
3820 trans = btrfs_start_transaction(tree_root, 1);
3821 btrfs_commit_transaction(trans, tree_root);
3822
3823 btrfs_clean_old_snapshots(tree_root);
3824 4972
3825 btrfs_start_delalloc_inodes(root); 4973 __alloc_chunk_for_shrink(root, block_group, 0);
3826 btrfs_wait_ordered_extents(tree_root, 0); 4974 ret = relocate_one_extent(root, path, &key, block_group,
4975 reloc_inode, pass);
4976 BUG_ON(ret < 0);
3827 4977
3828 trans = btrfs_start_transaction(tree_root, 1); 4978 key.objectid = cur_byte;
3829 btrfs_commit_transaction(trans, tree_root); 4979 key.type = 0;
3830 mutex_lock(&root->fs_info->alloc_mutex); 4980 key.offset = 0;
3831 goto again;
3832 } 4981 }
3833 4982
3834 /* 4983 btrfs_release_path(root, path);
3835 * we've freed all the extents, now remove the block
3836 * group item from the tree
3837 */
3838 mutex_unlock(&root->fs_info->alloc_mutex); 4984 mutex_unlock(&root->fs_info->alloc_mutex);
3839 4985
3840 trans = btrfs_start_transaction(root, 1); 4986 if (pass == 0) {
3841 4987 btrfs_wait_ordered_range(reloc_inode, 0, (u64)-1);
3842 mutex_lock(&root->fs_info->alloc_mutex); 4988 invalidate_mapping_pages(reloc_inode->i_mapping, 0, -1);
3843 memcpy(&key, &shrink_block_group->key, sizeof(key)); 4989 WARN_ON(reloc_inode->i_mapping->nrpages);
3844
3845 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
3846 if (ret > 0)
3847 ret = -EIO;
3848 if (ret < 0) {
3849 btrfs_end_transaction(trans, root);
3850 goto out;
3851 } 4990 }
3852 4991
3853 spin_lock(&root->fs_info->block_group_cache_lock); 4992 if (total_found > 0) {
3854 rb_erase(&shrink_block_group->cache_node, 4993 printk("btrfs found %llu extents in pass %d\n",
3855 &root->fs_info->block_group_cache_tree); 4994 (unsigned long long)total_found, pass);
3856 spin_unlock(&root->fs_info->block_group_cache_lock); 4995 pass++;
3857 4996 goto again;
3858 ret = btrfs_remove_free_space(shrink_block_group, key.objectid,
3859 key.offset);
3860 if (ret) {
3861 btrfs_end_transaction(trans, root);
3862 goto out;
3863 } 4997 }
3864 /*
3865 memset(shrink_block_group, 0, sizeof(*shrink_block_group));
3866 kfree(shrink_block_group);
3867 */
3868 4998
3869 btrfs_del_item(trans, root, path); 4999 /* delete reloc_inode */
3870 btrfs_release_path(root, path); 5000 iput(reloc_inode);
3871 mutex_unlock(&root->fs_info->alloc_mutex); 5001
3872 btrfs_commit_transaction(trans, root); 5002 /* unpin extents in this range */
5003 trans = btrfs_start_transaction(info->tree_root, 1);
5004 btrfs_commit_transaction(trans, info->tree_root);
3873 5005
3874 mutex_lock(&root->fs_info->alloc_mutex); 5006 mutex_lock(&root->fs_info->alloc_mutex);
3875 5007
3876 /* the code to unpin extents might set a few bits in the free 5008 spin_lock(&block_group->lock);
3877 * space cache for this range again 5009 WARN_ON(block_group->pinned > 0);
3878 */ 5010 WARN_ON(block_group->reserved > 0);
3879 /* XXX? */ 5011 WARN_ON(btrfs_block_group_used(&block_group->item) > 0);
3880 ret = btrfs_remove_free_space(shrink_block_group, key.objectid, 5012 spin_unlock(&block_group->lock);
3881 key.offset); 5013 ret = 0;
3882out: 5014out:
3883 btrfs_free_path(path);
3884 mutex_unlock(&root->fs_info->alloc_mutex); 5015 mutex_unlock(&root->fs_info->alloc_mutex);
5016 btrfs_free_path(path);
3885 return ret; 5017 return ret;
3886} 5018}
3887 5019
@@ -3922,6 +5054,33 @@ out:
3922 return ret; 5054 return ret;
3923} 5055}
3924 5056
5057int btrfs_free_block_groups(struct btrfs_fs_info *info)
5058{
5059 struct btrfs_block_group_cache *block_group;
5060 struct rb_node *n;
5061
5062 mutex_lock(&info->alloc_mutex);
5063 spin_lock(&info->block_group_cache_lock);
5064 while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
5065 block_group = rb_entry(n, struct btrfs_block_group_cache,
5066 cache_node);
5067
5068 spin_unlock(&info->block_group_cache_lock);
5069 btrfs_remove_free_space_cache(block_group);
5070 spin_lock(&info->block_group_cache_lock);
5071
5072 rb_erase(&block_group->cache_node,
5073 &info->block_group_cache_tree);
5074 spin_lock(&block_group->space_info->lock);
5075 list_del(&block_group->list);
5076 spin_unlock(&block_group->space_info->lock);
5077 kfree(block_group);
5078 }
5079 spin_unlock(&info->block_group_cache_lock);
5080 mutex_unlock(&info->alloc_mutex);
5081 return 0;
5082}
5083
3925int btrfs_read_block_groups(struct btrfs_root *root) 5084int btrfs_read_block_groups(struct btrfs_root *root)
3926{ 5085{
3927 struct btrfs_path *path; 5086 struct btrfs_path *path;
@@ -4039,3 +5198,46 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
4039 5198
4040 return 0; 5199 return 0;
4041} 5200}
5201
5202int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
5203 struct btrfs_root *root, u64 group_start)
5204{
5205 struct btrfs_path *path;
5206 struct btrfs_block_group_cache *block_group;
5207 struct btrfs_key key;
5208 int ret;
5209
5210 BUG_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
5211 root = root->fs_info->extent_root;
5212
5213 block_group = btrfs_lookup_block_group(root->fs_info, group_start);
5214 BUG_ON(!block_group);
5215
5216 memcpy(&key, &block_group->key, sizeof(key));
5217
5218 path = btrfs_alloc_path();
5219 BUG_ON(!path);
5220
5221 btrfs_remove_free_space_cache(block_group);
5222 rb_erase(&block_group->cache_node,
5223 &root->fs_info->block_group_cache_tree);
5224 spin_lock(&block_group->space_info->lock);
5225 list_del(&block_group->list);
5226 spin_unlock(&block_group->space_info->lock);
5227
5228 /*
5229 memset(shrink_block_group, 0, sizeof(*shrink_block_group));
5230 kfree(shrink_block_group);
5231 */
5232
5233 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
5234 if (ret > 0)
5235 ret = -EIO;
5236 if (ret < 0)
5237 goto out;
5238
5239 ret = btrfs_del_item(trans, root, path);
5240out:
5241 btrfs_free_path(path);
5242 return ret;
5243}
diff --git a/fs/btrfs/root-tree.c b/fs/btrfs/root-tree.c
index e3984f902e71..0091c01abb06 100644
--- a/fs/btrfs/root-tree.c
+++ b/fs/btrfs/root-tree.c
@@ -210,7 +210,10 @@ again:
210 goto err; 210 goto err;
211 } 211 }
212 212
213 ret = btrfs_add_dead_root(dead_root, latest); 213 if (objectid == BTRFS_TREE_RELOC_OBJECTID)
214 ret = btrfs_add_dead_reloc_root(dead_root);
215 else
216 ret = btrfs_add_dead_root(dead_root, latest);
214 if (ret) 217 if (ret)
215 goto err; 218 goto err;
216 goto again; 219 goto again;
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 8c83cf464c83..444abe0796ae 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -477,6 +477,7 @@ static noinline int add_dirty_roots(struct btrfs_trans_handle *trans,
477 dirty = root->dirty_root; 477 dirty = root->dirty_root;
478 478
479 btrfs_free_log(trans, root); 479 btrfs_free_log(trans, root);
480 btrfs_free_reloc_root(root);
480 481
481 if (root->commit_root == root->node) { 482 if (root->commit_root == root->node) {
482 WARN_ON(root->node->start != 483 WARN_ON(root->node->start !=
@@ -855,6 +856,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
855 * with the tree-log code. 856 * with the tree-log code.
856 */ 857 */
857 mutex_lock(&root->fs_info->tree_log_mutex); 858 mutex_lock(&root->fs_info->tree_log_mutex);
859 /*
860 * keep tree reloc code from adding new reloc trees
861 */
862 mutex_lock(&root->fs_info->tree_reloc_mutex);
863
858 864
859 ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, 865 ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix,
860 &dirty_fs_roots); 866 &dirty_fs_roots);
@@ -865,6 +871,8 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
865 */ 871 */
866 btrfs_free_log_root_tree(trans, root->fs_info); 872 btrfs_free_log_root_tree(trans, root->fs_info);
867 873
874 btrfs_free_reloc_mappings(root);
875
868 ret = btrfs_commit_tree_roots(trans, root); 876 ret = btrfs_commit_tree_roots(trans, root);
869 BUG_ON(ret); 877 BUG_ON(ret);
870 878
@@ -910,10 +918,13 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
910 mutex_unlock(&root->fs_info->tree_log_mutex); 918 mutex_unlock(&root->fs_info->tree_log_mutex);
911 919
912 btrfs_finish_extent_commit(trans, root, pinned_copy); 920 btrfs_finish_extent_commit(trans, root, pinned_copy);
913 mutex_lock(&root->fs_info->trans_mutex);
914
915 kfree(pinned_copy); 921 kfree(pinned_copy);
916 922
923 btrfs_drop_dead_reloc_roots(root);
924 mutex_unlock(&root->fs_info->tree_reloc_mutex);
925
926 mutex_lock(&root->fs_info->trans_mutex);
927
917 cur_trans->commit_done = 1; 928 cur_trans->commit_done = 1;
918 root->fs_info->last_trans_committed = cur_trans->transid; 929 root->fs_info->last_trans_committed = cur_trans->transid;
919 wake_up(&cur_trans->commit_wait); 930 wake_up(&cur_trans->commit_wait);
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index ddf89626498a..51f113119b25 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1268,7 +1268,7 @@ int btrfs_relocate_chunk(struct btrfs_root *root,
1268 em_tree = &root->fs_info->mapping_tree.map_tree; 1268 em_tree = &root->fs_info->mapping_tree.map_tree;
1269 1269
1270 /* step one, relocate all the extents inside this chunk */ 1270 /* step one, relocate all the extents inside this chunk */
1271 ret = btrfs_shrink_extent_tree(extent_root, chunk_offset); 1271 ret = btrfs_relocate_block_group(extent_root, chunk_offset);
1272 BUG_ON(ret); 1272 BUG_ON(ret);
1273 1273
1274 trans = btrfs_start_transaction(root, 1); 1274 trans = btrfs_start_transaction(root, 1);
@@ -1308,15 +1308,18 @@ int btrfs_relocate_chunk(struct btrfs_root *root,
1308 BUG_ON(ret); 1308 BUG_ON(ret);
1309 } 1309 }
1310 1310
1311 ret = btrfs_remove_block_group(trans, extent_root, chunk_offset);
1312 BUG_ON(ret);
1313
1311 spin_lock(&em_tree->lock); 1314 spin_lock(&em_tree->lock);
1312 remove_extent_mapping(em_tree, em); 1315 remove_extent_mapping(em_tree, em);
1316 spin_unlock(&em_tree->lock);
1317
1313 kfree(map); 1318 kfree(map);
1314 em->bdev = NULL; 1319 em->bdev = NULL;
1315 1320
1316 /* once for the tree */ 1321 /* once for the tree */
1317 free_extent_map(em); 1322 free_extent_map(em);
1318 spin_unlock(&em_tree->lock);
1319
1320 /* once for us */ 1323 /* once for us */
1321 free_extent_map(em); 1324 free_extent_map(em);
1322 1325
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-11 20:48:18 -0500