l = path->nodes[0];
btrfs_item_key_to_cpu(l, &key, path->slots[0]);
- BUG_ON(key.objectid != bytenr);
+ if (key.objectid != bytenr) {
+ btrfs_print_leaf(root->fs_info->extent_root, path->nodes[0]);
+ printk("wanted %Lu found %Lu\n", bytenr, key.objectid);
+ BUG();
+ }
BUG_ON(key.type != BTRFS_EXTENT_ITEM_KEY);
item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
return 0;
}
-static int get_reference_status(struct btrfs_root *root, u64 bytenr,
- u64 parent_gen, u64 ref_objectid,
- u64 *min_generation, u32 *ref_count)
+int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
+ struct btrfs_root *root, u64 bytenr)
{
struct btrfs_root *extent_root = root->fs_info->extent_root;
struct btrfs_path *path;
struct btrfs_extent_ref *ref_item;
struct btrfs_key key;
struct btrfs_key found_key;
- u64 root_objectid = root->root_key.objectid;
- u64 ref_generation;
+ u64 ref_root;
+ u64 last_snapshot;
u32 nritems;
int ret;
if (ret < 0)
goto out;
BUG_ON(ret == 0);
- if (ret < 0 || path->slots[0] == 0)
+
+ ret = -ENOENT;
+ if (path->slots[0] == 0)
goto out;
path->slots[0]--;
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.objectid != bytenr ||
- found_key.type != BTRFS_EXTENT_ITEM_KEY) {
- ret = 1;
+ found_key.type != BTRFS_EXTENT_ITEM_KEY)
goto out;
- }
-
- *ref_count = 0;
- *min_generation = (u64)-1;
+ last_snapshot = btrfs_root_last_snapshot(&root->root_item);
while (1) {
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
ref_item = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref);
- ref_generation = btrfs_ref_generation(leaf, ref_item);
- /*
- * For (parent_gen > 0 && parent_gen > ref_generation):
- *
- * we reach here through the oldest root, therefore
- * all other reference from same snapshot should have
- * a larger generation.
- */
- if ((root_objectid != btrfs_ref_root(leaf, ref_item)) ||
- (parent_gen > 0 && parent_gen > ref_generation) ||
- (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID &&
- ref_objectid != btrfs_ref_objectid(leaf, ref_item))) {
- *ref_count = 2;
- break;
- }
-
- *ref_count = 1;
- if (*min_generation > ref_generation)
- *min_generation = ref_generation;
-
- path->slots[0]++;
- }
- ret = 0;
-out:
- btrfs_free_path(path);
- return ret;
-}
-
-int btrfs_cross_ref_exists(struct btrfs_trans_handle *trans,
- struct btrfs_root *root,
- struct btrfs_key *key, u64 bytenr)
-{
- struct btrfs_root *old_root;
- struct btrfs_path *path = NULL;
- struct extent_buffer *eb;
- struct btrfs_file_extent_item *item;
- u64 ref_generation;
- u64 min_generation;
- u64 extent_start;
- u32 ref_count;
- int level;
- int ret;
-
- BUG_ON(trans == NULL);
- BUG_ON(key->type != BTRFS_EXTENT_DATA_KEY);
- ret = get_reference_status(root, bytenr, 0, key->objectid,
- &min_generation, &ref_count);
- if (ret)
- return ret;
-
- if (ref_count != 1)
- return 1;
-
- old_root = root->dirty_root->root;
- ref_generation = old_root->root_key.offset;
-
- /* all references are created in running transaction */
- if (min_generation > ref_generation) {
- ret = 0;
- goto out;
- }
-
- path = btrfs_alloc_path();
- if (!path) {
- ret = -ENOMEM;
- goto out;
- }
-
- path->skip_locking = 1;
- /* if no item found, the extent is referenced by other snapshot */
- ret = btrfs_search_slot(NULL, old_root, key, path, 0, 0);
- if (ret)
- goto out;
-
- eb = path->nodes[0];
- item = btrfs_item_ptr(eb, path->slots[0],
- struct btrfs_file_extent_item);
- if (btrfs_file_extent_type(eb, item) != BTRFS_FILE_EXTENT_REG ||
- btrfs_file_extent_disk_bytenr(eb, item) != bytenr) {
- ret = 1;
- goto out;
- }
-
- for (level = BTRFS_MAX_LEVEL - 1; level >= -1; level--) {
- if (level >= 0) {
- eb = path->nodes[level];
- if (!eb)
- continue;
- extent_start = eb->start;
- } else
- extent_start = bytenr;
-
- ret = get_reference_status(root, extent_start, ref_generation,
- 0, &min_generation, &ref_count);
- if (ret)
+ ref_root = btrfs_ref_root(leaf, ref_item);
+ if (ref_root != root->root_key.objectid &&
+ ref_root != BTRFS_TREE_LOG_OBJECTID) {
+ ret = 1;
goto out;
-
- if (ref_count != 1) {
+ }
+ if (btrfs_ref_generation(leaf, ref_item) <= last_snapshot) {
ret = 1;
goto out;
}
- if (level >= 0)
- ref_generation = btrfs_header_generation(eb);
+
+ path->slots[0]++;
}
ret = 0;
out:
- if (path)
- btrfs_free_path(path);
+ btrfs_free_path(path);
return ret;
}
struct btrfs_root * root = orig_root->fs_info->extent_root;
u64 total_needed = num_bytes;
u64 *last_ptr = NULL;
+ u64 last_wanted = 0;
struct btrfs_block_group_cache *block_group = NULL;
int chunk_alloc_done = 0;
int empty_cluster = 2 * 1024 * 1024;
int allowed_chunk_alloc = 0;
struct list_head *head = NULL, *cur = NULL;
int loop = 0;
+ int extra_loop = 0;
struct btrfs_space_info *space_info;
WARN_ON(num_bytes < root->sectorsize);
if (data & BTRFS_BLOCK_GROUP_METADATA) {
last_ptr = &root->fs_info->last_alloc;
- empty_cluster = 256 * 1024;
+ empty_cluster = 64 * 1024;
}
if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD))
last_ptr = &root->fs_info->last_data_alloc;
if (last_ptr) {
- if (*last_ptr)
+ if (*last_ptr) {
hint_byte = *last_ptr;
- else
+ last_wanted = *last_ptr;
+ } else
empty_size += empty_cluster;
+ } else {
+ empty_cluster = 0;
}
search_start = max(search_start, first_logical_byte(root, 0));
search_start = max(search_start, hint_byte);
- total_needed += empty_size;
+ if (last_wanted && search_start != last_wanted) {
+ last_wanted = 0;
+ empty_size += empty_cluster;
+ }
+
+ total_needed += empty_size;
block_group = btrfs_lookup_block_group(root->fs_info, search_start);
+ if (!block_group)
+ block_group = btrfs_lookup_first_block_group(root->fs_info,
+ search_start);
space_info = __find_space_info(root->fs_info, data);
down_read(&space_info->groups_sem);
* group thats not of the proper type, while looping this
* should never happen
*/
- WARN_ON(!block_group);
+ if (empty_size)
+ extra_loop = 1;
+
+ if (!block_group)
+ goto new_group_no_lock;
+
mutex_lock(&block_group->alloc_mutex);
if (unlikely(!block_group_bits(block_group, data)))
goto new_group;
if (search_start + num_bytes > end)
goto new_group;
+ if (last_wanted && search_start != last_wanted) {
+ total_needed += empty_cluster;
+ empty_size += empty_cluster;
+ last_wanted = 0;
+ /*
+ * if search_start is still in this block group
+ * then we just re-search this block group
+ */
+ if (search_start >= start &&
+ search_start < end) {
+ mutex_unlock(&block_group->alloc_mutex);
+ continue;
+ }
+
+ /* else we go to the next block group */
+ goto new_group;
+ }
+
if (exclude_nr > 0 &&
(search_start + num_bytes > exclude_start &&
search_start < exclude_start + exclude_nr)) {
if (search_start >= start &&
search_start < end) {
mutex_unlock(&block_group->alloc_mutex);
+ last_wanted = 0;
continue;
}
}
new_group:
mutex_unlock(&block_group->alloc_mutex);
+new_group_no_lock:
+ /* don't try to compare new allocations against the
+ * last allocation any more
+ */
+ last_wanted = 0;
+
/*
* Here's how this works.
* loop == 0: we were searching a block group via a hint
if (loop == 0) {
head = &space_info->block_groups;
cur = head->next;
-
- if (last_ptr && *last_ptr) {
- total_needed += empty_cluster;
- *last_ptr = 0;
- }
loop++;
} else if (loop == 1 && cur == head) {
+ int keep_going;
+
+ /* at this point we give up on the empty_size
+ * allocations and just try to allocate the min
+ * space.
+ *
+ * The extra_loop field was set if an empty_size
+ * allocation was attempted above, and if this
+ * is try we need to try the loop again without
+ * the additional empty_size.
+ */
+ total_needed -= empty_size;
+ empty_size = 0;
+ keep_going = extra_loop;
+ loop++;
+
if (allowed_chunk_alloc && !chunk_alloc_done) {
up_read(&space_info->groups_sem);
ret = do_chunk_alloc(trans, root, num_bytes +
if (ret < 0)
break;
down_read(&space_info->groups_sem);
- loop++;
head = &space_info->block_groups;
- cur = head->next;
+ /*
+ * we've allocated a new chunk, keep
+ * trying
+ */
+ keep_going = 1;
chunk_alloc_done = 1;
} else if (!allowed_chunk_alloc) {
space_info->force_alloc = 1;
- break;
+ }
+ if (keep_going) {
+ cur = head->next;
+ extra_loop = 0;
} else {
break;
}
struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
struct extent_map *em;
+ u64 start = extent_key->objectid - offset;
+ u64 end = start + extent_key->offset - 1;
em = alloc_extent_map(GFP_NOFS);
BUG_ON(!em || IS_ERR(em));
- em->start = extent_key->objectid - offset;
+ em->start = start;
em->len = extent_key->offset;
em->block_len = extent_key->offset;
em->block_start = extent_key->objectid;
set_bit(EXTENT_FLAG_PINNED, &em->flags);
/* setup extent map to cheat btrfs_readpage */
- mutex_lock(&BTRFS_I(reloc_inode)->extent_mutex);
+ lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
while (1) {
int ret;
spin_lock(&em_tree->lock);
free_extent_map(em);
break;
}
- btrfs_drop_extent_cache(reloc_inode, em->start,
- em->start + em->len - 1, 0);
+ btrfs_drop_extent_cache(reloc_inode, start, end, 0);
}
- mutex_unlock(&BTRFS_I(reloc_inode)->extent_mutex);
+ unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
- return relocate_inode_pages(reloc_inode, extent_key->objectid - offset,
- extent_key->offset);
+ return relocate_inode_pages(reloc_inode, start, extent_key->offset);
}
struct btrfs_ref_path {
next:
level--;
btrfs_release_path(extent_root, path);
- if (need_resched()) {
- cond_resched();
- }
+ cond_resched();
}
/* reached lowest level */
ret = 1;
}
btrfs_release_path(extent_root, path);
- if (need_resched()) {
- cond_resched();
- }
+ cond_resched();
}
/* reached max tree level, but no tree root found. */
BUG();
exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
fi);
- WARN_ON(exts[nr].offset > 0);
- WARN_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
+ BUG_ON(exts[nr].offset > 0);
+ BUG_ON(exts[nr].compression || exts[nr].encryption);
+ BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
cur_pos += exts[nr].num_bytes;
nr++;
u32 nritems;
int nr_scaned = 0;
int extent_locked = 0;
+ int extent_type;
int ret;
memcpy(&key, leaf_key, sizeof(key));
* the file extent item was modified by someone
* before the extent got locked.
*/
- mutex_unlock(&BTRFS_I(inode)->extent_mutex);
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
extent_locked = 0;
}
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
- if ((btrfs_file_extent_type(leaf, fi) !=
- BTRFS_FILE_EXTENT_REG) ||
+ extent_type = btrfs_file_extent_type(leaf, fi);
+ if ((extent_type != BTRFS_FILE_EXTENT_REG &&
+ extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
(btrfs_file_extent_disk_bytenr(leaf, fi) !=
extent_key->objectid)) {
path->slots[0]++;
lock_start = key.offset;
lock_end = lock_start + num_bytes - 1;
} else {
- BUG_ON(lock_start != key.offset);
- BUG_ON(lock_end - lock_start + 1 < num_bytes);
+ if (lock_start > key.offset ||
+ lock_end + 1 < key.offset + num_bytes) {
+ unlock_extent(&BTRFS_I(inode)->io_tree,
+ lock_start, lock_end, GFP_NOFS);
+ extent_locked = 0;
+ }
}
if (!inode) {
if (ordered)
btrfs_put_ordered_extent(ordered);
- mutex_lock(&BTRFS_I(inode)->extent_mutex);
extent_locked = 1;
continue;
}
if (nr_extents == 1) {
/* update extent pointer in place */
- btrfs_set_file_extent_generation(leaf, fi,
- trans->transid);
btrfs_set_file_extent_disk_bytenr(leaf, fi,
new_extents[0].disk_bytenr);
btrfs_set_file_extent_disk_num_bytes(leaf, fi,
new_extents[0].disk_num_bytes);
- btrfs_set_file_extent_ram_bytes(leaf, fi,
- new_extents[0].ram_bytes);
- ext_offset += new_extents[0].offset;
- btrfs_set_file_extent_offset(leaf, fi, ext_offset);
btrfs_mark_buffer_dirty(leaf);
btrfs_drop_extent_cache(inode, key.offset,
btrfs_release_path(root, path);
key.offset += num_bytes;
} else {
+ BUG_ON(1);
+#if 0
u64 alloc_hint;
u64 extent_len;
int i;
break;
}
BUG_ON(i >= nr_extents);
+#endif
}
if (extent_locked) {
- mutex_unlock(&BTRFS_I(inode)->extent_mutex);
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
extent_locked = 0;
if (inode) {
mutex_unlock(&inode->i_mutex);
if (extent_locked) {
- mutex_unlock(&BTRFS_I(inode)->extent_mutex);
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
}
lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
key.offset + num_bytes - 1, GFP_NOFS);
- mutex_lock(&BTRFS_I(inode)->extent_mutex);
btrfs_drop_extent_cache(inode, key.offset,
key.offset + num_bytes - 1, 1);
- mutex_unlock(&BTRFS_I(inode)->extent_mutex);
unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
key.offset + num_bytes - 1, GFP_NOFS);
cond_resched();
ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
- btrfs_set_file_extent_generation(leaf, fi, trans->transid);
- btrfs_set_file_extent_ram_bytes(leaf, fi,
- new_extent->ram_bytes);
btrfs_set_file_extent_disk_bytenr(leaf, fi,
new_extent->disk_bytenr);
btrfs_set_file_extent_disk_num_bytes(leaf, fi,
new_extent->disk_num_bytes);
- new_extent->offset += btrfs_file_extent_offset(leaf, fi);
- btrfs_set_file_extent_offset(leaf, fi, new_extent->offset);
btrfs_mark_buffer_dirty(leaf);
ret = btrfs_inc_extent_ref(trans, root,
*/
if (!new_extents) {
u64 group_start = group->key.objectid;
+ new_extents = kmalloc(sizeof(*new_extents),
+ GFP_NOFS);
+ nr_extents = 1;
ret = get_new_locations(reloc_inode,
extent_key,
- group_start, 0,
+ group_start, 1,
&new_extents,
&nr_extents);
- if (ret < 0)
+ if (ret)
goto out;
}
btrfs_record_root_in_trans(found_root);
btrfs_set_inode_generation(leaf, item, 1);
btrfs_set_inode_size(leaf, item, size);
btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
- btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NODATASUM);
+ btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NODATASUM |
+ BTRFS_INODE_NOCOMPRESS);
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(root, path);
out:
struct inode *reloc_inode;
struct btrfs_block_group_cache *block_group;
struct btrfs_key key;
+ u64 skipped;
u64 cur_byte;
u64 total_found;
u32 nritems;
btrfs_start_delalloc_inodes(info->tree_root);
btrfs_wait_ordered_extents(info->tree_root, 0);
again:
+ skipped = 0;
total_found = 0;
progress = 0;
key.objectid = block_group->key.objectid;
ret = relocate_one_extent(root, path, &key, block_group,
reloc_inode, pass);
BUG_ON(ret < 0);
+ if (ret > 0)
+ skipped++;
key.objectid = cur_byte;
key.type = 0;
printk("btrfs found %llu extents in pass %d\n",
(unsigned long long)total_found, pass);
pass++;
+ if (total_found == skipped && pass > 2) {
+ iput(reloc_inode);
+ reloc_inode = create_reloc_inode(info, block_group);
+ pass = 0;
+ }
goto again;
}
while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
block_group = rb_entry(n, struct btrfs_block_group_cache,
cache_node);
-
- spin_unlock(&info->block_group_cache_lock);
- btrfs_remove_free_space_cache(block_group);
- spin_lock(&info->block_group_cache_lock);
-
rb_erase(&block_group->cache_node,
&info->block_group_cache_tree);
+ spin_unlock(&info->block_group_cache_lock);
+
+ btrfs_remove_free_space_cache(block_group);
down_write(&block_group->space_info->groups_sem);
list_del(&block_group->list);
up_write(&block_group->space_info->groups_sem);
kfree(block_group);
+
+ spin_lock(&info->block_group_cache_lock);
}
spin_unlock(&info->block_group_cache_lock);
return 0;
projects / linux-2.6-block.git / blobdiff