* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 021110-1307, USA.
*/
-
#include <linux/sched.h>
-#include <linux/crc32c.h>
#include <linux/pagemap.h>
+#include <linux/writeback.h>
#include "hash.h"
+#include "crc32c.h"
#include "ctree.h"
#include "disk-io.h"
#include "print-tree.h"
btrfs_root *extent_root);
static int del_pending_extents(struct btrfs_trans_handle *trans, struct
btrfs_root *extent_root);
-int btrfs_make_block_group(struct btrfs_trans_handle *trans,
- struct btrfs_root *root, u64 bytes_used,
- u64 type, u64 chunk_tree, u64 chunk_objectid,
- u64 size);
static int cache_block_group(struct btrfs_root *root,
if (!cache)
goto out;
+
total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
free_space_cache = &root->fs_info->free_space_cache;
goto out;
last = max(search_start, cache->key.objectid);
- if (!block_group_bits(cache, data)) {
+ if (!block_group_bits(cache, data) || cache->ro) {
goto new_group;
}
continue;
}
spin_unlock_irq(&free_space_cache->lock);
+ if (cache->ro)
+ goto new_group;
if (start + num > cache->key.objectid + cache->key.offset)
goto new_group;
if (start + num > total_fs_bytes)
int bit;
int ret;
int full_search = 0;
- int factor = 8;
+ int factor = 10;
block_group_cache = &info->block_group_cache;
total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
- if (!owner)
- factor = 8;
+ if (data & BTRFS_BLOCK_GROUP_METADATA)
+ factor = 9;
bit = block_group_state_bits(data);
if (search_start && search_start < total_fs_bytes) {
struct btrfs_block_group_cache *shint;
shint = btrfs_lookup_block_group(info, search_start);
- if (shint && block_group_bits(shint, data)) {
+ if (shint && block_group_bits(shint, data) && !shint->ro) {
used = btrfs_block_group_used(&shint->item);
if (used + shint->pinned <
div_factor(shint->key.offset, factor)) {
}
}
}
- if (hint && block_group_bits(hint, data) &&
+ if (hint && !hint->ro && block_group_bits(hint, data) &&
hint->key.objectid < total_fs_bytes) {
used = btrfs_block_group_used(&hint->item);
if (used + hint->pinned <
if (cache->key.objectid > total_fs_bytes)
break;
- if (block_group_bits(cache, data)) {
+ if (!cache->ro && block_group_bits(cache, data)) {
if (full_search)
free_check = cache->key.offset;
else
u32 high_crc = ~(u32)0;
u32 low_crc = ~(u32)0;
__le64 lenum;
-
lenum = cpu_to_le64(root_objectid);
- high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
+ high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
lenum = cpu_to_le64(ref_generation);
- low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
+ low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
if (owner >= BTRFS_FIRST_FREE_OBJECTID) {
lenum = cpu_to_le64(owner);
- low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
+ low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
lenum = cpu_to_le64(owner_offset);
- low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
+ low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
}
return ((u64)high_crc << 32) | (u64)low_crc;
}
if (!path)
return -ENOMEM;
- path->reada = 0;
+ path->reada = 1;
key.objectid = bytenr;
btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
key.offset = num_bytes;
btrfs_release_path(root->fs_info->extent_root, path);
- path->reada = 0;
+ path->reada = 1;
ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root,
path, bytenr, root_objectid,
ref_generation, owner, owner_offset);
WARN_ON(num_bytes < root->sectorsize);
path = btrfs_alloc_path();
- path->reada = 0;
+ path->reada = 1;
key.objectid = bytenr;
key.offset = num_bytes;
btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
ret = get_state_private(block_group_cache, start, &ptr);
if (ret)
break;
-
cache = (struct btrfs_block_group_cache *)(unsigned long)ptr;
err = write_one_cache_group(trans, root,
path, cache);
if (found) {
found->total_bytes += total_bytes;
found->bytes_used += bytes_used;
+ found->full = 0;
WARN_ON(found->total_bytes < found->bytes_used);
*space_info = found;
return 0;
{
u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID10 |
BTRFS_BLOCK_GROUP_DUP);
if (extra_flags) {
if (flags & BTRFS_BLOCK_GROUP_DATA)
}
}
+static u64 reduce_alloc_profile(u64 flags)
+{
+ if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
+ (flags & (BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID10)))
+ flags &= ~BTRFS_BLOCK_GROUP_DUP;
+
+ if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
+ (flags & BTRFS_BLOCK_GROUP_RAID10))
+ flags &= ~BTRFS_BLOCK_GROUP_RAID1;
+
+ if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
+ ((flags & BTRFS_BLOCK_GROUP_RAID1) |
+ (flags & BTRFS_BLOCK_GROUP_RAID10) |
+ (flags & BTRFS_BLOCK_GROUP_DUP)))
+ flags &= ~BTRFS_BLOCK_GROUP_RAID0;
+ return flags;
+}
+
+
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root, u64 alloc_bytes,
u64 flags)
u64 num_bytes;
int ret;
+ flags = reduce_alloc_profile(flags);
+
space_info = __find_space_info(extent_root->fs_info, flags);
if (!space_info) {
ret = update_space_info(extent_root->fs_info, flags,
BUG_ON(ret);
ret = btrfs_make_block_group(trans, extent_root, 0, flags,
- extent_root->fs_info->chunk_root->root_key.objectid,
- start, num_bytes);
+ BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes);
BUG_ON(ret);
- set_avail_alloc_bits(extent_root->fs_info, flags);
return 0;
}
if (!path)
return -ENOMEM;
- path->reada = 0;
+ path->reada = 1;
ret = lookup_extent_backref(trans, extent_root, path,
bytenr, root_objectid,
ref_generation,
error:
return ret;
}
+
/*
* finds a free extent and does all the dirty work required for allocation
* returns the key for the extent through ins, and a tree buffer for
*/
int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
- u64 num_bytes, u64 root_objectid, u64 ref_generation,
+ u64 num_bytes, u64 min_alloc_size,
+ u64 root_objectid, u64 ref_generation,
u64 owner, u64 owner_offset,
u64 empty_size, u64 hint_byte,
- u64 search_end, struct btrfs_key *ins, int data)
+ u64 search_end, struct btrfs_key *ins, u64 data)
{
int ret;
int pending_ret;
u64 super_used;
u64 root_used;
u64 search_start = 0;
- u64 new_hint;
u64 alloc_profile;
u32 sizes[2];
struct btrfs_fs_info *info = root->fs_info;
info->metadata_alloc_profile;
data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
}
-
+again:
+ data = reduce_alloc_profile(data);
if (root->ref_cows) {
if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
ret = do_chunk_alloc(trans, root->fs_info->extent_root,
BUG_ON(ret);
}
- new_hint = max(hint_byte, root->fs_info->alloc_start);
- if (new_hint < btrfs_super_total_bytes(&info->super_copy))
- hint_byte = new_hint;
-
WARN_ON(num_bytes < root->sectorsize);
ret = find_free_extent(trans, root, num_bytes, empty_size,
search_start, search_end, hint_byte, ins,
trans->alloc_exclude_start,
trans->alloc_exclude_nr, data);
+
+ if (ret == -ENOSPC && num_bytes > min_alloc_size) {
+ num_bytes = num_bytes >> 1;
+ num_bytes = max(num_bytes, min_alloc_size);
+ goto again;
+ }
+ if (ret) {
+ printk("allocation failed flags %Lu\n", data);
+ }
BUG_ON(ret);
if (ret)
return ret;
int ret;
struct extent_buffer *buf;
- ret = btrfs_alloc_extent(trans, root, blocksize,
+ ret = btrfs_alloc_extent(trans, root, blocksize, blocksize,
root_objectid, ref_generation,
level, first_objectid, empty_size, hint,
(u64)-1, &ins, 0);
}
btrfs_set_header_generation(buf, trans->transid);
clean_tree_block(trans, root, buf);
- wait_on_tree_block_writeback(root, buf);
btrfs_set_buffer_uptodate(buf);
if (PageDirty(buf->first_page)) {
set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
buf->start + buf->len - 1, GFP_NOFS);
- set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->io_tree,
- buf->start, buf->start + buf->len - 1,
- EXTENT_CSUM, GFP_NOFS);
- buf->flags |= EXTENT_CSUM;
if (!btrfs_test_opt(root, SSD))
btrfs_set_buffer_defrag(buf);
trans->blocks_used++;
return 0;
}
+static unsigned long calc_ra(unsigned long start, unsigned long last,
+ unsigned long nr)
+{
+ return min(last, start + nr - 1);
+}
+
static int noinline relocate_inode_pages(struct inode *inode, u64 start,
u64 len)
{
u64 page_start;
u64 page_end;
- u64 delalloc_start;
- u64 existing_delalloc;
unsigned long last_index;
unsigned long i;
struct page *page;
struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
struct file_ra_state *ra;
+ unsigned long total_read = 0;
+ unsigned long ra_pages;
ra = kzalloc(sizeof(*ra), GFP_NOFS);
i = start >> PAGE_CACHE_SHIFT;
last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
+ ra_pages = BTRFS_I(inode)->root->fs_info->bdi.ra_pages;
+
file_ra_state_init(ra, inode->i_mapping);
- btrfs_force_ra(inode->i_mapping, ra, NULL, i, last_index);
- kfree(ra);
for (; i <= last_index; i++) {
+ if (total_read % ra_pages == 0) {
+ btrfs_force_ra(inode->i_mapping, ra, NULL, i,
+ calc_ra(i, last_index, ra_pages));
+ }
+ total_read++;
page = grab_cache_page(inode->i_mapping, i);
if (!page)
goto out_unlock;
goto out_unlock;
}
}
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+ ClearPageDirty(page);
+#else
+ cancel_dirty_page(page, PAGE_CACHE_SIZE);
+#endif
+ wait_on_page_writeback(page);
+ set_page_extent_mapped(page);
page_start = (u64)page->index << PAGE_CACHE_SHIFT;
page_end = page_start + PAGE_CACHE_SIZE - 1;
lock_extent(io_tree, page_start, page_end, GFP_NOFS);
- delalloc_start = page_start;
- existing_delalloc = count_range_bits(io_tree,
- &delalloc_start, page_end,
- PAGE_CACHE_SIZE, EXTENT_DELALLOC);
-
+ set_page_dirty(page);
set_extent_delalloc(io_tree, page_start,
page_end, GFP_NOFS);
unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
- set_page_dirty(page);
unlock_page(page);
page_cache_release(page);
+ balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
}
out_unlock:
+ kfree(ra);
mutex_unlock(&inode->i_mutex);
return 0;
}
goto out;
}
relocate_inode_pages(inode, ref_offset, extent_key->offset);
- /* FIXME, data=ordered will help get rid of this */
- filemap_fdatawrite(inode->i_mapping);
iput(inode);
mutex_lock(&extent_root->fs_info->fs_mutex);
} else {
return ret;
}
-int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size)
+static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
+{
+ u64 num_devices;
+ u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
+ BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
+
+ num_devices = btrfs_super_num_devices(&root->fs_info->super_copy);
+ if (num_devices == 1) {
+ stripped |= BTRFS_BLOCK_GROUP_DUP;
+ stripped = flags & ~stripped;
+
+ /* turn raid0 into single device chunks */
+ if (flags & BTRFS_BLOCK_GROUP_RAID0)
+ return stripped;
+
+ /* turn mirroring into duplication */
+ if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
+ BTRFS_BLOCK_GROUP_RAID10))
+ return stripped | BTRFS_BLOCK_GROUP_DUP;
+ return flags;
+ } else {
+ /* they already had raid on here, just return */
+ if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
+ (flags & BTRFS_BLOCK_GROUP_RAID1)) {
+ }
+ if (flags & stripped)
+ return flags;
+
+ stripped |= BTRFS_BLOCK_GROUP_DUP;
+ stripped = flags & ~stripped;
+
+ /* switch duplicated blocks with raid1 */
+ if (flags & BTRFS_BLOCK_GROUP_DUP)
+ return stripped | BTRFS_BLOCK_GROUP_RAID1;
+
+ /* turn single device chunks into raid0 */
+ return stripped | BTRFS_BLOCK_GROUP_RAID0;
+ }
+ return flags;
+}
+
+int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start)
{
struct btrfs_trans_handle *trans;
struct btrfs_root *tree_root = root->fs_info->tree_root;
struct btrfs_path *path;
u64 cur_byte;
u64 total_found;
+ u64 shrink_last_byte;
+ u64 new_alloc_flags;
+ struct btrfs_block_group_cache *shrink_block_group;
struct btrfs_fs_info *info = root->fs_info;
- struct extent_io_tree *block_group_cache;
struct btrfs_key key;
struct btrfs_key found_key;
struct extent_buffer *leaf;
int ret;
int progress = 0;
- btrfs_set_super_total_bytes(&info->super_copy, new_size);
- clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1,
- GFP_NOFS);
- block_group_cache = &info->block_group_cache;
+ shrink_block_group = btrfs_lookup_block_group(root->fs_info,
+ shrink_start);
+ BUG_ON(!shrink_block_group);
+
+ shrink_last_byte = shrink_start + shrink_block_group->key.offset;
+
+ shrink_block_group->space_info->total_bytes -=
+ shrink_block_group->key.offset;
path = btrfs_alloc_path();
root = root->fs_info->extent_root;
path->reada = 2;
again:
+ if (btrfs_block_group_used(&shrink_block_group->item) > 0) {
+ trans = btrfs_start_transaction(root, 1);
+ new_alloc_flags = update_block_group_flags(root,
+ shrink_block_group->flags);
+ do_chunk_alloc(trans, root->fs_info->extent_root,
+ btrfs_block_group_used(&shrink_block_group->item) +
+ 2 * 1024 * 1024, new_alloc_flags);
+ btrfs_end_transaction(trans, root);
+ }
+ shrink_block_group->ro = 1;
+
total_found = 0;
- key.objectid = new_size;
+ key.objectid = shrink_start;
key.offset = 0;
key.type = 0;
cur_byte = key.objectid;
ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
if (ret < 0)
goto out;
+
if (ret == 0) {
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
- if (found_key.objectid + found_key.offset > new_size) {
+ if (found_key.objectid + found_key.offset > shrink_start &&
+ found_key.objectid < shrink_last_byte) {
cur_byte = found_key.objectid;
key.objectid = cur_byte;
}
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+ if (found_key.objectid >= shrink_last_byte)
+ break;
+
if (progress && need_resched()) {
memcpy(&key, &found_key, sizeof(key));
mutex_unlock(&root->fs_info->fs_mutex);
goto again;
}
+ /*
+ * we've freed all the extents, now remove the block
+ * group item from the tree
+ */
trans = btrfs_start_transaction(root, 1);
- key.objectid = new_size;
- key.offset = 0;
- key.type = 0;
- while(1) {
- u64 ptr;
+ memcpy(&key, &shrink_block_group->key, sizeof(key));
- ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
- if (ret < 0)
- goto out;
-
- leaf = path->nodes[0];
- nritems = btrfs_header_nritems(leaf);
-bg_next:
- if (path->slots[0] >= nritems) {
- ret = btrfs_next_leaf(root, path);
- if (ret < 0)
- break;
- if (ret == 1) {
- ret = 0;
- break;
- }
- leaf = path->nodes[0];
- btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
-
- /*
- * btrfs_next_leaf doesn't cow buffers, we have to
- * do the search again
- */
- memcpy(&key, &found_key, sizeof(key));
- btrfs_release_path(root, path);
- goto resched_check;
- }
+ ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+ if (ret > 0)
+ ret = -EIO;
+ if (ret < 0)
+ goto out;
- btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
- if (btrfs_key_type(&found_key) != BTRFS_BLOCK_GROUP_ITEM_KEY) {
- printk("shrinker found key %Lu %u %Lu\n",
- found_key.objectid, found_key.type,
- found_key.offset);
- path->slots[0]++;
- goto bg_next;
- }
- ret = get_state_private(&info->block_group_cache,
- found_key.objectid, &ptr);
- if (!ret)
- kfree((void *)(unsigned long)ptr);
+ leaf = path->nodes[0];
+ nritems = btrfs_header_nritems(leaf);
+ btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
+ kfree(shrink_block_group);
- clear_extent_bits(&info->block_group_cache, found_key.objectid,
- found_key.objectid + found_key.offset - 1,
- (unsigned int)-1, GFP_NOFS);
+ clear_extent_bits(&info->block_group_cache, found_key.objectid,
+ found_key.objectid + found_key.offset - 1,
+ (unsigned int)-1, GFP_NOFS);
- key.objectid = found_key.objectid + 1;
- btrfs_del_item(trans, root, path);
- btrfs_release_path(root, path);
-resched_check:
- if (need_resched()) {
- mutex_unlock(&root->fs_info->fs_mutex);
- cond_resched();
- mutex_lock(&root->fs_info->fs_mutex);
- }
- }
- clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1,
+ btrfs_del_item(trans, root, path);
+ clear_extent_dirty(&info->free_space_cache,
+ shrink_start, shrink_last_byte - 1,
GFP_NOFS);
btrfs_commit_transaction(trans, root);
out:
return ret;
}
-int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans,
- struct btrfs_root *root, u64 new_size)
-{
- btrfs_set_super_total_bytes(&root->fs_info->super_copy, new_size);
- return 0;
-}
-
int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path,
struct btrfs_key *key)
{
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
- cache = kmalloc(sizeof(*cache), GFP_NOFS);
+ cache = kzalloc(sizeof(*cache), GFP_NOFS);
if (!cache) {
ret = -ENOMEM;
break;
btrfs_item_ptr_offset(leaf, path->slots[0]),
sizeof(cache->item));
memcpy(&cache->key, &found_key, sizeof(found_key));
- cache->cached = 0;
- cache->pinned = 0;
key.objectid = found_key.objectid + found_key.offset;
btrfs_release_path(root, path);
int btrfs_make_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytes_used,
- u64 type, u64 chunk_tree, u64 chunk_objectid,
+ u64 type, u64 chunk_objectid, u64 chunk_offset,
u64 size)
{
int ret;
extent_root = root->fs_info->extent_root;
block_group_cache = &root->fs_info->block_group_cache;
- cache = kmalloc(sizeof(*cache), GFP_NOFS);
+ cache = kzalloc(sizeof(*cache), GFP_NOFS);
BUG_ON(!cache);
- cache->key.objectid = chunk_objectid;
+ cache->key.objectid = chunk_offset;
cache->key.offset = size;
- cache->cached = 0;
- cache->pinned = 0;
+
btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY);
memset(&cache->item, 0, sizeof(cache->item));
btrfs_set_block_group_used(&cache->item, bytes_used);
- btrfs_set_block_group_chunk_tree(&cache->item, chunk_tree);
btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
cache->flags = type;
btrfs_set_block_group_flags(&cache->item, type);
&cache->space_info);
BUG_ON(ret);
- if (type & BTRFS_BLOCK_GROUP_DATA) {
- bit = BLOCK_GROUP_DATA;
- } else if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
- bit = BLOCK_GROUP_SYSTEM;
- } else if (type & BTRFS_BLOCK_GROUP_METADATA) {
- bit = BLOCK_GROUP_METADATA;
- }
- set_extent_bits(block_group_cache, chunk_objectid,
- chunk_objectid + size - 1,
+ bit = block_group_state_bits(type);
+ set_extent_bits(block_group_cache, chunk_offset,
+ chunk_offset + size - 1,
bit | EXTENT_LOCKED, GFP_NOFS);
- set_state_private(block_group_cache, chunk_objectid,
- (unsigned long)cache);
+ set_state_private(block_group_cache, chunk_offset,
+ (unsigned long)cache);
ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
sizeof(cache->item));
BUG_ON(ret);
finish_current_insert(trans, extent_root);
ret = del_pending_extents(trans, extent_root);
BUG_ON(ret);
+ set_avail_alloc_bits(extent_root->fs_info, type);
return 0;
}
projects / deliverable / linux.git / blobdiff