From: Miao Xie <miaox@cn.fujitsu.com>
To: linux-btrfs@vger.kernel.org
Subject: [PATCH] Btrfs: allocate the free space by the existed max extent size when ENOSPC
Date: Thu, 29 Aug 2013 13:47:38 +0800 [thread overview]
Message-ID: <1377755258-2722-1-git-send-email-miaox@cn.fujitsu.com> (raw)
By the current code, if the requested size is very large, and all the extents
in the free space cache are small, we will waste lots of the cpu time to cut
the requested size in half and search the cache again and again until it gets
down to the size the allocator can return. In fact, we can know the max extent
size in the cache after the first search, so we needn't cut the size in half
repeatedly, and just use the max extent size directly. This way can save
lots of cpu time and make the performance grow up when there are only fragments
in the free space cache.
According to my test, if there are only 4KB free space extents in the fs,
and the total size of those extents are 256MB, we can reduce the execute
time of the following test from 5.4s to 1.4s.
dd if=/dev/zero of=<testfile> bs=1MB count=1 oflag=sync
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
fs/btrfs/extent-tree.c | 29 ++++++++++++++-------
fs/btrfs/free-space-cache.c | 61 +++++++++++++++++++++++++++++++--------------
fs/btrfs/free-space-cache.h | 5 ++--
3 files changed, 65 insertions(+), 30 deletions(-)
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 0236de7..d634dbc 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -6065,10 +6065,13 @@ enum btrfs_loop_type {
/*
* walks the btree of allocated extents and find a hole of a given size.
* The key ins is changed to record the hole:
- * ins->objectid == block start
+ * ins->objectid == start position
* ins->flags = BTRFS_EXTENT_ITEM_KEY
- * ins->offset == number of blocks
+ * ins->offset == the size of the hole.
* Any available blocks before search_start are skipped.
+ *
+ * If there is no suitable free space, we will record the max size of
+ * the free space extent currently.
*/
static noinline int find_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *orig_root,
@@ -6082,6 +6085,7 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_block_group_cache *block_group = NULL;
struct btrfs_block_group_cache *used_block_group;
u64 search_start = 0;
+ u64 max_extent_size = 0;
int empty_cluster = 2 * 1024 * 1024;
struct btrfs_space_info *space_info;
int loop = 0;
@@ -6239,7 +6243,10 @@ have_block_group:
btrfs_get_block_group(used_block_group);
offset = btrfs_alloc_from_cluster(used_block_group,
- last_ptr, num_bytes, used_block_group->key.objectid);
+ last_ptr,
+ num_bytes,
+ used_block_group->key.objectid,
+ &max_extent_size);
if (offset) {
/* we have a block, we're done */
spin_unlock(&last_ptr->refill_lock);
@@ -6302,8 +6309,10 @@ refill_cluster:
* cluster
*/
offset = btrfs_alloc_from_cluster(block_group,
- last_ptr, num_bytes,
- search_start);
+ last_ptr,
+ num_bytes,
+ search_start,
+ &max_extent_size);
if (offset) {
/* we found one, proceed */
spin_unlock(&last_ptr->refill_lock);
@@ -6344,7 +6353,8 @@ unclustered_alloc:
spin_unlock(&block_group->free_space_ctl->tree_lock);
offset = btrfs_find_space_for_alloc(block_group, search_start,
- num_bytes, empty_size);
+ num_bytes, empty_size,
+ &max_extent_size);
/*
* If we didn't find a chunk, and we haven't failed on this
* block group before, and this block group is in the middle of
@@ -6451,7 +6461,8 @@ loop:
ret = 0;
}
out:
-
+ if (ret == -ENOSPC)
+ ins->offset = max_extent_size;
return ret;
}
@@ -6517,8 +6528,8 @@ again:
hint_byte, ins, flags);
if (ret == -ENOSPC) {
- if (!final_tried) {
- num_bytes = num_bytes >> 1;
+ if (!final_tried && ins->offset) {
+ num_bytes = min(num_bytes >> 1, ins->offset);
num_bytes = round_down(num_bytes, root->sectorsize);
num_bytes = max(num_bytes, min_alloc_size);
if (num_bytes == min_alloc_size)
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 7517285..8498355 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -1434,11 +1434,16 @@ static void bitmap_set_bits(struct btrfs_free_space_ctl *ctl,
ctl->free_space += bytes;
}
+/*
+ * If we can not find suitable extent, we will use bytes to record
+ * the size of the max extent.
+ */
static int search_bitmap(struct btrfs_free_space_ctl *ctl,
struct btrfs_free_space *bitmap_info, u64 *offset,
u64 *bytes)
{
unsigned long found_bits = 0;
+ unsigned long max_bits = 0;
unsigned long bits, i;
unsigned long next_zero;
@@ -1449,9 +1454,12 @@ static int search_bitmap(struct btrfs_free_space_ctl *ctl,
for_each_set_bit_from(i, bitmap_info->bitmap, BITS_PER_BITMAP) {
next_zero = find_next_zero_bit(bitmap_info->bitmap,
BITS_PER_BITMAP, i);
- if ((next_zero - i) >= bits) {
- found_bits = next_zero - i;
+ i = next_zero - i;
+ if (i >= bits) {
+ found_bits = i;
break;
+ } else if (i > max_bits) {
+ max_bits = i;
}
i = next_zero;
}
@@ -1462,38 +1470,41 @@ static int search_bitmap(struct btrfs_free_space_ctl *ctl,
return 0;
}
+ *bytes = (u64)(max_bits) * ctl->unit;
return -1;
}
+/* Cache the size of the max extent in bytes */
static struct btrfs_free_space *
find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
- unsigned long align)
+ unsigned long align, u64 *max_extent_size)
{
struct btrfs_free_space *entry;
struct rb_node *node;
- u64 ctl_off;
u64 tmp;
u64 align_off;
int ret;
if (!ctl->free_space_offset.rb_node)
- return NULL;
+ goto out;
entry = tree_search_offset(ctl, offset_to_bitmap(ctl, *offset), 0, 1);
if (!entry)
- return NULL;
+ goto out;
for (node = &entry->offset_index; node; node = rb_next(node)) {
entry = rb_entry(node, struct btrfs_free_space, offset_index);
- if (entry->bytes < *bytes)
+ if (entry->bytes < *bytes) {
+ if (entry->bytes > *max_extent_size)
+ *max_extent_size = entry->bytes;
continue;
+ }
/* make sure the space returned is big enough
* to match our requested alignment
*/
if (*bytes >= align) {
- ctl_off = entry->offset - ctl->start;
- tmp = ctl_off + align - 1;;
+ tmp = entry->offset - ctl->start + align - 1;
do_div(tmp, align);
tmp = tmp * align + ctl->start;
align_off = tmp - entry->offset;
@@ -1506,10 +1517,15 @@ find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
continue;
if (entry->bitmap) {
- ret = search_bitmap(ctl, entry, &tmp, bytes);
+ u64 size = *bytes;
+
+ ret = search_bitmap(ctl, entry, &tmp, &size);
if (!ret) {
*offset = tmp;
+ *bytes = size;
return entry;
+ } else if (size > *max_extent_size) {
+ *max_extent_size = size;
}
continue;
}
@@ -1518,7 +1534,7 @@ find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
*bytes = entry->bytes - align_off;
return entry;
}
-
+out:
return NULL;
}
@@ -2120,7 +2136,8 @@ void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
}
u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes, u64 empty_size)
+ u64 offset, u64 bytes, u64 empty_size,
+ u64 *max_extent_size)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_free_space *entry = NULL;
@@ -2131,7 +2148,7 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
spin_lock(&ctl->tree_lock);
entry = find_free_space(ctl, &offset, &bytes_search,
- block_group->full_stripe_len);
+ block_group->full_stripe_len, max_extent_size);
if (!entry)
goto out;
@@ -2141,7 +2158,6 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
if (!entry->bytes)
free_bitmap(ctl, entry);
} else {
-
unlink_free_space(ctl, entry);
align_gap_len = offset - entry->offset;
align_gap = entry->offset;
@@ -2155,7 +2171,6 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
else
link_free_space(ctl, entry);
}
-
out:
spin_unlock(&ctl->tree_lock);
@@ -2210,7 +2225,8 @@ int btrfs_return_cluster_to_free_space(
static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
struct btrfs_free_cluster *cluster,
struct btrfs_free_space *entry,
- u64 bytes, u64 min_start)
+ u64 bytes, u64 min_start,
+ u64 *max_extent_size)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
int err;
@@ -2222,8 +2238,11 @@ static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
search_bytes = bytes;
err = search_bitmap(ctl, entry, &search_start, &search_bytes);
- if (err)
+ if (err) {
+ if (search_bytes > *max_extent_size)
+ *max_extent_size = search_bytes;
return 0;
+ }
ret = search_start;
__bitmap_clear_bits(ctl, entry, ret, bytes);
@@ -2238,7 +2257,7 @@ static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
*/
u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
struct btrfs_free_cluster *cluster, u64 bytes,
- u64 min_start)
+ u64 min_start, u64 *max_extent_size)
{
struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
struct btrfs_free_space *entry = NULL;
@@ -2258,6 +2277,9 @@ u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
entry = rb_entry(node, struct btrfs_free_space, offset_index);
while(1) {
+ if (entry->bytes < bytes && entry->bytes > *max_extent_size)
+ *max_extent_size = entry->bytes;
+
if (entry->bytes < bytes ||
(!entry->bitmap && entry->offset < min_start)) {
node = rb_next(&entry->offset_index);
@@ -2271,7 +2293,8 @@ u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
if (entry->bitmap) {
ret = btrfs_alloc_from_bitmap(block_group,
cluster, entry, bytes,
- cluster->window_start);
+ cluster->window_start,
+ max_extent_size);
if (ret == 0) {
node = rb_next(&entry->offset_index);
if (!node)
diff --git a/fs/btrfs/free-space-cache.h b/fs/btrfs/free-space-cache.h
index 894116b..a701000 100644
--- a/fs/btrfs/free-space-cache.h
+++ b/fs/btrfs/free-space-cache.h
@@ -94,7 +94,8 @@ void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl);
void btrfs_remove_free_space_cache(struct btrfs_block_group_cache
*block_group);
u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
- u64 offset, u64 bytes, u64 empty_size);
+ u64 offset, u64 bytes, u64 empty_size,
+ u64 *max_extent_size);
u64 btrfs_find_ino_for_alloc(struct btrfs_root *fs_root);
void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
u64 bytes);
@@ -106,7 +107,7 @@ int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster);
u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
struct btrfs_free_cluster *cluster, u64 bytes,
- u64 min_start);
+ u64 min_start, u64 *max_extent_size);
int btrfs_return_cluster_to_free_space(
struct btrfs_block_group_cache *block_group,
struct btrfs_free_cluster *cluster);
--
1.8.3.1
next reply other threads:[~2013-08-29 5:46 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-08-29 5:47 Miao Xie [this message]
2013-08-29 12:45 ` [PATCH] Btrfs: allocate the free space by the existed max extent size when ENOSPC David Sterba
2013-08-30 10:58 ` Miao Xie
2013-08-29 19:34 ` Josef Bacik
-- strict thread matches above, loose matches on Subject: below --
2013-08-30 10:35 Miao Xie
2013-09-06 13:47 ` Stefan Behrens
2013-09-09 6:21 ` Miao Xie
2013-09-09 9:06 ` Stefan Behrens
2013-09-17 13:13 ` David Sterba
2013-09-18 4:04 ` Miao Xie
2013-09-20 9:25 ` David Sterba
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1377755258-2722-1-git-send-email-miaox@cn.fujitsu.com \
--to=miaox@cn.fujitsu.com \
--cc=linux-btrfs@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).