mballoc small file block allocation use per cpu prealloc
space. Use goal block when searching for the right prealloc
space. Also make sure ext4_da_writepages tries to write
all the pages for small files in single attempt
Without fix inodes reported by e2fsck -E fragcheck
21156 inodes used (0.47%)
180 non-contiguous inodes (0.9%)
# of inodes with ind/dind/tind blocks: 3/3/3
581215 blocks used (3.24%)
debugfs: ncheck 12 13 17 25 132 133 134
Inode Pathname
12 /bin
13 /bin/lpstat
17 /bin/xvminitoppm
25 /bin/scsiinfo
132 /bin/gpgkey2ssh
133 /bin/nwfstime
134 /bin/pbmtoln03
...
..
With fix. The fragmented inodes are all directories.
for which we don't use preallocation.
21156 inodes used (0.47%)
152 non-contiguous inodes (0.7%)
# of inodes with ind/dind/tind blocks: 6/6/6
581217 blocks used (3.24%)
debugfs: ncheck 1987 11602 7657 14279 20615
Inode Pathname
1987 /lib
11602 /lib/X11/xserver
7657 /lib/python2.5
14279 /lib/locale
20615 /lib/ooo-2.0/program
Signed-off-by: Aneesh Kumar K.V <[email protected]>
---
fs/ext4/inode.c | 21 +++++++++++++++------
fs/ext4/mballoc.c | 44 +++++++++++++++++++++++++++++++++++++-------
fs/ext4/super.c | 1 +
3 files changed, 53 insertions(+), 13 deletions(-)
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index e144896..0b34998 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -2318,13 +2318,12 @@ static int ext4_writepages_trans_blocks(struct inode *inode)
static int ext4_da_writepages(struct address_space *mapping,
struct writeback_control *wbc)
{
- struct inode *inode = mapping->host;
handle_t *handle = NULL;
- int needed_blocks;
- int ret = 0;
- long to_write;
loff_t range_start = 0;
- long pages_skipped = 0;
+ struct inode *inode = mapping->host;
+ int needed_blocks, ret = 0, nr_to_writebump = 0;
+ long to_write, pages_skipped = 0;
+ struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
/*
* No pages to write? This is mainly a kludge to avoid starting
@@ -2333,6 +2332,16 @@ static int ext4_da_writepages(struct address_space *mapping,
*/
if (!mapping->nrpages || !mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
return 0;
+ /*
+ * Make sure nr_to_write is >= sbi->s_mb_stream_request
+ * This make sure small files blocks are allocated in
+ * single attempt. This ensure that small files
+ * get less fragmented.
+ */
+ if (wbc->nr_to_write < sbi->s_mb_stream_request) {
+ nr_to_writebump = sbi->s_mb_stream_request - wbc->nr_to_write;
+ wbc->nr_to_write = sbi->s_mb_stream_request;
+ }
if (!wbc->range_cyclic)
/*
@@ -2413,7 +2422,7 @@ static int ext4_da_writepages(struct address_space *mapping,
}
out_writepages:
- wbc->nr_to_write = to_write;
+ wbc->nr_to_write = to_write - nr_to_writebump;
wbc->range_start = range_start;
return ret;
}
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index b14a7c7..1afcb11 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -3286,6 +3286,29 @@ static void ext4_mb_use_group_pa(struct ext4_allocation_context *ac,
mb_debug("use %u/%u from group pa %p\n", pa->pa_lstart-len, len, pa);
}
+static struct ext4_prealloc_space *
+ext4_mb_check_group_pa(ext4_fsblk_t goal_block,
+ struct ext4_prealloc_space *pa,
+ struct ext4_prealloc_space *cpa)
+{
+ ext4_fsblk_t cur_distance, new_distance;
+
+ if (cpa == NULL) {
+ atomic_inc(&pa->pa_count);
+ return pa;
+ }
+ cur_distance = abs(goal_block - cpa->pa_pstart);
+ new_distance = abs(goal_block - pa->pa_pstart);
+
+ if (cur_distance < new_distance)
+ return cpa;
+
+ /* drop the previous reference */
+ atomic_dec(&cpa->pa_count);
+ atomic_inc(&pa->pa_count);
+ return pa;
+}
+
/*
* search goal blocks in preallocated space
*/
@@ -3295,7 +3318,8 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
int order, i;
struct ext4_inode_info *ei = EXT4_I(ac->ac_inode);
struct ext4_locality_group *lg;
- struct ext4_prealloc_space *pa;
+ struct ext4_prealloc_space *pa, *cpa = NULL;
+ ext4_fsblk_t goal_block;
/* only data can be preallocated */
if (!(ac->ac_flags & EXT4_MB_HINT_DATA))
@@ -3338,6 +3362,10 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
/* The max size of hash table is PREALLOC_TB_SIZE */
order = PREALLOC_TB_SIZE - 1;
+ goal_block = ac->ac_g_ex.fe_group * EXT4_BLOCKS_PER_GROUP(ac->ac_sb) +
+ ac->ac_g_ex.fe_start +
+ le32_to_cpu(EXT4_SB(ac->ac_sb)->s_es->s_first_data_block);
+
for (i = order; i < PREALLOC_TB_SIZE; i++) {
rcu_read_lock();
list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[i],
@@ -3345,17 +3373,19 @@ ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
spin_lock(&pa->pa_lock);
if (pa->pa_deleted == 0 &&
pa->pa_free >= ac->ac_o_ex.fe_len) {
- atomic_inc(&pa->pa_count);
- ext4_mb_use_group_pa(ac, pa);
- spin_unlock(&pa->pa_lock);
- ac->ac_criteria = 20;
- rcu_read_unlock();
- return 1;
+
+ cpa = ext4_mb_check_group_pa(goal_block,
+ pa, cpa);
}
spin_unlock(&pa->pa_lock);
}
rcu_read_unlock();
}
+ if (cpa) {
+ ext4_mb_use_group_pa(ac, cpa);
+ ac->ac_criteria = 20;
+ return 1;
+ }
return 0;
}
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index a23e4d3..ed77786 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -568,6 +568,7 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
#endif
ei->i_block_alloc_info = NULL;
ei->vfs_inode.i_version = 1;
+ ei->vfs_inode.i_data.writeback_index = 0;
memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
INIT_LIST_HEAD(&ei->i_prealloc_list);
spin_lock_init(&ei->i_prealloc_lock);
--
1.6.0.rc0.42.g186458.dirty
I've tested v3 version of your patch, and it fixes it for me as well.
Using the "compilebench -i 10 -r 0" test, the number of fragmented
files went from 16352 (7.3%) to 32 (0.0%), and the bandwidth went up
by a tiny amount (32.93 MB/s to 33.12 MB/s) --- not enough for me to
be entirely sure that it's outside of measurement error, but I'll take
it.
I've added it to the patch queue. It's behind mingming's DIO credit
patches, and your ext4_da_writepages rework, since the patch seems to
depend (at least syntatically) on the other patches. I assume our
best strategy is to try to get those patches cleaned up and moved to
stable, so we can push the whole lot to Linux within the next few
days.
Thanks for working on this problem!
- Ted