[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20190603100545.GC27933@quack2.suse.cz>
Date: Mon, 3 Jun 2019 12:05:45 +0200
From: Jan Kara <jack@...e.cz>
To: Chengguang Xu <cgxu519@...o.com.cn>
Cc: jack@...e.com, linux-ext4@...r.kernel.org
Subject: Re: [PATCH] ext2: start from parent blockgroup when trying linear
search for a free inode
On Sat 01-06-19 16:49:41, Chengguang Xu wrote:
> Start from parent blockgroup when trying linear search
> for a free indoe because for non directory inode it's
> better to keep in same blockgroup with parent.
>
> Signed-off-by: Chengguang Xu <cgxu519@...o.com.cn>
Thanks for the patch! As much as I agree that incrementing group before
calling ext2_get_group_desc() was probably an oversight I don't think that
changing that at this point is good. For ext2 driver we don't really care
about performance too much anymore and also existing filesystems have
inodes for files in a directory allocated in parent_group + 1 and this
change would spread them also to parent_group effectively worsening the
situation.
Honza
> ---
> fs/ext2/ialloc.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/fs/ext2/ialloc.c b/fs/ext2/ialloc.c
> index fda7d3f5b4be..435463a255e6 100644
> --- a/fs/ext2/ialloc.c
> +++ b/fs/ext2/ialloc.c
> @@ -411,11 +411,11 @@ static int find_group_other(struct super_block *sb, struct inode *parent)
> */
> group = parent_group;
> for (i = 0; i < ngroups; i++) {
> - if (++group >= ngroups)
> - group = 0;
> desc = ext2_get_group_desc (sb, group, NULL);
> if (desc && le16_to_cpu(desc->bg_free_inodes_count))
> goto found;
> + if (++group >= ngroups)
> + group = 0;
> }
>
> return -1;
> --
> 2.17.2
>
>
>
--
Jan Kara <jack@...e.com>
SUSE Labs, CR
Powered by blists - more mailing lists