Re: [PATCH 2/2] ext4: optimize ext4_force_commit

[Theodore Ts'o <tytso@mit.edu>]

On Mon, Apr 01, 2013 at 01:03:57PM +0400, Dmitry Monakhov wrote:
> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
> index e3e6a06..280a918 100644
> --- a/fs/ext4/super.c
> +++ b/fs/ext4/super.c
> @@ -4436,13 +4436,14 @@ static void ext4_clear_journal_err(struct super_block *sb,
>   */
>  int ext4_force_commit(struct super_block *sb)
>  {
> -	journal_t *journal;
> +	tid_t target;
>  
>  	if (sb->s_flags & MS_RDONLY)
>  		return 0;
>  
> -	journal = EXT4_SB(sb)->s_journal;
> -	return ext4_journal_force_commit(journal);
> +	if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target))
> +		return jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);

EXT4_SB(sb)->s_journal is NULL in no-journal mode.  So you need to
check for this --- this is something which is done in
ext4_journal_force_commit().

Since this is the only user of ext4_journal_force_commit(), you might
as well get rid of ext4_journal_force_commit() as part of making this
change.

The other possibility is we perhaps we should just change
jbd2_journal_force_commit() to call jbd2_journal_start_commit() and
jbd2_log_wait_commit(); the only other caller of
jbd2_journal_force_commit() is ocfs2_sync_file(), and it would benefit
from this change as well.

					- Ted