quota: Remove locking for writing to the old quota format

The old quota quota format has fixed offset in quota file based on ID so
there's no locking needed against concurrent modifications of the file
(locking against concurrent IO on the same dquot is still provided by
dq_lock).
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Jan Kara <jack@suse.cz>

fs/quota/quota_v1.c

@@ -83,9 +83,7 @@ static int v1_commit_dqblk(struct dquot *dquot)
 	short type = dquot->dq_id.type;
 	ssize_t ret;
 	struct v1_disk_dqblk dqblk;
-	struct quota_info *dqopt = sb_dqopt(dquot->dq_sb);
 
-	down_write(&dqopt->dqio_sem);
 	v1_mem2disk_dqblk(&dqblk, &dquot->dq_dqb);
 	if (((type == USRQUOTA) && uid_eq(dquot->dq_id.uid, GLOBAL_ROOT_UID)) ||
 	    ((type == GRPQUOTA) && gid_eq(dquot->dq_id.gid, GLOBAL_ROOT_GID))) {
@@ -99,7 +97,6 @@ static int v1_commit_dqblk(struct dquot *dquot)
 		ret = dquot->dq_sb->s_op->quota_write(dquot->dq_sb, type,
 			(char *)&dqblk, sizeof(struct v1_disk_dqblk),
 			v1_dqoff(from_kqid(&init_user_ns, dquot->dq_id)));
-	up_write(&dqopt->dqio_sem);
 	if (ret != sizeof(struct v1_disk_dqblk)) {
 		quota_error(dquot->dq_sb, "dquota write failed");
 		if (ret >= 0)