Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Sat, 5 Dec 2020 14:32:50 -0800
From:   Jakub Kicinski <kuba@kernel.org>
To:     linux-sparse@vger.kernel.org
Cc:     linux-kernel@vger.kernel.org, edwin.peer@broadcom.com,
        Zhang Changzhong <zhangchangzhong@huawei.com>
Subject: sparse annotation for error types?
Message-ID: <20201205143250.2378b9f9@kicinski-fedora-pc1c0hjn.DHCP.thefacebook.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi!

Recently we've been getting a steady stream of patches from Changzhong
to fix missing assignment to error variables before jumping to error
cases.

I wonder if for new code it'd make sense to add an annotation for a type
which has to be returned non-zero?

What I have in mind is the following common flow:

int do_a_thing(struct my_obj *obj, int param)
{
	int err;

	err = first_step(obj, 1);
	if (err)
		return err;

	if (some_check(obj)) {
		err = -EINVAL; /* need explicit error set! */
		goto err_undo_1s;
	}

	err = second_step(obj, param);
	if (err)
		goto err_undo_1s;

	err = third_step(obj, 0);
	if (err)
		goto err_undo_2s;

	return 0;

err_undo_2s:
	second_undo(obj);
err_undo_1s:
	first_undo(obj);
	return err;
}


The variable err should never be returned when it's equal to 0.
So if we annotate it, let's say as:

	int __nzret err;

could sparse then warn if we forgot to assign it after
"if (some_check(obj))"? 

Am I the only one who thinks this would be a good idea?