Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Michael Ellerman <mpe@ellerman.id.au>
To:     Hangyu Hua <hbh25y@gmail.com>, benh@kernel.crashing.org,
        paulus@samba.org
Cc:     linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org,
        Hangyu Hua <hbh25y@gmail.com>
Subject: Re: [PATCH] powerpc: kernel: fix a refcount leak in format_show()
In-Reply-To: <20220228091103.39749-1-hbh25y@gmail.com>
References: <20220228091103.39749-1-hbh25y@gmail.com>
Date:   Tue, 01 Mar 2022 23:55:35 +1100
Message-ID: <87o82pomrc.fsf@mpe.ellerman.id.au>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Hangyu Hua <hbh25y@gmail.com> writes:
> node needs to be dropped when of_property_read_string fails. So an earlier call
> to of_node_put is required here.

That's true but ...

> diff --git a/arch/powerpc/kernel/secvar-sysfs.c b/arch/powerpc/kernel/secvar-sysfs.c
> index a0a78aba2083..cd0fa7028d86 100644
> --- a/arch/powerpc/kernel/secvar-sysfs.c
> +++ b/arch/powerpc/kernel/secvar-sysfs.c
> @@ -30,13 +30,12 @@ static ssize_t format_show(struct kobject *kobj, struct kobj_attribute *attr,
>  		return -ENODEV;

There's also a reference leak there ^

So if you're going to touch this code I'd like you to fix both reference
leaks in a single patch please.

Having the error cases set rc and then goto "out" which does the
of_node_put() is the obvious solution I think.

cheers

>  	rc = of_property_read_string(node, "format", &format);
> +	of_node_put(node);
>  	if (rc)
>  		return rc;
>  
>  	rc = sprintf(buf, "%s\n", format);
>  
> -	of_node_put(node);
> -
>  	return rc;
>  }
>  
> -- 
> 2.25.1