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;
MIME-Version: 1.0
References: <20221230075159.1482626-1-linmq006@gmail.com>
In-Reply-To: <20221230075159.1482626-1-linmq006@gmail.com>
From:   "Rafael J. Wysocki" <rafael@kernel.org>
Date:   Fri, 20 Jan 2023 14:43:57 +0100
Message-ID: <CAJZ5v0iPMyQUibJk5S=ZdnsU43yBqyDAVCFkbmmwd_LVZjR=5w@mail.gmail.com>
Subject: Re: [PATCH] powercap/dtpm_cpu: Fix refcount leak in __dtpm_cpu_setup
To:     Miaoqian Lin <linmq006@gmail.com>
Cc:     Daniel Lezcano <daniel.lezcano@kernel.org>,
        "Rafael J. Wysocki" <rafael@kernel.org>,
        Lukasz Luba <lukasz.luba@arm.com>, linux-pm@vger.kernel.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Dec 30, 2022 at 8:52 AM Miaoqian Lin <linmq006@gmail.com> wrote:
>
> The policy return by cpufreq_cpu_get() should be released with
> cpufreq_cpu_put() to balance the reference counter.
> Add missing cpufreq_cpu_put() to fix this.
>
> Fixes: 0e8f68d7f048 ("powercap/drivers/dtpm: Add CPU energy model based support")
> Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
> ---
>  drivers/powercap/dtpm_cpu.c | 17 ++++++++++++-----
>  1 file changed, 12 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/powercap/dtpm_cpu.c b/drivers/powercap/dtpm_cpu.c
> index 2ff7717530bf..3083c6b45c90 100644
> --- a/drivers/powercap/dtpm_cpu.c
> +++ b/drivers/powercap/dtpm_cpu.c
> @@ -195,12 +195,16 @@ static int __dtpm_cpu_setup(int cpu, struct dtpm *parent)
>                 return 0;
>
>         pd = em_cpu_get(cpu);
> -       if (!pd || em_is_artificial(pd))
> -               return -EINVAL;
> +       if (!pd || em_is_artificial(pd)) {
> +               ret = -EINVAL;
> +               goto out_put_policy;
> +       }
>
>         dtpm_cpu = kzalloc(sizeof(*dtpm_cpu), GFP_KERNEL);
> -       if (!dtpm_cpu)
> -               return -ENOMEM;
> +       if (!dtpm_cpu) {
> +               ret = -ENOMEM;
> +               goto out_put_policy;
> +       }
>
>         dtpm_init(&dtpm_cpu->dtpm, &dtpm_ops);
>         dtpm_cpu->cpu = cpu;
> @@ -220,6 +224,8 @@ static int __dtpm_cpu_setup(int cpu, struct dtpm *parent)
>         if (ret)
>                 goto out_dtpm_unregister;

The part of the patch above is valid, but I don't think that the
policy reference counter can be dropped below, because that may allow
the policy to go away and this driver will be still using it, won't
it?

>
> +       cpufreq_cpu_put(policy);
> +
>         return 0;
>
>  out_dtpm_unregister:
> @@ -230,7 +236,8 @@ static int __dtpm_cpu_setup(int cpu, struct dtpm *parent)
>         for_each_cpu(cpu, policy->related_cpus)
>                 per_cpu(dtpm_per_cpu, cpu) = NULL;
>         kfree(dtpm_cpu);
> -
> +out_put_policy:
> +       cpufreq_cpu_put(policy);
>         return ret;
>  }
>
> --
> 2.25.1
>