[PATCH] acpi-cpufreq: Skip cleanup if initialization didn't occur

[PATCH] acpi-cpufreq: Skip cleanup if initialization didn't occur

[kyle.meyer]

From: Kyle Meyer <kyle.meyer@hpe.com>

acpi-cpufreq is loaded without performing initialization when a cpufreq
driver exists.

If initialization didn't occur then skip cleanup in acpi_cpufreq_exit().
This prevents unnecessary freeing and unregistering when the module is
unloaded.

Reported-by: Takashi Iwai <tiwai@suse.com>
Signed-off-by: Kyle Meyer <kyle.meyer@hpe.com>
---
 drivers/cpufreq/acpi-cpufreq.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/drivers/cpufreq/acpi-cpufreq.c b/drivers/cpufreq/acpi-cpufreq.c
index 7e7450453714..8d425f14c267 100644
--- a/drivers/cpufreq/acpi-cpufreq.c
+++ b/drivers/cpufreq/acpi-cpufreq.c
@@ -1042,8 +1042,19 @@ static int __init acpi_cpufreq_init(void)
 
 static void __exit acpi_cpufreq_exit(void)
 {
+	const char *current_driver;
+
 	pr_debug("%s\n", __func__);
 
+	/*
+	 * If another cpufreq_driver was loaded, preventing acpi-cpufreq from
+	 * registering, there's no need to unregister it.
+	 */
+	current_driver = cpufreq_get_current_driver();
+	if (!current_driver ||
+	    strncmp(current_driver, acpi_cpufreq_driver.name, strlen(acpi_cpufreq_driver.name)))
+		return;
+
 	acpi_cpufreq_boost_exit();
 
 	cpufreq_unregister_driver(&acpi_cpufreq_driver);
-- 
2.25.1

Re: [PATCH] acpi-cpufreq: Skip cleanup if initialization didn't occur

[Viresh Kumar]

On 04-06-21, 12:05, kyle.meyer@hpe.com wrote:
> From: Kyle Meyer <kyle.meyer@hpe.com>
> 
> acpi-cpufreq is loaded without performing initialization when a cpufreq
> driver exists.
> 
> If initialization didn't occur then skip cleanup in acpi_cpufreq_exit().
> This prevents unnecessary freeing and unregistering when the module is
> unloaded.
> 
> Reported-by: Takashi Iwai <tiwai@suse.com>
> Signed-off-by: Kyle Meyer <kyle.meyer@hpe.com>
> ---
>  drivers/cpufreq/acpi-cpufreq.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/drivers/cpufreq/acpi-cpufreq.c b/drivers/cpufreq/acpi-cpufreq.c
> index 7e7450453714..8d425f14c267 100644
> --- a/drivers/cpufreq/acpi-cpufreq.c
> +++ b/drivers/cpufreq/acpi-cpufreq.c
> @@ -1042,8 +1042,19 @@ static int __init acpi_cpufreq_init(void)
>  
>  static void __exit acpi_cpufreq_exit(void)
>  {
> +	const char *current_driver;
> +
>  	pr_debug("%s\n", __func__);
>  
> +	/*
> +	 * If another cpufreq_driver was loaded, preventing acpi-cpufreq from
> +	 * registering, there's no need to unregister it.
> +	 */
> +	current_driver = cpufreq_get_current_driver();
> +	if (!current_driver ||
> +	    strncmp(current_driver, acpi_cpufreq_driver.name, strlen(acpi_cpufreq_driver.name)))
> +		return;
> +
>  	acpi_cpufreq_boost_exit();
>  
>  	cpufreq_unregister_driver(&acpi_cpufreq_driver);

Looks like some misunderstanding here, this shouldn't happen. If
initialization didn't occur, then exit shall never be called.

-- 
viresh

Re: [PATCH] acpi-cpufreq: Skip cleanup if initialization didn't occur

[Takashi Iwai]

On Mon, 07 Jun 2021 05:25:50 +0200,
Viresh Kumar wrote:
> 
> On 04-06-21, 12:05, kyle.meyer@hpe.com wrote:
> > From: Kyle Meyer <kyle.meyer@hpe.com>
> > 
> > acpi-cpufreq is loaded without performing initialization when a cpufreq
> > driver exists.
> > 
> > If initialization didn't occur then skip cleanup in acpi_cpufreq_exit().
> > This prevents unnecessary freeing and unregistering when the module is
> > unloaded.
> > 
> > Reported-by: Takashi Iwai <tiwai@suse.com>
> > Signed-off-by: Kyle Meyer <kyle.meyer@hpe.com>
> > ---
> >  drivers/cpufreq/acpi-cpufreq.c | 11 +++++++++++
> >  1 file changed, 11 insertions(+)
> > 
> > diff --git a/drivers/cpufreq/acpi-cpufreq.c b/drivers/cpufreq/acpi-cpufreq.c
> > index 7e7450453714..8d425f14c267 100644
> > --- a/drivers/cpufreq/acpi-cpufreq.c
> > +++ b/drivers/cpufreq/acpi-cpufreq.c
> > @@ -1042,8 +1042,19 @@ static int __init acpi_cpufreq_init(void)
> >  
> >  static void __exit acpi_cpufreq_exit(void)
> >  {
> > +	const char *current_driver;
> > +
> >  	pr_debug("%s\n", __func__);
> >  
> > +	/*
> > +	 * If another cpufreq_driver was loaded, preventing acpi-cpufreq from
> > +	 * registering, there's no need to unregister it.
> > +	 */
> > +	current_driver = cpufreq_get_current_driver();
> > +	if (!current_driver ||
> > +	    strncmp(current_driver, acpi_cpufreq_driver.name, strlen(acpi_cpufreq_driver.name)))
> > +		return;
> > +
> >  	acpi_cpufreq_boost_exit();
> >  
> >  	cpufreq_unregister_driver(&acpi_cpufreq_driver);
> 
> Looks like some misunderstanding here, this shouldn't happen. If
> initialization didn't occur, then exit shall never be called.

The missing key information is that it's a fix for the recent change
for 5.14, i.e. 
Fixes: c1d6d2fd2f64 ("cpufreq: acpi-cpufreq: Skip initialization if cpufreq driver is present")

The change made the module left even if it exits before registering
the cpufreq driver object.


Takashi

Re: [PATCH] acpi-cpufreq: Skip cleanup if initialization didn't occur

[Viresh Kumar]

On 07-06-21, 09:13, Takashi Iwai wrote:
> The missing key information is that it's a fix for the recent change
> for 5.14, i.e. 
> Fixes: c1d6d2fd2f64 ("cpufreq: acpi-cpufreq: Skip initialization if cpufreq driver is present")
> 
> The change made the module left even if it exits before registering
> the cpufreq driver object.

The original patch looks buggy to me, I was never able to review it :(

I have replied on the original thread instead.

-- 
viresh

Re: [PATCH] acpi-cpufreq: Skip cleanup if initialization didn't occur

[Rafael J. Wysocki]

On Mon, Jun 7, 2021 at 9:26 AM Viresh Kumar <viresh.kumar@linaro.org> wrote:
>
> On 07-06-21, 09:13, Takashi Iwai wrote:
> > The missing key information is that it's a fix for the recent change
> > for 5.14, i.e.
> > Fixes: c1d6d2fd2f64 ("cpufreq: acpi-cpufreq: Skip initialization if cpufreq driver is present")
> >
> > The change made the module left even if it exits before registering
> > the cpufreq driver object.
>
> The original patch looks buggy to me, I was never able to review it :(
>
> I have replied on the original thread instead.

Well, thanks, but that confused me a bit.

Given the above, I'm going to drop the original patch.

Kyle, please reconsider this approach.