If cpufreq_register_driver() fails we would free the acpi driver
related structures but not free the ones allocated
by acpi_cpufreq_boost_init() function. This meant that as
the driver error-ed out and a CPU online/offline event came
we would crash and burn as one of the CPU notifiers would point
to garbage.
This fixes a regression that commit cfc9c8ed03e4d908f2388af8815f44c87b503aaf
"acpi-cpufreq: Adjust the code to use the common boost attribute"
introduced.
CC: Lukasz Majewski <[email protected]>
CC: Myungjoo Ham <[email protected]>
CC: Viresh Kumar <[email protected]>
CC: Rafael J. Wysocki <[email protected]>
CC: Boris Ostrovsky <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
---
drivers/cpufreq/acpi-cpufreq.c | 5 +++--
1 files changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/cpufreq/acpi-cpufreq.c b/drivers/cpufreq/acpi-cpufreq.c
index 79e5608..3e856d6 100644
--- a/drivers/cpufreq/acpi-cpufreq.c
+++ b/drivers/cpufreq/acpi-cpufreq.c
@@ -969,9 +969,10 @@ static int __init acpi_cpufreq_init(void)
acpi_cpufreq_boost_init();
ret = cpufreq_register_driver(&acpi_cpufreq_driver);
- if (ret)
+ if (ret) {
free_acpi_perf_data();
-
+ acpi_cpufreq_boost_exit();
+ }
return ret;
}
--
1.7.7.6
Hi Konrad,
> If cpufreq_register_driver() fails we would free the acpi driver
> related structures but not free the ones allocated
> by acpi_cpufreq_boost_init() function. This meant that as
> the driver error-ed out and a CPU online/offline event came
> we would crash and burn as one of the CPU notifiers would point
> to garbage.
>
Thanks for spotting this bug. My bad.
> This fixes a regression that commit
> cfc9c8ed03e4d908f2388af8815f44c87b503aaf "acpi-cpufreq: Adjust the
> code to use the common boost attribute" introduced.
>
> CC: Lukasz Majewski <[email protected]>
> CC: Myungjoo Ham <[email protected]>
> CC: Viresh Kumar <[email protected]>
> CC: Rafael J. Wysocki <[email protected]>
> CC: Boris Ostrovsky <[email protected]>
> Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
> ---
> drivers/cpufreq/acpi-cpufreq.c | 5 +++--
> 1 files changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/cpufreq/acpi-cpufreq.c
> b/drivers/cpufreq/acpi-cpufreq.c index 79e5608..3e856d6 100644
> --- a/drivers/cpufreq/acpi-cpufreq.c
> +++ b/drivers/cpufreq/acpi-cpufreq.c
> @@ -969,9 +969,10 @@ static int __init acpi_cpufreq_init(void)
> acpi_cpufreq_boost_init();
>
> ret = cpufreq_register_driver(&acpi_cpufreq_driver);
> - if (ret)
> + if (ret) {
> free_acpi_perf_data();
> -
> + acpi_cpufreq_boost_exit();
> + }
> return ret;
> }
>
Acked-by: Lukasz Majewski <[email protected]>
--
Best regards,
Lukasz Majewski
Samsung R&D Institute Poland (SRPOL) | Linux Platform Group
On 28 January 2014 09:28, Konrad Rzeszutek Wilk <[email protected]> wrote:
> If cpufreq_register_driver() fails we would free the acpi driver
> related structures but not free the ones allocated
> by acpi_cpufreq_boost_init() function. This meant that as
> the driver error-ed out and a CPU online/offline event came
> we would crash and burn as one of the CPU notifiers would point
> to garbage.
>
> This fixes a regression that commit cfc9c8ed03e4d908f2388af8815f44c87b503aaf
> "acpi-cpufreq: Adjust the code to use the common boost attribute"
> introduced.
>
> CC: Lukasz Majewski <[email protected]>
> CC: Myungjoo Ham <[email protected]>
> CC: Viresh Kumar <[email protected]>
> CC: Rafael J. Wysocki <[email protected]>
> CC: Boris Ostrovsky <[email protected]>
> Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
> ---
> drivers/cpufreq/acpi-cpufreq.c | 5 +++--
> 1 files changed, 3 insertions(+), 2 deletions(-)
Acked-by: Viresh Kumar <[email protected]>