Thread (3 messages) 3 messages, 2 authors, 2021-12-30

Re: drivers/acpi/processor_thermal.c: avoid cpufreq_get_policy()

From: "Rafael J. Wysocki" <rafael@kernel.org>
Date: 2021-12-30 16:23:26
Also in: lkml

On Wed, Dec 22, 2021 at 8:14 PM Manfred Spraul [off-list ref] wrote:
quoted hunk ↗ jump to hunk
cpu_has_cpufreq() stores a 'struct cpufreq_policy' on the stack.
Unfortunately, with debugging options enabled, the structure can be
larger than 1024 bytes, which causes a compiler warning/error.

(actually observed: 1184 bytes).

Therefore: Switch to cpufreq_cpu_get().

Signed-off-by: Manfred Spraul <redacted>

---
 drivers/acpi/processor_thermal.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)
diff --git a/drivers/acpi/processor_thermal.c b/drivers/acpi/processor_thermal.c
index a3d34e3f9f94..74210d63f62c 100644
--- a/drivers/acpi/processor_thermal.c
+++ b/drivers/acpi/processor_thermal.c
@@ -53,10 +53,19 @@ static int phys_package_first_cpu(int cpu)

 static int cpu_has_cpufreq(unsigned int cpu)
 {
-       struct cpufreq_policy policy;
-       if (!acpi_processor_cpufreq_init || cpufreq_get_policy(&policy, cpu))
+       struct cpufreq_policy *policy;
+       int retval;
Why is this needed?
+
+       if (!acpi_processor_cpufreq_init)
                return 0;
-       return 1;
+
+       retval = 0;
+       policy = cpufreq_cpu_get(cpu);
+       if (policy) {
+               cpufreq_cpu_put(policy);
return 1;
+               retval = 1;
+       }
+       return retval;
return 0;
 }

 static int cpufreq_get_max_state(unsigned int cpu)
--
2.33.1
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help