Please ignore this email thread. I'll re-create a new email thread based on Rafael's suggestion.
-- Adrian
-----Original Message-----
From: Huang, Adrian (ISS Linux TW)
Sent: Monday, October 21, 2013 12:18 AM
To: 'Rafael J. Wysocki'
Cc: Viresh Kumar; cpufreq@xxxxxxxxxxxxxxx; linux-pm@xxxxxxxxxxxxxxx; Knippers, Linda; Huang, Adrian (ISS Linux TW)
Subject: RE: [PATCH 1/1] Skip the driver if ACPI has power mgmt option
Hi Rafael,
Thanks for the suggestion. The code snippet was re-written accordingly. Please have a review.
Signed-off-by: Adrian Huang <adrian.huang@xxxxxx>
---
drivers/cpufreq/intel_pstate.c | 80 ++++++++++++++++++++++++++++++++++++++++++
1 file changed, 80 insertions(+)
diff --git a/drivers/cpufreq/intel_pstate.c b/drivers/cpufreq/intel_pstate.c index badf620..c03915e 100644
--- a/drivers/cpufreq/intel_pstate.c
+++ b/drivers/cpufreq/intel_pstate.c
@@ -26,6 +26,8 @@
#include <linux/fs.h>
#include <linux/debugfs.h>
#include <trace/events/power.h>
+#include <linux/acpi.h>
+#include <acpi/processor.h>
#include <asm/div64.h>
#include <asm/msr.h>
@@ -129,6 +131,18 @@ static struct perf_limits limits = {
.max_sysfs_pct = 100,
};
+struct hw_vendor_info {
+ u16 valid;
+ char oem_id[ACPI_OEM_ID_SIZE];
+ char oem_table_id[ACPI_OEM_TABLE_ID_SIZE];
+};
+
+/* Hardware vendor-specific info that has its own power management
+modes */ static struct hw_vendor_info vendor_info[] = {
+ {1, "HP ", "ProLiant"},
+ {0, "", ""},
+};
+
static inline void pid_reset(struct _pid *pid, int setpoint, int busy,
int deadband, int integral) {
pid->setpoint = setpoint;
@@ -700,6 +714,63 @@ static int intel_pstate_msrs_not_valid(void)
return 0;
}
+
+static bool intel_pstate_no_acpi_pss(void) {
+ struct acpi_processor *pr;
+ union acpi_object *pss = NULL;
+ int i;
+
+ for_each_possible_cpu(i) {
+
+ pr = per_cpu(processors, i);
+
+ if (pr) {
+ struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER,
+ NULL };
+
+ if (ACPI_SUCCESS(acpi_evaluate_object(pr->handle,
+ "_PSS", NULL, &buffer))) {
+
+ pss = buffer.pointer;
+
+ if (pss && pss->type == ACPI_TYPE_PACKAGE) {
+ kfree(buffer.pointer);
+ return false;
+ }
+
+ kfree(buffer.pointer);
+ }
+ }
+ }
+
+ return true;
+}
+
+static bool intel_pstate_platform_pwr_mgmt_exists(void)
+{
+ struct acpi_table_header hdr;
+ struct hw_vendor_info *v_info;
+
+ if (acpi_get_table_header(ACPI_SIG_FADT, 0, &hdr) == AE_OK) {
+ for (v_info = vendor_info; v_info->valid; v_info++) {
+ /*
+ * Check if the hardware/platform is in the
+ * predefined vendor data.
+ */
+ if (!strncmp(hdr.oem_id, v_info->oem_id,
+ ACPI_OEM_ID_SIZE) &&
+ !strncmp(hdr.oem_table_id, v_info->oem_table_id,
+ ACPI_OEM_TABLE_ID_SIZE)) {
+ if (intel_pstate_no_acpi_pss())
+ return true;
+ }
+ }
+ }
+
+ return false;
+}
+
static int __init intel_pstate_init(void) {
int cpu, rc = 0;
@@ -708,6 +779,15 @@ static int __init intel_pstate_init(void)
if (no_load)
return -ENODEV;
+ if (!acpi_disabled) {
+ /*
+ * Check if the platform has its own power management modes.
+ * If so, the pstate cpufreq driver will be ignored.
+ */
+ if (intel_pstate_platform_pwr_mgmt_exists())
+ return 0;
+ }
+
id = x86_match_cpu(intel_pstate_cpu_ids);
if (!id)
return -ENODEV;
--
1.8.1.2
-----Original Message-----
From: Rafael J. Wysocki [mailto:rjw@xxxxxxx]
Sent: Friday, October 18, 2013 5:21 AM
To: Huang, Adrian (ISS Linux TW)
Cc: Viresh Kumar; cpufreq@xxxxxxxxxxxxxxx; linux-pm@xxxxxxxxxxxxxxx; Knippers, Linda
Subject: Re: [PATCH 1/1] Skip the driver if ACPI has power mgmt option
On Friday, September 20, 2013 03:32:48 PM Adrian Huang wrote:
> I re-sent this email becuase my system time was incorrect for my first submission (The sent time of the first submission showed "2012/8/16 08:12PM"). Sorry for any inconvenience.
>
> Do not load the Intel pstate driver if the platform firmware (ACPI
> BIOS) supports the power management alternatives.
> The ACPI BIOS indicates that the OS control mode can be used if the
> _PSS (Performance Supported States) is defined in ACPI table. For the
> OS control mode, the Intel pstate driver will be loaded.
>
> Signed-off-by: Adrian Huang <adrian.huang@xxxxxx>
> ---
> drivers/cpufreq/intel_pstate.c | 84
> ++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 84 insertions(+)
>
> diff --git a/drivers/cpufreq/intel_pstate.c
> b/drivers/cpufreq/intel_pstate.c index 6efd96c..c181109 100644
> --- a/drivers/cpufreq/intel_pstate.c
> +++ b/drivers/cpufreq/intel_pstate.c
> @@ -25,6 +25,9 @@
> #include <linux/types.h>
> #include <linux/fs.h>
> #include <linux/debugfs.h>
> +#include <linux/acpi.h>
> +#include <acpi/processor.h>
> +
> #include <trace/events/power.h>
>
> #include <asm/div64.h>
> @@ -129,6 +132,18 @@ static struct perf_limits limits = {
> .max_sysfs_pct = 100,
> };
>
> +struct hw_vendor_info {
> + u16 valid;
> + char oem_id[ACPI_OEM_ID_SIZE];
> + char oem_table_id[ACPI_OEM_TABLE_ID_SIZE];
> +};
> +
> +/* Hardware vendor-specific info that has its own power management
> +modes */ static struct hw_vendor_info vendor_info[] = {
> + {1, "HP ", "ProLiant"},
> + {0, "", ""},
> +};
> +
> static inline void pid_reset(struct _pid *pid, int setpoint, int busy,
> int deadband, int integral) {
> pid->setpoint = setpoint;
> @@ -692,6 +707,66 @@ static int intel_pstate_msrs_not_valid(void)
>
> return 0;
> }
> +
> +static bool intel_pstate_no_acpi_pss(void) {
> + struct acpi_processor *pr;
> + acpi_status status = AE_OK;
> + union acpi_object *pss = NULL;
> + struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
> + int i;
> +
> + for_each_possible_cpu(i) {
> + pr = per_cpu(processors, i);
> +
> + if (pr) {
> + status = acpi_evaluate_object(pr->handle, "_PSS",
> + NULL, &buffer);
> +
> + pss = buffer.pointer;
> +
> + if (ACPI_SUCCESS(status) && pss) {
> + /* Found _PSS */
> + if (pss->type == ACPI_TYPE_PACKAGE)
> + return false;
> + }
buffer.pointer has been dynamically allocated by acpi_evaluate_object(), so it should be freed here. Besides, you need to define the buffer variable inside the loop so that it is initialized on every iteration and not just once.
> + }
> + }
> +
> + /* No _PSS */
> + return true;
> +}
> +
> +static bool intel_pstate_check_platform_pwr_mgmt(void)
> +{
> + struct acpi_table_header hdr;
> + struct hw_vendor_info *v_info;
> +
> + if (acpi_get_table_header(ACPI_SIG_FADT, 0, &hdr) == AE_OK) {
> + for (v_info = vendor_info; v_info->valid; v_info++) {
> + /*
> + * Check if the hardware/platform is in the
> + * predefined vendor data.
> + */
> + if (!strncmp(hdr.oem_id, v_info->oem_id,
> + ACPI_OEM_ID_SIZE) &&
> + !strncmp(hdr.oem_table_id, v_info->oem_table_id,
> + ACPI_OEM_TABLE_ID_SIZE)) {
> +
> + /*
> + * No _PSS defintion in ACPI BIOS means that
> + * the platform firmware has its own power
> + * management modes.
> + */
> + if (intel_pstate_no_acpi_pss())
> + return true;
> + }
> + }
> + }
> +
> + return false;
> +}
> +
> static int __init intel_pstate_init(void) {
> int cpu, rc = 0;
> @@ -700,6 +775,15 @@ static int __init intel_pstate_init(void)
> if (no_load)
> return -ENODEV;
>
> + if (!acpi_disabled) {
> + /*
> + * Check if the platform has its own power management modes.
> + * If so, the pstate cpufreq driver will be ignored.
> + */
> + if (intel_pstate_check_platform_pwr_mgmt())
> + return 0;
> + }
> +
> id = x86_match_cpu(intel_pstate_cpu_ids);
> if (!id)
> return -ENODEV;
>
Thanks!
--
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
��.n��������+%������w��{.n��������^n�r������&��z�ޗ�zf���h���~����������_��+v���)ߣ�
