[PATCH 2/2] acpi: method: add _PSS test
Alex Hung
alex.hung at canonical.com
Thu Apr 19 10:26:09 UTC 2012
On 04/19/2012 04:37 PM, Colin King wrote:
> From: Colin Ian King<colin.king at canonical.com>
>
> Signed-off-by: Colin Ian King<colin.king at canonical.com>
> ---
> src/acpi/method/method.c | 135 +++++++++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 133 insertions(+), 2 deletions(-)
>
> diff --git a/src/acpi/method/method.c b/src/acpi/method/method.c
> index 6797868..91b027d 100644
> --- a/src/acpi/method/method.c
> +++ b/src/acpi/method/method.c
> @@ -102,7 +102,7 @@
> * _PSD 8.4.4.5 N
> * _PSL 11.4.8 N
> * _PSR 10.3.1 Y
> - * _PSS 8.4.4.2 N
> + * _PSS 8.4.4.2 Y
> * _PSV 11.4.9 Y
> * _PSW 7.2.12 Y
> * _PTC 8.4.3.1 N
> @@ -307,7 +307,7 @@ static int method_name_check(fwts_framework *fw)
> int failed = 0;
>
> if ((methods = fwts_method_get_names()) != NULL) {
> - fwts_log_info(fw, "Found %d Methods\n", methods->len);
> + fwts_log_info(fw, "Found %d Objects\n", methods->len);
>
> fwts_list_foreach(item, methods) {
> char *ptr;
> @@ -1846,6 +1846,134 @@ static int method_test_UID(fwts_framework *fw)
> }
>
>
> +/* Section 8.4 */
> +
> +static void method_test_PSS_return(fwts_framework *fw, char *name, ACPI_BUFFER *buf, ACPI_OBJECT *obj, void *private)
> +{
> + int i;
> + bool failed = false;
> + uint32_t max_freq = 0;
> + uint32_t prev_power = 0;
> + bool max_freq_valid = false;
> +
> + if (method_check_type(fw, name, buf, ACPI_TYPE_PACKAGE) != FWTS_OK)
> + return;
> +
> + /* Something is really wrong if we don't have any elements in _PSS */
> + if (obj->Package.Count< 1) {
> + fwts_failed(fw, LOG_LEVEL_MEDIUM, "Method_PSSElementCount",
> + "_PSS should return package of at least 1 element, "
> + "got %d elements instead.",
> + obj->Package.Count);
> + fwts_tag_failed(fw, FWTS_TAG_ACPI_METHOD_RETURN);
> + return;
> + }
> +
> + for (i=0; i< obj->Package.Count; i++) {
> + ACPI_OBJECT *pstate;
> +
> + if (obj->Package.Elements[i].Type != ACPI_TYPE_PACKAGE) {
> + fwts_failed(fw, LOG_LEVEL_MEDIUM, "Method_PSSElementType",
> + "_PSS package element %d was not a package.", i);
> + fwts_tag_failed(fw, FWTS_TAG_ACPI_METHOD_RETURN);
> + failed = true;
> + continue; /* Skip processing sub-package */
> + }
> +
> + pstate =&obj->Package.Elements[i];
> + if (pstate->Package.Count != 6) {
> + fwts_failed(fw, LOG_LEVEL_MEDIUM, "Method_PSSSubPackageElementCount",
> + "_PSS P-State sub-package %d was expected to have "
> + "6 elements, got %d elements instead.",
> + i, obj->Package.Count);
> + fwts_tag_failed(fw, FWTS_TAG_ACPI_METHOD_RETURN);
> + failed = true;
> + continue; /* Skip processing sub-package */
> + }
> +
> + /* Elements need to be all ACPI integer types */
> + if ((pstate->Package.Elements[0].Type != ACPI_TYPE_INTEGER) ||
> + (pstate->Package.Elements[1].Type != ACPI_TYPE_INTEGER) ||
> + (pstate->Package.Elements[2].Type != ACPI_TYPE_INTEGER) ||
> + (pstate->Package.Elements[3].Type != ACPI_TYPE_INTEGER) ||
> + (pstate->Package.Elements[4].Type != ACPI_TYPE_INTEGER) ||
> + (pstate->Package.Elements[5].Type != ACPI_TYPE_INTEGER)) {
> + fwts_failed(fw, LOG_LEVEL_MEDIUM, "Method_PSSSubPackageElementType",
> + "_PSS P-State sub-package %d was expected to have "
> + "6 Integer elements but didn't", i);
> + failed = true;
> + continue;
> + }
> +
> + fwts_log_info(fw, "P-State %d: CPU %ld Mhz, %lu mW, latency %lu us, bus master latency %lu us.",
> + i,
> + (unsigned long)pstate->Package.Elements[0].Integer.Value,
> + (unsigned long)pstate->Package.Elements[1].Integer.Value,
> + (unsigned long)pstate->Package.Elements[2].Integer.Value,
> + (unsigned long)pstate->Package.Elements[3].Integer.Value);
> +
> + /*
> + * Collect maximum frequency. The sub-packages are sorted in
> + * descending power dissipation order, so one would assume that
> + * the highest frequency is first. However, it is not clear
> + * from the ACPI spec that this is necessarily an assumption we
> + * should make, so instead we should probably scan through all
> + * the valid sub-packages and find thed highest frequency.
> + */
> + if (max_freq< pstate->Package.Elements[0].Integer.Value) {
> + max_freq = pstate->Package.Elements[0].Integer.Value;
> + max_freq_valid = true;
> + }
> +
> + /* Sanity check descending power dissipation levels */
> + if ((i> 0)&& (prev_power != 0)&&
> + (pstate->Package.Elements[1].Integer.Value>= prev_power)) {
> + fwts_failed(fw, LOG_LEVEL_MEDIUM, "Method_PSSSubPackagePowerNotDecending",
> + "_PSS P-State sub-package %d has a larger power dissipation "
> + "setting than the previous sub-package.", i);
> + fwts_advice(fw, "_PSS P-States must be ordered in decending order of "
> + "power dissipation, so that the zero'th entry has the highest "
> + "power dissipation level and the Nth has the lowest.");
> + failed = true;
> + }
> + prev_power = pstate->Package.Elements[1].Integer.Value;
> + }
> +
> + /*
> + * Sanity check maximum frequency. We could also check the DMI data
> + * for a BIOS date (but this can be wrong) or check the CPU identity
> + * (which requires adding in new CPU identity checks) to make a decision
> + * on when it is reasonable to assume a CPU is modern and hence clocked
> + * incorrectly. For now, just flag up a low level error that the
> + * frequency looks rather low rather than try to be intelligent (and
> + * possibly make a mistake). I'd rather flag up a few false positives
> + * on older machines than miss flagging up bad _PSS settings on new
> + * machines.
> + */
> + if (max_freq_valid&& max_freq< 1000) {
> + fwts_failed(fw, LOG_LEVEL_LOW, "Method_PSSSubPackageLowFreq",
> + "Maximum CPU frequency is %dHz and this is low for "
> + "a modern processor. This may indicate the _PSS P-States "
> + "are incorrect\n", max_freq);
> + fwts_advice(fw,
> + "The _PSS P-States are used by the Linux CPU frequency "
> + "driver to set the CPU frequencies according to system "
> + "load. Sometimes the firmware sets these incorrectly "
> + "and the machine runs at a sub-optimal speed. One can "
> + "view the firmware defined CPU frequencies via "
> + "/sys/devices/system/cpu/cpu*/cpufreq/scaling_available_frequencies");
> + failed = true;
> + }
> +
> + if (!failed)
> + fwts_passed(fw, "_PSS correctly returned sane looking package.");
> +}
> +
> +static int method_test_PSS(fwts_framework *fw)
> +{
> + return method_evaluate_method(fw, METHOD_OPTIONAL, "_PSS", NULL, 0, method_test_PSS_return, NULL);
> +}
> +
> /* Tests */
>
> static fwts_framework_minor_test method_tests[] = {
> @@ -1973,6 +2101,9 @@ static fwts_framework_minor_test method_tests[] = {
> { method_test_ON, "Check _ON (Set resource on)." },
> { method_test_OFF, "Check _OFF (Set resource off)." },
>
> + /* Section 8.4 */
> + { method_test_PSS, "Check _PSS (Performance Supported States)." },
> +
> /* Appendix B, ACPI Extensions for Display Adapters */
>
> { method_test_DOS, "Check _DOS (Enable/Disable Output Switching)." },
Acked-by: Alex Hung <alex.hung at canonical.com>
More information about the fwts-devel
mailing list