FWTS fails method and mpcheck tests on Dell T20 server

[Jeff Lane ]

Running this against a new low-end server and found both 'method' and 'mpcheck' tests failed. I've attached the fwts results log because the individual test results for these (for method at least) are way to long to include here in a comment.

I need you guys to PLEASE look into this and let me know if these should gate a certificate for this machine. I've never seen these tests fail so I am rather concerned that they did.

[Colin Ian King (colin-king)]

Errors:

1: _PLD

ACPICA Exception AE_AML_NO_RETURN_VALUE during execution of method _PLD
FAILED [HIGH] AEAMLNoReturnValue: Test 8, Detected error 'No return value' when
evaluating '\_SB_.PCI0.XHC_.RHUB.HS10._PLD'.
ACPICA Exception AE_AML_NO_RETURN_VALUE during execution of method _PLD
FAILED [HIGH] AEAMLNoReturnValue: Test 8, Detected error 'No return value' when
evaluating '\_SB_.PCI0.XHC_.RHUB.HS11._PLD'.
ACPICA Exception AE_AML_NO_RETURN_VALUE during execution of method _PLD
FAILED [HIGH] AEAMLNoReturnValue: Test 8, Detected error 'No return value' when
evaluating '\_SB_.PCI0.XHC_.RHUB.HS12._PLD'.
ACPICA Exception AE_AML_NO_RETURN_VALUE during execution of method _PLD
FAILED [HIGH] AEAMLNoReturnValue: Test 8, Detected error 'No return value' when
evaluating '\_SB_.PCI0.XHC_.RHUB.HS13._PLD'.
ACPICA Exception AE_AML_NO_RETURN_VALUE during execution of method _PLD
FAILED [HIGH] AEAMLNoReturnValue: Test 8, Detected error 'No return value' when
evaluating '\_SB_.PCI0.XHC_.RHUB.HS14._PLD'.
ACPICA Exception AE_AML_NO_RETURN_VALUE during execution of method _PLD
FAILED [HIGH] AEAMLNoReturnValue: Test 8, Detected error 'No return value' when
evaluating '\_SB_.PCI0.XHC_.RHUB.HS15._PLD'.

If these refer to a USB device then because they don't return a value usb_acpi_check_port_connect_type() will assume that the devices are not available and -ENODEV is returned. At a hunch, these refere to root hub devices, but without the ACPI tables I can't easily confirm that.

Perhaps you can attach the output from:

sudo acpidump

2: _PCT

Test 73 of 155: Test _PCT (Performance Control).
FAILED [MEDIUM] MethodReturnNullObj: Test 73, \_PR_.CPU0._PCT returned a NULL
object, and did not return ACPI_TYPE_PACKAGE.

This is a broken processor performance control, which makes ACPI based processor performance control unavailable. I think this needs checking. The firmware does support _PPC (Performance Present Capabilities) which generally implies _PCT should be valid. So I think this is an issue that need fixing the firmware.

3: mpcheck

I think this needs checking. Can you supply the kernel log when the machine is cleanly booted to see what the kernel reports.

[Kent Baxley (kentb)]

Thanks, Colin. The server was on a very brief loan to us, but, I will find out about possibly getting another sample machine so I can supply you with more info.

[Samantha Jian-Pielak (samantha-jian)]

Hi Colin,

I have a system with similar mpcheck errors. I hope it's ok to tag along this bug. Attached is the fwts test results log, kernel logs, and acpidump.

[Colin Ian King (colin-king)]

Hi there,

I've had a look at this, and according to the documentation I can find, the MP tables are being ignored since this machine has ACPI support for SMP CPU configuration.

In fact, the kernel checks if ACPI lapic and ACPI ioapic support is enabled and returns early without doing any MP table parsing or configuration, cf default_get_smp_config():

        /*
         * MPS doesn't support hyperthreading, aka only have
         * thread 0 apic id in MPS table
         */
        if (acpi_lapic && acpi_ioapic)
                return;

So this is a non-issue. I'll file a bug against fwts to add notes about the mpcheck test if the machine supports ACPI in preference to MP tables.

[Colin Ian King (colin-king)]

So, just to clarify, this is a non-issue, and hence I will mark it "Won't Fix" against this bug and I'll file a bug against fwts to make sure the mpcheck tests are less aggressive in failing if the machine supports ACPI.

[Colin Ian King (colin-king)]

FYI: A fix to make mpcheck smarter in detecting a system that does not use MP tables and uses ACPI instead has been sent to the fwts-devel mailing list. See bug 1373340