Re: CALL FOR TESTING

[Matthew Wilcox <willy-8fiUuRrzOP0dnm+yROfE0A@public.gmane.org>]

On Thu, Sep 19, 2002 at 11:47:53AM +0200, Federico Di Gregorio wrote:
> everything seems to work fine. how can i say if an interrupt has been
> wrongly assigned? anyway the only problem has always been there but i
> don't know if it is about irq:

So the problem exists in both the vanilla Marcelo kernel and in the
ACPI-patched kernel?  That doesn't seem like a reason to exclude the
ACPI patch from going in.

> ACPI: PCI Interrupt Routing Table [\_SB_.PCI0._PRT]
> ACPI: PCI Interrupt Routing Table [\_SB_.PCI0.PCI1._PRT]
> pci_bind-0194 [05] acpi_pci_bind         : Device 00:00:1f.03 not
> present in PCI namespace
> pci_bind-0194 [05] acpi_pci_bind         : Device 00:00:1f.06 not
> present in PCI namespace

Interesting ... could you send the output from /sbin/lspci -s 00: I
suspect a bug in your machine's _PRT -- it's describing routing for a
device which doesn't exist.  Maybe we should silently ignore that.

> ACPI: Power Resource [PFAN] (off)
> 
> and later:
> 
> acpi_processor-2115 [07] acpi_processor_get_inf: Invalid PBLK length [5]

Interesting.  The code does:

        else if (object.processor.pblk_length < 6)
                ACPI_DEBUG_PRINT((ACPI_DB_ERROR, "Invalid PBLK length [%d]\n",
                        object.processor.pblk_length));
        else {
                pr->throttling.address = object.processor.pblk_address;
                pr->throttling.duty_offset = acpi_fadt.duty_offset;
                pr->throttling.duty_width = acpi_fadt.duty_width;
                pr->power.states[ACPI_STATE_C2].address =
                        object.processor.pblk_address + 4;
                pr->power.states[ACPI_STATE_C3].address =
                        object.processor.pblk_address + 5;
        }

so you're one element short, and that element is what tells us the address
to call to enter C3.  Clearly this is a broken BIOS, but I wonder if we
couldn't hande this more cleanly ...

	else if (object.processor.pblk_length < 4)
		ACPI_DEBUG_PRINT((ACPI_DB_ERROR, "Invalid PBLK length [%d]\n",
			object.processor.pblk_length));
	else {
		pr->throttling.address = object.processor.pblk_address;
		pr->throttling.duty_offset = acpi_fadt.duty_offset;
		pr->throttling.duty_width = acpi_fadt.duty_width;
		if (object.processor.pblk_length >= 5)
			pr->power.states[ACPI_STATE_C2].address =
				object.processor.pblk_address + 4;
		if (object.processor.pblk_length >= 6)
			pr->power.states[ACPI_STATE_C3].address =
				object.processor.pblk_address + 5;
	}

Andy, what do you think?

-- 
Revolutions do not require corporate support.


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf