PERFORCE change 36551 for review

[John Baldwin]

On 21-Aug-2003 Marcel Moolenaar wrote:
> On Thu, Aug 21, 2003 at 10:32:25AM -0700, Marcel Moolenaar wrote:
>> 
>> > If the UART devices raise an ISA interrupt, then by my reading,
>> > the ACPI resource should specify the ISA interrupt number (0-15),
>> > and the MADT should include a source override that maps that
>> > ISA interrupt number to a global interrupt number of 66 or
>> > whatever (which maps to a SAPIC:intpin).
>> 
>> This makes sense. It's however not how it is (unfortunately).
> 
> The updated SPPA specification (HP's ia64 platform) has a section
> devoted to the interrupt polarity and mode of the UART. It basicly
> says this:
> o  The DIG64 HCDP table [supported] or the Mcrosoft SPCR table
>    [unsupported] tells whether the UART is a PCI device or not.
> o  PCI UARTs have level triggered, active low interrupts. They
>    are not described in ACPI then (reminder: this is SPPA).
> o  Non-PCI UARTs described in the ACPI namespace have interrupt
>    polarity and mode as described by _CRS in the device object!
> o  Non-PCI devices that are not decribed in the ACPI namespace
>    can still be mentioned in the HCDP table and we [FreeBSD]
>    will use the UART as console. Interrupt polarity and mode
>    should be assumed active low, level sensitive.
>    Currently we will panic the moment we try to go single-user
>    or multi-user because there will not be a device major number
>    assigned to the console. We need to catch this case someday.
> 
> So: It appears that we need to interpret the _CRS method, field
> or whatever. Especially the Interrupt Descriptor.
> 
> Going to the source: in acpi_parse_resources() we need to create
> a callback to MD code to tell it about polarity and mode. This
> means tweaking the ACPI_RSTYPE_IRQ or ACPI_RSTYPE_EXT_IRQ cases.
> Better would be to create bus methods for this (see for example
> acpi_res_set_irq()).

Yes.  For i386 definitely it would make sense to have a bus method
that bubbles back up to the nexus(4) and eventually calls the
MD interrupt code.  Maybe some kind of interrupt properties kobj
interface:

INTERFACE interrupt_properties

#
# Set the polarity to one of three values:
# - conforming (conform to the bus attached to, the bus can set
#   this on the way up through the chain maybe?)
# - active high
# - active low
#
METHOD int set_polarity {
        device_t dev;
        device_t child;
        struct resource *irq;
        int polarity;
};

#
# Set the trigger mode to one of three values:
# - conforming (conform to the bus attached to, the bus can set
#   this on the way up through the chain maybe?)
# - edge triggered
# - level triggered
#
METHOD int set_trigger_mode {
        device_t dev
        device_t child;
        struct resource *irq;
        int trigger_mode;
};

Or I guess we could just add these to the bus interface.

What do you think we should do?

-- 

John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve!"  -  http://www.FreeBSD.org/