On Fri, 20 Dec 2002 10:38:45 -0600 (CST), Scott Garfinkle wrote:
>"who really cares" refers to whether there is a specific need for the functionality.
>Since there does
>appear to be one, that's fine. At this point, providing general-purpose support on
>OS/2 is often not
>too important! As for KEE, it was implemented mainly to support JFS and the 32 bit
>tcpip stack, not
>out of some sense that OS/2 just ought to be able to do 32 bit DevHelps. Once it was
>implemented,
>it certainly made sense to make it available to others; however, in most cases, there
>really isn't
>anything to be gained (and often much efficiency to be lost) in doing 32 bit OS/2 dd
>code.
I know about the efficiency of thunking :-)
But like e.g. the Soundblaster Live! drivers. If they would use GCONFIG-API, one would
have to support both,
16bit and 32-bit, otherwise Sander could dump the 32-bit version or regenerate
GCONFIG-APIs internally.
That's why I'm doing 32-bit. I won't implement them at first and the whole detection
code is 16-bit mode. I'm
just designing the detection area to allow access from 32-bit code later. Also device
attach/detach are
handled by 16-bit code. 32-bit only contains the API, which includes device search
functions, specific
functions like e.g. PCI-space access and hot-plug callback API.
It's meant so one can write such driver like SBLive, so that it's possible to write a
16-bit and 32-bit version of it
w/o modifying any code (at least for GCONFIG-calling).
Example of some GCONFIG-API code:
HWDEVICE MyDevice;
USHORT MyIRQ;
if (MyDevice = PCI_FindDevice(0x1073, -1, PDD_FIND_STRAIGHT, 0)) {
MyIRQ = PDD_GetDeviceIRQ(MyDevice,0);
PrintF ("Yamaha found on %lx using %d!", &MyDevice, MyIRQ);
PCI_WriteConfigDWord (MyDevice, 64, 0x1000);
}
That's all the code needed to find the first PCI-device of vendor 0x1073 and getting
the IRQ line that is
currently assigned by the device, printing it out and set register 64 (dword) to
0x1000. Print/PrintF are also part
of GCONFIG-API. That's the needed of it. And this code will run on ACPI and normal PCI
systems. And the
PDD_x functions are generic (!) which means they will work on any device. PnP, PCI,
whatever. And all take
HWDeviceID as the device specifier. There are also APIs to reconfigure a device (those
are nearly of the
same simplicity).
cu, Kiewitz
-----------
To unsubscribe yourself from this list, send the following message
to [EMAIL PROTECTED]
unsubscribe acpi-os2
end
