OSDev.org

Hi out there.
Does anyone out there know how programming of a chipset is done?
Thanks.

Well Mr Vague..

Can you be a little more... specific?

Ok then,
I wish to use USB on my OS and also directly probe the memory available in the system.But i don't want to go through the BIOS.
I know i can program the chipset registers for that.
How to proceed is the question.

You cannot do that without having a driver for each chipset. Memory detection depends heavily on your chipset, there is no common way to probe for useable memory without using the BIOS.

So how do i write a driver for a chipset?
There is this old version of linux working with some sort of "generic" chipset driver.
Thats exactly what i need.

einsteinjunior wrote:Ok then,
I wish to use USB on my OS and also directly probe the memory available in the system.But i don't want to go through the BIOS.
I know i can program the chipset registers for that.
How to proceed is the question.

einstein shame on you direct probing is very hazardous for you system. Hence the int $15/$E820 function in the BIOS which is very reliable BTW.

There is no shame on me!!!
Direct probing can be done SAFELY using the CHIPSET.
Got that?

Typically a small EEPROM on SDRAM modules is interpreted by the BIOS, I'm assuming that's how the BIOS detects available RAM these days.. (Not direct probing..)

http://en.wikipedia.org/wiki/Serial_Presence_Detect

I do not think Linux uses a chipset driver to detect available memory. I guess Linux as well as Windows uses Int 0x15/ Function 0xE820 to detect memory (and then falls back to some other BIOS calls). There is no way to write a generic chipset driver that is able to detect every kind of memory that is present in the system.

Hi,

einsteinjunior wrote:I wish to use USB on my OS and also directly probe the memory available in the system.But i don't want to go through the BIOS.
I know i can program the chipset registers for that.
How to proceed is the question.

The USB controller is the same as any other PCI card - use PCI configuration space to detect it, find/configure the resources it's using, etc. It's not "chipset specific" (even when there's USB controllers built into the chipset they behave exactly the same as a USB controller in a plug-in PCI card).

For memory detection, every chipset does it differently and it's impossible to detect RAM using chipset registers while your code is running from RAM because you need to severely mess with the RAM chips. The BIOS can do it because the BIOS is designed specifically for the chipset and runs from ROM.

Rather than using chipset registers for memory detection, you could manually probe for RAM (e.g. write a value at an address and read it back to see if it was stored). This is a highly dodgy method with many problems. For example, you can't tell the different between RAM and memory mapped I/O (e.g. video card display memory). It's also very slow - e.g. for a 3 GHz CPU with 4 GB of RAM it could take 16 seconds.

Lastly, detecting "present RAM" is mostly useless anyway - you need to detect "usable/free RAM". This is partly because the BIOS uses some RAM for other things, like SMM (System Management Mode) that the OS has no control over, and partly because the BIOS can put tables in RAM for the OS to use that your OS might want later.

Basically, if your OS can't use BIOS functions (or firmware) for memory detection then it has severe design flaws - fixing those design flaws makes a lot more sense than using any unreliable hack.


Cheers,

Brendan

I am primarily not interested in direct memory probing using chipset since i can do it using BIOS.What i really need is a chipset driver for USB.
Thanks.

Hi,

einsteinjunior wrote:I am primarily not interested in direct memory probing using chipset since i can do it using BIOS.What i really need is a chipset driver for USB.

Hmm...

Are you sure you don't need a generic USB driver for PCI, or does it have to be for a specific chipset (e.g. a chipset with broken/dodgy USB hardware that doesn't follow standards)?


Cheers,

Brendan

Are you sure what you're looking for isn't BIOS calls? (Still possibly in protected mode using v86 emulation).

We're in OS Design and Theory right now? You know, the place without hardware-hacking newbies asking utterly underspecified questions?

Let me fix that then. USB implementation is only barely theory or design, mostly... implementation...