Re: NONE for ALL, and ALL for NONE

On December 5, 2018 8:55:52 PM GMT+05:30, "Cherry G. Mathew" <[email protected]> wrote:
>Manuel Bouyer <[email protected]> writes:
>
>> On Wed, Dec 05, 2018 at 01:40:58AM +0530, Cherry G.Mathew wrote:
>>> Hello,
>>> 
>>> I was paring down GENERIC to see if I could build a kernel without
>>> ACPI/MPBIOS - and voila - it's still possible! (with a few nits here
>and
>>> there).
>>> 
>>> In the spirit of yin/yang, I thought this was a fun provide a
>minimalist
>>> configuration without these, and the following suggestion seemed
>>> inspiring:
>>> 
>>> <jak> well, we do have a ALL, maybe we need a NONE
>>> 
>>> (disclaimer: <jak> may or may not resemble anyone in real life, with
>or
>>> without above opinion).
>>> 
>>> One interesting outcome of this exercise was to tease out cross
>subsystem
>>> dependencies that have slowly crept in - for eg: PCI vs. MSI.
>>> 
>>> I've added a patch below for your entertainment, but more seriously,
>I'm
>>> wondering what people think about having a "NONE" kernel - something
>>> with just the very basic code needed to boot on the least common
>>> denominator x86 platform - real or emulated.
>>
>> I think you could take one more step and get rid of PCI entirely.
>> pciide should show up as wdc0 at isa, and vga should show up at isa
>too
>
>I left it there as a test for platform PCI support (without ACPI or
>MPBIOS). It's possible to just comment it out.
>
>I suppose a lot more can be culled. Thoughts ?

Btw, isn't PCI supported in the basic 'PC architecture'? My memory is that there were 4 interrupt lines per slot mapped by 'BIOS' to irq numbers and that they were labeled alphabetically (INT#A etc.)

If that were the case it would be interesting to include pci in the basic config just as we'd have isa(4).

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.