Re: [Altosdev-kernel] Drivers (more)

[Benjamin J. <bsj...@mi...>]

I'm awake.  I'm just kind of waiting for people to give me stuff to do...

Ben

Kernel Panic wrote:

> Hello, AltOsDev.
>
> So.. I've been thinking on the drivers model for quite a while, but
> haven't come up with anything. How are we going to write drivers??
>
> I suggest something like what I've seen in FreeBSD (don't know about
> Linux). A driver is declared by some magic macro (like
> DRIVER_MODULE([...])) right in the C-source. I have no clue how it
> works, but somehow it does =) . (If anyone is interested, I can post a
> link to FreeBSD drivers programming guide.) The thing is, that we can
> define this macro in such a way, so that it produces an exportable
> (global) symbol. So, when we load the driver, we can easily find the
> its description. This description (defined by the above-mentioned
> DRIVER_MODULE or something) is basically a structure which lists the
> name of the driver and pointers to the list of methods it allows
> (open, close, ioctl, etc.). This seems rather easy.
>
> Well, now the hardest part starts. First, how do we organise the info
> about drivers inside the kernel (in run-time, I mean)? Say, we have
> loaded a PCI bus driver, which has found that there are 5 devices
> attached to it. How is this information going to be passed to the
> kernel? (It's also clear now the we have to maintain a tree-like
> structure for the list of drivers. Or am I wrong?) Then.. suppose we
> have a bunch of drivers loaded now, that represent the physical
> devices. But we also need virtual devices (I mean, non-physical, or
> not strictly phisical), like a console, various partitions on the
> harddrive etc. How are they going to be loaded and assigned to the
> correct device? I guess we wouldn't like to reinvent a wheel, so if
> you are familiar with any existing implementation of this, please tell
> me. New ideas are also welcome.
>
> Also.. I have no clue of how to support non-self-enumerating buses
> (like ISA), but I don't think that's a problem right now, as we don't
> have any drivers at all.
>
> Then.. I guess we would have to write a simple statically-linked C
> library for the drivers (because drivers are executables, and cannot
> import anything from the kernel). It would also provide the means of
> IPC transparent to the driver.
>
> Suggestions are welcome..
>
> HEY, GUYS, WAKE UP!!!!!!!! =))
>
> --
> Regards,
>  Kernel Panic                          mailto:rzh...@he...
>
> _______________________________________________
> altOSdev-kernel mailing list
> alt...@li...
> https://lists.sourceforge.net/lists/listinfo/altosdev-kernel