Aaron Griffin wrote:
> On 3/2/06, Dimitrios Apostolou <[EMAIL PROTECTED]> wrote:
> 
>>My primary concern is that necessary modules should be autoloaded by
>>default. Since module autoloading is enabled in the kernel insmod should
>>be redundant for any of them. I have two theories:
>>1) Modules can't be autoloaded with a static /dev so we need udev in the
>>initrd too. However, I'm pretty sure I remember module autoloading was
>>doable in the old days too, with a static /dev.
>>2) The modprobe utility is missing from the initrd, so module
>>autoloading can't happen.
> 
> 
> There are alot of reasons for all this stuff, however, before I go
> that way, let me ask what you're trying to accomplish:
> Yes, you want less modules loaded.  We get that - that is the reason
> for this page:
> http://wiki.archlinux.org/index.php/Initrd
> 
> Autoloading can be done, and it's probably preferrable, but at this
> point in time, it might not be the best move - udev and modprobe in
> the initrd image will (a) require alot of testing and (b) require
> glibc out of the box (or possibly static compiles, or uClibc) - either
> way, it's alot of stuff that would require much testing.

I see now, thank you for your thourough answer. Academically speaking, 
do you think that by only statically compiling modprobe (while keeping 
the static /dev) and inserting it in the initrd would work?
> 
> In the near future, klibc will be integrated with the kernel, making
> the initrd redundant.  I have been, off-and-on, doing some initramfs
> testing with klibc.  Module autoloading and all the stuff you speak of
> is almost trivial when using kinit.  The only thing holding me back is
> the fact that kinit is rather young, and just recently added support
> for raid devices, so I'm letting it mature.

Hmmm, I don't have a clue about klibc/kinit. I'd better google a bit...
> 
> I do have a plan in mind, I am just waiting on things outside of my
> control.  When klibc/kinit become a bit more mature, I will experiment
> a tad more on a wider breadth of hardware.
> 
> For now, though, I would recommend following the wiki.  You will
> achieve exactly what you want without massive changes to the initrd
> process which may be going away in the near future.

Actually I had this problem solved before by fine-tuning myself the 
initrd. I just wanted sometime to post my questions, since I couldn't 
exactly understand the situation.

Thanks for explaining the situation,
Dimitris

_______________________________________________
arch mailing list
[email protected]
http://www.archlinux.org/mailman/listinfo/arch

Reply via email to