Re: [dev] suckless distro

From: Weldon Goree <>
Date: Wed, 25 Jun 2014 13:49:58 +0530

On 06/24/2014 11:58 AM, Markus Teich wrote:
> I've built me a hardware tailored kernel, containing only the drivers, my laptop
> needs and mostly statically linked. Only a few drivers (UMTS modem, wifi, audio)
> are built as modules for convenience reasons, so I don't have to reboot if one
> of them fails, but can just reload the module.

Also, some of the drivers really "want" to be modules. Not so much the
literal hardware drivers as the "driver drivers" like snd, drm, *-agp,
various wireless frameworks, etc. You can do them as built-ins, but it's
rarely worth the headache (and requires a
rebuild/re-bootloader-configure/reboot when you inevitably screw
something up). Though some of that headache is mitigated by using a
slightly more sane hotplug than udev (which would be just about anything).

Personally I tend to use module-less kernels, but then I also still like
static /dev file trees, so I'm kind of a dinosaur about this. But my
side hobby of making a distro has made me more sympathetic to the ideas
behind modules -- once "it works on my machines" is no longer enough,
modules start to be more useful.

That said, even for a live system as long as you don't mind loading an
immense kernel (Slackware's "huge" is 6.4M, for instance; twice the size
of "generic", and "huge" still has some modules) I don't think there's
ever a situation where you would _need_ a modular kernel. The
"run-anywhere" kernel I've been working on has basically every possible
hardware selection built-in, and modules for behavior or features. A
modular kernel in principle seems to suck less than a big static one,
but in practice it ties you down to whatever daemon out there in
userland is telling the kernel what to do with hardware (mdev,
fortunately, sucks less than udev, but I'd still rather just have the
kernel deal with stuff; if we could get rid of people's obsession with
hotplugging hardware we wouldn't even need that...)

Received on Wed Jun 25 2014 - 10:19:58 CEST

This archive was generated by hypermail 2.3.0 : Tue Jun 24 2014 - 10:24:06 CEST