On Tue, 28 Jul 2020 at 13:22, Lennart Poettering <mzerqung(a)0pointer.de>
wrote:
[..]
> Looks like all other pseudo filesystems are mounted over .mount
units
with
> probably one exception .. binfmt_misc.
> Probably .. because as I've pointed there are two units for that fs.
Only binfmt_misc is typically a kernel module of its own. For stuff
that is built-in it's pointless trying to avoid module loading.
https://www.investopedia.com/terms/d/death-1000-cuts.asp
Alternatively it is possible to use here KISS or Okhan Razor.
If in case of things like full scale distribution you don't care about
details sooner or later bad things will start.
I'm not talking about modularisation of the binfmt_misc (which IIRC is
possible to compile as the module).
I'm talking about currently compiled into the kernel autofs support (and
building it as the loadable module).
In this case it looks like already mounting binfmt_misc over the existing
(dist) .mount unit is in place.
And going back to the original question.
Why not provide the kernel with all possible to use file systems which are
possible to use as root fs as modules?
kloczek
--
Tomasz Kłoczko | LinkedIn:
http://lnkd.in/FXPWxH