Re: List of modules required by linux-modules

["dabbede@gmail.com" <dabbede@gmail.com>]

Yes yes yes!
That's the tip I was hoping for! I've missed CONFIG_SCSI_LOWLEVEL
because I was looking for answers here:
https://cateee.net/lkddb/web-lkddb/SCSI_ISCI.html.
It's the first time that that database gives me wrong information...
Conversely, this website is correct:
https://www.kernelconfig.io/config_scsi_isci?q=&kernelversion=5.4.230&arch=x86.

And for the module list, I was blaming linux-module.scm because the
system reconfigure was failing at that point. Looking in the code I
knew that the list was generated elsewhere, but I couldn't find where,
nor how it is passed to linux-modules... I admit I've just wet my toe
in Guix (and guile programming) and the codebase is huge...

Thank you very much. Now my system reconfigure ended correctly! Time
for latency tests... :-D

On Fri, Jan 27, 2023 at 11:07 AM Tobias Geerinckx-Rice <me@tobias.gr> wrote:
>
> Hi,
>
> dabbede@gmail.com 写道：
> > I've stopped the kernel build after the config phase and I've
> > inspected the resulting .config file. There's no evidence of
> > CONFIG_SCSI_ISCI in there, not even "# CONFIG_SCSI_ISCI is not
> > set".
>
> CONFIG_SCSI_ISCI depends on CONFIG_SCSI_LOWLEVEL.  Is it set?
>
> That's the only possibility I see, apart from the obvious
> CONFIG_SCSI.
>
> Otherwise, share your (Guix-generated, final) .config.
>
> > It's not clear why this happens with my custom build and not
> > with linux-libre...
> > Can you help me debug this issue?
> > Moreover, where is created the list of module required by
> > "linux-modules" ?
>
> I'm confused by your explicit reference to
> gnu/build/linux-modules[.scm], which provides a set of mechanisms
> to work with modules.  It *should* not set policies or ‘require’
> any ones in particular.  Does it?
>
> Do you mean ‘%base-initrd-modules’ in
> <https://git.savannah.gnu.org/cgit/guix.git/tree/gnu/system/linux-initrd.scm#n342>?
>
> This static list can be frustrating, but we can easily remove
> elements from it or even replace it entirely:
>
>   (operating-system
>     …
>     (initrd-modules
>       ((@ (rnrs lists) remove) "isci" %base-initrd-modules))
>     …)
>
> What I suggest you do instead (because it's what I do to my
> kernels :-) is simply ship a ‘%xenomai-initrd-modules’ variable
> along with your ‘xenomai-linux’ package, keep them in perfect
> harmony yourself, and ask users to use both.
>
> Kind regards,
>
> T G-R