GNU bug report logs - #41066
[PATCH] gnu: grub: Support for chain loading.

Previous Next

Full log

Message #94 received at 41066 <at> debbugs.gnu.org (full text, mbox):

From: Danny Milosavljevic <dannym <at> scratchpost.org>
To: Stefan <stefan-guix <at> vodafonemail.de>
Cc: 41066 <at> debbugs.gnu.org, Mathieu Othacehe <othacehe <at> gnu.org>,
 Efraim Flashner <efraim <at> flashner.co.il>,
 Maxim Cournoyer <maxim.cournoyer <at> gmail.com>
Subject: Re: [PATCH] gnu: bootloader: Support for chain loading.
Date: Wed, 18 Nov 2020 19:05:59 +0100

[Message part 1 (text/plain, inline)]

Hi Stefan,

On Tue, 17 Nov 2020 21:27:42 +0100
Stefan <stefan-guix <at> vodafonemail.de> wrote:

> For example the profile contains at least these folders from the grub package, plus the collection/:
> 
> /gnu/store/…-bootloader-profile/collection
> /gnu/store/…-bootloader-profile/etc
> /gnu/store/…-bootloader-profile/share
> /gnu/store/…-bootloader-profile/lib
> /gnu/store/…-bootloader-profile/bin
> /gnu/store/…-bootloader-profile/sbin

Ohhh, that's right.  It actually contains the grub installer--which it got from
the grub-efi derivation.  The actual boot EFI file is not in the grub-efi
derivation.

I wonder if it is possible (one day!) to have a package that actually contains
the EFI boot file that the grub installer generated.  That would be a lot
nicer--though I'm not sure whether that stuff depends on the system and thus
cannot be "pre-generated".

> But the /boot/ folder finally contains something like this, with most of it being content from the collection/, where the GRUB files get installed belaw /boot/efi/:
> 
> /boot/bcm2710-rpi-3-b.dtb
> /boot/bootcode.bin
> /boot/bootloader.txt
> /boot/config.txt
> /boot/custom.txt
> /boot/efi/
> /boot/fixup.dat
> /boot/gnu/
> /boot/grub/
> /boot/overlays/
> /boot/u-boot.bin
> 
> Actually one could say that the profile hook is a kind of ‘pre-installer’ for everything not related to GRUB, able to prepare everything, but unable to write into the /boot folder.
> 
> > If there are no good reasons to do it like that, I'd just put the
> > profile into $output directly instead--it's easier to understand, and also it's
> > how other profiles are being used.  
> 
> Not having the collection folder would mean that the installer would need to assume much more about the result of the profile hook, to copy the right files to /boot.

I agree.  Let's leave it inside subdir "collection".

> > (2) The COPY-FILES? flag is kinda weird.
> > I would prefer if INSTALLER just defaulted to a procedure that: does copy
> > files, and then calls the final bootloader installer.
> > If the user doesn't want it then the user could still pass a INSTALLER
> > that doesn't (for example the user could pass #:installer
> > (bootloader-installer final-bootloader)).  
> 
> Agreed.

> Another possibility would be to remove the collection folder within a hook.

I don't like that that much because it's mutating too much and thus
composability goes down.

> > I would prefer if the user would just change the INSTALLER in the case he wants
> > to not use the profile (which is kinda weird?!) or pack it or whatever.  
> 
> OK, I see, in case of a custom installer we can skip the copying completely. That makes sense. 

Could you send an updated patch that does it like that?  Or do you rather want
the old variant ?

> > (3) Why isn't the final bootloader installed last?  I would have expected
> > it to be installed last so that if it does packing of the profile contents
> > in order to quickly find it at boot, it would have to have all the files
> > of the profiles already, no?  
> 
> I thought about the order as well. My conclusion was that a file from the collection should be able to overwrite a file installed from final-bootloader, for example to install own device-tree files.

Yeah, that makes sense--if a little unusual (thus should be documented).

[Message part 2 (application/pgp-signature, inline)]

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.