bug-hurd
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Boot proposition for Debian Hurd on USB keys


From: Samuel Thibault
Subject: Re: Boot proposition for Debian Hurd on USB keys
Date: Fri, 5 Feb 2021 20:03:02 +0100
User-agent: NeoMutt/20170609 (1.8.3)

Paul Dufresne, le ven. 05 févr. 2021 12:22:33 -0500, a ecrit:
> Because Debian Hurd have no support for USB yet. So it cannot read the files.

Yes, and I believe that's what we should really fix, and currently being
fixed by the rumpdisk translator.

Departing from the current way d-i works is asking for maintenance
burden.

> That's a bit weird for people, because it did read the files from the USB key
> up to the installer searching packages...

Another way would be to just include all required d-i packages in the
d-i initrd, making it bigger. The rest can be fetched from the internet.

> I think we can have a not so small installer.ext2 file that would be loaded in
> memory by Grub, working around the lack of USB drivers.
[...]
> So clearly, for now, this system must ... basically be the installer.

That is *exactly* what the initrd already is.
The fact that the installer loads .udebs on demand is just a way to
support lots of situations without loading everything altogether in
memory.

> Also, currently, Debian Hurd boot an ext2 filesystem, that breaks everytime 
> the
> system is not cleanly shut down.

What do you mean by "break"?

When the filesystem is not properly shut down, one has to run e2fsck,
yes, that's handled by the init scripts, is that not the case on your
boxes?

> and not return to Grub after fsck step.

Returning to grub is a requirement when it is the root filesystem that
got mangled by fsck: the existing ext2fs has to reload all of the data
since it was mangled. Just rebooting is way simpler than trying to track
everything that should getting re-read.

Samuel



reply via email to

[Prev in Thread] Current Thread [Next in Thread]