[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20181121100603.5afd53ed@endymion>
Date: Wed, 21 Nov 2018 10:06:03 +0100
From: Jean Delvare <jdelvare@...e.de>
To: Masahiro Yamada <yamada.masahiro@...ionext.com>
Cc: Michal Marek <michal.lkml@...kovi.net>,
Linux Kbuild mailing list <linux-kbuild@...r.kernel.org>,
martin.wilck@...e.com,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: Backed up kernels
Hi Masahiro,
On Wed, 21 Nov 2018 15:59:49 +0900, Masahiro Yamada wrote:
> On Tue, Nov 20, 2018 at 10:40 PM Jean Delvare <jdelvare@...e.de> wrote:
> > Therefore I am asking, can we change "make install" so that it does NOT
> > create a backup copy of an existing kernel?
>
> I think your suggestion makes sense,
> but "make install" is basically implemented
> by arch-specific shell script.
> (For example, arch/x86/boot/install.sh)
Thanks for the pointer. However I have a hard time believing that the
script above is what is run when I call "make install". It looks pretty
old, doesn't support kernel files with version strings, and only knows
of lilo as a boot loader.
But I see there is a hook at the beginning for a user or distribution
provided install script:
if [ -x ~/bin/${INSTALLKERNEL} ]; then exec ~/bin/${INSTALLKERNEL} "$@"; fi
if [ -x /sbin/${INSTALLKERNEL} ]; then exec /sbin/${INSTALLKERNEL} "$@"; fi
So I guess that what I really care about is the /sbin/installkernel
script on my system, which is part of the dracut package. Which means I
must talk to the dracut package maintainer of my distribution.
> Will you talk to the maintainers
> of architecture you are interested in?
>
> (or send it to linux-arch@...r.kernel.org)
It doesn't seem x86-specific, as apparently a lot of code was
copy-and-pasted across architectures over time. It probably doesn't
make sense to change it on one architecture and not on the others.
Also, if anyone is using these basic kernel-provided installation
scripts, then keeping a backup may actually make sense, because the
kernel files have no version strings, so a new kernel would always
overwrite the previous one, only leaving one kernel in place. If that
kernel doesn't boot for whatever reason, then game over.
So I think we should leave things as is on the kernel side.
Thanks again,
--
Jean Delvare
SUSE L3 Support
Powered by blists - more mailing lists