[arch-releng] grub install not listing partitions
Keshav P R
the.ridikulus.rat at gmail.com
Fri Dec 16 08:15:33 EST 2011
On Fri, Dec 16, 2011 at 18:39, Keshav P R <the.ridikulus.rat at gmail.com>wrote:
> On Fri, Dec 16, 2011 at 15:04, Eric Fernandez <zeb at zebulon.org.uk> wrote:
>> 2011/12/16 Matthew Gyurgyik <pyther at pyther.net>:
>> > I agree with Thomas.
>> > 1.) There are better alternatives to installing grub on a partition (see
>> > Thomas' mail).
>> > 3.) Don't use grub-legacy (not supported upstream or maintained). Use
>> > syslinux (installer) or grub2 (manually).
>> I appreciate these comments and gather syslinux might be a better
>> solution than grub, and will look into it.
>> However, I would like to point out that grub (legacy) is still the
>> recommended bootloader solution in the beginner's guide, and in the
>> installer isos. Furthermore, grub legacy is the one in core, whereas
>> grub2 is in extra. I think that it is prematurate to cripple the
>> installation of a bootloader just because it is not maintained
>> upstream, until we have a better replacement. If really grub legacy is
>> bad, then we should phase it out and replace it with a better one,
>> which should be the new default. Why not replace grub legacy with
>> grub2 in the core repository and in the installer?
> I suggest someone should replace grub-legacy with syslinux as the
> recomended bootloader in the Beginner's Guide (I do have a wiki account but
> skeptical about editing the "Guide" wiki article myself). There is already
> a bug report to remove grub-legacy from base
> https://bugs.archlinux.org/task/26187 . But if you really want people to
> stop using grub-legacy (which is logical given that it is surviving with
> the help of many patches and still doesn't support many newer technologies
> like GPT, BTRFS etc.), then you should stop providing "grub" as an option
> in AIF and force them to do it manually similar to grub2, and then drop it
> from core.
> Grub2 is not KISS, but technically much better than grub-legacy (upstream
> considers grub2 1.99 to be more stable than grub-legacy 0.97 minus the
> patches), with support of virtually all the newer technologies out there.
> It takes a lot of work to make Grub2 work with AIF. Also the upstream is in
> high flux (even command options are changing between releases). I know this
> because I am tracking the grub2 bzr development. Even the extra grub2
> packages sometimes use bzr snapshots as source to include upstream bug
> So instead of bothering about rapidly changing grub2 and dead grub-legacy,
> I would suggest making syslinux as the default (and for now the only
> bootloader) in AIF.
> > Installing grub to a partition is an uncommon setup and used by few
>> > Those who really want to install grub to a partition can do so manually.
>> Well, that is against the KISS principle. The fact it is uncommon is
>> irrelevant for a distro like Arch, which is not suppose to hide
>> options for their own good (especially when users report having used
>> the partition installation with no problem).
> I don't see any issue with this. Sometimes scripting bootloader
> installation is a pain compared to directly installing manually. Matthew
> simply pointed out that supporting grub installation to a partition might
> be difficult from a shell-scripting/automation point of view (which is what
> happens in the installer) as compared to telling the user to do it manually
> (this is my opinion, I haven't used grb-legacy to comment on the difficulty
> of installing it to a partition, I use grub2-bios). He is not hiding
> anything. Users by themselves should be able to identify if it is possible
> to install the bootloader to the partition instead of the MBR (in BIOS
> systems), and should be able to do it themselves.
> Not everything (read use-cases) can be done by the installer script
> itself. I did shell scripting for Archboot installer grub2-bios and
> grub2-uefi support and it is not so simple to install a bootloader,
> especially when you have very less/minimum user input, as bootloader,
> firmware and the partitioning scheme are highly integrated and all of them
> should be considered. For example in case of grub2-bios in MBR it requires
> atleast 1 MiB gap after the 512-byte MBR gap, while in GPT it requires a 1
> MiB bios_grub partition. In case of grub2-uefi (or any UEFI bootloader) a
> GPT disk with atleast one >=200 MiB FAT32 UEFISYS partition is required.
> Supporting grub2 in the installer is much more than running grub-install as
> these pre-requisites should be satisfied for which proper partitioning
> tools must be used. For both MBR and GPT GNU Parted is better when compared
> to util-linux sfdisk/cfdisk which support only MBR.
I also want to point out the grub2 upstream does not recommend (but
supports) installing grub2 to a partition. I don't whether the same applies
to grub-legacy. Syslinux by itself is simply installed to a partition with
a small code in the MBR which chainloads the syslinux partition. Syslinux
does not access files outside the partition in which it was instakked
(exceot maybe when chainloading using chain.c32), so ideally grub-legacy in
a partition can be replaced by syslinux (minus the MBR code). Users who
want a bootloader that can access (directly) files from multiple partitions
like grub-legacy does should go for grub2 as syslinux does not support that.
More information about the arch-releng