Re: Which partitioning schemes should be supported by GRUB?

[Vladimir 'φ-coder/phcoder' Serbinenko]

On 06/06/2010 08:02 PM, Grégoire Sutre wrote:
> Hi,
>
> Tests of GRUB on NetBSD (and FreeBSD) have raised several issues (most
> of them reported on the list) regarding partition detection.  However,
> I have the feeling that some of these issues are not considered as real
> issues since the test configuration is not supported by GRUB.  This
> surprises me since I naively thought that most user configurations
> should be supported.
>
> So I ask the question: Which partitioning schemes are (or shall be)
> supported by GRUB on i386-pc (with standard BIOS)?
There are two parts of this question:
1) Which partition schemes should GRUB be able to read modules and
payloads from ? It's platform-indepedent and 2 conditions apply:
- Usage. There are OS which are able to boot from such OS and such
configuration isn't considered obscure by them.
- Non-confusability. The risk of false positive of this partition config
which would prevent normal function is small.
If at least one condition is met it's worth considering. If both
conditions are met it should be supported.
2) Support for embedding.
Embedding is a potentially dangerous operation so we have to be
cautious. Using a dedicated embedding partition if it can be
unambiguously identified as such is a sane solution.
>
> To start the discussion, I'll focus on a few examples (the list is
> surely not exhaustive).  Maybe some configurations simply cannot exist,
> in which case please let me know.
>
> 1. A single top-level partition map
>    (a) MS-DOS
>    (b) GPT
>    (c) BSD disklabel
>    (d) Apple partition map
>    (e) Sun label
>
> 2. Hybrid: top-level MS-DOS + another *top-level* partition map
>    (a) MS-DOS + GPT
>        (i.e. GPT + at-least one non 0xEE MS-DOS partition)
>    (b) MS-DOS + BSD
>    (c) ...
>
> If I read the code correctly, grub-setup (on i386-pc) only supports
> 1(a) and 1(b).  However, on NetBSD, 1(c) is very common, and 2(b) is
> not rare.  Also, some OSes are fine with 2(a), e.g. FreeBSD.
>
> Personally, I would rather support all possible configurations, unless
> some technical reason prevents it.  So grub-setup would not test for
> some specific configurations, but would instead use a generic
> (and simple) approach.  If it fails, it should be for a good reason,
> and not because "No DOS-style partitions [were] found".
>
> What's your opinion?
>
> Grégoire
>
>
> _______________________________________________
> Grub-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/grub-devel
>


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko

signature.asc
Description: OpenPGP digital signature