Proposal for fixing bootinfoscript when grub v2.00 sits on the MBR

Proposal for fixing bootinfoscript when grub v2.00 sits on the MBR

[Arbiel (gmx)]

Hi

Apparently, you are in charge of maintaining bootinfoscript.

It turns out that the script does not correctly discriminate grub2's
versions 1.99 and 2.00.

To understand why, I've studied the release available for download at
http://sourceforge.net/p/bootinfoscript/code/ci/master/tree/
and used by "boot-repair", a boot repairing tool written by YannUbuntu.

I found out that the mistake comes from the fact that bootinfoscript
looks in core.img for a 6-byte value of "d1 e9 df fe ff ff" and, when
found, looks at the following two bytes for either "66 90" or "8d b6" to
deduce that grub2's version is 2.00. For any other value, it deduces the
version to be 1.99, whereas the latter should be deduced only from a "00
00" value, if the comments inside the script's code are reliable.

I've installed grub2 v2.00 from a Ubuntu distribution and it happens
that the two bytes following the "d1 e9 df fe ff ff" token hold a third
value, namely "8d 76". I obviously do not know whether this third value
is a "good" value to be considered as a grub v2.00 token or if the
search for "d1 e9 df fe ff ff" should be continued until it reaches a
"good" "00 00", "66 90" or "8d b6" value into the two following bytes.

Could you please investigate this issue, and possibly update
bootinfoscript ?

Thank's in advance.

Arbiel

Re: Proposal for fixing bootinfoscript when grub v2.00 sits on the MBR

[Andrey Borzenkov]

В Sat, 15 Feb 2014 15:48:03 +0100
"Arbiel (gmx)" <arbiel.perlacremaz@gmx.fr> пишет:

> Hi
> 
> Apparently, you are in charge of maintaining bootinfoscript.
> 
> It turns out that the script does not correctly discriminate grub2's
> versions 1.99 and 2.00.
> 

Botinfoscript is unrelated to grub development.

You can find version with my fixes submitted upstream but never applied
at https://github.com/arvidjaar/bootinfoscript

It should correctly parse current master as well as offer some sugar on
top (like parsing embedded prefix or config). I may add listing of
modules as well.

> To understand why, I've studied the release available for download at
> http://sourceforge.net/p/bootinfoscript/code/ci/master/tree/
> and used by "boot-repair", a boot repairing tool written by YannUbuntu.
> 
> I found out that the mistake comes from the fact that bootinfoscript
> looks in core.img for a 6-byte value of "d1 e9 df fe ff ff" and, when
> found, looks at the following two bytes for either "66 90" or "8d b6" to
> deduce that grub2's version is 2.00. For any other value, it deduces the
> version to be 1.99, whereas the latter should be deduced only from a "00
> 00" value, if the comments inside the script's code are reliable.
> 
> I've installed grub2 v2.00 from a Ubuntu distribution and it happens
> that the two bytes following the "d1 e9 df fe ff ff" token hold a third
> value, namely "8d 76". I obviously do not know whether this third value
> is a "good" value to be considered as a grub v2.00 token or if the
> search for "d1 e9 df fe ff ff" should be continued until it reaches a
> "good" "00 00", "66 90" or "8d b6" value into the two following bytes.
> 
> Could you please investigate this issue, and possibly update
> bootinfoscript ?
> 
> Thank's in advance.
> 
> Arbiel
> 
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel