A New Release Of GRUB 2.0 Fixes Security Issue

[phoronix]

Phoronix: A New Release Of GRUB 2.0 Fixes Security Issue

It was just about two weeks ago that GRUB 2.0 moved closer to release with version 1.97 finally making it out the door after being just shy of two years without a new release. GRUB 1.97 brought Mac OS X kernel support, EXT4 file-system support, support for RAID 4/6/10, high-resolution timer support, support for loading FreeBSD/NetBSD/OpenBSD kernels, and many other changes...

http://www.phoronix.com/vr.php?view=NzY4NQ

[faemir]

GNU/Hurd support, just what I need!

But no, good thing it got patched swiftly for that fun exploit

[admax88]

Correction for the article

log-in just entering "p" or "nix" or any other part of the string -- even just one character

The bug report says differently.

GRUB accepts user input as valid password as long as user enters some first
characters of password correctly.

Not any part of the string will do, only those which are a valid prefix of the string are valid input.

Ex, if the password is phoronix,
valid input is
phoronix
phoroni
phoron
...
p

but not "nix" or "ron" or any other part.

[Peter_Cordes]

Worst strcmp implementation ever

I read this story, and though, ok, who thought it was a good idea to write their own strcmp. And how do you make that kind of mistake? Well, it turns out you make that kind of mistake by starting with a really convoluted loop in the first place:

Here's the first commit that tried to fix the bug, where you can see the old
version.

http://bzr.savannah.gnu.org/lh/grub/trunk/grub/revision/1805?start_revid=1822

corrected to, since it was using ptr2 uninitialized.

http://bzr.savannah.gnu.org/lh/grub/trunk/grub/revision/1807

The loop looks like its trying to do something if the user types more characters than the stored password. I still don't see how anyone came up with that mess, and haven't made the effort to figure out exactly how it behaves for every input. I'm a fan of the trinary operator, but that loop just made my eyes glaze over.

Fortunately Robert Millan rewrote Vladimir Serbinenko's grub_auth_strcmp function the sane way

http://bzr.savannah.gnu.org/lh/grub/trunk/grub/revision/1809?start_revid=1822

while adding a delay to defend against passwd guessing

So now it's just basically

return strcmp(input, template);

like it should have been in the first place! Don't write complicated logic when you can use standard string functions. Even in a stand-alone program like GRUB where you have to provide your own strcmp.