Re: Re[2]: 'password' command in GRUB 2?

[Michal Suchanek]

2009/8/25 Vladimir 'phcoder' Serbinenko <address@hidden>:
>> Does it has the same problem as CVE-2008-3896 published for grub-legacy?
> It's completely different concern. Actually BIOS keyboard buffer
> shouldn't be a problem since only root can read raw memory and if user
> is a root he can just kexec any kernel he wants.
> I could add keyboard buffer wiping to my sendkey work but it only
> offsets the problem since same info is stored in RAM by grb anyway.
> The only solution I see for second problem is to make grub_free shred
> the memory and ensuring all sensitive fields are dynamically allocated
> and free'ed before boot (last part makes code cleaner too). Actually I
> have done some experiments with replacing grub_sprintf with
> grub_asprintf which revealed many spot of suboptimal code too.
> After all I think this is worth to do (wiping keyboard buffer and
> making grub_free wipe the memory). But it doesn't destroy the info
> which was in memory before grub booted. Doing so may take significant
> booting time (to be tested) but may be desirable in some cases.

Ever tried memtest?

Depending on the speed of the memory bus and memory size I would
expect doing a single write over the whole ram would take seconds to a
few tens of seconds.

Not unreasonable if you don't boot often but unless it's in the range
of seconds I probably would not use such feature.

However, that CVE is about grub leaving its passwords in memory.
Wiping memory used by grub should be fast - orders of magnitude faster
than loading the OS kernel for example.

Thanks

Michal

Thanks

Michal