FYI:
https://bugzilla.kernel.org/show_bug.cgi?id=42630
On Sat, Jan 21, 2012 at 00:34, Maxim Kammerer <mk@???> wrote:
> Well, it seems that even with the reminder sent about 8 days ago, no
> one on the linux-mm mailing list cares. I tried OFTC/#mm several times
> as well, but never saw any response. I think I will stay with the
> 32/64-bit KEXEC split in the meanwhile. Will also ask memtest86+
> author about the difficulties of adapting it for the task.
>
> Best regards,
> Maxim
>
> On Mon, Dec 26, 2011 at 04:18, Maxim Kammerer <mk@???> wrote:
>> 1. On 32-bit x86, memtest=n tests only LOWMEM memory (~ 895 MiB),
>> HIGHMEM is ignored
>>
>> 2. On 3.0.4-hardened-r5, HIGHMEM memory (HIGHMEM64G in my tests) is
>> apparently ignored during memtest. Looking at arch/x86/mm/memtest.c,
>> no special mapping is performed (kmap/kunmap?), so it seems that at
>> most ~895 MiB can be tested in 32-bit x86 kernels. This might not
>> appear like an important issue (as there are other memory testing
>> tools available), but memtest is extremely useful for anti-forensic
>> memory wiping on shutdown/reboot in security-oriented distributions
>> like Liberté Linux and Tails, and there is no other good substitute.
>> See, for instance, some background in Debian bug
>> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=646361.
>>
>> 3. Keywords: memtest, highmem, mm, security
>>
>> 4. Kernel version: 3.0.4-hardened-r5 (Gentoo) x86 32-bit with PAE
--
Maxim Kammerer
Liberté Linux (discussion / support:
http://dee.su/liberte-contribute)
