Doomsdrzej wrote:
rOn Thu, 21 Dec 2017 13:31:17 -0000 (UTC), John Doe
wrote:

Doomsdrzej wrote:

Paul wrote:
Doomsdrzej wrote:
https://www.phoronix.com/scan.php?pa..._item&px=Ubunt
u-17.10-BIOS-Corrupter

Linux corrupts your BIOS. Brilliant stuff.
Before high-fiving yourself, also consider that UEFI has a
checkered history.

https://mjg59.dreamwidth.org/25091.html

"Samsung can end up bricked

And I remember when I thought UEFI would make our lives
simpler and reduce the amount of ridiculous hacks we
needed. Sigh."

There was also a case in Windows, where some Windows thing
modifies a setting in the BIOS - and later when the user
gets into the BIOS, there *no* way to put it back. So even
Windows has a means of triggering trap-door behavior, by
flipping something in the BIOS, that cannot be corrected
from a BIOS setup screen.
UEFI is just a bad bad idea. That's the message.
I have no doubt of it. I tend to disable it entirely
whenever I install Linux on this machine. It's always more
trouble than it's worth.

I *think* I'm covered, because my computer has a special
USB port, with a flasher function built right in. You plug
in a USB stick with a named BIOS file on it, push a button,
and the image is loaded into the BIOS (there is a separate
chip on the motherboard handling this). The function works
so well, that even if the CPU is not in the CPU socket and
it's "just a motherboard", the flashing function still
works. You can buy my motherboard, connect an ATX power
supply, plug in a USB stick, flash the BIOS, power off...
and insert a previously-unsupported CPU and have it work.

I haven't needed to use that, but that's my "insurance
policy" in times of trouble.
A very cool feature, I have to admit. If anything, it's
protection against malware like the dreaded CIH virus of
1999.

The traditional BIOS, also used to write to itself, but not
anywhere nearly as badly designed as UEFI. It has a
microcode cache (I loaded mine manually on my P2B-S), it
has DMI/ESCD (mostly innocuous). Whereas UEFI can be
bricked, just via the "NVRAM storage in flash" feature, an
area used to hold various variables.
I had a motherboard a few years ago which had two BIOS
chips. If the first one gets corrupted, the second is used
as a failsafe. I believe the motherboard's brand was DFI. I
don't know if their motherboards are any good but that
feature, at the very least, was very welcome.
That is GIGABYTE DualBIOS™

After losing my last motherboard to a BIOS problem, I made it a
point to get one. I suppose Paul's version would work too.

Maybe Gigabyte does it too, but I checked and the model I had was a
DFI LANParty for AMD Athlon XP processors. I ran an Athlon 3500+ and
then 4800+ on it.

Gigabyte invented the Dual BIOS. Not Diamond Flower.
AFAIK, the Gigabyte usage pre-dates DFI usage.

One thing you should know about the Gigabyte dual BIOS, is
it's one boot block, and two main code blocks. If the
boot block gets zorched, she ain't gonna boot. It needs
an intact (single) boot block, to select between two
main code block images. That means it is not fully
redundant in the way that people might expect.
Normally, main code blocks have a checksum, so the
boot block can tell whether a main code block is
bad or not.

I don't know if DFI used the same implementation or not.

Generally speaking, with a computer industry implementation,
there's still the possibility of brick-age. The computer
industry doesn't like to add a lot of custom logic
for features like this, for cost reasons. If a feature
exists in a motherboard LSI device, fine, you get it for
free. If the logic must be provided "out-board", then
nobody will do that. You generally don't find PALs
or FPGAs on consumer motherboards.

We used dual BIOS at work, however ours had two
full images in separate flash devices, and it had
a hardware arbiter to switch from one image to
the other. After two reboots, you would have tested
both of them, if one was in a failure state and
wouldn't boot. (This is in a piece of communications
equipment, not a PC Compatible computer.) I couldn't
tell you who has a patent on that idea. Hell, it might
have been used in a spacecraft in the 1960's
for all I know.

If Gigabyte had a patent on theirs, it's hard to see
how DFI would have gotten away with it. Or vice versa.

Paul