[Libre-soc-dev] Raptor announces FPGA BMC

[Luke Kenneth Casson Leighton]

On Friday, January 22, 2021, Lauri Kasanen <cand at gmx.com> wrote:

>
> A BMC runs as long as the computer is plugged in, even when the main
> computer is off. Besides boot it provides remote management etc.

yehyeh.  i have been speaking offlist with Tim about us doing a BMC ASIC,
to replace the ASpeed 2600.  OpenBMC has the firmware implementing IMPI etc.

my point is that these are not performance critical tasks, and they are
managing multiple 200W+ POWER9 processors on easily 1000W systems where
even a 5W BMC solution in FPGA is a drop in the ocean.

much more interestingly from a commercial perspective is that Supermicro
motherboards were rumoured by bloomberg to contain a pullup resistor
replaced by a specialist IC.

(this does NOT mean that it was Supermicro who installed it: see "CIA
Inside" powerpoint slides of the CIA intercepting CISCO routers en-route to
customers and installing spying firmware, AFTER the routers left CISCO
premises)

a proof of concept was developed that confirmed that replacing a tiny
pullup resistor on the BMC PCB was sufficient to "override" the line driver
and compromise the BMC firmware.

having a replacement that a *justifiably* paranoid customer can replace on
their own cognizance by following instructions using an off-the-shelf
commodity FPGA board is extremely valuable.

not only that but if we are approached by such customers who say "prove to
us your solution is secure, we will not buy until you do" we can say to
them, "go get the FPGA version, don't trust us, do the proof yourself oh
and incidentally the FPGA version actually does the job, you can run that
*right now*, buying you time to do the evaluation of our ASIC".

this alone will give them 95% confidence that we are not bullshitters just
mouthing the word "trust" in order to get their money.

l.






-- 
---
crowd-funded eco-conscious hardware: https://www.crowdsupply.com/eoma68