Spectre & Meltdown

Fred Cisin cisin at xenosoft.com
Sat Jan 13 17:24:31 CST 2018


On Sat, 13 Jan 2018, Murray McCullough via cctalk wrote:
> I wrote about Spectre and Meltdown recently: INTEL took its time to inform
> the world! Did it inform the world back in earlier days about potential
> flaws? Not to blame INTEL only: What about Zilog, etc.? Or did pre-Internet
> era protect us computer-classic users? What about running emulation
> software as I’ve been doing with ADAM?
> Happy computing!

Few emulations are exact enough to duplicate all bugs.
Q: Should an emulator do an exact imitation, or should it work the way 
that it is s'posed to?   (behavior? or specs?)

Pre-internet protected against most web based malware.  But, there are 
instances of virus software ever since people exchanged files and disks.
(I'm unaware of any punch-card attacks, but trojans were possible when 
people used prior subroutines)
Most prevalent were boot-sector virus attacks and executable file virus 
attacks.  As software became too eager to help provide dancing kangaroos 
and yodelling jellyfish, harmful macros in "productivity software" macro 
capabilities also started to surface.

Internet made it much easier to acquire a trojan that would mess you up.
Although reduction in sneaker-net has virtually eliminated boot-sector 
spread.


How fast SHOULD the public response be?
If they become aware of that kind of flaw, and can delay public knowledge 
until they have patches, they significantly reduce the risk of actual 
instances of malware using the exploits.
Note: AFAIK, no examples of actual use of Spectre nor Meltdown have yet 
been encountered.
If Microsoft had been in less of a rush, would they still have shipped 
patches that gave a BSOD with AMD processors?

After public announcement, there ARE people actively working on developing 
malware using it.

Similarly, after the Michelangelo Virus media panic, one of the variants 
later encountered was a fairly obvious "wannabe" consisting of "Stoned" 
patched to behave like the publicized Michelangelo behavior.  The 
"thousands or millions of computers will be destroyed" was bogus.
(BTW, the name "Michelangelo" was based on looking at a calendar to see 
what was special about March 6.  If McAfee had had a Texas calendar, 
instead of a KQED (PBS) one, then it would have been named "Alamo")


Intel made some mistakes in handling the FDIV bug. First, they made the 
assumption that the bug would be amazingly rarely encountered due to their 
calculations of probability of randomly hitting "winning" combinations of 
numerator and denominator, but failed to allow for any of the "winning" 
numbers happening to be more commonly used.

THEN, they offered replacements to anybody who could PROVE that it 
actually affected their use of the machine.  A more appropriate response 
would have been, "We WILL replace all affected processors!  BUT, there 
aren't enough in stock right now to handle all immediately, so we will 
START by replacing those for all who can prove that they are affected, and 
then get to all others as we can manufacture more suitable replacements."
(Perhaps the majority of people would have already replaced their machine 
before their turn came around!  What is it? "a new machine every 18 
months"?)

Many of the general public had been led to believe that it would produce 
completely WRONG results, rather than the LOW ORDER bits of the mantissa 
being incorrect.  No, it was not capable of "causing the wrong amount of 
sales tax to be charged!"



More information about the cctalk mailing list