Re: Meltdown and spectre

  • From: Hans Forbrich <fuzzy.graybeard@xxxxxxxxx>
  • To: oracle-l@xxxxxxxxxxxxx
  • Date: Fri, 5 Jan 2018 16:51:19 -0700

On 2018-01-05 2:33 PM, Reen, Elizabeth (Redacted sender elizabeth.reen for DMARC) wrote:

I have a background in system engineering.  I don’t get how a chip can be exploited.  What code can be hacked there?

For speculative execution, a command is executed that MIGHT be required.  That command might ask to move stuff into some portion of memory, or need a specific page moved in. If that command is then rolled back, what happens to the memory that it just filled?  (Hint: it's still filled in, perhaps with a password.)  Back in the day (early 90s) when this stuff was dreamt up, the idea of flushing that memory on command rollback would not have been a concern - hacking was for fun, not profit, in those days.  It's not actually the code being hacked, as much as a side effect that is not properly handled.

It wasn't just the hardware guys, either.  We s/w devs were pretty sloppy about things like end-of-arrays and random pointers in our code, and few people worried about (or even understood) what happened at the chip level.  (Remember why Java came into being?)


Other related posts: