I was trying an alternate Spectre/Meltdown fix a while back, in which instead of manipulating the TLB, I just ran the kernel on one core and userspace on the other, using MONITOR/MWAIT to swap between them. Looked promising! I think I got 12ns context switching?https://twitter.com/Nick_Craver/status/1009822483686584320 …
-
Show this thread
-
To be clear, one virtual core vs. the other in HT, on a Xeon Broadwell arch. And I don't think I got MONITOR/MWAIT working precisely, I think I was just polling. Anyway that's by far the lowest latency cross-process monitor I saw so of course there's Spectre/Meltdown bugs.
1 reply 2 retweets 3 likesShow this thread -
You should just assume, for at least the next year or two, if you *really* want two security domains, get two computers. But don't be absolutist either, which is a common failure mode itself. And of course there's huge perf implications to disabling HT, as
@erratarob mentions.2 replies 3 retweets 5 likesShow this thread
Replying to @dakami @ErrataRob
And the cloud cried.
12:54 AM - 23 Jun 2018
0 replies
0 retweets
0 likes
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.