So the #meldown steps are:
1. Load a byte of memory from kernel. This crashes.
2. Use that byte to load one of 256 cache-lines. This happens before the crash is registered, so while the data is discarded, the data is still cached.
3. Measure which of the 256 cache-lines are fast
-
Show this thread
-
Replying to @ErrataRob
1. Shouldn't crash if you do it in the speculative context, i.e. branch misprediction. And it should still work. This is what I was trying and failing on, but should it work
@tehjh ?1 reply 2 retweets 2 likes -
Replying to @scarybeasts @ErrataRob
yes, putting a mispredicted branch in front should work (and both my PoC and the test harness that
@dougallj posted do that)1 reply 3 retweets 3 likes
Replying to @tehjh @scarybeasts and
FWIW, "load one of 256 cachelines" never worked quite right for me, although maybe I was just missing something - always just mapping to one of two cachelines worked better
3:08 AM - 4 Jan 2018
0 replies
3 retweets
3 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.