7 ms·
Maybe I'm misunderstanding something. I thought the attack was from returning from a fault and reading back cached data from before the fault. So the branch p
by CoolGuySteve 9y ago
Maybe I'm misunderstanding something. I thought the attack was from returning from a fault and reading back cached data from before the fault.
So the branch predictor only needs to be cleared after any interrupt or syscall return.
- kbwt 9y agoRead up on the "out of bounds" and "indirect branch" variants. Spectre doesn't have much to do with mode switches.
- bzbarsky 9y agoThere are at least three separate attacks involved, and the reporting has been pretty confusing. See http://robert.ocallahan.org/2018/01/meltdownspectre-needs-better-disclosure.html http://robert.ocallahan.org/2018/01/meltdownspectre-needs-be...