Allow me to summarize x86 side channel attacks: Spectre v1: speculation is insecure by design Spectre v2: secure branch prediction matters Meltdown: Intel are dumbasses L1TF: Intel are monumental, inexcusable dumbasses PortSmash: hyperthreading is insecure by design
-
-
Ultimately, computing needs to come to terms with the fact that any time there is *any* shared resource, that's ripe for side channels. This includes main memory and peripheral devices. It's worse and more insidious the lower level and more tightly coupled you make it, like HT.
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
NUMA NUMA yay?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
They arn’t going away as in future processors might still be susceptible?
-
As in they *will* be susceptible. It's inherent. They cannot be fixed without killing performance.
End of conversation
New conversation -
-
-
But the Intel decided DDIO allowing the NIC to write directly into L3 cache was a good idea… ¯\_(ツ)_/¯
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Sparc has (had ?) up to 8 threads per core.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Well, "should" means nobody could proof it in any way yet, and could also mean they are not affected at all.
-
No, they *will* be affected. It's fundamental to how HT works.
End of conversation
New conversation -
-
-
I wonder if this could be adopted to the Niagara T1/T2. So far they were unaffected by every Spectre and similar variant. That would be the first thing.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.