Hector Martin

Rowhammer is Intel's fault too: they're largely the reason why desktop systems do not use ECC memory (because they market it as premium). It is utter insanity that at current DRAM densities we pretend memory is flawless. Every other storage tech has used ECC for decades.

(I know ECC as it exists today for DRAM does not completely mitigate rowhammer, but it certainly helps a lot)

Rowhammer has been solved, BTW. There are algorithms that can be used to proactively refresh vulnerable rows. Combined with faster refresh intervals, it's an effective mitigation.

if you mean TRR, then, that is trivial to bypass (we did early this year) we even bypassed it while running with doubled refresh rate... so, maybe it is solved, but not the way you describe here ;)

Yeah I am not sure how we solve rowhammer within current economic constraints. Memory controller is probably the best place to do it.

Exactly what I've been talking about in quite a few of my talks this year: Rowhammer is an optimization problem. Similar for other HW/SW issues. We want computers to be fast, cheap, and reliable at the same time. Naturally we must run into edge cases from time to time.

Right, we *understand* what causes the disturbances, so we can counter it. But you also need a backbone of ECC to at least have something against random errors and a chance at detecting targeted attacks even if you can't correct them.

We do not understand the effects well enough. Else we wouldn't come up with things like TRR. ECC is much more effective than TRR or any similar mechanism to refresh "nearby" rows. While surely possible, I've never seen any bit flip on ECC memory.