Re:The mysterious case of the Linux Page Table Isolation pat...
by fulafel on Monday January 01, 2018 @02:28PM
(#16047569)
This is far away from "hardware fix is impossible" though. In reputable hardware, hardware vendors are expected to maintain correctness in spite of performance advances.
Also, DDR3/4 DRAM is glacially slow in latency terms, it's far from clear that there would be appreciable slowdown. There are already big latency compromises in the standardized JEDEC protocols that are not inherent in DRAM - it would be very two-faced for DRAM vendors to say they'll only trade off latency over backward compatibility or tiny cost savings, but not over correctness.
Re:The mysterious case of the Linux Page Table Isolation pat...
by olliej on Monday January 01, 2018 @04:33PM
(#16048168)
The trick I heard was to literally throttle the writes if you see repeated parallel writes or similar -- as row hammer depends on rapid writes that should be sufficient. But I also have no real idea how hardware actually works :) (magnets? how do they work?)
Re:The mysterious case of the Linux Page Table Isolation pat...
by Dylan16807 on Tuesday January 02, 2018 @06:03AM
(#16051179)
You don't need to go that far at all. Just refresh neighbors occasionally when there are tons of accesses to the same place.
ECC helps, and can be done at full rate, but isn't a complete solution for all possible problems. And anything you do in hardware at full RAM speed is expensive.