Partial charge restoration technique to improve the performance of rowhammer mitigation mechanisms

Read disturbance in modern DRAM chips is a widespread weakness that is used for breaking memory isolation, one of the fundamental building blocks of security and privacy in memory. RowHammer is a prime example of read disturbance in DRAM where repeatedly accessing (hammering) a row of DRAM cells (DRAM row) induces bitflips in physically nearby DRAM rows (victim rows). Unfortunately, shrinking technology node size exacerbates RowHammer and as such, significantly fewer accesses can induce bitflips with newer DRAM chip generations. To ensure reliable DRAM operation, state-of-the-art mitigation mechanisms restore the charge in potential victim rows (i.e., preventive refresh or charge restoration). With newer DRAM chip generations, these mechanisms perform preventive refresh more aggressively and cause larger performance overheads. To reduce this overhead without sacrificing reliability, security, and safety, our goal is to reduce time spent for preventive refreshes and understand this reduced time's impact on RowHammer. To this end, we present the first rigorous experimental study on the interactions between data retention time, refresh, and RowHammer characteristics in real DRAM chips. We test 388 DDR4 DRAM chips from three major manufacturers and observe that a preventive refresh operation's latency can be significantly reduced (by 64%) at the expense of requiring slightly more (by 0.54%) preventive refresh operations. To leverage this observation, we propose Partial Charge Restoration for Aggressive Mitigation (PaCRAM), a memory controller-based low-cost mechanism that dynamically tunes the refresh latency and adjusts the aggressiveness of existing RowHammer solutions. PaCRAM significantly improves system performance (energy efficiency) by reducing the overhead induced by five RowHammer solutions by 18.95% (14.59%), 12.28% (11.56%), 2.07% (1.15%), 2.56% (2.18%), and 5.37% (4.50%), on average.