Dual-Level Memory Refresh for Targeted Rowhammer Mitigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory technologies lack effective, scalable, and efficient methods to mitigate rowhammer attacks, which corrupt data in adjacent memory rows due to insufficient tracking and resource-intensive broad swath refreshing.
Innovation Solution
Implementing dual-level Refresh Management (RFM) that tracks and refreshes memory rows and banks separately, using access counts to identify and reset affected rows and banks, with threshold-based refresh commands to mitigate rowhammer attacks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If broad swath refreshing is performed to mitigate rowhammer attacks, then data integrity is improved, but resource consumption and time loss increase
Solution Approach 1:
The patent applies local quality by performing refresh operations only on specific victim rows adjacent to attacked rows rather than refreshing entire memory banks. The memory controller identifies attacked rows through access pattern monitoring and selectively refreshes only the necessary adjacent rows, making the refresh operation localized to the affected area and reducing overall resource consumption.
Solution Approach 2:
The patent segments the memory refresh operation into targeted row-level refreshes based on attack identification. Instead of treating the entire memory bank as a single unit requiring refresh, the system divides the refresh operation into discrete segments corresponding to specific victim rows, allowing selective refreshing that reduces the scope of the operation and conserves resources.
2Reliability
If broad swath refreshing is performed to mitigate rowhammer attacks, then data integrity is improved, but operation time increases
Solution Approach 1:
The patent reduces operation time by applying local quality to the refresh process. Instead of refreshing entire memory banks, the system identifies specific victim rows adjacent to attacked rows and performs refresh operations only on those localized areas. This selective approach significantly reduces the time required compared to comprehensive bank-level refresh operations.
Solution Approach 2:
The patent implements preliminary action by monitoring and identifying attacked rows before the actual refresh operation. The memory controller continuously tracks access patterns and identifies rows under attack, allowing it to prepare and execute targeted refresh operations on victim rows immediately after detection, rather than waiting for broader conditions to trigger a comprehensive refresh.
3Measurement precision
If access tracking is implemented to identify victim rows, then refresh precision is improved, but device complexity increases
Solution Approach 1:
The patent employs feedback mechanisms where the memory controller monitors row access patterns and uses this information to identify attacked rows. The access count information feeds back to the controller, which then determines which rows are victims and requires refresh. This feedback loop enables precise identification of victim rows without requiring overly complex external tracking mechanisms.
Solution Approach 2:
The memory system performs self-service through built-in access pattern monitoring capabilities within the memory controller. The controller inherently tracks row access counts and identifies attacked rows using its own internal resources, eliminating the need for separate external tracking devices or complex additional circuitry. The system uses its existing control infrastructure to perform the tracking function.
Data Source
AI summary
Methods, systems, and devices related to determining whether a target address of a memory array associated with an access request is stored in a CAM. If the target address is stored in the CAM, the CAM may be updated to increment an access count of a target row corresponding to the target address. If the target row exceeds a first threshold value, rows of the memory array directly adjacent to the target row may be refreshed. If the target address is not stored in the CAM, the target address may be written to the CAM. The CAM may be updated to increment an access count of an address of a bank including the target row corresponding to the target address.