Patent — Utilizing Host Memory Buffers for Storage Device Recoveries
Recoverable SSDs after brick events · USPTO US11487439B1
When an SSD bricks, it usually goes in the trash. This patent puts the recovery key somewhere else.
The problem
When an SSD bricks in the field, critical state — like RPMB (Replay Protected Memory Block) and FFU (Field Firmware Update) data — is locked behind the drive's own controller. If that controller is unresponsive, the drive is gone.
For consumer drives this is a return. For the industry it's a recurring cost.
The approach
Co-invented and patented a recovery mechanism that stashes a copy of the critical drive state in the Secure Host Memory Buffer (HMB) — memory the host already exposes to the drive over PCIe, but with appropriate cryptographic protection.
When the SSD goes unresponsive, the host can drive a recovery path using the HMB-resident state to bring the drive back, instead of writing it off.
Granted as USPTO US11487439B1 — Utilizing Host Memory Buffers for Storage Device Recoveries.
Stack
Embedded C++ · NVMe firmware · HMB protocol design
Outcomes
- Drives become recoverable after brick events that would previously be terminal.
- Granted as USPTO US11487439B1.
What I learned
The most useful state to copy out of the drive isn't necessarily the largest — it's the state whose loss makes the rest unrecoverable. Identifying that minimal critical set was most of the design work.
Public
Hiring for something like this?
Tell me what you're building