SSD Bad Sector Repair via Cache Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In SSDs, when an uncorrectable error occurs across an entire page, existing repair methods fail to successfully write data back to the affected sector, leading to RAID degradation and system reliability issues due to inconsistent write units between the host and NAND flash.
Innovation Solution
A method and apparatus that allocate cache storage space matching the physical page size, write preset data when reading fails, update this data according to the sector's logical block address, and write it back to the SSD, setting flags to indicate data validity, allowing for successful sector repair even if the error spans an entire page.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a pre-padding solution is used to write data to NAND flash, then data can be successfully written to the flash chip, but the minimum write unit inconsistency between host (sector) and NAND flash (page) causes repair failures when UNC occurs in an entire page
Solution Approach 1:
The patent segments the page-level repair problem into sector-level operations by allocating separate cache storage spaces for each sector within a page. When a UNC error occurs, the system can repair individual sectors independently rather than treating the entire page as a single unit, thus resolving the write unit inconsistency between host and flash memory.
Solution Approach 2:
The patent introduces a cache memory as an intermediary layer between the host and NAND flash. The cache stores page data and allows the system to perform sector-level read/write operations on cached data even when the original page in flash memory has UNC errors, effectively mediating the write unit mismatch between host (sector) and flash (page).
2Measurement precision
If repair is performed in sector units when UNC occurs in an entire page, then sector-level precision is achieved, but reading the entire page fails causing the write operation to quit and reply writing success falsely
Solution Approach 1:
The patent divides the page into multiple sector-level cache storage spaces, allowing the system to read and repair only the specific sector that contains the UNC error rather than attempting to read the entire page. This segmentation enables precise sector-level repair while avoiding the failure that would occur when trying to read the corrupted entire page.
Solution Approach 2:
The patent performs preliminary actions by pre-allocating cache storage spaces for all sectors and pre-loading page data into the cache before repair operations begin. This preliminary preparation ensures that when a UNC error occurs, the system can immediately perform sector-level repairs on cached data without needing to read from the corrupted flash memory during the repair process.
3Measurement precision
If the host performs read operation to verify repair success, then repair verification is achieved, but the host determines the disk is faulty and removes it causing RAID degradation
Solution Approach 1:
The cache memory serves as an intermediary that allows the host to verify repair success by reading from the cached sector data rather than directly from the flash memory. Since the cache contains the successfully repaired data, the host can accurately verify repair success without being misled by the corrupted state of the original flash memory, thus avoiding false failure detection and RAID degradation.
Data Source
AI summary
A bad sector repair method includes receiving a first write operation instruction sent by a host, reading data of a physical page including the first sector according to a logical block address (LBA) of a first sector included in the first write operation instruction, allocating storage space with a same size as the physical page in a cache and setting data in the storage space to preset data when reading the data of the physical page fails, updating preset data of a storage area corresponding to the first sector and in the storage space with data that needs to be written to the first sector, and writing data in the storage space to the physical page to repair the first sector.