Buggy SSD Hard Drives Could Self-Destruct in 3 Years

A hard drive manufacturer has warned its solid state drives could effectively self-destruct after a little less than four years of total operation. HPE (Hewlett Packard Enterprise) says users need to apply a firmware fix immediately, while it's another reminder of the importance of backups.

20 different HPE models are affected, all in the SAS range, which are hard drives typically meant for enterprise. All are solid state drives, which have no moving parts. This makes them more expensive and much faster than traditional hard drives. (Source: hpe.com)

The problem isn't with the actual drives itself, but rather the firmware, which is a small program that allows the device to communicate with the computer. HPE says it has now discovered a bug that means the SSD will irretrievably fail after 32,768 total hours of operation (3 years, 270 days and eight hours.) Once the drive fails, the data can't be recovered.

Firmware Updates Coming in Two Stages

HPE is rolling out firmware updates for the 20 models, with some starting this week and others delayed until December 9, 2019. It says those models getting the later update were released recently enough that it's not possible for them to hit their inadvertent expiration date.

According to HPE, "By disregarding this notification and not performing the recommended resolution, the customer accepts the risk of incurring future related errors." Whether that's a legally valid disclaimer is open to debate.

Clock Runs Out Of Time

HPE hasn't explained exactly why the bug has occurred, but tech analysts believe the number is no coincidence. It appears almost certain that the firmware is using a short signed integer. That means one that can be expressed in binary code using only 16 characters (bits) and allows for both positive and negative numbers. (Source: mql4.com)

The lowest possible short signed integer is -32,768. That means the firmware would hit that limit if it was trying to track how many hours it had been operating and expressed it as "how many hours ago it started."

It's a similar problem to one that will hit 32-bit computers in 2038. That will be 2,147,483,647 seconds after 1 January 1970, the reference point for calculating date and time on such computers. The problem is that number is the biggest that can be stored in the space allotted for time, which means computers will begin reading the number as negative and screwing up the date calculation.

Fortunately, unlike with HEP's drives, there's enough time to mitigate the problem and it's likely most such computers will no longer be in use at that point.

What's Your Opinion?

Do you have an HPE drive? Are you surprised the firmware developers have made such an apparent error? Do you think HPE should be held responsible for any data loss?