TRIM bug in Samsung firmware?

[HammerSandwich]

Just saw the news at TechReport. If accurate, oh shit!

 

[evilsofa]

This seems to be for Linux; I've seen at least one comment that Windows implementation of TRIM may be somehow different and unaffected by this (with no explanation). You'd think Samsung 840 Pro users would have complained about this by now if it was such a serious problem.

Would there be a process for Windows Samsung SSD users to check for files silently corrupted this way?

 

[NoOther]

It is a lot of panicky noises about a fairly unique situation. The problem seems to be caused by aggressive use of TRIM in a production/enterprise Linux server environment, where data is being constantly updated. A lot of people are freaking out about their home systems which will likely not see the same problems, especially if they are not constantly writing/re-writing data on their drives and using TRIM as a consistent cron job.

For enterprise companies trying to get cheap drives to do enterprise level work, sure this can be an issue. But then one has to ask, why are they buying consumer drives to do enterprise work?

 

[olavgg]

This is not about system load. This is about unqueued trim, which Windows 7 & 8 doesn't support. Affected systems are all modern operating systems, and that may also include vmware.

 

[Quartz-1]

For enterprise companies trying to get cheap drives to do enterprise level work, sure this can be an issue.

They're using Samsung's enterprise SSDs.

 

[mwroobel]

The problem seems to be at least reduced if you add the specific drives to the blacklist and recompile.

 

[AlienTech]

Some Crucial models also have this bug so it is not unique. Also is there data loss? Other wise it is just a speed issue. Samsung seems to be like apple, lots of groupies with hyped up expectations. Good for samsung but not so good for anyone else. Their drives are also priced higher than anything else. If samsung wanted to make a difference they would price their stuff like the others instead of depending on their hype and charging more. They charge as much for TLC as others do for MLC. And yet this is what was so hyped about for lower prices and higher capacity drives.

 

[evilsofa]

This is not about system load. This is about unqueued trim, which Windows 7 & 8 doesn't support. Affected systems are all modern operating systems, and that may also include vmware.

You've got it backwards. Windows and older Linux builds use unqueued TRIM. Queued TRIM is the new thing. This problem was reported on Linux systems using unqueued TRIM - see the update in the blogpost.

Yet, there are no reports of any such issues with TRIM with Samsung SSDs on Windows, and the Samsung 840 Pro was extremely popular for a very long time with a lot of very saavy users who would most certainly have noticed such a catastrophic file corruption issue by now. So something's different about the way Linux - or, at least, that particular Linux build - is using TRIM. There's going to be a bunch of arguments and debate and fingerpointing until they sort this out, but there's no reason for Windows users to panic about this.

 

[PliotronX]

They charge as much for TLC as others do for MLC. And yet this is what was so hyped about for lower prices and higher capacity drives.

I thought I was the only one who saw that unfolding. Cost savings and new tech almost never go hand-in-hand.

 

[NoOther]

They're using Samsung's enterprise SSDs.

I was not aware that the 840 Pro and 850 Pro were enterprise drives? That seems to be what most users are worried about here. There are also a number of issues with what they were doing and how they were doing it.

 

[evilsofa]

Update in the blog post shows Samsung is very interested in finding out what happened:

"UPDATE June 19:
On Monday June 22, the engineering team from Samsung is going analyze one of our servers in Singapore and if nothing will be found on-site, the server will travel to Samsung HQ in Korea for further analysis."