[PATCH] raid0: data corruption when using trim

This is by way of follow-up to the post last month (https://news.ycombinator.com/item?id=9723066) from the Algolia blog about data loss on some SSDs (notably Samsung).

The original Algolia article has since been updated to document the work that Samsung has done to identify and rectify the issue that was originally attributed to their SSD firmware. The root cause now appears to be a Linux kernel issue that only appeared with very heavy discard loads, and from my reading of the patch discussion it also would only apply when LVM-based RAID was in use over the SSDs too. The linked article is the resulting patch submission from a Samsung Linux kernel developer.

There's still the kernel blacklist of devices that don't handle queued-TRIM, but that's unrelated to this issue.

[I have no relationship with any of the parties involved - I was just looking to buy a new SSD yesterday and thought I'd take another look at the issues to see which brands were considered 'safe'.]