[PATCH 2/2] nvme: avoid possible double completions for the same request

[Jens Axboe]

On 6/8/26 9:17 AM, Keith Busch wrote:
> However, I'm really skeptical controllers actually behave this way.
> You wouldn't have a viable storage device if it was doing this.

Same, and I would be strongly opposed to including theoretical
work-arounds like this that just have fast path overhead for zero gain
on a valid device. If your device is broken, ask the manufacturer to fix
it. Let's please not burden the kernel with catering to broken
implementations.

-- 
Jens Axboe