Re: [PATCH v5] nvme: Skip trace complete_rq on host path error

[전민식]

On Thu, Mar 26, 2026 at 03:43PM +0100, hch@xxxxxx wrote:
> Yeah, the spec is not very clear on their use. It just defines the codes
> and a one sentence description, and then never mentioned again. I think
> it allows the possibility for the controller to internally complete a
> command with such status from some undefined OOB mechanism. I'm not sure
> why the host needs to have their own self-generated status codes anyway
> since it can already attach whatever arbitrary flags it wants to its
> private data, like how we use NVME_REQ_CANCELLED.
> 
> Anyway if a controller did return it for whatever reason, even if it is
> a bug, we'd want to see it in the trace.


I've confirmed what you discussed, and it seems like the approach you
proposed is a better one.
Can I send patch v7 as above?

Regrad,
Minsik Jeon