From: Stefan Sperling <stsp@stsp.name>
Subject: Re: The 6th batch of qwz commits
To: tech@openbsd.org
Date: Thu, 1 Oct 2026 13:43:35 +0200

(moved to tech@ from ports@)

On Thu, Oct 01, 2026 at 01:29:40PM +0200, Kirill A. Korinsky wrote:
> Stefan,
> 
> here the 6th batch of qwz commits.
> 
> It is quite mechanical with exception of one commit where I fix REO layout
> to that I had discovered in ath12k, and the next one fixes its lifetime.
> 
> It was tested on my device and makes it works better, but I may make typos.
> 
> Ok?

All OK by me, expect the last one of them:

> Subject: [PATCH 7/7] sys/qwz: fix REO queue lifetime
> 
> Track REO completions before publication and wait for peer unmap,
> deletion, and cache flushes before reusing queue DMA. Submission errors
> and timeouts retain ownership; hardware failures block reuse until cold
> cleanup. HAL error conventions and flush semantics follow ath12k;
> tracking and the reuse barrier adapt qwz's retained pool.

This diff is too complicated for me to properly review right now.
Since you have tested it and it seems to be a needed improvement to get
this driver to work properly, I won't object to it going in without my OK.
But I cannot fully endorse this diff.

I don't like the conversion of some "return errno" to "return -errno".
The Linux project uses this negative error code convention, but we do not.