Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.33 as permitted sender) client-ip=23.128.96.33;
Date:   Wed, 6 Dec 2023 09:54:36 +0100
From:   Christoph Hellwig <hch@lst.de>
To:     Daniel Wagner <dwagner@suse.de>
Cc:     linux-nvme@lists.infradead.org, linux-kernel@vger.kernel.org,
        Keith Busch <kbusch@kernel.org>,
        Christoph Hellwig <hch@lst.de>,
        Sagi Grimberg <sagi@grimberg.me>,
        Hannes Reinecke <hare@suse.de>
Subject: Re: [PATCH v3 3/4] nvme: move ns id info to struct nvme_ns_head
Message-ID: <20231206085436.GB24484@lst.de>
References: <20231206081244.32733-1-dwagner@suse.de> <20231206081244.32733-4-dwagner@suse.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20231206081244.32733-4-dwagner@suse.de>
User-Agent: Mutt/1.5.17 (2007-11-01)
Precedence: bulk

On Wed, Dec 06, 2023 at 09:12:43AM +0100, Daniel Wagner wrote:
> Move the namesapce info to struct nvme_ns_head, because it's the same
> for all associated namespaces.
> 
> The head pointer is accessible from the ns pointer so we could just
> update all places with ns->x to ns->head->x. While this is okay for the
> slow path,

Do you have any data to show that it matters?  All the I/O command
setup functions already access the ns_head for ->ns_id, so looking
at more fields can't really make any difference.

If we have a good argument about reducing the pointer chasing I'm all
for it, but please as a separate, well-documented commit that also
explains the tradeoffs for the newly added lookups this adds in a few
places.