Re: [PATCH] nvme: remove multipath module parameter

[Sagi Grimberg]

On 13/02/2025 22:37, John Meneghini wrote:
> Keith, Christoph and Sagi,
>
> This patch has been fully tested and analyzed by Red Hat's QA group 
> and no
> unexpected side effects or regressions have been found. Both NVMe/FC 
> and NVMe/TCP
> have been tested. Our QE engineer has asked me to report this upstream.
>
> Tested-by: Marco Patalano <mpatalan@xxxxxxxxxx>
>
> Fixes: 32acab3181c7 ("nvme: implement multipath access to nvme 
> subsystems")
>
> As discussed in previous email threads, the nvme.core_multipath 
> parameter was included
> by Christoph in the original implementation of nvme/host/multipath.c 
> back in 2017
> at the request of Red Hat. At that time Red Hat was only supporting 
> DMMP multipath
> with NVMe and we needed a way to disable core native nvme multipathing 
> without
> reconfiging the kernel.
>
> The nvme.core_multipath parameter has been used by Red Hat, together 
> with some additional
> out-of-tree changes to nvme/host/core.c, to support DMMP multipath 
> with NVMe since RHEL-8.
> However, the plan all along has been to deprecate and remove support 
> for DMMP multipath with NVMe
> in RHEL and to remove this parameter. This change has been advertised 
> and communicated to our
> customers and partners, beginning with RHEL-9, through release notes 
> and other communications.
>
> In RHEL-9 we changed the default setting of this parameter from "N" to 
> "Y" to match the upstream
> kernel and the move our customers to native nvme multipath. DMMP 
> multipath w/NVMe was deprecated
> in RHEL-9 but still supported. Customers were still able to optionally 
> enabled DMMP multipath
> with NVMe if they wanted, by changing this parameter, and many of them 
> do.
>
> In the (soon to be released) RHEL-10 release DMMP multipath is longer 
> supported with NVMe and
> the out-of-tree patches needed to support DMMP multipath with NVMe 
> have been removed.
>
> Red Hat is proposing this patch to remove nvme.core_multipath 
> parameter from the kernel as we believe
> leaving this non-supported parameter in the kernel for future releases 
> will only lead to confusion.
>
> We plan to ship this patch with RHEL-10. So it would be really good if 
> we could get this
> change accepted and merged upstream, perhaps into v6.15.

Hey John,

This looks fine to me, I'm assuming this was also tested with 
CONFIG_NVME_MULTIPATH=n ?