Re: [PATCH-for-9.1] rdma: Remove RDMA subsystem and pvrdma device

[Philippe Mathieu-Daudé]

On 28/3/24 10:06, Daniel P. Berrangé wrote:
> On Wed, Mar 27, 2024 at 11:55:48AM +0100, Philippe Mathieu-Daudé wrote:
> > The whole RDMA subsystem was deprecated in commit e9a54265f5
> > ("hw/rdma: Deprecate the pvrdma device and the rdma subsystem")
> > released in v8.2. Time to remove it.
> >
> > Keep the RAM_SAVE_FLAG_HOOK definition since it might appears
> > in old migration streams.
> >
> > Remove the dependencies on libibumad and libibverbs.
>
> > Remove the generated vmw_pvrdma/ directory from linux-headers.
> >
> > Remove RDMA handling from migration.
> >
> > Remove RDMA handling in GlusterFS block driver.
>
> The RDMA support in GlusterFS is completely opaque to QEMU.
> All we have there is the CLI syntax to enable use of the
> RDMA support inside libglusterfs. I'm not convinced that
> the justification for deprecation (lack of maintanier)
> applies to this scenario.

I'll quote commit 0552ff2465 from 2016 then:

    block/gluster: deprecate rdma support

    gluster volfile server fetch happens through unix and/or tcp,
    it doesn't support volfile fetch over rdma. The rdma code may
    actually mislead, so to make sure things do not break, for now
    we fallback to tcp when requested for rdma, with a warning.

    If you are wondering how this worked all these days, its the gluster
    libgfapi code which handles anything other than unix transport as
    socket/tcp, sad but true.

> > Remove rdmacm-mux tool from contrib/.
> >
> > Remove PVRDMA device.
>
> I agree with Thomas that each functional area listed here
> should be handld in a separate patch, since they're all
> independant.
>
> With regards,
> Daniel