Re: [PATCH RFC] RDMA/core: Store zero GIDs in some cases

[Chuck Lever III <chuck.lever@oracle.com>]

> On Apr 28, 2023, at 9:47 AM, Jason Gunthorpe <jgg@ziepe.ca> wrote:
> 
> On Fri, Apr 28, 2023 at 01:42:24PM +0000, Chuck Lever III wrote:
>> 
>> 
>>> On Apr 28, 2023, at 9:39 AM, Jason Gunthorpe <jgg@ziepe.ca> wrote:
>>> 
>>> On Thu, Apr 27, 2023 at 01:14:43PM -0400, Chuck Lever wrote:
>>>> From: Bernard Metzler <bmt@zurich.ibm.com>
>>>> 
>>>> Tunnel devices have zero GIDs, so skip the zero GID check when
>>>> setting up soft iWARP over a tunnel device.
>>> 
>>> Huh? Why? How does that make any sense?
>> 
>> Read it as a cry for help.
>> 
>> The scenario is attempting to set up a soft iWARP device
>> with a slave that is a tunnel device. The set up seems to
>> work, but when connecting, the ULP gets an ADDR_ERROR
>> because the setup did not add an entry to the GID table.
> 
> Don't assign a 0 IP to the tunnel?

That's a little cryptic... can you expand?

Right now I have a Tailscale VPN device with assigned IP
addresses:

3: tailscale0: <POINTOPOINT,MULTICAST,NOARP,UP,LOWER_UP> mtu 1280 qdisc fq_codel state UNKNOWN group default qlen 500
    link/none      inet 100.64.0.16/32 scope global tailscale0
       valid_lft forever preferred_lft forever
    inet6 fd7a:115c:a1e0::10/128 scope global         valid_lft forever preferred_lft forever
    inet6 fe80::725c:1b6d:60ed:fce4/64 scope link stable-privacy         valid_lft forever preferred_lft forever

And after that i/f is UP, I've done this:

 $ sudo rdma link add siw0 type siw netdev tailscale0

With the patch I sent, I can do NFS/RDMA via soft iWARP through
the tunnel. I'm not at all claiming that's a good fix, but only
that this scenario is supposed to work, but currently doesn't.


--
Chuck Lever