* nvme: Duplicate cntlid error
@ 2020-03-30 11:44 Alex Talker
2020-03-31 6:45 ` Sagi Grimberg
0 siblings, 1 reply; 4+ messages in thread
From: Alex Talker @ 2020-03-30 11:44 UTC (permalink / raw)
To: linux-nvme
Using Fedora 31 as a host I'm trying to connect to 2 nodes that host
subsystem with one and the same NQN.
So when I call:
nvme connect -t rdma -a 10.21.0.1 -n nqn.2020-03.com.example:subsystem0
device nvme0 successfully allocated but when then I call:
nvme connect -t rdma -a 10.20.0.1 -n nqn.2020-03.com.example:subsystem0
I get an error:
[239355.661527] nvme nvme1: Duplicate cntlid 1 with nvme0, rejecting
what could be the problem?
Few months before I checked connection to a single node via few ports
and it worked out just fine
but attempt of H.A. setup fails here.
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: nvme: Duplicate cntlid error
2020-03-30 11:44 nvme: Duplicate cntlid error Alex Talker
@ 2020-03-31 6:45 ` Sagi Grimberg
2020-03-31 10:37 ` Alex Talker
0 siblings, 1 reply; 4+ messages in thread
From: Sagi Grimberg @ 2020-03-31 6:45 UTC (permalink / raw)
To: Alex Talker, linux-nvme
> Using Fedora 31 as a host I'm trying to connect to 2 nodes that host
> subsystem with one and the same NQN.
> So when I call:
>
> nvme connect -t rdma -a 10.21.0.1 -n nqn.2020-03.com.example:subsystem0
>
> device nvme0 successfully allocated but when then I call:
>
> nvme connect -t rdma -a 10.20.0.1 -n nqn.2020-03.com.example:subsystem0
>
> I get an error:
>
> [239355.661527] nvme nvme1: Duplicate cntlid 1 with nvme0, rejecting
>
> what could be the problem?
> Few months before I checked connection to a single node via few ports
> and it worked out just fine
> but attempt of H.A. setup fails here.
Your subsystem present two controllers with the same controller id, that
is a violation of the spec, and we properly reject such controllers.
In upstream code, you can specify a cntlid min/max range such that if
you are trying to play games with different linux targets as a single
subsystem you won't violate the spec.
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: nvme: Duplicate cntlid error
2020-03-31 6:45 ` Sagi Grimberg
@ 2020-03-31 10:37 ` Alex Talker
2020-03-31 13:24 ` Christoph Hellwig
0 siblings, 1 reply; 4+ messages in thread
From: Alex Talker @ 2020-03-31 10:37 UTC (permalink / raw)
To: linux-nvme; +Cc: maxg, Sagi Grimberg
3/31/2020 9:45 AM, Sagi Grimberg wrote:
>
>> Using Fedora 31 as a host I'm trying to connect to 2 nodes that host
>> subsystem with one and the same NQN.
>> So when I call:
>> nvme connect -t rdma -a 10.21.0.1 -n
>> nqn.2020-03.com.example:subsystem0
>> device nvme0 successfully allocated but when then I call:
>> nvme connect -t rdma -a 10.20.0.1 -n
>> nqn.2020-03.com.example:subsystem0
>> I get an error:
>> [239355.661527] nvme nvme1: Duplicate cntlid 1 with nvme0, rejecting
>> what could be the problem?
>> Few months before I checked connection to a single node via few ports
>> and it worked out just fine
>> but attempt of H.A. setup fails here.
>
> Your subsystem present two controllers with the same controller id, that
> is a violation of the spec, and we properly reject such controllers.
>
> In upstream code, you can specify a cntlid min/max range such that if
> you are trying to play games with different linux targets as a single
> subsystem you won't violate the spec.
To summarize problem is that cntlid is automatically allocated inside
the target driver
per-connection but since there's no synchronization between the nodes,
each node counts separately
and that's why the numbers may clash, right?
I've checked and the patch
https://github.com/torvalds/linux/commit/94a39d61f80fcd679debda11e1ca02b88d90e67e
is in fact available only upstream
but not in any MLNX OFED driver back-port. Does Mellanox has plans to
include this feature in next release?
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2020-03-31 15:45 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2020-03-30 11:44 nvme: Duplicate cntlid error Alex Talker
2020-03-31 6:45 ` Sagi Grimberg
2020-03-31 10:37 ` Alex Talker
2020-03-31 13:24 ` Christoph Hellwig
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox