NVMe/FC compatible hardware

NVMe/FC compatible hardware

[Talker Alex]

Hi,

Could anyone kindly clarify if NVMe/FC Linux drivers would work with any FibreChannel HBA?

I've connected ATTO Celerity FC-162E and ATTO Celerity FC-82EN cards(on subsystem & host respectively) and successfully managed to setup a port
but I struggle to connect to it, issuing the following command:

> nvme discover -t fc -a nn-0x20000010860370a4:pn-0x10000010860370a4  -w nn-0x2000001086036870:pn-0x2100001086036870

I get this error:

> [4164487.114514] (NULL device *): queue 0 connect admin queue failed (-111).
> [4164487.114519] nvme nvme0: NVME-FC{0}: reset: Reconnect attempt failed (-111)                                                                                           
> [4164487.114522] nvme nvme0: NVME-FC{0}: Reconnect attempt in 10 seconds
> [4164487.114535] nvme nvme0: NVME-FC{0}: new ctrl: NQN "nqn.2014-08.org.nvmexpress.discovery"
> [4164487.114672] nvme nvme0: Removing ctrl: NQN "nqn.2014-08.org.nvmexpress.discovery"

and nothing on subsystem(controller?) side.

I expect that if addr_tractive attribute of the port in ConfigFS equals 1, it means that selected port/HBA capable in utilization for NVMe-oF.
Am I wrong in that too?

Would anyone kindly provide hardware list where the FC driver works properly?

------------------
Best regards,
Alex

_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

Re: NVMe/FC compatible hardware

[James Smart]

On 2/6/2020 10:28 AM, Talker Alex wrote:
> Hi,
>
> Could anyone kindly clarify if NVMe/FC Linux drivers would work with any FibreChannel HBA?
>
> I've connected ATTO Celerity FC-162E and ATTO Celerity FC-82EN cards(on subsystem & host respectively) and successfully managed to setup a port
> but I struggle to connect to it, issuing the following command:
For drivers within the kernel, Emulex hbas support NVME initiator or 
target operation, and afaik, Qlogic hbas are NVME initiator only.  I 
assume you're working on an upstream kernel, so no tricks are needed to 
configure the drivers to enable nvme (initiator mode), but it may still 
require the hba fw to be nvme enabled for the drivers to register.  
There isn't much that identifies when an hba has registered with nvme.  
You should look at your log and look at what driver attached (or didn't) 
to the ATTO cards. All depends on their PCI  IDs.

>
>> nvme discover -t fc -a nn-0x20000010860370a4:pn-0x10000010860370a4  -w nn-0x2000001086036870:pn-0x2100001086036870
> I get this error:
>
>> [4164487.114514] (NULL device *): queue 0 connect admin queue failed (-111).
>> [4164487.114519] nvme nvme0: NVME-FC{0}: reset: Reconnect attempt failed (-111)
>> [4164487.114522] nvme nvme0: NVME-FC{0}: Reconnect attempt in 10 seconds
>> [4164487.114535] nvme nvme0: NVME-FC{0}: new ctrl: NQN "nqn.2014-08.org.nvmexpress.discovery"
>> [4164487.114672] nvme nvme0: Removing ctrl: NQN "nqn.2014-08.org.nvmexpress.discovery"

Well, by the fact you got into the nvme-fc transport,  and that you 
received a -111 status (-ECONNREFUSED) it says you passed all the checks 
for finding a hostport to match the -w/--host-traddr info and a 
remoteport on that hostport that matches the -a/-traddr info and started 
the controller creation. So that's pretty far.

But two things are bothersome:
- The NULL device * : which makes me think that the hostport mapped to 
something based on fcloop - not to a real pci-based hba. You should have 
the devices base name.
- The -111 status - which is occurs if the transport tries to send the 
Create Association LS but it fails as the remote port isn't online - 
meaning the driver has unregistered the remote port but it has yet to be 
fully removed yet (waiting on references).  This of course stops any LS 
from being transmit.   This is odd as it's usually only a state that 
would be seen if there was a logout with the target and to have you hit 
it right as you type the discover command would be a mighty coincidence.


> and nothing on subsystem(controller?) side.
>
> I expect that if addr_tractive attribute of the port in ConfigFS equals 1, it means that selected port/HBA capable in utilization for NVMe-oF.
> Am I wrong in that too?
I don't know what this attribute is - what's it's path ?

feel free to send me email offline if you want to send more logs and 
discuss this further.

-- james


_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme