NVMeoF driver & Reservations

NVMeoF driver & Reservations

[Talker Alex]

Hi!

Is there plans to implement Reservations support in target driver(from NVMe v1.4 part 8.8)?
I see that host support is already implemented in nvme utility & driver level.
However, I haven't found hardware on which it works - target driver doesn't support this command and PCIe drives available to me also.

So, I'm wondering if anyone begin to implement this feature at least on single target level.

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

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

Re: NVMeoF driver & Reservations

[Chaitanya Kulkarni]

On 12/01/2019 02:19 AM, Talker Alex wrote:
> Hi!
>
> Is there plans to implement Reservations support in target driver(from NVMe v1.4 part 8.8)?
> I see that host support is already implemented in nvme utility & driver level.
> However, I haven't found hardware on which it works - target driver doesn't support this command and PCIe drives available to me also.
>
> So, I'm wondering if anyone begin to implement this feature at least on single target level.

Yes there is and I also have testcases to test this feature from host side.

for target I did start working on this but stopped it since I didn't
find proper use-case (or a client) for it to justify the effort and
code getting into kernel. But if you have a legit scenario or a
use-case I'll be happy to work on this.


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

Re: NVMeoF driver & Reservations

[Talker Alex]

> But if you have a legit scenario or a
> use-case I'll be happy to work on this.

I think in SCSI subsystems this support is usually made to pass-by Microsoft Windows Server Failover Clustering requirements.
Regarding the use-case I think this is as usual an effort to create
a lock mechanism in order to synchronize access to the same device without an additional layer on client side.

I don't know whether NVMeoF works on Windows at all by now, haven't checked yet.
Anyway, I'd like to see your code if you published it anywhere.
Unlikely with SCSI, NVMe Reservations seems to be easier to implement with dedicated HostNQN identifier,
digging through the specification conditions.

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

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

Re: NVMeoF driver & Reservations

[Bart Van Assche]

On 12/2/19 2:07 PM, Chaitanya Kulkarni wrote:
> On 12/01/2019 02:19 AM, Talker Alex wrote:
>> Hi!
>>
>> Is there plans to implement Reservations support in target driver(from NVMe v1.4 part 8.8)?
>> I see that host support is already implemented in nvme utility & driver level.
>> However, I haven't found hardware on which it works - target driver doesn't support this command and PCIe drives available to me also.
>>
>> So, I'm wondering if anyone begin to implement this feature at least on single target level.
> 
> Yes there is and I also have testcases to test this feature from host side.
> 
> for target I did start working on this but stopped it since I didn't
> find proper use-case (or a client) for it to justify the effort and
> code getting into kernel. But if you have a legit scenario or a
> use-case I'll be happy to work on this.

Hi Chaitanya,

If you would be considering to develop test code for NVMe reservations, 
you may want to have a look at the SCSI persistent reservation tests in 
the libiscsi test-tool 
(https://github.com/sahlberg/libiscsi/tree/master/test-tool).

Bart.

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

Re: NVMeoF driver & Reservations

[Chaitanya Kulkarni]

Indeed I have looked into this and it's very helpful. 
Thanks for reminding me again :). 
> On Dec 3, 2019, at 8:08 AM, Bart Van Assche <bvanassche@acm.org> wrote:
> 
> On 12/2/19 2:07 PM, Chaitanya Kulkarni wrote:
>>> On 12/01/2019 02:19 AM, Talker Alex wrote:
>>> Hi!
>>> 
>>> Is there plans to implement Reservations support in target driver(from NVMe v1.4 part 8.8)?
>>> I see that host support is already implemented in nvme utility & driver level.
>>> However, I haven't found hardware on which it works - target driver doesn't support this command and PCIe drives available to me also.
>>> 
>>> So, I'm wondering if anyone begin to implement this feature at least on single target level.
>> Yes there is and I also have testcases to test this feature from host side.
>> for target I did start working on this but stopped it since I didn't
>> find proper use-case (or a client) for it to justify the effort and
>> code getting into kernel. But if you have a legit scenario or a
>> use-case I'll be happy to work on this.
> 
> Hi Chaitanya,
> 
> If you would be considering to develop test code for NVMe reservations, you may want to have a look at the SCSI persistent reservation tests in the libiscsi test-tool (https://github.com/sahlberg/libiscsi/tree/master/test-tool).
> 
> Bart.
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme