nguid for nvme over fabrics

nguid for nvme over fabrics

[Narasimhan V]

Hi,

I am trying nvme over fabrics, and found this: nguid is set to 0s by 
default for all devices.

So, nguid can be same (0s in this case) for multiple devices. Is this 
valid ?
Also, same nguid for multiple devices could be set, by the above 
scenario.
No errors/warnings from either the nvme initiator/target.

 From spec, I see this:
Namespace Globally Unique Identifier (NGUID): This field contains a 
128-bit value that is
globally  unique and assigned to  the  namespace  when  the  namespace  
is  created.  This  field
remains fixed throughout the life of the namespace and is preserved 
across namespace and
controller operations (e.g., controller reset,  namespace format, etc.).


-- 
Regards
Narasimhan V

nguid for nvme over fabrics

[Sagi Grimberg]

> Hi,
> 
> I am trying nvme over fabrics, and found this: nguid is set to 0s by 
> default for all devices.
> 
> So, nguid can be same (0s in this case) for multiple devices. Is this 
> valid ?
> Also, same nguid for multiple devices could be set, by the above scenario.
> No errors/warnings from either the nvme initiator/target.
> 
>  From spec, I see this:
> Namespace Globally Unique Identifier (NGUID): This field contains a 
> 128-bit value that is
> globally? unique and assigned to? the? namespace? when? the? namespace 
> is? created.? This? field
> remains fixed throughout the life of the namespace and is preserved 
> across namespace and
> controller operations (e.g., controller reset,? namespace format, etc.).

We generate a UUID (which can be overwritten by the user)

nguid for nvme over fabrics

[Narasimhan V]

On 2018-03-23 00:32, Sagi Grimberg wrote:
>> Hi,
>> 
>> I am trying nvme over fabrics, and found this: nguid is set to 0s by 
>> default for all devices.
>> 
>> So, nguid can be same (0s in this case) for multiple devices. Is this 
>> valid ?
>> Also, same nguid for multiple devices could be set, by the above 
>> scenario.
>> No errors/warnings from either the nvme initiator/target.
>> 
>>  From spec, I see this:
>> Namespace Globally Unique Identifier (NGUID): This field contains a 
>> 128-bit value that is
>> globally? unique and assigned to? the? namespace? when? the? 
>> namespace is? created.? This? field
>> remains fixed throughout the life of the namespace and is preserved 
>> across namespace and
>> controller operations (e.g., controller reset,? namespace format, 
>> etc.).
> 
> We generate a UUID (which can be overwritten by the user)

Shouldn't it be unique ?
I have instances where for NVMef, multiple devices can have same nguids.
Which component should be responsible to generate nguid in NVMef ?


-- 
Regards
Narasimhan V

nguid for nvme over fabrics

[Christoph Hellwig]

On Fri, Mar 23, 2018@12:59:53PM +0530, Narasimhan V wrote:
> Shouldn't it be unique ?
> I have instances where for NVMef, multiple devices can have same nguids.
> Which component should be responsible to generate nguid in NVMef ?

You must only usea nguid if you have a ieee id, and then it is your
responsibility to hand out ids based on that.

Which means that you generally should not set a nguid unless you
are building an appliance and really know what you are doing.

nguid for nvme over fabrics

[Narasimhan V]

On 2018-03-28 13:57, Christoph Hellwig wrote:
> On Fri, Mar 23, 2018@12:59:53PM +0530, Narasimhan V wrote:
>> Shouldn't it be unique ?
>> I have instances where for NVMef, multiple devices can have same 
>> nguids.
>> Which component should be responsible to generate nguid in NVMef ?
> 
> You must only usea nguid if you have a ieee id, and then it is your
> responsibility to hand out ids based on that.

Thnaks Chris,
I get it now.

> 
> Which means that you generally should not set a nguid unless you
> are building an appliance and really know what you are doing.
> 

For my confirmation, if we are setting duplicate nguids, it is user's
responsibility, and system does not give any warning / error ?

> _______________________________________________
> Linux-nvme mailing list
> Linux-nvme at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-nvme

-- 
Regards
Narasimhan V