Re: [PATCH 2/5] hw/nvme: always set eui64

[Klaus Jensen <its@irrelevant.dk>]

[-- Attachment #1: Type: text/plain, Size: 2148 bytes --]

On Apr 20 07:30, Christoph Hellwig wrote:
> On Tue, Apr 19, 2022 at 02:10:36PM +0200, Klaus Jensen wrote:
> > From: Klaus Jensen <k.jensen@samsung.com>
> > 
> > Unconditionally set an EUI64 for namespaces. The nvme-ns device defaults
> > to auto-generating a persistent EUI64 if not specified, but for single
> > namespace setups (-device nvme,drive=...), this does not happen.
> > 
> > Since the EUI64 has previously been zeroed it is not considered valid,
> > so it should be safe to add this now.
> > 
> > The generated EUI64 is of the form 52:54:00:<namespace counter>. Note,
> > this is NOT the namespace identifier since that is not unique across
> > subsystems; it is a global namespace counter. This has the effect that
> > the value of this auto-generated EUI64 is dependent on the order with
> > which the namespaces are created. If a more flexible setup is required,
> > the eui64 namespace parameter should be explicitly set. Update the
> > documentation to make this clear.
> 
> How is this actually globally unique given that it uses a start value
> that is incremented for each created namespace?

I think it is as good as we can do when we cannot store the EUI64
persistently anywhere. The EUI64s will be unique to a single QEMU
instance. If someone wants to simulate a fabrics setup or something like
that, then, as per the documentation, set the EUI explicitly.

> Also EUI64 values are based on a OUI, while NVME_EUI64_DEFAULT seems
> to have the OUI values cleared to all zero as far as I can tell.
> 

It really should be a u8 array, yes, but won't the integer approach
work? The "template" is byte swapped to big endian, or am I off here?

> I would strongly advise againt autogenerating eui64 values.  They are
> small and have little entropy, and require at least three bytes (for
> new allocations more) to be set to a IEEE assigned OUI.

52:54:00 is a "private" OUI if I am not mistaken (something about some
bit being 1 or 0, cant remember the specifics) and is what QEMU uses
when an IEEE OUI is needed (MAC-addresses etc.). This is also what the
device uses in the IEEE field.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]