Re: [PATCH 2/2] nvme: add emulation for zone-append

[Javier Gonzalez]

On 19.08.2020 12:43, Christoph Hellwig wrote:
> On Wed, Aug 19, 2020 at 09:14:13AM +0000, Damien Le Moal wrote:
> > While defining a zone append command for SCSI/ZBC is possible (using sense data
> > for returning the written offset), there is no way to define zone append for
> > SATA/ZAC without entirely breaking the ATA command model. This is why we went
> > after an emulation implementation instead of trying to standardized native
> > commands. That implementation does not have any performance impact over regular
> > writes *and* zone write locking does not in general degrade HDD write
> > performance (only a few corner cases suffer from it). Comparing things equally,
> > the same could be said of NVMe drives that do not have zone append native
> > support: performance will be essentially the same using regular writes and
> > emulated zone append. But mq-deadline and zone write locking will significantly
> > lower performance for emulated zone append compared to a native zone append
> > support by the drive.
>
> And to summarize the most important point - Zone Append doesn't exist
> in ZAC/ABC.  For people that spent the last years trying to make zoned
> storage work, the lack of such a primite has been the major pain point.
> That's why I came up with the Zone Append design in response to a
> request for such an operation from another company that is now heavily
> involved in both Linux development and hosting Linux VMs.  For ZAC and
> ZBC the best we can do is to emulate the approach in the driver, but
> for NVMe we can do it.  ZNS until just before the release had Zone
> Append mandatory, and it did so for a very good reason.  While making
> it optional allows OEMs to request drives without it, I fundamentally
> think we should not support that in Linux and request vendors do
> implement writes to zones the right way.

Ok. We will just pursue Linux support for the ZNS following the append
model.

> And just as some OEMs can request certain TPs or optional features to
> be implemented, so can Linux.  Just to give an example from the zone
> world - Linux requires uniform and power of two zone sizes, which in
> ZAC and ZBC are not required.