Re: [PATCHv8 2/6] block: use generic u16 for write hints

[Christoph Hellwig <hch@lst.de>]

On Thu, Oct 24, 2024 at 01:45:02PM -0600, Keith Busch wrote:
> I'm not sure I follow this feedback. The SCSI feature is defined as a
> lifetime stream association in SBC-5. So it's still a stream for SCSI,
> but you want to call it "WRITE_HINT", which is not a term used in the
> SCSI spec for this feature. But, you want to call it STREAM_SEPARATION
> for NVMe only, even though the FDP spec doesn't use that term? What's
> wrong with just calling it a generic hint support feature?

The "Constrained Streams with Data Lifetimes" are called streams for
political reasons but they are not.  They are buckets of different data
lifetimes.

> I also don't see why SCSI couldn't use per-io hints just like this
> enables for NVMe. The spec doesn't limit SCSI to just 5 streams, so this
> provides a way to access them all through the raw block device.

I don't mind passing per-I/O temperature hints to SCSI block devices.

But we should not confuse a streams/FDP like streams that are different
context which are assumed to be discarded together and have a concept of
Stream Granularity Size or Reclaim Unit size with the data temperature
hints that are at the storage level fundamentally per-I/O and just bucket
into temperature group without any indication of data locality.