Martin, Christoph
On 29/08/24 06:59PM, Anuj gupta wrote:
On Thu, Aug 29, 2024 at 8:47 AM Martin K. Petersen
<martin.petersen@xxxxxxxxxx> wrote:
Kanchan,
> With Guard/Reftag/Apptag, we get 6 combinations. For NVMe, all can be
> valid. For SCSI, maximum 4 can be valid. And we factor the pi-type in
> while listing what all is valid. For example: 010 or 001 is not valid
> for SCSI and should not be shown by this.
I thought we had tentatively agreed to let the block layer integrity
flags only describe what the controller should do? And then let sd.c
decide what to do about RDPROTECT/WRPROTECT (since host-to-target is a
different protection envelope anyway). That is kind of how it works
already.
Do you see that this patch (and this set of flags) are fine?
If not, which specific flags do you suggest should be introduced?
While other things are sorted for next iteration, it's not fully clear
what are we missing for this part. Can you comment on the above?
