Re: fine-grained PI control

[Christoph Hellwig <hch@lst.de>]

On Fri, Jul 26, 2024 at 03:51:56PM +0530, Anuj Gupta wrote:
> 
> I was thinking something like below patch[*] could help us get rid of the
> BIP_USER_CHECK_FOO flags, and also driver can now check flags passed by block
> layer instead of checking if it's user passthrough data. Haven't plumbed the
> scsi side of things, but do you think it can work with scsi? 
> 
> Subject: [PATCH] block: introduce BIP_CHECK_GUARD/REFTAG/APPTAG bip_flags
> 
> This patch introduces BIP_CHECK_GUARD/REFTAG/APPTAG bip_flags which
> indicate how the hardware should check the payload. The driver can now
> just rely on block layer flags, and doesn't need to know the integrity
> source. Submitter of PI chooses which tags to check. This would also
> give us a unified interface for user and kernel generated integrity.

This looks reasonably to me for the in-kernel interface.  We'll still
need to deal with the fact that SCSI is a bit selective in what
combination of these flags it actually allows.