From: Borislav Petkov <bp@alien8.de> Sent: Thursday, February 23, 2023 2:45 AM

On Wed, Feb 22, 2023 at 05:21:27PM -0800, Sean Christopherson wrote:

quoted

All I'm advocating is that for determining whether or not a device should be mapped
private vs. shared, provide an API so that the hypervisor-specific enlightened code
can manage that insanity without polluting common code.  If we are ever fortunate
enough to have common enumeration, e.g. through ACPI or something, the enlightened
code can simply reroute to the common code.  This is a well established pattern for
many paravirt features, I don't see why it wouldn't work here.

Yah, that would be good.

Just so I'm clear, are you saying you are good with the proposal that Sean
sketched out with code here? [1]   With his proposal, the device driver
is not involved in deciding whether to map encrypted or decrypted.  That
decision is made in the hypervisor-specific callback function based on
the physical address.   The callback has to be made in two places in common
code.  But then as Sean said, " the hypervisor-specific enlightened code
can manage that insanity without polluting common code". :-)

I like Sean's proposal, so if you are good with it, I'll do v6 of the patch
set with that approach.

Dave Hansen:  Are you also OK with Sean's proposal?  Looking for consensus
here ....

If the device can know upfront how it needs to
ioremap its address range, then that is fine - we already have
ioremap_encrypted() for example.

Again, the device driver would not be involved in Sean's proposal.

Michael

[1] https://lore.kernel.org/linux-hyperv/Y+bXjxUtSf71E5SS@google.com/ (local)