[PATCH v1 1/1] riscv: sbi: Introduce system suspend support

[Conor Dooley]

Hey Drew, Anup,

On Thu, Oct 19, 2023 at 10:36:50AM +0200, Andrew Jones wrote:

> I was just reviewing the DBCN series and was about to raise my thoughts
> there again about not being so strict about SBI implementation versions,
> but then I thought some more about it. I guess we should be strict about
> the version since it's otherwise not possible to any confidence that the
> extension advertised is the extension described in the frozen/ratified
> version of the spec (there could be implementations of draft spec versions
> which are not compatible with the final version and those will have the
> same extension ID).

TBH, I don't think that having the version check is a sufficient test to
stop things like that happening, but being insufficient does not make it
not worthwhile.

> The most confidence Linux can have in an SBI
> extension's implementation being what it expects to be will come from
> both the extension's presence and the SBI implementation's version
> being at least as big as the version in which the extension was frozen.

I am not sure what the status of the patches are on the OpenSBI side,
but this is kinda how I feel about the Andestech PMU series. They're
intending advertising having the PMU extension using the same interface
as the standard PMU stuff - at least, that is the basis on which the
kernel patches worked. The software doing the ecall to probe for support
does really need to be able to trust that when it is told the extension
is present that it is in fact the standard, or at least standard-compatible,
extension.

<20230906111455.4161641-1-peterlin at andestech.com> is the OpenSBI series
implementing this and I voiced my objection to the kernel patches at
https://lore.kernel.org/linux-riscv/20230911-worry-reformat-fbb5c473085d@wendy

> Long story short, I'll change the above condition to look for 2.0 and
> ask QEMU folks to put an OpenSBI binary which advertises 2.0 into its
> repo as soon as possible.

Cool, thanks.


BTW Anup, do you think we could get the OpenSBI mailing lists archived
on lore.kernel.org?

Cheers,
Conor.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-riscv/attachments/20231019/65be880c/attachment.sig>