[PATCH] Documentation: RISC-V: Define Xlinuxs{s,m}aia

[Palmer Dabbelt]

On Thu, 02 Feb 2023 23:32:47 PST (-0800), Conor Dooley wrote:
>
>
> On 3 February 2023 00:12:01 GMT, Palmer Dabbelt <palmer at rivosinc.com> wrote:
>>The AIA specification was only partially frozen, but provides no way to
>>refer to the subset of behavior that has been frozen.
>
> Ahh, wonderful.
>
>>It seems like
>>there's not a whole lot of interest in the non-frozen behavior, so let's
>>just define an extension that only consists of the frozen behavior
>
>>
>>Signed-off-by: Palmer Dabbelt <palmer at rivosinc.com>
>>---
>> Documentation/riscv/extensions.rst | 10 ++++++++++
>> 1 file changed, 10 insertions(+)
>> create mode 100644 Documentation/riscv/extensions.rst
>>
>>diff --git a/Documentation/riscv/extensions.rst b/Documentation/riscv/extensions.rst
>>new file mode 100644
>>index 000000000000..c12bd3780520
>>--- /dev/null
>>+++ b/Documentation/riscv/extensions.rst
>>@@ -0,0 +1,10 @@
>>+.. SPDX-License-Identifier: GPL-2.0
>>+
>>+Custom RISC-V Extensions
>>+========================
>>+
>>+* The `Xlinuxsmaia` and `Xlinuxssaia` extensions coorespond to the standard
>
> Typo: correspond
>
>>+  `Smaia` and `Ssaia` extensions as defined by version 1.0-RC2 of the RISC-V
>>+  Advanced Interrupt Architecture (AIA) specification, with the difference that
>>+  chapters 5 (Duo-PLIC) and 9 (IOMMU Support for MSIs to Virtual Machines) do
>>+  not exist.
>
> Perhaps this document needs an intro section specifying what you expect to go in it.
> Is this for any custom stuff, or just places where Linux has to work around RVI?

>From talking on the patchwork call, it sounds like 1.0.0-rc2 may not be 
what's going to be ratified.  If the plan really is to remove the draft 
chapters when the spec is ratified, which does sometimes happen, then 
those might just solve the problem -- but sometimes things marked as 
drafts get ratified, so we really just have no idea what the spec is 
going to be.  It's going to be pretty much imposible to get everyone on 
the same page until we can see the spec that's going to be ratified.  So 
I think until that gets sorted out, there's not really a whole lot more 
we can do here.

The other option would be to just properly commit to moving away from 
ISA strings in the DT interface.  We've sort of soft-forked already, but 
we're at least trying to keep things looking similar.  Properly forking 
would really make everyone happy, as we'd no longer need to worry about 
the specs getting the details right.