Re: [PATCH v8 01/11] riscv: Add B to hwcap and hwprobe

[Guodong Xu <guodong.xu@oss.qualcomm.com>]

Hi Radim,

On Tue, 29 Sep 2026 15:12:36 +0000, Radim Krcmar wrote:
> Are we sure that B will always be listed in the ISA string when Zba,
> Zbb, and Zbs are present?

For DT, yes. The binding requires b whenever zba, zbb and zbs are
listed, since commit 0cdb7fc1879b ("dt-bindings: riscv: Add B ISA
extension description").

For ACPI there is no such check. But B is a ratified RVI extension
and RVA23U64 lists it as mandatory [1], so any serious hardware vendor
should respect it. Right?

If the firmware does not advertise B, why should the kernel correct that
for them? This came up in v2 [2], and maybe here I should quote what Conor
said in [3], which I agree with:

" Part of me wants to know why we should be going to a bunch of effort
  to cover up for inaccurate devicetrees etc. Garbage in -> garbage out
  and let them keep the pieces."

[1] https://github.com/riscv/riscv-profiles/releases/tag/rva23-rvb23-ratified
[2] https://lore.kernel.org/linux-riscv/CAH1PCMa-5W9PsX8cDLUk6-MkcM53HOz2QtaxCHd+XOr7DgH5+w@mail.gmail.com/
[3] https://lore.kernel.org/all/20260604-decency-salaried-4cb69e1c9fa9@spud/

Guodong