Re: [PATCH v1] riscv: remove RISCV_ISA_EXT_BUNDLE[_VALIDATE]

From: Paul Walmsley

Date: Tue Nov 17 2026 - 22:04:27 EST


On Fri, 6 Nov 2026, Conor Dooley wrote:

> From: Conor Dooley <conor.dooley@xxxxxxxxxxxxx>
>
> RISCV_ISA_EXT_BUNDLE actually made sense when it was originally added,
> since it presented a simple interface where users (be that in kernel or
> userspace via things they are presented by the kernel) would check the
> exact extension needed for a feature, rather than muddying the water
> with parent extensions. As a result, the kernel does not track the
> status of the parent extension in the bundle, nor expose it to
> userspace.
>
> This is in comparison to RISCV_ISA_EXT_SUPERSET,
> where the parent extension provided some explicit value, and would have
> to be checked for itself for some potential feature, and therefore we
> actually track the status of both child and parent extensions, and let
> userspace detect them independently.
>
> Since we added this stuff, there's been some retroactive extension
> releases by RVI that have made previous monolithic extensions into pure
> wrappers that we intended using RISCV_ISA_EXT_BUNDLE for. Since we
> previously tracked that as one extension, and exported it to users as
> one extension, we can't just use RISCV_ISA_EXT_BUNDLE, and have to use
> RISCV_ISA_EXT_SUPERSET instead and nobody seems to know what to use each
> actually for anymore.
>
> Just delete RISCV_ISA_EXT_BUNDLE for the sake of consistency between
> things that were always a bundle from ratification and for those
> retroactively made into bundles, with the side benefit of being less
> confusing for those adding extensions to the kernel.
>
> Signed-off-by: Conor Dooley <conor.dooley@xxxxxxxxxxxxx>

Thanks, queued for v7.5.


- Paul