On Mon, Jun 26, 2023 at 05:08:28PM +0100, Conor Dooley wrote:
> On Mon, Jun 26, 2023 at 05:19:08PM +0200, Andrew Jones wrote:
> > On Mon, Jun 26, 2023 at 12:19:40PM +0100, Conor Dooley wrote:
> > > isa_ext_arr cannot be empty, as some of the extensions within it are
> > > always built into the kernel.
> >
> > This is only true since commit 07edc32779e3 ("RISC-V: always report
> > presence of extensions formerly part of the base ISA"), right? If
> > so, it might be nice to call that commit out in this commit message.
>
> Per my last mail, where I commented on the origins of some of this code,
> there were no multi-letter extensions when this code was first added.
> When the first multi-letter ones did get added, it was Sscofpmf - that
> doesn't have a Kconfig symbol to disable it, so I think this has been
> redundant for a long time.
>
> Apart from the ones I recently added, there's a fair few others that
> are not gated & should always be present.
> It's probably not clear from the comment, but this check is for whether
> the kernel supports extensions, not whether the system it is running on
> does. I guess I should expand on that in my commit message.
That part I understood, but I was thinking it'd be nice to call out when
the first extension was added which cannot be disabled by a config to
provide extra evidence that it's safe to remove the check.
Thanks,
drew
