Re: [External] [PATCH RFC/RFT v2 3/4] riscv: Stop emitting preventive sfence.vma for new vmalloc mappings

[Alexandre Ghiti]

On Tue, Jun 4, 2024 at 10:52 AM Conor Dooley <conor@xxxxxxxxxx> wrote:
>
> On Tue, Jun 04, 2024 at 09:17:26AM +0200, Alexandre Ghiti wrote:
> > On Tue, Jun 4, 2024 at 9:15 AM Alexandre Ghiti <alexghiti@xxxxxxxxxxxx> wrote:
> > > On Tue, Jun 4, 2024 at 8:21 AM yunhui cui <cuiyunhui@xxxxxxxxxxxxx> wrote:
> > > >
> > > > As for the current status of the patch, there are two points that can
> > > > be optimized:
> > > > 1. Some chip hardware implementations may not cache TLB invalid
> > > > entries, so it doesn't matter whether svvptc is available or not. Can
> > > > we consider adding a CONFIG_RISCV_SVVPTC to control it?
> >
> > That would produce a non-portable kernel. But I'm not opposed to that
> > at all, let me check how we handle other extensions. Maybe @Conor
> > Dooley has some feedback here?
>
> To be honest, not really sure what to give feedback on. Could you
> elaborate on exactly what the option is going to do? Given the
> portability concern, I guess you were proposing that the option would
> remove the preventative fences, rather than your current patch that
> removes them via an alternative?

No no, I won't do that, we need a generic kernel for distros so that's
not even a question. What Yunhui was asking about (to me) is: can we
introduce a Kconfig option to always remove the preventive fences,
bypassing the use of alternatives altogether?

To me, it won't make a difference in terms of performance. But if we
already offer such a possibility for other extensions, well I'll do
it. Otherwise, the question is: should we start doing that?

> I don't think we have any extension
> related options that work like that at the moment, and making that an
> option will just mean that distros that look to cater for multiple
> platforms won't be able to turn it on.
>
> Thanks,
> Conor.