[PATCH] arm64: Disable fine grained traps on boot

[Marc Zyngier]

On Thu, 01 Apr 2021 16:48:05 +0100,
Mark Brown <broonie at kernel.org> wrote:
> 
> On Thu, Apr 01, 2021 at 04:14:35PM +0100, Marc Zyngier wrote:
> > On Thu, 01 Apr 2021 13:49:36 +0100,
> > Mark Brown <broonie at kernel.org> wrote:
> 
> > > +	msr_s	SYS_HFGRTR_EL2, xzr
> > > +	msr_s	SYS_HFGITR_EL2, xzr
> > > +	msr_s	SYS_HFGWTR_EL2, xzr
> 
> > nit: consider grouping SYS_HFGWTR_EL2 and SYS_HFGRTR_EL2 together,
> > since they affect the same registers.
> 
> Ack, I'd sorted them numerically IIRC.
> 
> > > +	mrs	x1, id_aa64pfr0_el1		// AMU traps UNDEF without AMU
> > > +	ubfx	x1, x1, #ID_AA64PFR0_AMU_SHIFT, #4
> > > +	cbz	x1, .Lskip_fgt_\@
> 
> > > +	msr_s	SYS_HAFGRTR_EL2, xzr
> 
> > Do we need to document the need for SCR_EL3.FGTEn to be set so that
> > these register accesses don't trap? That'd be consistent with what we
> > do for other features (PtrAuth, AMU...).
> 
> Yes, good point.  We should if this is going in - this is the result of
> some other stuff I've been sending to document disabling these on boot,
> Will wanted to have code doing that, so I'll pull that into there.  It
> does end up being a bit weird that we demand this be enabled so we can
> explicitly not use it, but better to do that so we can in future decide
> to take advantage of it.

I definitely have plans for this feature right now, as some of the
traps that KVM uses could be minimised using FGT, so I don't believe
the feature will stay unused for too long.

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.