Re: [PATCH v3 02/12] KVM: arm64: selftests: Add write_sysreg_s and read_sysreg_s

From: Andrew Jones <drjones@redhat.com>
To: Oliver Upton <oupton@google.com>
Cc: Raghavendra Rao Ananta <rananta@google.com>,
	kvm@vger.kernel.org, Will Deacon <will@kernel.org>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Peter Shier <pshier@google.com>,
	linux-kernel@vger.kernel.org, Marc Zyngier <maz@kernel.org>,
	Paolo Bonzini <pbonzini@redhat.com>,
	kvmarm@lists.cs.columbia.edu,
	linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH v3 02/12] KVM: arm64: selftests: Add write_sysreg_s and read_sysreg_s
Date: Thu, 2 Sep 2021 14:31:10 +0200	[thread overview]
Message-ID: <20210902123110.royrzw4dsykkrcjx@gator> (raw)
In-Reply-To: <YTAHYrQslkY12715@google.com>

On Wed, Sep 01, 2021 at 11:06:10PM +0000, Oliver Upton wrote:
> On Wed, Sep 01, 2021 at 03:48:40PM -0700, Raghavendra Rao Ananta wrote:
> > On Wed, Sep 1, 2021 at 3:08 PM Oliver Upton <oupton@google.com> wrote:
> > >
> > > On Wed, Sep 01, 2021 at 09:28:28PM +0000, Oliver Upton wrote:
> > > > On Wed, Sep 01, 2021 at 09:14:02PM +0000, Raghavendra Rao Ananta wrote:
> > > > > For register names that are unsupported by the assembler or the ones
> > > > > without architectural names, add the macros write_sysreg_s and
> > > > > read_sysreg_s to support them.
> > > > >
> > > > > The functionality is derived from kvm-unit-tests and kernel's
> > > > > arch/arm64/include/asm/sysreg.h.
> > > > >
> > > > > Signed-off-by: Raghavendra Rao Ananta <rananta@google.com>
> > > >
> > > > Would it be possible to just include <asm/sysreg.h>? See
> > > > tools/arch/arm64/include/asm/sysreg.h
> > >
> > > Geez, sorry for the noise. I mistakenly searched from the root of my
> > > repository, not the tools/ directory.
> > >
> > No worries :)
> > 
> > > In any case, you could perhaps just drop the kernel header there just to
> > > use the exact same source for kernel and selftest.
> > >
> > You mean just copy/paste the entire header? There's a lot of stuff in
> > there which we
> > don't need it (yet).
> 
> Right. It's mostly register definitions, which I don't think is too high
> of an overhead. Don't know where others stand, but I would prefer a
> header that is equivalent between kernel & selftests over a concise
> header.
>

Until now we haven't needed the sys_reg(...) type of definitions for
sysregs in selftests. In case we did, we defined the registers we
needed for get/set_one_reg by their parts, e.g.

 #define ID_AA64DFR0_EL1 3, 0,  0, 5, 0

allowing us to choose how we use them, ARM64_SYS_REG(...) vs.
sys_reg(...).

Bringing over sysreg.h is probably a good idea though. If we do, then
I'd suggest we define a new macro that allows us to convert a SYS_*
register definition from sysreg.h into an ARM64_SYS_REG definition
for get/set_one_reg in order to avoid redundant definitions.

Thanks,
drew