Re: [PATCH v4] arm64: run-time detection for aarch32 support

From: Mark Rutland <mark.rutland@arm.com>
To: "Suzuki K. Poulose" <Suzuki.Poulose@arm.com>
Cc: Yury Norov <ynorov@caviumnetworks.com>,
	Catalin.Marinas@arm.com, klimov.linux@gmail.com,
	ddaney.cavm@gmail.com, ard.biesheuvel@linaro.org,
	will.deacon@arm.com, linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4] arm64: run-time detection for aarch32 support
Date: Fri, 18 Dec 2015 17:46:08 +0000	[thread overview]
Message-ID: <20151218174607.GE30229@leverpostej> (raw)
In-Reply-To: <56743C4F.7060201@arm.com>

On Fri, Dec 18, 2015 at 05:03:11PM +0000, Suzuki K. Poulose wrote:
> On 18/12/15 16:00, Yury Norov wrote:
> >Kernel option COMPAT defines the ability of executing aarch32 binaries.
> >Some platforms does not support aarch32 mode, and so cannot execute that
> >binaries. But we cannot just disable COMPAT for them because the same
> >kernel binary may be used by multiple platforms.
> 
> 
> >diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h
> >index 8f271b8..781a2f7 100644
> >--- a/arch/arm64/include/asm/cpufeature.h
> >+++ b/arch/arm64/include/asm/cpufeature.h
> >@@ -184,6 +184,13 @@ static inline bool system_supports_mixed_endian_el0(void)
> >  	return id_aa64mmfr0_mixed_endian_el0(read_system_reg(SYS_ID_AA64MMFR0_EL1));
> >  }
> >
> >+static inline bool system_supports_aarch32_el0(void)
> >+{
> >+	u64 pfr0 = read_system_reg(SYS_ID_AA64PFR0_EL1);
> >+	return ((pfr0 >> ID_AA64PFR0_EL0_SHIFT) & ID_AA64PFR0_ELx_MASK)
> >+						!= ID_AA64PFR0_EL0_64BIT_ONLY;
> 
> Could you please use
> 
> cpuid_feature_extract_field(pfr0, ID_AA64PFR0_EL0_SHIFT) != ID_AA64PFR0_EL0_64BIT_ONLY
> 
> instead and
> 
> >--- a/arch/arm64/include/asm/sysreg.h
> >+++ b/arch/arm64/include/asm/sysreg.h
> >@@ -102,6 +102,7 @@
> >  #define ID_AA64PFR0_EL2_SHIFT		8
> >  #define ID_AA64PFR0_EL1_SHIFT		4
> >  #define ID_AA64PFR0_EL0_SHIFT		0
> >+#define ID_AA64PFR0_ELx_MASK		0xf
> 
> get rid of ^ ?
> 
> As per ARM ARM, AArch32 only ID register values are unknown if AArch32 is
> not implemented. So I think we need to skip accessing the AArch32 ID registers
> everywhere (feature tracking), if the CPU doesn't supports it, to avoid
> unnecessary SANITY failures and TAINTing the kernel.

That all sounds good to me.

After boot-time we should also fail hotplug of a CPU that doesn't
support AArch32, if we decided at boot-time that AArch32 was supported
accross the system. That should probably be added to your early cpu
feature verification [1].

Thanks,
Mark.

[1] http://lists.infradead.org/pipermail/linux-arm-kernel/2015-December/392237.html