[PATCH v2] ARM: Add check for Cortex-A15 errata 798181 ECO

Will Deacon will.deacon at arm.com
Mon Jul 29 05:32:22 EDT 2013


Hi Rob,

On Sun, Jul 28, 2013 at 10:14:08PM +0100, Rob Herring wrote:
> From: Rob Herring <rob.herring at calxeda.com>
> 
> The work-around for A15 errata 798181 is not needed if appropriate ECO
> fixes have been applied to r3p2 and earlier core revisions. This can be
> checked by reading REVIDR register bits 4 and 9. If only bit 4 is set,
> then the IPI broadcast can be skipped.
> 
> Signed-off-by: Rob Herring <rob.herring at calxeda.com>
> ---
> v2:
> - Determine the work-around needed and save in a static varible instead
>   of re-reading the ID registers.
> 
>  arch/arm/include/asm/cputype.h |  1 +
>  arch/arm/kernel/smp_tlb.c      | 31 +++++++++++++++++++++++--------
>  2 files changed, 24 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/arm/include/asm/cputype.h b/arch/arm/include/asm/cputype.h
> index 8c25dc4..e5952b7 100644
> --- a/arch/arm/include/asm/cputype.h
> +++ b/arch/arm/include/asm/cputype.h
> @@ -10,6 +10,7 @@
>  #define CPUID_TLBTYPE	3
>  #define CPUID_MPUIR	4
>  #define CPUID_MPIDR	5
> +#define CPUID_REVIDR	6
>  
>  #ifdef CONFIG_CPU_V7M
>  #define CPUID_EXT_PFR0	0x40
> diff --git a/arch/arm/kernel/smp_tlb.c b/arch/arm/kernel/smp_tlb.c
> index a98b62d..da20db1 100644
> --- a/arch/arm/kernel/smp_tlb.c
> +++ b/arch/arm/kernel/smp_tlb.c
> @@ -73,12 +73,29 @@ static inline void ipi_flush_bp_all(void *ignored)
>  #ifdef CONFIG_ARM_ERRATA_798181
>  static int erratum_a15_798181(void)
>  {
> -	unsigned int midr = read_cpuid_id();
> -
> -	/* Cortex-A15 r0p0..r3p2 affected */
> -	if ((midr & 0xff0ffff0) != 0x410fc0f0 || midr > 0x413fc0f2)
> -		return 0;
> -	return 1;
> +	static int errata_fix_needed = -1;
> +
> +	if (unlikely(errata_fix_needed == -1)) {
> +		unsigned int midr = read_cpuid_id();
> +		unsigned int revidr = read_cpuid(CPUID_REVIDR);
> +
> +		/* Cortex-A15 r0p0..r3p2 w/o ECO fix affected */
> +		if ((midr & 0xff0ffff0) != 0x410fc0f0 || midr > 0x413fc0f2 ||
> +		    (revidr & 0x210) == 0x210) {
> +			errata_fix_needed = 0;	/* No work-around needed */
> +			return 0;
> +		}
> +		if (revidr & 0x10)
> +			errata_fix_needed = 1;	/* Only TLB flush needed */
> +		else
> +			errata_fix_needed = 2;	/* IPI broadcast needed */
> +	}
> +
> +	if (errata_fix_needed)
> +		dummy_flush_tlb_a15_erratum();
> +
> +	/* Return 1 if IPI broadcast to other cores is needed */
> +	return (errata_fix_needed == 2) ? 1: 0;

Perhaps this would be cleaner if we had a separate function to initialise an
appropriate function pointer for the workaround instead? It might also be
faster for non-affected cores (depending on what the above ends up compiling
to -- I don't know what ARM GCC does with __builtin_expect).

Will



More information about the linux-arm-kernel mailing list