Re: [PATCH] ARM: pm: fix HYP/SVC mode mismatch when MCPM is used

[Russell King - ARM Linux admin <linux@armlinux.org.uk>]

On Thu, Feb 14, 2019 at 03:31:14PM +0100, Marek Szyprowski wrote:
> MCPM does a soft reset of the CPUs and uses common cpu_resume() routine to
> perform low-level platform initialization. This results in a try to install
> HYP stubs for the second time for each CPU and results in false HYP/SVC
> mode mismatch detection. The HYP stubs are already installed at the
> beginning of the kernel initialization on the boot CPU (head.S) or in the
> secondary_startup() for other CPUs. To fix this issue MCPM code should use
> a cpu_resume() routine without HYP stubs installation.
> 
> This change fixes HYP/SVC mode mismatch on Samsung Exynos5422-based Odroid
> XU3/XU4/HC1 boards.
> 
> Fixes: 3721924c8154 ("ARM: 8081/1: MCPM: provide infrastructure to allow for MCPM loopback")
> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> ---
>  arch/arm/common/mcpm_entry.c   |  2 +-
>  arch/arm/include/asm/suspend.h |  1 +
>  arch/arm/kernel/sleep.S        | 11 +++++++++++
>  3 files changed, 13 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/arm/common/mcpm_entry.c b/arch/arm/common/mcpm_entry.c
> index ad574d20415c..1b1b82b37ce0 100644
> --- a/arch/arm/common/mcpm_entry.c
> +++ b/arch/arm/common/mcpm_entry.c
> @@ -381,7 +381,7 @@ static int __init nocache_trampoline(unsigned long _arg)
>  	unsigned int cluster = MPIDR_AFFINITY_LEVEL(mpidr, 1);
>  	phys_reset_t phys_reset;
>  
> -	mcpm_set_entry_vector(cpu, cluster, cpu_resume);
> +	mcpm_set_entry_vector(cpu, cluster, cpu_resume_no_hyp);
>  	setup_mm_for_reboot();
>  
>  	__mcpm_cpu_going_down(cpu, cluster);
> diff --git a/arch/arm/include/asm/suspend.h b/arch/arm/include/asm/suspend.h
> index 452bbdcbcc83..506314265c6f 100644
> --- a/arch/arm/include/asm/suspend.h
> +++ b/arch/arm/include/asm/suspend.h
> @@ -10,6 +10,7 @@ struct sleep_save_sp {
>  };
>  
>  extern void cpu_resume(void);
> +extern void cpu_resume_no_hyp(void);
>  extern void cpu_resume_arm(void);
>  extern int cpu_suspend(unsigned long, int (*)(unsigned long));
>  
> diff --git a/arch/arm/kernel/sleep.S b/arch/arm/kernel/sleep.S
> index a8257fc9cf2a..b856d183691e 100644
> --- a/arch/arm/kernel/sleep.S
> +++ b/arch/arm/kernel/sleep.S
> @@ -122,6 +122,11 @@ ENDPROC(cpu_resume_after_mmu)
>  
>  #ifdef CONFIG_MMU
>  	.arm
> +#ifdef CONFIG_MCPM
> +ENTRY(cpu_resume_no_hyp)
> +ARM_BE8(setend be)			@ ensure we are in BE mode
> +	b	0f

What if the kernel is built for Thumb?  You'll be branching to thumb
code at the '0' label - don't you need to do the same that
cpu_resume_arm does below?

> +#endif
>  ENTRY(cpu_resume_arm)
>   THUMB(	badr	r9, 1f		)	@ Kernel is entered in ARM.
>   THUMB(	bx	r9		)	@ If this is a Thumb-2 kernel,
> @@ -135,6 +140,9 @@ ARM_BE8(setend be)			@ ensure we are in BE mode
>  	bl	__hyp_stub_install_secondary
>  #endif
>  	safe_svcmode_maskall r1
> +#ifdef CONFIG_MCPM
> +0:
> +#endif

You don't need to conditionalise this.  I'd also use a symbol rather
than a numeric label given that safe_svcmode_maskall is a macro that
also uses numeric labels.

>  	mov	r1, #0
>  	ALT_SMP(mrc p15, 0, r0, c0, c0, 5)
>  	ALT_UP_B(1f)
> @@ -163,6 +171,9 @@ ENDPROC(cpu_resume)
>  
>  #ifdef CONFIG_MMU
>  ENDPROC(cpu_resume_arm)
> +#endif
> +#ifdef CONFIG_MCPM
> +ENDPROC(cpu_resume_no_hyp)
>  #endif
>  
>  	.align 2
> -- 
> 2.17.1
> 
> 

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up
According to speedtest.net: 11.9Mbps down 500kbps up