[PATCH/RFC] i.MX31 and i.MX35 : fix errate TLSbo65953 and ENGcm09472

[Uwe Kleine-König]

On Tue, Oct 05, 2010 at 11:25:39AM +0200, Eric Bénard wrote:
> Without this exiting WFI can result in cache corruption.
> Code taken from Freescale's 2.6.27 BSP and tested on i.MX35
> 
> Signed-off-by: Eric Bénard <eric at eukrea.com>
> ---
>  arch/arm/plat-mxc/include/mach/system.h |   26 ++++++++++++++++++++++++--
>  1 files changed, 24 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/arm/plat-mxc/include/mach/system.h b/arch/arm/plat-mxc/include/mach/system.h
> index 4acd114..1661b29 100644
> --- a/arch/arm/plat-mxc/include/mach/system.h
> +++ b/arch/arm/plat-mxc/include/mach/system.h
> @@ -1,7 +1,7 @@
>  /*
>   *  Copyright (C) 1999 ARM Limited
>   *  Copyright (C) 2000 Deep Blue Solutions Ltd
> - *  Copyright 2004-2007 Freescale Semiconductor, Inc. All Rights Reserved.
> + *  Copyright 2004-2008 Freescale Semiconductor, Inc. All Rights Reserved.
>   *
>   * This program is free software; you can redistribute it and/or modify
>   * it under the terms of the GNU General Public License as published by
> @@ -22,14 +22,36 @@
>  
>  static inline void arch_idle(void)
>  {
> +#if (defined (CONFIG_ARCH_MX35) || defined (CONFIG_ARCH_MX31))
> +	/* fix i.MX31 errata TLSbo65953 and i.MX35 errata ENGcm09472 */
> +	unsigned long reg = 0;
> +	__asm__ __volatile__(
> +		"mrc p15, 0, %0, c1, c0, 0\n"
> +		"bic %0, %0, #0x00001000\n"
> +		"bic %0, %0, #0x00000004\n"
> +		"mcr p15, 0, %0, c1, c0, 0\n"
> +		"mov %0, #0\n"
> +		"mcr p15, 0, %0, c7, c5, 0\n"
> +		"mov %0, #0\n"
> +		"mcr p15, 0, %0, c7, c14, 0\n"
> +		"mov %0, #0\n"
> +		"mcr p15, 0, %0, c7, c0, 4\n"
> +		"nop\n" "nop\n" "nop\n" "nop\n"
> +		"nop\n" "nop\n" "nop\n"
> +		"mrc p15, 0, %0, c1, c0, 0\n"
> +		"orr %0, %0, #0x00001000\n"
> +		"orr %0, %0, #0x00000004\n"
> +		"mcr p15, 0, %0, c1, c0, 0\n"
> +		:: "r" (reg));
Isn't it possible to just say

		: "=r" (reg));

here?  (That is only a single : to make reg an output register and add
the = flag.)

> +#else
>  #ifdef CONFIG_ARCH_MXC91231
>  	if (cpu_is_mxc91231()) {
>  		/* Need this to set DSM low-power mode */
>  		mxc91231_prepare_idle();
>  	}
>  #endif
> -
>  	cpu_do_idle();
> +#endif
No.  When we start to have multi-soc support for imx and a kernel for
both mx35 and mxc91231 is compiled mxc91231_prepare_idle isn't called
anymore for mxc91231.  Ah, and cpu_do_idle isn't called anymore at all.

This needs to read:

	#if defined (CONFIG_ARCH_MX31) || defined (CONFIG_ARCH_MX35)
		if (cpu_is_mx31() || cpu_is_mx35()) {
			unsigned long reg = 0;
			__asm__ __volatile__(...);
		} else
	#endif
		{
		
#ifdef CONFIG_ARCH_MXC91231
			if (cpu_is_mxc91231()) {
				/* Need this to set DSM low-power mode */
				mxc91231_prepare_idle();
			}
#endif
			cpu_do_idle();
		}

Looks ugly.  If you come up with nice code (that still does the correct
thing) you get five extra points.  (Maybe add another function for the
last block.  I couldn't come up with a nice name though.)

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |