[PATCH] ARM: poison_init_mem: fixed underflow in loop

[sboyd@codeaurora.org (Stephen Boyd)]

On 2/20/2013 3:52 AM, Enrico Scholz wrote:
> When initmemory ends at an odd address, the loop variable in
> poison_init_mem() will underflow which causes overriding of valid
> memory. Such situations are not a theoretical issue only but
> exist in practice:
>
> | [    0.000000]       .init : 0xc0634000 - 0xc0c215ab   (6070 kB)
>
> Patch makes a simple division instead of DIV_ROUND_UP() because latter
> can override some bytes after .init.
>
> Signed-off-by: Enrico Scholz <enrico.scholz@sigma-chemnitz.de>
> ---
>  arch/arm/mm/init.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c
> index ad722f1..c431a2d 100644
> --- a/arch/arm/mm/init.c
> +++ b/arch/arm/mm/init.c
> @@ -449,7 +449,8 @@ static inline int free_area(unsigned long pfn, unsigned long end, char *s)
>  static inline void poison_init_mem(void *s, size_t count)
>  {
>  	u32 *p = (u32 *)s;
> -	for (; count != 0; count -= 4)
> +	count /= sizeof(*p);
> +	for (; count != 0; count--)
>  		*p++ = 0xe7fddef0;
>  }
>  

Sigh. I really screwed that up badly. I wonder if we should instead
PAGE_ALIGN() the end address similar to what we're doing for the initrd
and then put a BUG_ON() inside poison_init_mem() that catches none page
aligned sizes.

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation