Re: [oprof-cvs] [PATCH] oprofile: use KM_NMI slot for kmap_atomic

[Robert R. <rob...@am...>]

On 22.08.12 10:21:07, Junxiao Bi wrote:
> If one kernel path is using KM_USER0 slot and is interrupted by
> the oprofile nmi, then in copy_from_user_nmi(), the KM_USER0 slot
> will be overwrite and cleared to zero at last, when the control
> return to the original kernel path, it will access an invalid
> virtual address and trigger a crash.
> 
> Cc: Robert Richter <rob...@am...>
> Cc: Greg KH <gr...@li...>
> Cc: st...@vg...
> Signed-off-by: Junxiao Bi <jun...@or...>
> 
> Hi, Please review this patch.
> 
> It is for linux-2.6.32.y stable branch not for mainline.

I am not sure if there will be any .32 stable release in the future,
but this could be at least for .34 or if there is one for .27 and .35.

> 
> Thanks,
> Junxiao.
> ---
>  arch/x86/oprofile/backtrace.c |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)

We should implement the perf version here, which does a:

	int type = in_nmi() ? KM_NMI : KM_IRQ0;

See arch/x86/kernel/cpu/perf_event.c.

-Robert

> 
> diff --git a/arch/x86/oprofile/backtrace.c b/arch/x86/oprofile/backtrace.c
> index 829edf0..b50a280 100644
> --- a/arch/x86/oprofile/backtrace.c
> +++ b/arch/x86/oprofile/backtrace.c
> @@ -71,9 +71,9 @@ copy_from_user_nmi(void *to, const void __user *from, unsigned long n)
>  		offset = addr & (PAGE_SIZE - 1);
>  		size = min(PAGE_SIZE - offset, n - len);
>  
> -		map = kmap_atomic(page, KM_USER0);
> +		map = kmap_atomic(page, KM_NMI);
>  		memcpy(to, map+offset, size);
> -		kunmap_atomic(map, KM_USER0);
> +		kunmap_atomic(map, KM_NMI);
>  		put_page(page);
>  
>  		len  += size;
> -- 
> 1.7.9.5
> 
> 

-- 
Advanced Micro Devices, Inc.
Operating System Research Center