lkml.org 
[lkml]   [2008]   [Jan]   [31]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] [1/9] Handle kernel near memory hole in clear_kernel_mapping
On Tue, 29 Jan 2008, Andi Kleen wrote:
> This was a long standing obscure problem in the relocatable kernel. The
> AMD GART driver needs to unmap part of the GART in the kernel direct mapping to
> prevent cache corruption. With the relocatable kernel it is in theory possible
> that the separate kernel text mapping straddles that area too.
>
> Normally it should not happen because GART tends to be >= 2GB, and the kernel
> is normally not loaded that high, but it is possible in theory.

This smells fishy.

>
....
>
> +#define overlaps(as, ae, bs, be) ((ae) >= (bs) && (as) <= (be))

inline function please and a bit more intuituive arrangement of
arguments.

> +void __init clear_kernel_mapping(unsigned long address, unsigned long size)
> +{
> + int sh = PMD_SHIFT;
> + unsigned long kernel = __pa(__START_KERNEL_map);
> +
> + /*
> + * Note that we cannot unmap the kernel itself because the unmapped
> + * holes here are always at least 2MB aligned.

That's not enforced. The unmap code just does not split pages.

> + * This just applies to the trailing areas of the 40MB kernel mapping.

How is this ensured, that it only affects the end of the 40MB mapping ?

> + */
> + if (overlaps(kernel >> sh, (kernel + KERNEL_TEXT_SIZE) >> sh,
> + __pa(address) >> sh, __pa(address + size) >> sh)) {

This checks:

(kernel_end + 1) >= gart_start && kernel_start <= gart_end

One off error: kernel + KERNEL_TEXT_SIZE
needs to be: kernel + KERNEL_TEXT_SIZE - 1

Also there is no sanity check, whether the area is inside real kernel text.

> + printk(KERN_WARNING
> + "Kernel mapping at %lx within 2MB of memory hole\n",
> + kernel);


> + __clear_kernel_mapping(__START_KERNEL_map+__pa(address), size);

Doh! This is unmapping the wrong place. According to __phys_addr():

paddr = vaddr - __START_KERNEL_map + phys_base;

Transformation:

vaddr = padd + __START_KERNEL_map - phys_base;

Sigh. The clear_kernel_mapping() functionality is just another wreckaged
implementation of CPA.

We really should use CPA and have a function to mark entries not
present. Actually the DEBUG_PAGEALLOC code has one already. There are
sanity checks in the CPA code as well, which can be extended to cover
the GART unmap in a safe way.

Thanks,
tglx


\
 
 \ /
  Last update: 2008-01-31 16:51    [W:0.409 / U:0.132 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site