On Wed, Jan 04, 2012 at 04:29:19PM -0500, David Miller wrote:
> From: Sam Ravnborg <sam@xxxxxxxxxxxx>
> Date: Tue, 3 Jan 2012 21:45:16 +0100
>
> > - kern_addr_valid is (temporary) disabled
> > I could not grock the code allocating sparc_valid_addr_bitmap
> > and I noticed most archs do not implement this check
>
> Is it hard to grok where the memory is allocated from or the
> calculation of the bitmap memory's size? :-)
i = last_valid_pfn >> ((20 - PAGE_SHIFT) + 5);
i += 1;
sparc_valid_addr_bitmap = (unsigned long *)
__alloc_bootmem(i << 2, SMP_CACHE_BYTES, 0UL);
I failed to find out how we calculate the amount
of memory to allocate.
last_valid_pfn is the last page of the available memory.
So last_valid_pfn >> (20 - PAGE_SHIFT) is the amount of
memory expressed in MB.
But then we shifted an additional "+ 5".
This looks like we go for 32 MB.
+ 1 is just to make sure we cover the last page I think.
And when we allocate we do " i << 2 " so we are down to
8 MB.
When we use it we do:
#define kern_addr_valid(addr) \
(test_bit(__pa((unsigned long)(addr))>>20, sparc_valid_addr_bitmap))
so here it looks like we have one bit for each MB.
I also looked at:
static void __init taint_real_pages(void)
{
int i;
for (i = 0; sp_banks[i].num_bytes; i++) {
unsigned long start, end;
start = sp_banks[i].base_addr;
end = start + sp_banks[i].num_bytes;
while (start < end) {
set_bit(start >> 20, sparc_valid_addr_bitmap);
start += PAGE_SIZE;
}
}
}
Here I could see that we set a bit for each MB
in sparc_valid_addr_bitmap and it is not limited
to the kernel address range.
In the end I got confused and dropped it because I could see that
lots of other archs did not implment this check either.
I will capture all this info in the commit which drop this
for later reference.
Sam
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
