[kernel-hardening] Re: [PATCH 6/6] x86: kaslr: relocate base offset at boot

["H. Peter Anvin" <hpa@zytor.com>]

On 04/15/2013 03:38 PM, Yinghai Lu wrote:
> On Mon, Apr 15, 2013 at 3:07 PM, Kees Cook <keescook@chromium.org> wrote:
>> On Mon, Apr 15, 2013 at 3:00 PM, Yinghai Lu <yinghai@kernel.org> wrote:
>>> also do not overlap with boot_param, command_line, and initrd.
>>>
>>> and need to double check setup_header.init_size to make sure bss and
>>> etc will not
>>> fall into memory hole or reserved area in e820.
>>>
>>> also may need to setup page table for target position as bootloader may only
>>> has ident mapping only for loaded bzImage 64 areas.
>>>
>>> looks you are trying redo the work for bootloader to pick loaded phys addr.
>>
>> aslr.S's select_aslr_address uses z_extract_offset as the upper bound.

That seems really, really odd.

> so the decompressed image is not moved high?

This really seems wrong in so many ways.

If you relocate pre-decompression the amount of memory you need is given
by init_size:

#define ZO_INIT_SIZE    (ZO__end - ZO_startup_32 + ZO_z_extract_offset)
#define VO_INIT_SIZE    (VO__end - VO__text)
#if ZO_INIT_SIZE > VO_INIT_SIZE
#define INIT_SIZE ZO_INIT_SIZE
#else
#define INIT_SIZE VO_INIT_SIZE
#endif
init_size:              .long INIT_SIZE         # kernel initialization size

In real life, VO_INIT_SIZE dominates almost every time.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.