RE: page allocation/attributes question (i386/x86_64 specific)

[Stuart_Hayes]

Ingo Molnar wrote:
> * Stuart_Hayes@xxxxxxxx <Stuart_Hayes@xxxxxxxx> wrote:
> 
>> Ingo Molnar wrote:
>>> there's one problem with the patch: it breaks things that need the
>>> low 1MB executable (e.g. APM bios32 calls). It would at a minimum be
>>> needed to exclude the BIOS area in 0xd0000-0xfffff.
>>> 
>>> 	Ingo
>> 
>> I wrote it to make everything below 1MB executable, if it isn't RAM
>> according to the e820 map, which should include the BIOS area.  This
>> includes 0xd0000-0xffff on my system.  Do you think I should
>> explicity make 0xd0000-0xfffff executable regardless of the e820 map?
> 
> hm ... which portion does this? I'm looking at fixnx2.patch. I
> definitely saw a APM bootup crash due to this, but that was on a
> 2.4-ish backport of the patch.  
> 
> 	Ingo

Oh, sorry, we're talking about two different patches.  I sent in a
different patch yesterday, because Andi Kleen didn't seem very
enthusiastic about fixnx2.patch.  Here's the patch that I sent yesterday
(attached as file init.c.patch).

Thanks
Stuart

Attachment: init.c.patch
Description: init.c.patch