Re: [PATCH] x86: vm86 - hide X86_VM_MASK from userland programs v2

[Cyrill Gorcunov <gorcunov@gmail.com>]

[H. Peter Anvin - Tue, Apr 29, 2008 at 08:46:09AM -0700]
> Cyrill Gorcunov wrote:
>> | | with this patch we have <asm/vm86.h> included only
>> | if __i386__ defined *BUT* X86_VM_MASK is tried to be
>> | used *without* __386__ being checked (as example - 
>> ptrace.h:user_mode_vm).
>> | I'm not sure how to properly hanle this situation. But will take a look.
>> | | So I suggest you drop my last patch (which moves X86_VM_MASK
>> | into __KERNEL__ section) for a while. I'll recheck all
>> | this stuff later (too busy now).
>> | | 		- Cyrill -
>> Thomas, could you take a look please - is my suspicious wrong?
>> 		- Cyrill -
>
> X86_VM_MASK should be defined to zero on x86-64.  Part of the reason for 
> this symbol is so we don't have to put #ifdef around its uses.
>
> 	-hpa
>

Hi Peter,

yes, we already have X86_VM_MASK defined to 0 in vm86.h on 64bits
cpu - the only question - why this file was not included in ptrace.h
even the machine was 32bit cpu configured (according to Ingo's config).
I've been suspecting that it's __i386__ who is responsible for that but
I was wrong - it's just alias for CONFIG_X86_32. Interesting... ;)
Need time to investigate. /And sorry Thomas, I was wrong about your
commit/

		- Cyrill -