Re: [PATCH 1/2 v2] x86: add max_addr boot option

[Wen Congyang <wency@cn.fujitsu.com>]

At 06/13/2012 11:29 AM, H. Peter Anvin Wrote:
> On 06/12/2012 07:21 PM, Kamezawa Hiroyuki wrote:
>>
>> But now, we know mem= boot option is buggy....it acts as max_addr=
>> option, we have concerns that 'someone may fix mem= option as sane as ia64. because
>> it's buggy".
>>
>> We'd like to fix mem= boot option by ourselves and preserve old behavior
>> with max_addr= boot option, which ia64 has.
>>
> 
> Now I'm *really* confused.
> 
> Realistically, there is no point in the old mem= behavior of assuming a
> contiguous chunk of memory up to that point; it simply doesn't match how
> modern hardware is constructed.  Your notion that ia64 is "sane" is
> probably more of "outdated" in my opinion.
> 
> As such, the current behavior for mem= seems like the right thing and
> the change was intentional (not to mention has been in place since
> kernel 2.5.65, back in 2003); it also solves your requirements.  If you
> are concerned about it, it would make more sense to make sure it is
> documented as intentional.


Here is the document(Documentation/kernel-parameters.txt):

mem=nn[KMG]	[KNL,BOOT] Force usage of a specific amount of memory

The implementation of mem= on ia64 is the same as the description in the document, but
the implementation of mem= on x86 box is not the same as the descrition.

Now, which should we fix? Document or the implementition?

Another problem is: the mem= cannot work if the user specifies add_efi_memmap
option. I think we should also fix this problem.

Thanks
Wen Congyang
> 
> In fact, it looks like IA64 introduced a divergence when the max_addr=
> patch was introduced in 2004.  You're basically proposing the same
> divergence for x86 now; talk about having the tail wag the dog.
> 
> Sorry.  NAK.
> 
> 	-hpa
> 
>