Re: [Patch 6/8] powerpc: add CONFIG_KEXEC_AUTO_RESERVE

[Amerigo Wang <amwang@redhat.com>]

Michael Ellerman wrote:
> On Fri, 2009-08-21 at 02:55 -0400, Amerigo Wang wrote:
>   
>> Introduce a new config option KEXEC_AUTO_RESERVE for powerpc.
>>
>> Index: linux-2.6/arch/powerpc/Kconfig
>> ===================================================================
>> --- linux-2.6.orig/arch/powerpc/Kconfig
>> +++ linux-2.6/arch/powerpc/Kconfig
>> @@ -346,6 +346,17 @@ config KEXEC
>>  	  support.  As of this writing the exact hardware interface is
>>  	  strongly in flux, so no good recommendation can be made.
>>  
>> +config KEXEC_AUTO_RESERVE
>> +	bool "automatically reserve memory for kexec kernel"
>> +	depends on KEXEC
>> +	default y
>> +	---help---
>> +	  Automatically reserve memory for a kexec kernel, so that you don't
>> +	  need to specify numbers for the "crashkernel=X@Y" boot option,
>> +	  instead you can use "crashkernel=auto". To make this work, you need
>> +	  to have more than 4G memory. On PPC, 256M is reserved, 1/32 memory
>> +	  on PPC64, but it will not exceed 1T/32.
>>     
>
> To be honest I don't see why this logic goes in the kernel. It seems to
> me that it's policy how much memory you devote to the crash kernel vs
> the production kernel. It depends on what kind of crash kernel you're
> loading, a minimal UP dump kernel, or a full-featured SMP behemoth, An
> it depends on how much memory you're willing to leave idle in the
> off-chance you crash.
>   

True, but since in the crash kernel, we have very little memory, so 
probably loading a full-featured SMP kernel doesn't make much sense...

And in patch 1/8, I introduced a way to free the reserved memory at 
run-time.

> That aside, I don't see how this will be useful in practice, if it only
> works for memory sizes over 4G? Or are we saying that people with less
> than 4G don't need crash kernels? If we're not saying that, those users,
> or those users' distros, still need to do some logic to work out if they
> have < 4GB of memory and if so pick a crash kernel size. So why can't
> they pick the size in the > 4GB case also?
>   

No, we set 4G as a threshold because we only want this work when have 
have enough memory which is defined as 4G currently... This can be 
changed to arch-dependent, e.g. ppc. I am very open to this.


> Also the numbers seem a bit arbitrary. 4GB ? 256M ? 1/32?  I don't think
> we really want to be blowing 32GB on a crash kernel, even if we do have
> 1T of RAM :)
>   

Ah, maybe, to be honest, I am not familiar with ppc at all.

Please feel free to suggest other numbers for ppc (or other algorithms 
to reserve memory automatically for ppc).

Thanks!