[PATCH] x86: Find offset for crashkernel reservation automatically

[Yinghai Lu]

On Mon, Jul 14, 2008 at 2:44 AM, Eric W. Biederman
<ebiederm at xmission.com> wrote:
> Bernhard Walle <bwalle at suse.de> writes:
>
>> * Yinghai Lu [2008-07-14 00:11]:
>>>
>>> should use min_t(u64, 1ULL<<32, max_low_pfn<<PAGE_SHIFT) replace ULONG_MAX
>>
>> Shouldn't we use  min_t(u64, ULLONG_MAX, max_low_pfn<<PAGE_SHIFT), i.e.
>> should we really limit the crashkernel to a 32 bit address on a 64 bit
>> system?
>
> We should use the lowest physical address that meets our size and
> alignment constraints.  However there is no reason to make this be <
> 4G or even < 1G.  The worst case is that we reserve an area the kdump
> kernel can't run out of.  However /sbin/kexec should check for that as
> the kexec on panic code is not necessarily a linux kernel.

for 64bit, kdump can start from address above 4g with bzImage?

YH