kdump regression: /proc/vmcore is 0 after kdump

[Luming Yu]

On Sat, Jun 14, 2008 at 12:22 AM, Luming Yu <luming.yu at gmail.com> wrote:
> On Thu, Jun 5, 2008 at 4:16 PM, Simon Horman <horms at verge.net.au> wrote:
>> On Thu, Jun 05, 2008 at 03:58:40PM +0800, Luming Yu wrote:
>>> On Tue, Jun 3, 2008 at 2:02 PM, Luming Yu <luming.yu at gmail.com> wrote:
>>> >> I'm trying upstream kdump to get more detailed data for help analysis
>>> >> the scenario..
>>> > Upstream kdump doesn't work! don't know what cause the regression,
>>> > possibly kexec-tools...
>>> > Need investigation here!
>>> >
>>>
>>> I tested mainline kernel from 2.6.20 to 2.6.26-rc4 on my tiger4..
>>> whatever I try, /proc/vmcore is always 0 !
>>>
>>> Then I managed to get Nan hai's original patch against 2.6.19..
>>> http://marc.info/?l=linux-ia64&m=116521104831321&w=2
>>> This patch doesn't have zero-vmcore problem.
>>>
>>> Given the age of the kdump feature and so many upstream releases since
>>> 2.6.20, I'm in doubt why I don't see this kind of bug report on ia64
>>> mailing list before?
>>>
>>> PS. Now I can start to capture info for the strace problem with 2.6.19
>>> kernel +Nahai's 2.6.19 kdump patch, hope the problem is still
>>> reproducible.
>>
>> Which version of kexec-tools are you using?
>> If you haven't done so already, could you try the latest version, 20080324.
>>
>> http://www.eu.kernel.org/pub/linux/kernel/people/horms/kexec-tools/
>>
>
> Yes, this resolves one of my kdump problem.
> But there seems to have a regression introduced into upstream probably
> starting from 2.6.25.
> The last kernel with kdump working on this box is 2.6.24. I don't have
> time to start bisection to find out which patch causes that regression
> right now.Maybe someone on this list can give me some hint..

My git bisection result indicates the following commit causes the
kdump regression in 2.6.25-rc1:
commit 3c75e23784e6ed5f4841de43d0750fd9b37bafcb
Author: Andrew Patterson <andrew.patterson at hp.com>
Date:   Tue Jan 22 17:18:27 2008 -0700