Re: [RESEND makedumpfile PATCH 0/2] Add 5-level paging support

[Dou Liyang <douly.fnst@cn.fujitsu.com>]

Hi Baoquan,

At 02/28/2018 04:25 PM, Baoquan He wrote:
> On 02/28/18 at 04:12pm, Dou Liyang wrote:
>> Hi Baoquan,
>>
>> At 02/28/2018 04:00 PM, Baoquan He wrote:
>>> Hi dou,
>>>
>>> On 02/28/18 at 03:13pm, Dou Liyang wrote:
>>>> this patchset adds 5-level paging support in makedumpfile
>>>>
>>>>    -the 1th patch does some cleanup for the preparation of 5-level page tables support
>>>>    -the 2th patch add the 5-level page tables support
>>>>
>>>> Dou Liyang (2):
>>>>     arch/x86_64: Cleanup the address translation of the 4-level page
>>>>       tables
>>>>     arch/x86_64: Add 5-level paging support
>>>
>>> It doesn't work on my qemu+la57, kernel 4.16-rc1+. Not sure what kernel
>>> you are testing, it truly works on kernel 4.16-rc1+ when cpu la57 is not
>>
>> I tested in Linux v4.15+ and v4.16-rc2+.
>>
>> The following commits in linux kernel changed mem_section array and
>> it causes this problem in makedumpfile.
>>
>> commit 83e3c48729d9ebb7af5a31a504f3fd6aff0348c4
>> Author: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
>> Date:   Fri Sep 29 17:08:16 2017 +0300
>>
>> commit 629a359bdb0e0652a8227b4ff3125431995fec6e
>> Author: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
>> Date:   Tue Nov 7 11:33:37 2017 +0300
>>
>> So when I tested makedumpfile, I reverted these commits from kernel
>> temporarily.
> 
> Hmm, how much memory do you reserve for crashkernel? I am afraid both of

768M as you suggested.

> them need be kept in kernel, since makedumpfile need be adapted to kernel,
> might be not the opposite. My personal opinion.

Yes, I agree, revert them just for temporarily tests.

> 
> So what's your plan about these two reverted commits? Will you change
> it in kernel, or just leave it as is and may post another patch to fix
> it?
> 

I have no idea about it. IMO, It need to be fixed in makedumpfile
tools.

Resently, I found Thadeu posted a patch about the mem_section.

   [PATCH makedumpfile] handle mem_section as either a pointer or an array

I tested makedumpfile with this patch, also can't fix the problem

BTW, it seems I misunderstood you comments, the problem I met only
happened with the following command, which we discussed before.

    /makedumpfile -x vmlinux  vmcore  vmcore_file

Using the makedumpfile as core_collector in /etc/kdump.conf is OK for
me. So, can you gave me your error log?

I will retest this patch without reverting this two commits tomorrow.

Thanks
	dou

> I will take time to check this issue, maybe later. Not sure what
> makedumpfile maintainers plan to do about these?
> 
> Thanks
> Baoquan
> 
>>
>>> supported?
>>>
>>> Thanks
>>> Baoquan
>>>
>>>>
>>>>    arch/x86_64.c  | 131 +++++++++++++++++++++++++++++++++++++++++----------------
>>>>    makedumpfile.h |  38 +++++++++++------
>>>>    2 files changed, 120 insertions(+), 49 deletions(-)
>>>>
>>>> -- 
>>>> 2.14.3
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> kexec mailing list
>>>> kexec@lists.infradead.org
>>>> http://lists.infradead.org/mailman/listinfo/kexec
>>>
>>>
>>>
>>
>>
>>
>> _______________________________________________
>> kexec mailing list
>> kexec@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/kexec
> 
> 
> 



_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec