On 17.01.20 10:49, Pingfan Liu wrote:
> On Fri, Jan 17, 2020 at 3:47 PM Michal Hocko <mhocko@xxxxxxxx> wrote:
>>
>> On Thu 16-01-20 23:14:02, Dan Williams wrote:
>>> On Thu, Jan 16, 2020 at 10:23 PM Pingfan Liu <kernelfans@xxxxxxxxx> wrote:
>>>>
>>>> On Thu, Jan 16, 2020 at 3:50 PM Michal Hocko <mhocko@xxxxxxxx> wrote:
>>>>>
>>>>> On Thu 16-01-20 11:01:08, Pingfan Liu wrote:
>>>>>> When fully deactivated, it is meaningless to keep the value of a section's
>>>>>> mem_map. And its mem_map will be reassigned during re-added.
>>>>>>
>>>>>> Beside this, it breaks the user space tool "makedumpfile", which makes
>>>>>> assumption that a hot-removed section having mem_map as NULL.
>>>>>
>>>>> We used to do that before ba72b4c8cf60 ("mm/sparsemem: support
>>>>> sub-section hotplug"). Dan was this an intentional change?
>>>> I do not know the purpose of this. But the change just leave section
>>>> start pfn in fully deactivated section_mem_map, and not used any more.
>>>
>>> Not intentional, IIRC at the time I had convinced myself that the
>>> value would always be translated by sparse_decode_mem_map(), so I
>>> thought it could be hiding NULL de-references. I don't see any harm
>>> in the patch.
>>
>> Thanks for the confirmation. It would be great to have this in the
>> changelog.
> Should I post V2 with this commit log?
>
> Thanks,
> Pingfan
>
Id' say yes, and maybe also add some details why this makes makedumpfile
bail out now, so people looking at this commit only know what's happening.
--
Thanks,
David / dhildenb
