Re: [PATCH v2 2/8] s390x/mm: Implement arch_remove_memory()

[David Hildenbrand <david@redhat.com>]

On 07.05.19 22:57, Dan Williams wrote:
> On Tue, May 7, 2019 at 1:47 PM David Hildenbrand <david@redhat.com> wrote:
>>
>> On 07.05.19 22:46, Dan Williams wrote:
>>> On Tue, May 7, 2019 at 11:38 AM David Hildenbrand <david@redhat.com> wrote:
>>>>
>>>> Will come in handy when wanting to handle errors after
>>>> arch_add_memory().
>>>>
>>>> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
>>>> Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
>>>> Cc: Andrew Morton <akpm@linux-foundation.org>
>>>> Cc: Michal Hocko <mhocko@suse.com>
>>>> Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
>>>> Cc: David Hildenbrand <david@redhat.com>
>>>> Cc: Vasily Gorbik <gor@linux.ibm.com>
>>>> Cc: Oscar Salvador <osalvador@suse.com>
>>>> Signed-off-by: David Hildenbrand <david@redhat.com>
>>>> ---
>>>>  arch/s390/mm/init.c | 13 +++++++------
>>>>  1 file changed, 7 insertions(+), 6 deletions(-)
>>>>
>>>> diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c
>>>> index 31b1071315d7..1e0cbae69f12 100644
>>>> --- a/arch/s390/mm/init.c
>>>> +++ b/arch/s390/mm/init.c
>>>> @@ -237,12 +237,13 @@ int arch_add_memory(int nid, u64 start, u64 size,
>>>>  void arch_remove_memory(int nid, u64 start, u64 size,
>>>>                         struct vmem_altmap *altmap)
>>>>  {
>>>> -       /*
>>>> -        * There is no hardware or firmware interface which could trigger a
>>>> -        * hot memory remove on s390. So there is nothing that needs to be
>>>> -        * implemented.
>>>> -        */
>>>> -       BUG();
>>>> +       unsigned long start_pfn = start >> PAGE_SHIFT;
>>>> +       unsigned long nr_pages = size >> PAGE_SHIFT;
>>>> +       struct zone *zone;
>>>> +
>>>> +       zone = page_zone(pfn_to_page(start_pfn));
>>>
>>> Does s390 actually support passing in an altmap? If 'yes', I think it
>>> also needs the vmem_altmap_offset() fixup like x86-64:
>>>
>>>         /* With altmap the first mapped page is offset from @start */
>>>         if (altmap)
>>>                 page += vmem_altmap_offset(altmap);
>>>
>>> ...but I suspect it does not support altmap since
>>> arch/s390/mm/vmem.c::vmemmap_populate() does not arrange for 'struct
>>> page' capacity to be allocated out of an altmap defined page pool.
>>>
>>> I think it would be enough to disallow any arch_add_memory() on s390
>>> where @altmap is non-NULL. At least until s390 gains ZONE_DEVICE
>>> support and can enable the pmem use case.
>>>
>>
>> As far as I know, it doesn't yet, however I guess this could change once
>> virtio-pmem is supported?
> 
> I would expect and request virtio-pmem remain a non-starter on s390
> until s390 gains ZONE_DEVICE support. As it stands virtio-pmem is just
> another flavor of the general pmem driver and the pmem driver
> currently only exports ZONE_DEVICE pfns tagged by the PTE_DEVMAP
> pte-flag and PFN_DEV+PFN_MAP pfn_t-flags.

Yes, I think ZONE_DEVICE will be the way to go. On real HW, there will
never be anything mapped into the physical address space besides system
ram. However with virtio-pmem in virtual environments, we have the
option to change that.

-- 

Thanks,

David / dhildenb