On 08.07.20 09:38, Mike Rapoport wrote:
> On Wed, Jul 08, 2020 at 09:21:25AM +0200, David Hildenbrand wrote:
>> On 08.07.20 07:27, Mike Rapoport wrote:
>>> On Tue, Jul 07, 2020 at 03:05:48PM -0700, Dan Williams wrote:
>>>> On Tue, Jul 7, 2020 at 11:01 AM Mike Rapoport <rppt@xxxxxxxxxxxxx> wrote:
>>>>>
>>>>> On Tue, Jul 07, 2020 at 02:26:08PM +0200, David Hildenbrand wrote:
>>>>>> On 07.07.20 14:13, Mike Rapoport wrote:
>>>>>>> On Tue, Jul 07, 2020 at 01:54:54PM +0200, Michal Hocko wrote:
>>>>>>>> On Tue 07-07-20 13:59:15, Jia He wrote:
>>>>>>>>> This exports memory_add_physaddr_to_nid() for module driver to use.
>>>>>>>>>
>>>>>>>>> memory_add_physaddr_to_nid() is a fallback option to get the nid in case
>>>>>>>>> NUMA_NO_NID is detected.
>>>>>>>>>
>>>>>>>>> Suggested-by: David Hildenbrand <david@xxxxxxxxxx>
>>>>>>>>> Signed-off-by: Jia He <justin.he@xxxxxxx>
>>>>>>>>> ---
>>>>>>>>> arch/arm64/mm/numa.c | 5 +++--
>>>>>>>>> 1 file changed, 3 insertions(+), 2 deletions(-)
>>>>>>>>>
>>>>>>>>> diff --git a/arch/arm64/mm/numa.c b/arch/arm64/mm/numa.c
>>>>>>>>> index aafcee3e3f7e..7eeb31740248 100644
>>>>>>>>> --- a/arch/arm64/mm/numa.c
>>>>>>>>> +++ b/arch/arm64/mm/numa.c
>>>>>>>>> @@ -464,10 +464,11 @@ void __init arm64_numa_init(void)
>>>>>>>>>
>>>>>>>>> /*
>>>>>>>>> * We hope that we will be hotplugging memory on nodes we already know about,
>>>>>>>>> - * such that acpi_get_node() succeeds and we never fall back to this...
>>>>>>>>> + * such that acpi_get_node() succeeds. But when SRAT is not present, the node
>>>>>>>>> + * id may be probed as NUMA_NO_NODE by acpi, Here provide a fallback option.
>>>>>>>>> */
>>>>>>>>> int memory_add_physaddr_to_nid(u64 addr)
>>>>>>>>> {
>>>>>>>>> - pr_warn("Unknown node for memory at 0x%llx, assuming node 0\n", addr);
>>>>>>>>> return 0;
>>>>>>>>> }
>>>>>>>>> +EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
>>>>>>>>
>>>>>>>> Does it make sense to export a noop function? Wouldn't make more sense
>>>>>>>> to simply make it static inline somewhere in a header? I haven't checked
>>>>>>>> whether there is an easy way to do that sanely bu this just hit my eyes.
>>>>>>>
>>>>>>> We'll need to either add a CONFIG_ option or arch specific callback to
>>>>>>> make both non-empty (x86, powerpc, ia64) and empty (arm64, sh)
>>>>>>> implementations coexist ...
>>>>>>
>>>>>> Note: I have a similar dummy (return 0) patch for s390x lying around here.
>>>>>
>>>>> Then we'll call it a tie - 3:3 ;-)
>>>>
>>>> So I'd be happy to jump on the train of people wanting to export the
>>>> ARM stub for this (and add a new ARM stub for phys_to_target_node()),
>>>> but Will did have a plausibly better idea that I have been meaning to
>>>> circle back to:
>>>>
>>>> http://lore.kernel.org/r/20200325111039.GA32109@willie-the-truck
>>>>
>>>> ...i.e. iterate over node data to do the lookup. This would seem to
>>>> work generically for multiple archs unless I am missing something?
>>
>> IIRC, only memory assigned to/onlined to a ZONE is represented in the
>> pgdat node span. E.g., not offline memory blocks.
>>
>> Esp., when hotplugging + onlining consecutive memory, there won't really
>> be any intersections in most cases if I am not wrong. It would not be
>> "intersection" but rather "closest fit".
>>
>> With overlapping nodes it's even more unclear. Which one to pick?
>>
>>>
>>> I think it would work on arm64, power and, most propbably on s390
>>
>> With only a single dummy node I guess it should work (searching when
>> there is only a single node does not make too much sense).
>>
>>> (David?), but not on x86. x86 does not have reserved memory in pgdat,
>>> it's never memblock_add()'ed (see e820__memblock_setup()).
>>
>> Can you enlighten me why that is relevant for the memory hotplug path?
>> (or is it just a general comment to make the function as accurate as
>> possible for all addresses?)
>
> phys_to_target_node() on x86 falls back to numa_reserved_meminfo which
> holds memory that is never listed in a node.
>
Ah, I see - thanks.
--
Thanks,
David / dhildenb
