Re: [PATCH] ppc64/numa: consider the max numa node for migratable LPAR

[Laurent Dufour <ldufour@linux.ibm.com>]

Le 29/04/2021 à 21:29, Tyrel Datwyler a écrit :
> On 4/29/21 11:19 AM, Laurent Dufour wrote:
>> When a LPAR is migratable, we should consider the maximum possible NUMA
>> node instead the number of NUMA node from the actual system.
>>
>> The DT property 'ibm,current-associativity-domains' is defining the maximum
>> number of nodes the LPAR can see when running on that box. But if the LPAR
>> is being migrated on another box, it may seen up to the nodes defined by
>> 'ibm,max-associativity-domains'. So if a LPAR is migratable, that value
>> should be used.
>>
>> Unfortunately, there is no easy way to know if a LPAR is migratable or
>> not. The hypervisor is exporting the property 'ibm,migratable-partition' in
>> the case it set to migrate partition, but that would not mean that the
>> current partition is migratable.
> 
> Wording is a little hard to follow for me here. From PAPR the
> 'ibm,migratable-partition' property presence indicates that the platform
> supports the potential migration of the partition. I guess maybe the point is
> that all migratable partitions define 'ibm,migratable-partition', but all
> partitions that define 'ibm,migratable-partition' are not necessarily migratable.

That's what I meant.

Laurent

> -Tyrel
> 
>>
>> Without that patch, when a LPAR is started on a 2 nodes box and then
>> migrated to a 3 nodes box, the hypervisor may spread the LPAR's CPUs on the
>> 3rd node. In that case if a CPU from that 3rd node is added to the LPAR, it
>> will be wrongly assigned to the node because the kernel has been set to use
>> up to 2 nodes (the configuration of the departure node). With that patch
>> applies, the CPU is correctly added to the 3rd node.
>>
>> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
>> Signed-off-by: Laurent Dufour <ldufour@linux.ibm.com>
>> ---
>>   arch/powerpc/mm/numa.c | 14 +++++++++++---
>>   1 file changed, 11 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c
>> index f2bf98bdcea2..673fa6e47850 100644
>> --- a/arch/powerpc/mm/numa.c
>> +++ b/arch/powerpc/mm/numa.c
>> @@ -893,7 +893,7 @@ static void __init setup_node_data(int nid, u64 start_pfn, u64 end_pfn)
>>   static void __init find_possible_nodes(void)
>>   {
>>   	struct device_node *rtas;
>> -	const __be32 *domains;
>> +	const __be32 *domains = NULL;
>>   	int prop_length, max_nodes;
>>   	u32 i;
>>
>> @@ -909,9 +909,14 @@ static void __init find_possible_nodes(void)
>>   	 * it doesn't exist, then fallback on ibm,max-associativity-domains.
>>   	 * Current denotes what the platform can support compared to max
>>   	 * which denotes what the Hypervisor can support.
>> +	 *
>> +	 * If the LPAR is migratable, new nodes might be activated after a LPM,
>> +	 * so we should consider the max number in that case.
>>   	 */
>> -	domains = of_get_property(rtas, "ibm,current-associativity-domains",
>> -					&prop_length);
>> +	if (!of_get_property(of_root, "ibm,migratable-partition", NULL))
>> +		domains = of_get_property(rtas,
>> +					  "ibm,current-associativity-domains",
>> +					  &prop_length);
>>   	if (!domains) {
>>   		domains = of_get_property(rtas, "ibm,max-associativity-domains",
>>   					&prop_length);
>> @@ -920,6 +925,9 @@ static void __init find_possible_nodes(void)
>>   	}
>>
>>   	max_nodes = of_read_number(&domains[min_common_depth], 1);
>> +	printk(KERN_INFO "Partition configured for %d NUMA nodes.\n",
>> +	       max_nodes);
>> +
>>   	for (i = 0; i < max_nodes; i++) {
>>   		if (!node_possible(i))
>>   			node_set(i, node_possible_map);
>>
>