Re: [PATCH 2/2] selftests/mm: fix ksm NUMA merge test for systems with memoryless NUMA nodes

[Sayali Patil <sayalip@linux.ibm.com>]

On 26/06/26 21:25, David Hildenbrand (Arm) wrote:
>>   
>> +static int count_mem_nodes(void)
>> +{
>> +	int node, count = 0;
>> +
>> +	for (node = 0; node <= numa_max_node(); node++) {
>> +		if (numa_node_size(node, NULL) > 0)
>> +			count++;
>> +	}
>> +
>> +	return count;
>> +}
>> +
> 
> Can we instead build upon our existing helpers get_first_mem_node() +
> get_next_mem_node() ?
> 
>  From 432774fb50237519c1c041e402fddfdf4b35aa2c Mon Sep 17 00:00:00 2001
> From: "David Hildenbrand (Arm)" <david@kernel.org>
> Date: Fri, 26 Jun 2026 17:52:21 +0200
> Subject: [PATCH] tmp
> 
> Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
> ---
>   tools/testing/selftests/mm/ksm_tests.c | 16 +++++++++-------
>   1 file changed, 9 insertions(+), 7 deletions(-)
> 
> diff --git a/tools/testing/selftests/mm/ksm_tests.c b/tools/testing/selftests/mm/ksm_tests.c
> index a050f4840cfa3..2ebbb544c6711 100644
> --- a/tools/testing/selftests/mm/ksm_tests.c
> +++ b/tools/testing/selftests/mm/ksm_tests.c
> @@ -440,9 +440,9 @@ static int get_next_mem_node(int node)
>   		mem_node = i % (max_node + 1);
>   		node_size = numa_node_size(mem_node, NULL);
>   		if (node_size > 0)
> -			break;
> +			return mem_node;
>   	}
> -	return mem_node;
> +	return -ENODEV;
>   }
>   
>   static int get_first_mem_node(void)
> @@ -455,8 +455,8 @@ static int check_ksm_numa_merge(int merge_type, int mapping, int prot, int timeo
>   {
>   	void *numa1_map_ptr, *numa2_map_ptr;
>   	struct timespec start_time;
> +	int first_node, second_node;
>   	int page_count = 2;
> -	int first_node;
>   
>   	if (clock_gettime(CLOCK_MONOTONIC_RAW, &start_time)) {
>   		ksft_perror("clock_gettime");
> @@ -467,17 +467,19 @@ static int check_ksm_numa_merge(int merge_type, int mapping, int prot, int timeo
>   		ksft_print_msg("NUMA support not enabled\n");
>   		return KSFT_SKIP;
>   	}
> -	if (numa_num_configured_nodes() <= 1) {
> -		ksft_print_msg("At least 2 NUMA nodes must be available\n");
> +	first_node = get_first_mem_node();
> +	second_node = get_next_mem_node(first_node);
> +
> +	if (second_node < 0) {
> +		ksft_print_msg("At least 2 NUMA nodes with memory must be available\n");
>   		return KSFT_SKIP;
>   	}
>   	if (ksm_write_sysfs(KSM_FP("merge_across_nodes"), merge_across_nodes))
>   		return KSFT_FAIL;
>   
>   	/* allocate 2 pages in 2 different NUMA nodes and fill them with the same data */
> -	first_node = get_first_mem_node();
>   	numa1_map_ptr = numa_alloc_onnode(page_size, first_node);
> -	numa2_map_ptr = numa_alloc_onnode(page_size, get_next_mem_node(first_node));
> +	numa2_map_ptr = numa_alloc_onnode(page_size, second_node);
>   	if (!numa1_map_ptr || !numa2_map_ptr) {
>   		ksft_perror("numa_alloc_onnode");
>   		return KSFT_FAIL;

Hi David,

Thanks for review!

Yes, I updated the implementation in v2 to use the existing
get_first_mem_node() + get_next_mem_node() helper as suggested.

V2:https://lore.kernel.org/all/878cbc1dd24921d55049932dcd5fd6ee517557ca.1782811071.git.sayalip@linux.ibm.com/