Re: [PATCH 6/7] mm: hugeltb: check NUMA_NO_NODE in only_alloc_fresh_hugetlb_folio()

[Kefeng Wang <wangkefeng.wang@huawei.com>]

On 2025/8/7 4:43, jane.chu@oracle.com wrote:
> 
> 
> On 8/2/2025 12:31 AM, Kefeng Wang wrote:
>> Move the NUMA_NO_NODE check out of buddy and gigantic folio allocation
>> to cleanup code a bit.
>>
>> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
>> ---
>>   mm/hugetlb.c | 7 +++----
>>   1 file changed, 3 insertions(+), 4 deletions(-)
>>
>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>> index b16011c9645d..4f73b74a2cff 100644
>> --- a/mm/hugetlb.c
>> +++ b/mm/hugetlb.c
>> @@ -1479,8 +1479,6 @@ static struct folio *alloc_gigantic_folio(int 
>> order, gfp_t gfp_mask,
>>       struct folio *folio;
>>       bool retried = false;
>> -    if (nid == NUMA_NO_NODE)
>> -        nid = numa_mem_id();
>>   retry:
>>       folio = hugetlb_cma_alloc_folio(order, gfp_mask, nid, nodemask);
>>       if (!folio) {
>> @@ -1942,8 +1940,6 @@ static struct folio 
>> *alloc_buddy_hugetlb_folio(int order, gfp_t gfp_mask,
>>           alloc_try_hard = false;
>>       if (alloc_try_hard)
>>           gfp_mask |= __GFP_RETRY_MAYFAIL;
>> -    if (nid == NUMA_NO_NODE)
>> -        nid = numa_mem_id();
> 
> Looks like you're also fixing a bug, because the code block was
>          if (node_alloc_noretry && node_isset(nid, *node_alloc_noretry)) 
>   <--
>                  alloc_try_hard = false;
>          if (alloc_try_hard)
>                  gfp_mask |= __GFP_RETRY_MAYFAIL;
>          if (nid == NUMA_NO_NODE)
>                  nid = numa_mem_id();
> in theory, NUMA_NO_NODE (-1) could be passed as 'nid' to node_isset().

alloc_surplus_hugetlb_folio/hugetlb_hstate_alloc_pages_onenode call
alloc_buddy_hugetlb_folio with node_alloc_noretry = NULL, and
alloc_pool_huge_folio call alloc_buddy_hugetlb_folio  with nid != 
NUMA_NO_NODE, so no issue for now, but we move it ahead which avoid
the possible issue in the future.


> 
>>       folio = (struct folio *)__alloc_frozen_pages(gfp_mask, order, 
>> nid, nmask);
>>       if (folio) {
>> @@ -1976,6 +1972,9 @@ static struct folio 
>> *only_alloc_fresh_hugetlb_folio(struct hstate *h,
>>       struct folio *folio;
>>       int order = huge_page_order(h);
>> +    if (nid == NUMA_NO_NODE)
>> +        nid = numa_mem_id();
>> +
>>       if (order > MAX_PAGE_ORDER)
>>           folio = alloc_gigantic_folio(order, gfp_mask, nid, nmask);
>>       else
> Looks good, thanks!
> 
> Reviewed-by: Jane Chu <jane.chu@oracle.com>

Thanks.