Re: [PATCH v2] mm: remove migration for HugePage in isolate_single_pageblock()

[David Hildenbrand <david@redhat.com>]

On 20.08.24 05:26, Kefeng Wang wrote:
> The gigantic page size may larger than memory block size, so memory
> offline always fails in this case after commit b2c9e2fbba32 ("mm: make
> alloc_contig_range work at pageblock granularity"),
> 
> offline_pages
>    start_isolate_page_range
>      start_isolate_page_range(isolate_before=true)
>        isolate [isolate_start, isolate_start + pageblock_nr_pages)
>      start_isolate_page_range(isolate_before=false)
>        isolate [isolate_end - pageblock_nr_pages, isolate_end) pageblock
>         	__alloc_contig_migrate_range
>            isolate_migratepages_range
>              isolate_migratepages_block
>                isolate_or_dissolve_huge_page
>                  if (hstate_is_gigantic(h))
>                      return -ENOMEM;
> 
> [   15.815756] memory offlining [mem 0x3c0000000-0x3c7ffffff] failed due to failure to isolate range
> 
> Gigantic PageHuge is bigger than a pageblock, but since it is freed as
> order-0 pages, its pageblocks after being freed will get to the right
> free list. There is no need to have special handling code for them in
> start_isolate_page_range(). For both alloc_contig_range() and memory
> offline cases, the migration code after start_isolate_page_range() will
> be able to migrate gigantic PageHuge when possible.
> 
> Let's clean up start_isolate_page_range() and fix the aforementioned
> memory offline failure issue all together.
> 
> Fixes: b2c9e2fbba32 ("mm: make alloc_contig_range work at pageblock granularity")
> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
> ---
> v2:
> - update changelog, thanks Zi, David
> 
>   mm/page_isolation.c | 28 +++-------------------------
>   1 file changed, 3 insertions(+), 25 deletions(-)
> 
> diff --git a/mm/page_isolation.c b/mm/page_isolation.c
> index 39fb8c07aeb7..7e04047977cf 100644
> --- a/mm/page_isolation.c
> +++ b/mm/page_isolation.c
> @@ -403,30 +403,8 @@ static int isolate_single_pageblock(unsigned long boundary_pfn, int flags,
>   			unsigned long head_pfn = page_to_pfn(head);
>   			unsigned long nr_pages = compound_nr(head);
>   
> -			if (head_pfn + nr_pages <= boundary_pfn) {
> -				pfn = head_pfn + nr_pages;
> -				continue;
> -			}
> -
> -#if defined CONFIG_COMPACTION || defined CONFIG_CMA
> -			if (PageHuge(page)) {
> -				int page_mt = get_pageblock_migratetype(page);
> -				struct compact_control cc = {
> -					.nr_migratepages = 0,
> -					.order = -1,
> -					.zone = page_zone(pfn_to_page(head_pfn)),
> -					.mode = MIGRATE_SYNC,
> -					.ignore_skip_hint = true,
> -					.no_set_skip_hint = true,
> -					.gfp_mask = gfp_flags,
> -					.alloc_contig = true,
> -				};
> -				INIT_LIST_HEAD(&cc.migratepages);
> -
> -				ret = __alloc_contig_migrate_range(&cc, head_pfn,
> -							head_pfn + nr_pages, page_mt);
> -				if (ret)
> -					goto failed;
> +			if (head_pfn + nr_pages <= boundary_pfn ||
> +			    PageHuge(page)) {

I'm wondering if we should have here some kind of WARN_ON_ONCE if 
PageLRU + "spans more than a single pageblock" check.

Then we could catch whenever we would have !hugetlb LRU folios that span 
more than a single pageblock.

/*
  * We cannot currently handle movable (LRU) folios that span more than
  * a single pageblock. hugetlb folios are fine, though.
  */
WARN_ON_ONCE(PageLRU(page) && nr_pages > pageblock_nr_pages);


But now I realized something I previously missed: We are only modifying 
behavior of hugetlb folios ... stupid misleading "PageHuge" check :)

So that would be independent of this change.

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers,

David / dhildenb