From: Muchun Song <songmuchun@bytedance.com>
To: Andrew Morton <akpm@linux-foundation.org>,
David Hildenbrand <david@kernel.org>,
Oscar Salvador <osalvador@suse.de>,
Mike Rapoport <rppt@kernel.org>
Cc: "Liam R . Howlett" <liam@infradead.org>,
Vlastimil Babka <vbabka@kernel.org>,
Michal Hocko <mhocko@suse.com>,
Muchun Song <muchun.song@linux.dev>,
Li Zhe <lizhe.67@bytedance.com>,
Andrew Popple <apopple@nvidia.com>,
Balbir Singh <balbirs@nvidia.com>,
Lorenzo Stoakes <ljs@kernel.org>,
Suren Baghdasaryan <surenb@google.com>,
James Houghton <jthoughton@google.com>,
Lance Yang <lance.yang@linux.dev>, Qi Zheng <qi.zheng@linux.dev>,
linux-mm@kvack.org, linux-kernel@vger.kernel.org,
Muchun Song <songmuchun@bytedance.com>
Subject: [PATCH v3 07/11] mm/hugetlb: use common compound page initialization helper
Date: Fri, 20 Nov 2026 18:08:52 +0800 [thread overview]
Message-ID: <20261120100856.16929-8-songmuchun@bytedance.com> (raw)
In-Reply-To: <20261120100856.16929-1-songmuchun@bytedance.com>
Boot-time gigantic hugetlb folios currently use a separate path to
initialize their compound struct pages.
Use memmap_init_compound_page() for these folios so that HugeTLB and
ZONE_DEVICE share the common compound page initialization path. This
removes the HugeTLB-specific implementation and improves initialization
performance.
ZONE_DEVICE initializes struct pages by walking one contiguous memory
range. Boot-time HugeTLB instead initializes multiple disjoint ranges,
one for each contiguous hugepage. Measurements showed that memcpy()
performed better than memcpy_nontemporal() for the latter access pattern.
Therefore, use memcpy() for normal memory while retaining
memcpy_nontemporal() for ZONE_DEVICE.
The initialization time for 64 boot-time 1G hugepages with 4K base
pages was measured as follows:
arch without this patch with this patch reduction
x86-64 195.60 ms 170.40 ms 12.88%
arm64 374.40 ms 237.60 ms 36.54%
Co-developed-by: Li Zhe <lizhe.67@bytedance.com>
Signed-off-by: Li Zhe <lizhe.67@bytedance.com>
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
---
v3:
- Integrate the template-copy optimization into the common helper switch
(suggested by Mike Rapoport)
- Use memcpy() for normal memory and keep memcpy_nontemporal() for
ZONE_DEVICE
- Update performance results for the integrated implementation (measured
by Li Zhe)
- Add Co-developed-by for Li Zhe
---
mm/hugetlb.c | 52 ++--------------------------------------------------
mm/mm_init.c | 10 ++++++----
2 files changed, 8 insertions(+), 54 deletions(-)
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 2a94f8f60870..6a613d9c9332 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -3319,50 +3319,6 @@ static bool __init alloc_bootmem_huge_page(struct hstate *h, int nid)
return true;
}
-/* Initialize [start_page:end_page_number] tail struct pages of a hugepage */
-static void __init hugetlb_folio_init_tail_vmemmap(struct folio *folio,
- struct hstate *h,
- unsigned long start_page_number,
- unsigned long end_page_number)
-{
- enum zone_type zone = folio_zonenum(folio);
- int nid = folio_nid(folio);
- struct page *page = folio_page(folio, start_page_number);
- unsigned long head_pfn = folio_pfn(folio);
- unsigned long pfn, end_pfn = head_pfn + end_page_number;
- unsigned int order = huge_page_order(h);
-
- /*
- * As we marked all tail pages with memblock_reserved_mark_noinit(),
- * we must initialize them ourselves here.
- */
- for (pfn = head_pfn + start_page_number; pfn < end_pfn; page++, pfn++) {
- __init_single_page(page, pfn, zone, nid);
- prep_compound_tail(page, &folio->page, order);
- set_page_count(page, 0);
- }
-}
-
-static void __init hugetlb_folio_init_vmemmap(struct folio *folio,
- unsigned long pfn,
- struct hstate *h,
- enum zone_type zone, int nid,
- unsigned long nr_pages)
-{
- struct page *head = &folio->page;
-
- /*
- * This is an open-coded prep_compound_page() whereby we avoid
- * walking pages twice by initializing/preparing+freezing them in the
- * same go.
- */
- __init_single_page(head, pfn, zone, nid);
- __folio_set_head(folio);
- set_page_count(head, 0);
- hugetlb_folio_init_tail_vmemmap(folio, h, 1, nr_pages);
- prep_compound_head(head, huge_page_order(h));
-}
-
/*
* memblock-allocated pageblocks might not have the migrate type set
* if marked with the 'noinit' flag. Set it to the default (MIGRATE_MOVABLE)
@@ -3420,8 +3376,7 @@ static void __init gather_bootmem_prealloc_node(unsigned long nid)
struct folio *folio = (void *)page;
const unsigned long pfn = PHYS_PFN(__pa(m));
const unsigned long nr_pages = pages_per_huge_page(m->hstate);
- const enum zone_type zone = zone_idx(pfn_to_zone(pfn, nid));
- unsigned long nr_struct_pages = nr_pages;
+ const struct zone *zone = pfn_to_zone(pfn, nid);
h = m->hstate;
/*
@@ -3433,10 +3388,7 @@ static void __init gather_bootmem_prealloc_node(unsigned long nid)
prev_h = h;
VM_BUG_ON(!hstate_is_gigantic(h));
-
- if (vmemmap_optimizable_order(pfn_to_section_compound_order(pfn)))
- nr_struct_pages = VMEMMAP_OPTIMIZATION_NR_STRUCT_PAGES;
- hugetlb_folio_init_vmemmap(folio, pfn, h, zone, nid, nr_struct_pages);
+ memmap_init_compound_page(page, pfn, zone, huge_page_order(h));
init_new_hugetlb_folio(folio);
if (vmemmap_optimizable_order(pfn_to_section_compound_order(pfn)))
diff --git a/mm/mm_init.c b/mm/mm_init.c
index 3e645302377e..8bf350d5a695 100644
--- a/mm/mm_init.c
+++ b/mm/mm_init.c
@@ -722,16 +722,18 @@ static void __meminit memmap_init_page_from_template(struct page *page,
unsigned long pfn, const struct zone *zone, struct page *template)
{
/*
- * Refresh PFN-dependent fields in the template before the non-temporal
- * copy, so the destination page does not need cached stores after
- * memcpy_nontemporal().
+ * Refresh PFN-dependent fields in the template before copying it, so
+ * the destination page does not need follow-up stores.
*/
set_page_section_from_pfn(template, pfn);
#ifdef WANT_PAGE_VIRTUAL
if (!is_highmem_idx(zone_idx(zone)))
set_page_address(template, __va(pfn << PAGE_SHIFT));
#endif
- memcpy_nontemporal(page, template, sizeof(*page));
+ if (zone_is_zone_device(zone))
+ memcpy_nontemporal(page, template, sizeof(*page));
+ else
+ memcpy(page, template, sizeof(*page));
vmemmap_optimization_verify_zone(page);
}
--
2.54.0
next prev parent reply other threads:[~2026-11-20 10:10 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-20 10:08 [PATCH v3 00/11] mm: Unify memmap initialization for compound pages Muchun Song
2026-11-20 10:08 ` [PATCH v3 01/11] mm/mm_init: verify zones for template-initialized pages Muchun Song
2026-11-20 10:08 ` [PATCH v3 02/11] mm/mm_init: factor out common struct page initialization Muchun Song
2026-11-20 10:08 ` [PATCH v3 03/11] mm/mm_init: simplify zone device " Muchun Song
2026-11-20 10:08 ` [PATCH v3 04/11] mm/mm_init: factor out compound " Muchun Song
2026-11-20 10:08 ` [PATCH v3 05/11] mm/hugetlb: drop boot-time HVO handling for gigantic folios Muchun Song
2026-11-20 10:08 ` [PATCH v3 06/11] mm/hugetlb: initialize all struct pages for bootmem hugepages Muchun Song
2026-11-20 10:08 ` Muchun Song [this message]
2026-11-20 10:08 ` [PATCH v3 08/11] mm/mm_init: make __init_single_page() static Muchun Song
2026-11-20 10:08 ` [PATCH v3 09/11] mm/cma: centralize CMA pageblock initialization Muchun Song
2026-11-20 10:08 ` [PATCH v3 10/11] mm/cma: localize init_cma_pageblock() to cma.c Muchun Song
2026-11-20 10:08 ` [PATCH v3 11/11] mm/mm_init: move pageblock migratetype init to memmap helpers Muchun Song
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20261120100856.16929-8-songmuchun@bytedance.com \
--to=songmuchun@bytedance.com \
--cc=akpm@linux-foundation.org \
--cc=apopple@nvidia.com \
--cc=balbirs@nvidia.com \
--cc=david@kernel.org \
--cc=jthoughton@google.com \
--cc=lance.yang@linux.dev \
--cc=liam@infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=lizhe.67@bytedance.com \
--cc=ljs@kernel.org \
--cc=mhocko@suse.com \
--cc=muchun.song@linux.dev \
--cc=osalvador@suse.de \
--cc=qi.zheng@linux.dev \
--cc=rppt@kernel.org \
--cc=surenb@google.com \
--cc=vbabka@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®