From: Muchun Song <songmuchun@bytedance.com>
To: Muchun Song <muchun.song@linux.dev>,
	Oscar Salvador <osalvador@suse.de>,
	Andrew Morton <akpm@linux-foundation.org>
Cc: David Hildenbrand <david@kernel.org>,
	Kiryl Shutsemau <kas@kernel.org>,
	linux-mm@kvack.org, linux-kernel@vger.kernel.org,
	stable@vger.kernel.org, Muchun Song <songmuchun@bytedance.com>
Subject: [PATCH] mm/hugetlb_vmemmap: fix incorrect vmemmap restore in rollback
Date: Mon, 25 May 2026 10:52:13 +0800	[thread overview]
Message-ID: <20260525025213.2229628-1-songmuchun@bytedance.com> (raw)

vmemmap_restore_pte() rebuilds restored vmemmap pages from a
tail-page template derived from compound_head(). This is wrong when the
current PTE already maps a page whose contents are not tail-page
metadata.

In the rollback path of vmemmap_remap_free(), the first restored PTE is
backed by vmemmap_head and contains head-page metadata. Reconstructing
that page from a tail-page template overwrites the head-page state and
corrupts the restored vmemmap page.

Fix this by copying the full page from the page currently mapped by the
PTE. Also pass vmemmap_tail to the rollback walk so only PTEs backed by
the shared tail page are restored, while the head PTE remains mapped to
vmemmap_head. Add VM_WARN_ON_ONCE() checks for unexpected cases.

Fixes: c0b495b91a47 ("mm/hugetlb: refactor code around vmemmap_walk")
Cc: stable@vger.kernel.org
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
---
 mm/hugetlb_vmemmap.c | 36 ++++++++++++++++++------------------
 1 file changed, 18 insertions(+), 18 deletions(-)

diff --git a/mm/hugetlb_vmemmap.c b/mm/hugetlb_vmemmap.c
index 4a077d231d3a..133b46dfb09f 100644
--- a/mm/hugetlb_vmemmap.c
+++ b/mm/hugetlb_vmemmap.c
@@ -207,6 +207,8 @@ static void vmemmap_remap_pte(pte_t *pte, unsigned long addr,
 
 	/* Remapping the head page requires r/w */
 	if (unlikely(walk->nr_walked == 0 && walk->vmemmap_head)) {
+		VM_WARN_ON_ONCE(!PageHead((const struct page *)addr));
+
 		list_del(&walk->vmemmap_head->lru);
 
 		/*
@@ -218,6 +220,8 @@ static void vmemmap_remap_pte(pte_t *pte, unsigned long addr,
 
 		entry = mk_pte(walk->vmemmap_head, PAGE_KERNEL);
 	} else {
+		VM_WARN_ON_ONCE(!PageTail((const struct page *)addr));
+
 		/*
 		 * Remap the tail pages as read-only to catch illegal write
 		 * operation to the tail pages.
@@ -232,33 +236,28 @@ static void vmemmap_remap_pte(pte_t *pte, unsigned long addr,
 static void vmemmap_restore_pte(pte_t *pte, unsigned long addr,
 				struct vmemmap_remap_walk *walk)
 {
-	struct page *page;
-	struct page *from, *to;
-
-	page = list_first_entry(walk->vmemmap_pages, struct page, lru);
-	list_del(&page->lru);
+	struct page *src = pte_page(ptep_get(pte)), *dst;
 
 	/*
-	 * Initialize tail pages in the newly allocated vmemmap page.
-	 *
-	 * There is folio-scope metadata that is encoded in the first few
-	 * tail pages.
-	 *
-	 * Use the value last tail page in the page with the head page
-	 * to initialize the rest of tail pages.
+	 * When rolling back vmemmap_remap_free(), keep the copied head page
+	 * mapping and restore only PTEs currently pointing at the shared tail
+	 * page.
 	 */
-	from = compound_head((struct page *)addr) +
-		PAGE_SIZE / sizeof(struct page) - 1;
-	to = page_to_virt(page);
-	for (int i = 0; i < PAGE_SIZE / sizeof(struct page); i++, to++)
-		*to = *from;
+	if (walk->vmemmap_tail && walk->vmemmap_tail != src)
+		return;
+
+	VM_WARN_ON_ONCE(PageHead((const struct page *)addr));
+
+	dst = list_first_entry(walk->vmemmap_pages, struct page, lru);
+	list_del(&dst->lru);
+	copy_page(page_to_virt(dst), page_to_virt(src));
 
 	/*
 	 * Makes sure that preceding stores to the page contents become visible
 	 * before the set_pte_at() write.
 	 */
 	smp_wmb();
-	set_pte_at(&init_mm, addr, pte, mk_pte(page, PAGE_KERNEL));
+	set_pte_at(&init_mm, addr, pte, mk_pte(dst, PAGE_KERNEL));
 }
 
 /**
@@ -324,6 +323,7 @@ static int vmemmap_remap_free(unsigned long start, unsigned long end,
 	 */
 	walk = (struct vmemmap_remap_walk) {
 		.remap_pte	= vmemmap_restore_pte,
+		.vmemmap_tail	= vmemmap_tail,
 		.vmemmap_pages	= vmemmap_pages,
 		.flags		= 0,
 	};

base-commit: e98d21c170b01ddef366f023bbfcf6b31509fa83
-- 
2.54.0

next             reply	other threads:[~2026-05-25  2:52 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-25  2:52 Muchun Song [this message]
2026-05-25 15:52 ` [PATCH] mm/hugetlb_vmemmap: fix incorrect vmemmap restore in rollback Kiryl Shutsemau
2026-05-25 17:04 ` Oscar Salvador (SUSE)
2026-05-25 21:49 ` Andrew Morton
2026-05-26  2:01   ` 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=20260525025213.2229628-1-songmuchun@bytedance.com \
    --to=songmuchun@bytedance.com \
    --cc=akpm@linux-foundation.org \
    --cc=david@kernel.org \
    --cc=kas@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=muchun.song@linux.dev \
    --cc=osalvador@suse.de \
    --cc=stable@vger.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

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.