mm: hugetlb: introduce arch_free_vmemmap_page

We register bootmem info for vmemmap pages when boot on x86-64, so
the vmemmap pages must be freed by using free_bootmem_page(). But
on some other architectures, we do not need bootmem info. In this
case, free_reserved_page() is enough to free vmemmap pages.
Currently, only x86-64 need free_bootmem_page(), so introduce a
default arch_free_vmemmap_page() which use free_reserved_page()
to free vmemmap pages directly. On x86-64, we can implement
arch_free_vmemmap_page() to override the default behavior.

Signed-off-by: Muchun Song <songmuchun@bytedance.com>

arch/x86/mm/init_64.c

@@ -1622,6 +1622,11 @@ int __meminit vmemmap_populate(unsigned long start, unsigned long end, int node,
 }
 
 #ifdef CONFIG_HAVE_BOOTMEM_INFO_NODE
+void arch_free_vmemmap_page(struct page *page)
+{
+	free_bootmem_page(page);
+}
+
 void register_page_bootmem_memmap(unsigned long section_nr,
 				  struct page *start_page, unsigned long nr_pages)
 {

mm/sparse-vmemmap.c

@@ -156,16 +156,21 @@ static void vmemmap_remap_range(unsigned long start, unsigned long end,
 	flush_tlb_kernel_range(start + PAGE_SIZE, end);
 }
 
+void __weak arch_free_vmemmap_page(struct page *page)
+{
+	free_reserved_page(page);
+}
+
 /*
  * Free a vmemmap page. A vmemmap page can be allocated from the memblock
  * allocator or buddy allocator. If the PG_reserved flag is set, it means
  * that it allocated from the memblock allocator, just free it via the
- * free_bootmem_page(). Otherwise, use __free_page().
+ * arch_free_vmemmap_page(). Otherwise, use __free_page().
  */
 static inline void free_vmemmap_page(struct page *page)
 {
 	if (PageReserved(page))
-		free_bootmem_page(page);
+		arch_free_vmemmap_page(page);
 	else
 		__free_page(page);
 }