Skip to contentSkip to conversation

mm: fix accounting of memmap pages

PATCH Applied Sumanth Korikkar 4 participants 5 messages Aug 7, 2025 – Aug 21, 2025

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 mm: fix accounting of memmap pages

For !CONFIG_SPARSEMEM_VMEMMAP, memmap page accounting is currently done upfront in sparse_buffer_init(). However, sparse_buffer_alloc() may return NULL in failure scenario.

Also, memmap pages may be allocated either from the memblock allocator during early boot or from the buddy allocator. When removed via arch_remove_memory(), accounting of memmap pages must reflect the original allocation source.

To ensure correctness:
* Account memmap pages after successful allocation in sparse_init_nid() and section_activate().
* Account memmap pages in section_deactivate() based on allocation source.

Fixes 15995a352474 ("mm: report per-page metadata information")

---
v4:
* Add fixes and suggested-by.

v3:
* Account memmap pages for !CONFIG_SPARSEMEM_VMEMMAP and only when memmap allocation succeeds. Thanks Wei Yang.

v2:
* Account memmap pages for !CONFIG_SPARSEMEM_VMEMMAP in section_deactivate(). Thanks David.
* https://lore.kernel.org/all/[email protected]/

v1:
* Account memmap pages for early sections.
* https://lore.kernel.org/all/[email protected]/

 mm/sparse-vmemmap.c |  5 -----
 mm/sparse.c         | 15 +++++++++------
 2 files changed, 9 insertions(+), 11 deletions(-)
mm/sparse-vmemmap.c +0 -5
index fd2ab5118e13..41aa0493eb03 100644
@@ -578,11 +578,6 @@ struct page * __meminit __populate_section_memmap(unsigned long pfn,
 	if (r < 0)
 		return NULL;
 
-	if (system_state == SYSTEM_BOOTING)
-		memmap_boot_pages_add(DIV_ROUND_UP(end - start, PAGE_SIZE));
-	else
-		memmap_pages_add(DIV_ROUND_UP(end - start, PAGE_SIZE));
-
 	return pfn_to_page(pfn);
 }
 
mm/sparse.c +9 -6
index 066cbf82acb8..24323122f6cb 100644
@@ -454,9 +454,6 @@ static void __init sparse_buffer_init(unsigned long size, int nid)
 	 */
 	sparsemap_buf = memmap_alloc(size, section_map_size(), addr, nid, true);
 	sparsemap_buf_end = sparsemap_buf + size;
-#ifndef CONFIG_SPARSEMEM_VMEMMAP
-	memmap_boot_pages_add(DIV_ROUND_UP(size, PAGE_SIZE));
-#endif
 }
 
 static void __init sparse_buffer_fini(void)
@@ -567,6 +564,8 @@ static void __init sparse_init_nid(int nid, unsigned long pnum_begin,
 				sparse_buffer_fini();
 				goto failed;
 			}
+			memmap_boot_pages_add(DIV_ROUND_UP(PAGES_PER_SECTION * sizeof(struct page),
+							   PAGE_SIZE));
 			sparse_init_early_section(nid, map, pnum, 0);
 		}
 	}
@@ -680,7 +679,6 @@ static void depopulate_section_memmap(unsigned long pfn, unsigned long nr_pages,
 	unsigned long start = (unsigned long) pfn_to_page(pfn);
 	unsigned long end = start + nr_pages * sizeof(struct page);
 
-	memmap_pages_add(-1L * (DIV_ROUND_UP(end - start, PAGE_SIZE)));
 	vmemmap_free(start, end, altmap);
 }
 static void free_map_bootmem(struct page *memmap)
@@ -856,10 +854,14 @@ static void section_deactivate(unsigned long pfn, unsigned long nr_pages,
 	 * The memmap of early sections is always fully populated. See
 	 * section_activate() and pfn_valid() .
 	 */
-	if (!section_is_early)
+	if (!section_is_early) {
+		memmap_pages_add(-1L * (DIV_ROUND_UP(nr_pages * sizeof(struct page), PAGE_SIZE)));
 		depopulate_section_memmap(pfn, nr_pages, altmap);
-	else if (memmap)
+	} else if (memmap) {
+		memmap_boot_pages_add(-1L * (DIV_ROUND_UP(nr_pages * sizeof(struct page),
+							  PAGE_SIZE)));
 		free_map_bootmem(memmap);
+	}
 
 	if (empty)
 		ms->section_mem_map = (unsigned long)NULL;
@@ -904,6 +906,7 @@ static struct page * __meminit section_activate(int nid, unsigned long pfn,
 		section_deactivate(pfn, nr_pages, altmap);
 		return ERR_PTR(-ENOMEM);
 	}
+	memmap_pages_add(DIV_ROUND_UP(nr_pages * sizeof(struct page), PAGE_SIZE));
 
 	return memmap;
 }
Signature · 1 line
2.48.1
Reviewed-by on v4 1/1
#1 lore raw
Quoted 15 lines · Sumanth Korikkar

For !CONFIG_SPARSEMEM_VMEMMAP, memmap page accounting is currently done upfront in sparse_buffer_init(). However, sparse_buffer_alloc() may return NULL in failure scenario.

Also, memmap pages may be allocated either from the memblock allocator during early boot or from the buddy allocator. When removed via arch_remove_memory(), accounting of memmap pages must reflect the original allocation source.

To ensure correctness:
* Account memmap pages after successful allocation in sparse_init_nid() and section_activate().
* Account memmap pages in section_deactivate() based on allocation source.

Quoted 4 lines
Fixes 15995a352474 ("mm: report per-page metadata information")

LGTM

Signature · 2 lines
Wei Yang
Help you, Help me
#2 lore raw
In reply to Sumanth Korikkar
Quoted 71 lines · Sumanth Korikkar

For !CONFIG_SPARSEMEM_VMEMMAP, memmap page accounting is currently done upfront in sparse_buffer_init(). However, sparse_buffer_alloc() may return NULL in failure scenario.

Also, memmap pages may be allocated either from the memblock allocator during early boot or from the buddy allocator. When removed via arch_remove_memory(), accounting of memmap pages must reflect the original allocation source.

To ensure correctness:
* Account memmap pages after successful allocation in sparse_init_nid() and section_activate().
* Account memmap pages in section_deactivate() based on allocation source.

Fixes 15995a352474 ("mm: report per-page metadata information")

---
v4:
* Add fixes and suggested-by.

v3:
* Account memmap pages for !CONFIG_SPARSEMEM_VMEMMAP and only when memmap allocation succeeds. Thanks Wei Yang.

v2:
* Account memmap pages for !CONFIG_SPARSEMEM_VMEMMAP in section_deactivate(). Thanks David.
* https://lore.kernel.org/all/[email protected]/

v1:
* Account memmap pages for early sections.
* https://lore.kernel.org/all/[email protected]/

  mm/sparse-vmemmap.c |  5 -----
  mm/sparse.c         | 15 +++++++++------
  2 files changed, 9 insertions(+), 11 deletions(-)
mm/sparse-vmemmap.c +0 -5
index fd2ab5118e13..41aa0493eb03 100644
@@ -578,11 +578,6 @@ struct page * __meminit __populate_section_memmap(unsigned long pfn,
  	if (r < 0)
  		return NULL;
 
-	if (system_state == SYSTEM_BOOTING)
-		memmap_boot_pages_add(DIV_ROUND_UP(end - start, PAGE_SIZE));
-	else
-		memmap_pages_add(DIV_ROUND_UP(end - start, PAGE_SIZE));
-
  	return pfn_to_page(pfn);
  }
 
mm/sparse.c +0 -3
index 066cbf82acb8..24323122f6cb 100644
@@ -454,9 +454,6 @@ static void __init sparse_buffer_init(unsigned long size, int nid)
  	 */
  	sparsemap_buf = memmap_alloc(size, section_map_size(), addr, nid, true);
  	sparsemap_buf_end = sparsemap_buf + size;
-#ifndef CONFIG_SPARSEMEM_VMEMMAP
-	memmap_boot_pages_add(DIV_ROUND_UP(size, PAGE_SIZE));
-#endif
  }
 
  static void __init sparse_buffer_fini(void)

mm/sparse.c static void __init sparse_init_nid(int nid, unsigned long pnum_begin, · line 568

Quoted 5 lines
mm/sparse.c +2 -0
@@ -567,6 +564,8 @@ static void __init sparse_init_nid(int nid, unsigned long pnum_begin,
  				sparse_buffer_fini();
  				goto failed;
  			}
+			memmap_boot_pages_add(DIV_ROUND_UP(PAGES_PER_SECTION * sizeof(struct page),
+							   PAGE_SIZE));

IIRC, we can have partially populated boot sections, where only some subsections actually have a memmap ... so this calculation is possibly wrong in some cases.

Signature · 3 lines
Cheers,

David / dhildenb
maintainer
#3 lore raw
Quoted 10 lines · On Fri, 8 Aug 2025 16:23:52 +0200 David Hildenbrand <[email protected]> wrote:

mm/sparse.c static void __init sparse_init_nid(int nid, unsigned long pnum_begin, · line 568

Quoted 6 lines
mm/sparse.c +2 -0
@@ -567,6 +564,8 @@ static void __init sparse_init_nid(int nid, unsigned long pnum_begin,
  				sparse_buffer_fini();
  				goto failed;
  			}
+			memmap_boot_pages_add(DIV_ROUND_UP(PAGES_PER_SECTION * sizeof(struct page),
+							   PAGE_SIZE));

IIRC, we can have partially populated boot sections, where only some subsections actually have a memmap ... so this calculation is possibly wrong in some cases.

The patch (which has cc:stable!) has been in state "stuck" for 12 days due to "possibly wrong". Could someone please find a way to get it unstuck?

#4 lore raw
In reply to David Hildenbrand
Quoted 15 lines
Quoted 15 lines
mm/sparse.c +0 -3
index 066cbf82acb8..24323122f6cb 100644
@@ -454,9 +454,6 @@ static void __init sparse_buffer_init(unsigned long size, int nid)
  	 */
  	sparsemap_buf = memmap_alloc(size, section_map_size(), addr, nid, true);
  	sparsemap_buf_end = sparsemap_buf + size;
-#ifndef CONFIG_SPARSEMEM_VMEMMAP
-	memmap_boot_pages_add(DIV_ROUND_UP(size, PAGE_SIZE));
-#endif
  }
  static void __init sparse_buffer_fini(void)
@@ -567,6 +564,8 @@ static void __init sparse_init_nid(int nid, unsigned long pnum_begin,
  				sparse_buffer_fini();
Quoted 8 lines

mm/sparse.c static void __init sparse_init_nid(int nid, unsigned long pnum_begin, · line 568

Quoted 4 lines
mm/sparse.c +2 -0
  				goto failed;
  			}
+			memmap_boot_pages_add(DIV_ROUND_UP(PAGES_PER_SECTION * sizeof(struct page),
+							   PAGE_SIZE));

IIRC, we can have partially populated boot sections, where only some subsections actually have a memmap ... so this calculation is possibly wrong in some cases.

In section_activate():

/*
* The early init code does not consider partially populated initial
* sections, it simply assumes that memory will never be referenced. If
* we hot-add memory into such a section then we do not need to populate
* the memmap and can simply reuse what is already there.
*/
if (nr_pages < PAGES_PER_SECTION && early_section(ms))

        return pfn_to_page(pfn);

The patch ignores the accounting here, based on the comments described above for partially populated initial sections.

memmap = populate_section_memmap(pfn, nr_pages, nid, altmap, pgmap);

if (!memmap) {
        section_deactivate(pfn, nr_pages, altmap);
        return ERR_PTR(-ENOMEM);
}

memmap_pages_add(DIV_ROUND_UP(nr_pages * sizeof(struct page), PAGE_SIZE));

only bookkeeping for newly allocated memmap is performed.

Also before this patch, __populate_section_memmap() did memmap accounting for !NULL usecases. This patch also does similar change, but covers memmap accounting for both CONFIG_SPARSEMEM_VMEMMAP + !CONFIG_SPARSEMEM_VMEMMAP usecases and memmap accounting based on allocation resource.

Let me know, if this sounds right.

Thank you.