Skip to contentSkip to conversation

Fix corruption of memmap on IA64 SPARSEMEM when mem_section is not a power of 2

PATCH In review Mel Gorman 3 participants 7 messages Mar 13, 2007 – Jul 25, 2007

Subsystems (3)
lore
Share & tools
lore

Mel Gorman

author
#0 lore raw

PATCH 1/1 Fix corruption of memmap on IA64 SPARSEMEM when mem_section is not a power of 2

There are problems in the use of SPARSEMEM and pageblock flags that causes problems on ia64.

The first part of the problem is that units are incorrect in SECTION_BLOCKFLAGS_BITS computation. This results in a map_section's section_mem_map being treated as part of a bitmap which isn't good. This was evident with an invalid virtual address when mem_init attempted to free bootmem pages while relinquishing control from the bootmem allocator.

The second part of the problem occurs because the pageblock flags bitmap is be located with the mem_section. The SECTIONS_PER_ROOT computation using sizeof (mem_section) may not be a power of 2 depending on the size of the bitmap. This renders masks and other such things not power of 2 base. This issue was seen with SPARSEMEM_EXTREME on ia64. This patch moves the bitmap outside of mem_section and uses a pointer instead in the mem_section. The bitmaps are allocated when the section is being initialised.

Note that sparse_early_usemap_alloc() does not use alloc_remap() like sparse_early_mem_map_alloc(). The allocation required for the bitmap on x86, the only architecture that uses alloc_remap is typically smaller than a cache line. alloc_remap() pads out allocations to the cache size which would be a needless waste.

Credit to Bob Picco for identifying the original problem and effecting a fix for the SECTION_BLOCKFLAGS_BITS calculation. Credit to Andy Whitcroft for devising the best way of allocating the bitmaps only when required for the section.

 include/linux/mmzone.h |    6 ++++--
 mm/sparse.c            |   49 ++++++++++++++++++++++++++++++++++++++++++++++---
 2 files changed, 50 insertions(+), 5 deletions(-)
include/linux/mmzone.h +4 -2
diff -rup -X /usr/src/patchset-0.6/bin//dontdiff linux-2.6.21-rc2-mm2-clean/include/linux/mmzone.h linux-2.6.21-rc2-mm2-sparsemem_rootsize_fix/include/linux/mmzone.h
@@ -705,7 +705,7 @@ extern struct zone *next_zone(struct zon
 #define PAGE_SECTION_MASK	(~(PAGES_PER_SECTION-1))
 
 #define SECTION_BLOCKFLAGS_BITS \
-		((SECTION_SIZE_BITS - (MAX_ORDER-1)) * NR_PAGEBLOCK_BITS)
+		((1 << (PFN_SECTION_SHIFT - (MAX_ORDER-1))) * NR_PAGEBLOCK_BITS)
 
 #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
 #error Allocator MAX_ORDER exceeds SECTION_SIZE
@@ -726,7 +726,9 @@ struct mem_section {
 	 * before using it wrong.
 	 */
 	unsigned long section_mem_map;
-	DECLARE_BITMAP(pageblock_flags, SECTION_BLOCKFLAGS_BITS);
+
+	/* See declaration of similar field in struct zone */
+	unsigned long *pageblock_flags;
 };
 
 #ifdef CONFIG_SPARSEMEM_EXTREME
mm/sparse.c +46 -3

99 lines not on this page · raw message

maintainer
#1 lore raw
Quoted 23 lines · On Tue, 13 Mar 2007 10:42:02 +0000 [email protected] (Mel Gorman) wrote:

There are problems in the use of SPARSEMEM and pageblock flags that causes problems on ia64.

The first part of the problem is that units are incorrect in SECTION_BLOCKFLAGS_BITS computation. This results in a map_section's section_mem_map being treated as part of a bitmap which isn't good. This was evident with an invalid virtual address when mem_init attempted to free bootmem pages while relinquishing control from the bootmem allocator.

The second part of the problem occurs because the pageblock flags bitmap is be located with the mem_section. The SECTIONS_PER_ROOT computation using sizeof (mem_section) may not be a power of 2 depending on the size of the bitmap. This renders masks and other such things not power of 2 base. This issue was seen with SPARSEMEM_EXTREME on ia64. This patch moves the bitmap outside of mem_section and uses a pointer instead in the mem_section. The bitmaps are allocated when the section is being initialised.

Note that sparse_early_usemap_alloc() does not use alloc_remap() like sparse_early_mem_map_alloc(). The allocation required for the bitmap on x86, the only architecture that uses alloc_remap is typically smaller than a cache line. alloc_remap() pads out allocations to the cache size which would be a needless waste.

Quoted 4 lines

Credit to Bob Picco for identifying the original problem and effecting a fix for the SECTION_BLOCKFLAGS_BITS calculation. Credit to Andy Whitcroft for devising the best way of allocating the bitmaps only when required for the section.

SuperH allmodconfig blew up:

mm/sparse.c: In function `sparse_init':
mm/sparse.c:482: error: implicit declaration of function sparse_early_usemap_alloc' mm/sparse.c:482: warning: assignment makes pointer from integer without a cast mm/sparse.c: In function sparse_add_one_section':
mm/sparse.c:553: error: implicit declaration of function `__kmalloc_section_usemap' mm/sparse.c:553: warning: assignment makes pointer from integer without a cast

author
#2 lore raw

On (24/07/07 22:01), Andrew Morton didst pronounce:

Quoted 30 lines · On Tue, 13 Mar 2007 10:42:02 +0000 [email protected] (Mel Gorman) wrote:
Quoted 24 lines

There are problems in the use of SPARSEMEM and pageblock flags that causes problems on ia64.

The first part of the problem is that units are incorrect in SECTION_BLOCKFLAGS_BITS computation. This results in a map_section's section_mem_map being treated as part of a bitmap which isn't good. This was evident with an invalid virtual address when mem_init attempted to free bootmem pages while relinquishing control from the bootmem allocator.

The second part of the problem occurs because the pageblock flags bitmap is be located with the mem_section. The SECTIONS_PER_ROOT computation using sizeof (mem_section) may not be a power of 2 depending on the size of the bitmap. This renders masks and other such things not power of 2 base. This issue was seen with SPARSEMEM_EXTREME on ia64. This patch moves the bitmap outside of mem_section and uses a pointer instead in the mem_section. The bitmaps are allocated when the section is being initialised.

Note that sparse_early_usemap_alloc() does not use alloc_remap() like sparse_early_mem_map_alloc(). The allocation required for the bitmap on x86, the only architecture that uses alloc_remap is typically smaller than a cache line. alloc_remap() pads out allocations to the cache size which would be a needless waste.

Quoted 4 lines

Credit to Bob Picco for identifying the original problem and effecting a fix for the SECTION_BLOCKFLAGS_BITS calculation. Credit to Andy Whitcroft for devising the best way of allocating the bitmaps only when required for the section.

Quoted 8 lines

SuperH allmodconfig blew up:

mm/sparse.c: In function `sparse_init':
mm/sparse.c:482: error: implicit declaration of function sparse_early_usemap_alloc' mm/sparse.c:482: warning: assignment makes pointer from integer without a cast mm/sparse.c: In function sparse_add_one_section':
mm/sparse.c:553: error: implicit declaration of function `__kmalloc_section_usemap' mm/sparse.c:553: warning: assignment makes pointer from integer without a cast

I'm looking at this now and getting a superh cross-compiler built to build-test any fix. My first impression is that sparse_early_usemap_alloc() needs to be defined whether CONFIG_SPARSEMEM_VMEMMAP is set or not. Right now, sparse_early_usemap_alloc() is only defined when it is set and it's not clear why although "by accident" is the most likely explanation.

Signature · 3 lines
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab
maintainer
#3 lore raw
Quoted 7 lines · On Wed, 25 Jul 2007 13:55:02 +0100 [email protected] (Mel Gorman) wrote:
Quoted 6 lines

mm/sparse.c: In function `sparse_init':
mm/sparse.c:482: error: implicit declaration of function sparse_early_usemap_alloc' mm/sparse.c:482: warning: assignment makes pointer from integer without a cast mm/sparse.c: In function sparse_add_one_section':
mm/sparse.c:553: error: implicit declaration of function `__kmalloc_section_usemap' mm/sparse.c:553: warning: assignment makes pointer from integer without a cast

Quoted 7 lines

I'm looking at this now and getting a superh cross-compiler built to build-test any fix. My first impression is that sparse_early_usemap_alloc() needs to be defined whether CONFIG_SPARSEMEM_VMEMMAP is set or not. Right now, sparse_early_usemap_alloc() is only defined when it is set and it's not clear why although "by accident" is the most likely explanation.

Thanks.

There's an sh cross-compiler in http://userweb.kernel.org/~akpm/cross-compilers/,
btw.

author
#4 lore raw
In reply to Andrew Morton

On (24/07/07 22:01), Andrew Morton didst pronounce:

Quoted 7 lines · On Tue, 13 Mar 2007 10:42:02 +0000 [email protected] (Mel Gorman) wrote:
Quoted 5 lines

There are problems in the use of SPARSEMEM and pageblock flags that causes problems on ia64.

<SNIP>

Quoted 8 lines

SuperH allmodconfig blew up:

mm/sparse.c: In function `sparse_init':
mm/sparse.c:482: error: implicit declaration of function sparse_early_usemap_alloc' mm/sparse.c:482: warning: assignment makes pointer from integer without a cast mm/sparse.c: In function sparse_add_one_section':
mm/sparse.c:553: error: implicit declaration of function `__kmalloc_section_usemap' mm/sparse.c:553: warning: assignment makes pointer from integer without a cast

This error is due to a bad interaction between generic-virtual-memmap-support-for-sparsemem.patch and fix-corruption-of-memmap-on-ia64-sparsemem-when-mem_section-is-not-a-power-of-2.patch . Functions that are needed whether CONFIG_SPARSEMEM_VMEMMAP is set are not are depending on CONFIG_SPARSEMEM_VMEMMAP . This breaks on arch/sh for example where SPARSEMEM may be set but not NUMA.

---
 sparse.c |   58 +++++++++++++++++++++++++++++-----------------------------
 1 file changed, 29 insertions(+), 29 deletions(-)
mm/sparse.c +29 -29
diff -rup -X /usr/src/patchset-0.6/bin//dontdiff linux-2.6.23-rc1-mm1-clean/mm/sparse.c linux-2.6.23-rc1-mm1-005-sh-sparsemem-fix/mm/sparse.c
@@ -220,6 +220,35 @@ void *alloc_bootmem_high_node(pg_data_t 
 	return NULL;
 }
 
+static unsigned long usemap_size(void)
+{
+	unsigned long size_bytes;
+	size_bytes = roundup(SECTION_BLOCKFLAGS_BITS, 8) / 8;
+	size_bytes = roundup(size_bytes, sizeof(unsigned long));
+	return size_bytes;
+}
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+static unsigned long *__kmalloc_section_usemap(void)
+{
+	return kmalloc(usemap_size(), GFP_KERNEL);
+}
+#endif /* CONFIG_MEMORY_HOTPLUG */
+
+static unsigned long *sparse_early_usemap_alloc(unsigned long pnum)
+{
+	unsigned long *usemap;
+	struct mem_section *ms = __nr_to_section(pnum);
+	int nid = sparse_early_nid(ms);
+
+	usemap = alloc_bootmem_node(NODE_DATA(nid), usemap_size());
+	if (usemap)
+		return usemap;
+
+	printk(KERN_WARNING "%s: allocation failed\n", __FUNCTION__);
+	return NULL;
+}
+
 #ifdef CONFIG_SPARSEMEM_VMEMMAP
 /*
  * Virtual Memory Map support
@@ -254,35 +283,6 @@ void *alloc_bootmem_high_node(pg_data_t 
  * 	require one PTE/TLB per PAGE_SIZE chunk of the virtual memory map.
  */
 
-static unsigned long usemap_size(void)
-{
-	unsigned long size_bytes;
- 	size_bytes = roundup(SECTION_BLOCKFLAGS_BITS, 8) / 8;
-	size_bytes = roundup(size_bytes, sizeof(unsigned long));
-	return size_bytes;
-}
-
-#ifdef CONFIG_MEMORY_HOTPLUG
-static unsigned long *__kmalloc_section_usemap(void)
-{
-	return kmalloc(usemap_size(), GFP_KERNEL);
-}
-#endif /* CONFIG_MEMORY_HOTPLUG */
-
-static unsigned long *sparse_early_usemap_alloc(unsigned long pnum)
-{
-	unsigned long *usemap;
-	struct mem_section *ms = __nr_to_section(pnum);
-	int nid = sparse_early_nid(ms);
-
-	usemap = alloc_bootmem_node(NODE_DATA(nid), usemap_size());
-	if (usemap)
-		return usemap;
-
-	printk(KERN_WARNING "%s: allocation failed\n", __FUNCTION__);
-	return NULL;
-}
-
 /*
  * Allocate a block of memory to be used to back the virtual memory map
  * or to back the page tables that are used to create the mapping.
Signature · 3 lines
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab
author
#5 lore raw

On (25/07/07 15:49), Mel Gorman didst pronounce:

Quoted 9 lines
Quoted 8 lines

SuperH allmodconfig blew up:

mm/sparse.c: In function `sparse_init':
mm/sparse.c:482: error: implicit declaration of function sparse_early_usemap_alloc' mm/sparse.c:482: warning: assignment makes pointer from integer without a cast mm/sparse.c: In function sparse_add_one_section':
mm/sparse.c:553: error: implicit declaration of function `__kmalloc_section_usemap' mm/sparse.c:553: warning: assignment makes pointer from integer without a cast

Quoted 7 lines

This error is due to a bad interaction between generic-virtual-memmap-support-for-sparsemem.patch and fix-corruption-of-memmap-on-ia64-sparsemem-when-mem_section-is-not-a-power-of-2.patch . Functions that are needed whether CONFIG_SPARSEMEM_VMEMMAP is set are not are depending on CONFIG_SPARSEMEM_VMEMMAP . This breaks on arch/sh for example where SPARSEMEM may be set but not NUMA.

This patch only moves code and does not alter it. However, as SPARSEMEM is set without NUMA, the code move generates a warning on SuperH because NODE_DATA(nid) is a no-op. The following patch supresses the warning

---
 sparse.c |    3 +++
 1 file changed, 3 insertions(+)
mm/sparse.c +3 -0
diff -rup -X /usr/src/patchset-0.6/bin//dontdiff linux-2.6.23-rc1-mm1-005-sh-sparsemem-fix/mm/sparse.c linux-2.6.23-rc1-mm1-010-sh-sparsemem-fix-warning/mm/sparse.c
@@ -245,6 +245,9 @@ static unsigned long *sparse_early_usema
 	if (usemap)
 		return usemap;
 
+	/* Stupid: suppress gcc warning for SPARSEMEM && !NUMA */
+	nid = 0;
+
 	printk(KERN_WARNING "%s: allocation failed\n", __FUNCTION__);
 	return NULL;
 }
#6 lore raw
In reply to Mel Gorman
Quoted 95 lines · Mel Gorman

On (24/07/07 22:01), Andrew Morton didst pronounce:

Quoted 6 lines · On Tue, 13 Mar 2007 10:42:02 +0000 [email protected] (Mel Gorman) wrote:
Quoted 5 lines

There are problems in the use of SPARSEMEM and pageblock flags that causes problems on ia64.

<SNIP>

Quoted 8 lines

SuperH allmodconfig blew up:

mm/sparse.c: In function `sparse_init':
mm/sparse.c:482: error: implicit declaration of function sparse_early_usemap_alloc' mm/sparse.c:482: warning: assignment makes pointer from integer without a cast mm/sparse.c: In function sparse_add_one_section':
mm/sparse.c:553: error: implicit declaration of function `__kmalloc_section_usemap' mm/sparse.c:553: warning: assignment makes pointer from integer without a cast

This error is due to a bad interaction between generic-virtual-memmap-support-for-sparsemem.patch and fix-corruption-of-memmap-on-ia64-sparsemem-when-mem_section-is-not-a-power-of-2.patch . Functions that are needed whether CONFIG_SPARSEMEM_VMEMMAP is set are not are depending on CONFIG_SPARSEMEM_VMEMMAP . This breaks on arch/sh for example where SPARSEMEM may be set but not NUMA.

---
 sparse.c |   58 +++++++++++++++++++++++++++++-----------------------------
 1 file changed, 29 insertions(+), 29 deletions(-)

diff -rup -X /usr/src/patchset-0.6/bin//dontdiff linux-2.6.23-rc1-mm1-clean/mm/sparse.c linux-2.6.23-rc1-mm1-005-sh-sparsemem-fix/mm/sparse.c

mm/sparse.c +29 -22
@@ -220,6 +220,35 @@ void *alloc_bootmem_high_node(pg_data_t 
 	return NULL;
 }
 
+static unsigned long usemap_size(void)
+{
+	unsigned long size_bytes;
+	size_bytes = roundup(SECTION_BLOCKFLAGS_BITS, 8) / 8;
+	size_bytes = roundup(size_bytes, sizeof(unsigned long));
+	return size_bytes;
+}
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+static unsigned long *__kmalloc_section_usemap(void)
+{
+	return kmalloc(usemap_size(), GFP_KERNEL);
+}
+#endif /* CONFIG_MEMORY_HOTPLUG */
+
+static unsigned long *sparse_early_usemap_alloc(unsigned long pnum)
+{
+	unsigned long *usemap;
+	struct mem_section *ms = __nr_to_section(pnum);
+	int nid = sparse_early_nid(ms);
+
+	usemap = alloc_bootmem_node(NODE_DATA(nid), usemap_size());
+	if (usemap)
+		return usemap;
+
+	printk(KERN_WARNING "%s: allocation failed\n", __FUNCTION__);
+	return NULL;
+}
+
 #ifdef CONFIG_SPARSEMEM_VMEMMAP
 /*
  * Virtual Memory Map support
@@ -254,35 +283,6 @@ void *alloc_bootmem_high_node(pg_data_t 
  * 	require one PTE/TLB per PAGE_SIZE chunk of the virtual memory map.
  */
 
-static unsigned long usemap_size(void)
-{
-	unsigned long size_bytes;
- 	size_bytes = roundup(SECTION_BLOCKFLAGS_BITS, 8) / 8;
-	size_bytes = roundup(size_bytes, sizeof(unsigned long));
-	return size_bytes;
-}
-
-#ifdef CONFIG_MEMORY_HOTPLUG
-static unsigned long *__kmalloc_section_usemap(void)
-{
-	return kmalloc(usemap_size(), GFP_KERNEL);
-}
-#endif /* CONFIG_MEMORY_HOTPLUG */
-
-static unsigned long *sparse_early_usemap_alloc(unsigned long pnum)
-{
-	unsigned long *usemap;
-	struct mem_section *ms = __nr_to_section(pnum);
-	int nid = sparse_early_nid(ms);
-
-	usemap = alloc_bootmem_node(NODE_DATA(nid), usemap_size());

mm/sparse.c void *alloc_bootmem_high_node(pg_data_t · line 288

Quoted 10 lines
mm/sparse.c +0 -7
-	if (usemap)
-		return usemap;
-
-	printk(KERN_WARNING "%s: allocation failed\n", __FUNCTION__);
-	return NULL;
-}
-
 /*
  * Allocate a block of memory to be used to back the virtual memory map
  * or to back the page tables that are used to create the mapping.

This also affects NUMA-Q in some SPARSEMEM configurations. Applying this fixes it up.

Looks very much like diff has dropped the stuff within my #ifdef 'incorrectly' and as such is a merge error. This should probabally be considered a fix to this patch:

fix-corruption-of-memmap-on-ia64-sparsemem-when-mem_section-is-not-a-power-of-2.patch

-apw