Ensure pfn_to_nid() is always defined for i386

PATCH In review Matthew Dobson 3 participants 8 messages Dec 22, 2003 – Mar 29, 2004

Subsystems (2)

MM core MGLRU

lore

Share & tools

lore

Matthew Dobson

author

2003-12-22 19:48 UTC #0 lore raw

PATCH 1/1 Ensure pfn_to_nid() is always defined for i386

pfn_to_nid() is defined for *almost* all configurations for i386. There is a small bug in that for CONFIG_X86_PC it is not. This just sets up a generic definition so that this function is always defined in a reasonable manner.

Cheers!

-Matt

include/asm-i386/mmzone.h +1 -0

diff -Nurp --exclude-from=/home/mcd/.dontdiff linux-2.6.0-vanilla/include/asm-i386/mmzone.h linux-2.6.0-patched/include/asm-i386/mmzone.h
@@ -123,6 +123,7 @@ static inline struct pglist_data *pfn_to
 #include <asm/srat.h>
 #elif CONFIG_X86_PC
 #define get_zholes_size(n) (0)
+#define pfn_to_nid(pfn)		(0)
 #else
 #define pfn_to_nid(pfn)		(0)
 #endif /* CONFIG_X86_NUMAQ */

Martin J. Bligh

2003-12-22 23:55 UTC #1 lore raw

Quoted 4 lines

pfn_to_nid() is defined for *almost* all configurations for i386. There is a small bug in that for CONFIG_X86_PC it is not. This just sets up a generic definition so that this function is always defined in a reasonable manner.

I'd prefer to fix the current ungodly mess in the current fashion ... I know it's longer, but I'd prefer to clean it up more if we're going to touch it. This has been sitting in my tree for a while, waiting for the code freeze to slush out a bit.

include/asm-i386/mmzone.h +42 ~~-39~~

diff -purN -X /home/mbligh/.diff.exclude 276-per_node_rss/include/asm-i386/mmzone.h 278-pfn_to_nid/include/asm-i386/mmzone.h
@@ -10,7 +10,49 @@
 
 #ifdef CONFIG_DISCONTIGMEM
 
+#ifdef CONFIG_NUMA
+	#ifdef CONFIG_X86_NUMAQ
+		#include <asm/numaq.h>
+	#else	/* summit or generic arch */
+		#include <asm/srat.h>
+	#endif
+#else /* !CONFIG_NUMA */
+	#define get_memcfg_numa get_memcfg_numa_flat
+	#define get_zholes_size(n) (0)
+#endif /* CONFIG_NUMA */
+
 extern struct pglist_data *node_data[];
+#define NODE_DATA(nid)		(node_data[nid])
+
+/*
+ * generic node memory support, the following assumptions apply:
+ *
+ * 1) memory comes in 256Mb contigious chunks which are either present or not
+ * 2) we will not have more than 64Gb in total
+ *
+ * for now assume that 64Gb is max amount of RAM for whole system
+ *    64Gb / 4096bytes/page = 16777216 pages
+ */
+#define MAX_NR_PAGES 16777216
+#define MAX_ELEMENTS 256
+#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)
+
+extern u8 physnode_map[];
+
+static inline int pfn_to_nid(unsigned long pfn)
+{
+#ifdef CONFIG_NUMA
+	return(physnode_map[(pfn) / PAGES_PER_ELEMENT]);
+#else
+	return 0;
+#endif
+}
+
+static inline struct pglist_data *pfn_to_pgdat(unsigned long pfn)
+{
+	return(NODE_DATA(pfn_to_nid(pfn)));
+}
+
 
 /*
  * Following are macros that are specific to this numa platform.

@@ -43,11 +85,6 @@ extern struct pglist_data *node_data[];
  */
 #define kvaddr_to_nid(kaddr)	pfn_to_nid(__pa(kaddr) >> PAGE_SHIFT)
 
-/*
- * Return a pointer to the node data for node n.
- */
-#define NODE_DATA(nid)		(node_data[nid])
-
 #define node_mem_map(nid)	(NODE_DATA(nid)->node_mem_map)
 #define node_start_pfn(nid)	(NODE_DATA(nid)->node_start_pfn)
 #define node_end_pfn(nid)						\

@@ -93,40 +130,6 @@ extern struct pglist_data *node_data[];
  */ 
 #define pfn_valid(pfn)          ((pfn) < num_physpages)
 
-/*
- * generic node memory support, the following assumptions apply:
- *
- * 1) memory comes in 256Mb contigious chunks which are either present or not
- * 2) we will not have more than 64Gb in total
- *
- * for now assume that 64Gb is max amount of RAM for whole system
- *    64Gb / 4096bytes/page = 16777216 pages
- */
-#define MAX_NR_PAGES 16777216
-#define MAX_ELEMENTS 256
-#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)
-
-extern u8 physnode_map[];
-
-static inline int pfn_to_nid(unsigned long pfn)
-{
-	return(physnode_map[(pfn) / PAGES_PER_ELEMENT]);
-}
-static inline struct pglist_data *pfn_to_pgdat(unsigned long pfn)
-{
-	return(NODE_DATA(pfn_to_nid(pfn)));
-}
-
-#ifdef CONFIG_X86_NUMAQ
-#include <asm/numaq.h>
-#elif CONFIG_ACPI_SRAT
-#include <asm/srat.h>
-#elif CONFIG_X86_PC
-#define get_zholes_size(n) (0)
-#else
-#define pfn_to_nid(pfn)		(0)
-#endif /* CONFIG_X86_NUMAQ */
-
 extern int get_memcfg_numa_flat(void );
 /*
  * This allows any one NUMA architecture to be compiled

include/linux/mmzone.h +1 -1

diff -purN -X /home/mbligh/.diff.exclude 276-per_node_rss/include/linux/mmzone.h 278-pfn_to_nid/include/linux/mmzone.h
@@ -303,7 +303,7 @@ extern struct pglist_data contig_page_da
 #define NODE_DATA(nid)		(&contig_page_data)
 #define NODE_MEM_MAP(nid)	mem_map
 #define MAX_NODES_SHIFT		0
-
+#define pfn_to_nid(pfn)		(0)
 #else /* CONFIG_DISCONTIGMEM */
 
 #include <asm/mmzone.h>

Matthew Dobson

author

2004-01-05 21:40 UTC #2 lore raw

Ok... This looks good to me. It's definitely longer than my 1 line fix, but it's also more correct long-term solution.

Cheers!

-Matt

Quoted 129 lines · Martin J. Bligh

Quoted 1 line

pfn_to_nid() is defined for *almost* all configurations for i386. There is a small bug in that for CONFIG_X86_PC it is not. This just sets up a generic definition so that this function is always defined in a reasonable manner.

I'd prefer to fix the current ungodly mess in the current fashion ...
I know it's longer, but I'd prefer to clean it up more if we're going to touch it. This has been sitting in my tree for a while, waiting for the code freeze to slush out a bit.

diff -purN -X /home/mbligh/.diff.exclude 276-per_node_rss/include/asm-i386/mmzone.h 278-pfn_to_nid/include/asm-i386/mmzone.h

include/asm-i386/mmzone.h +42 ~~-39~~

@@ -10,7 +10,49 @@
 
 #ifdef CONFIG_DISCONTIGMEM
 
+#ifdef CONFIG_NUMA
+	#ifdef CONFIG_X86_NUMAQ
+		#include <asm/numaq.h>
+	#else	/* summit or generic arch */
+		#include <asm/srat.h>
+	#endif
+#else /* !CONFIG_NUMA */
+	#define get_memcfg_numa get_memcfg_numa_flat
+	#define get_zholes_size(n) (0)
+#endif /* CONFIG_NUMA */
+
 extern struct pglist_data *node_data[];
+#define NODE_DATA(nid)		(node_data[nid])
+
+/*
+ * generic node memory support, the following assumptions apply:
+ *
+ * 1) memory comes in 256Mb contigious chunks which are either present or not
+ * 2) we will not have more than 64Gb in total
+ *
+ * for now assume that 64Gb is max amount of RAM for whole system
+ *    64Gb / 4096bytes/page = 16777216 pages
+ */
+#define MAX_NR_PAGES 16777216
+#define MAX_ELEMENTS 256
+#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)
+
+extern u8 physnode_map[];
+
+static inline int pfn_to_nid(unsigned long pfn)
+{
+#ifdef CONFIG_NUMA
+	return(physnode_map[(pfn) / PAGES_PER_ELEMENT]);
+#else
+	return 0;
+#endif
+}
+
+static inline struct pglist_data *pfn_to_pgdat(unsigned long pfn)
+{
+	return(NODE_DATA(pfn_to_nid(pfn)));
+}
+
 
 /*
  * Following are macros that are specific to this numa platform.

@@ -43,11 +85,6 @@ extern struct pglist_data *node_data[];
  */
 #define kvaddr_to_nid(kaddr)	pfn_to_nid(__pa(kaddr) >> PAGE_SHIFT)
 
-/*
- * Return a pointer to the node data for node n.
- */
-#define NODE_DATA(nid)		(node_data[nid])
-
 #define node_mem_map(nid)	(NODE_DATA(nid)->node_mem_map)
 #define node_start_pfn(nid)	(NODE_DATA(nid)->node_start_pfn)
 #define node_end_pfn(nid)						\

@@ -93,40 +130,6 @@ extern struct pglist_data *node_data[];
  */ 
 #define pfn_valid(pfn)          ((pfn) < num_physpages)
 
-/*
- * generic node memory support, the following assumptions apply:
- *
- * 1) memory comes in 256Mb contigious chunks which are either present or not
- * 2) we will not have more than 64Gb in total
- *
- * for now assume that 64Gb is max amount of RAM for whole system
- *    64Gb / 4096bytes/page = 16777216 pages
- */
-#define MAX_NR_PAGES 16777216
-#define MAX_ELEMENTS 256
-#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)
-
-extern u8 physnode_map[];
-
-static inline int pfn_to_nid(unsigned long pfn)
-{
-	return(physnode_map[(pfn) / PAGES_PER_ELEMENT]);
-}
-static inline struct pglist_data *pfn_to_pgdat(unsigned long pfn)
-{
-	return(NODE_DATA(pfn_to_nid(pfn)));
-}
-
-#ifdef CONFIG_X86_NUMAQ
-#include <asm/numaq.h>
-#elif CONFIG_ACPI_SRAT
-#include <asm/srat.h>
-#elif CONFIG_X86_PC
-#define get_zholes_size(n) (0)
-#else
-#define pfn_to_nid(pfn)		(0)
-#endif /* CONFIG_X86_NUMAQ */
-
 extern int get_memcfg_numa_flat(void );
 /*
  * This allows any one NUMA architecture to be compiled

diff -purN -X /home/mbligh/.diff.exclude 276-per_node_rss/include/linux/mmzone.h 278-pfn_to_nid/include/linux/mmzone.h

include/linux/mmzone.h +1 -1

@@ -303,7 +303,7 @@ extern struct pglist_data contig_page_da
 #define NODE_DATA(nid)		(&contig_page_data)
 #define NODE_MEM_MAP(nid)	mem_map
 #define MAX_NODES_SHIFT		0
-
+#define pfn_to_nid(pfn)		(0)
 #else /* CONFIG_DISCONTIGMEM */
 
 #include <asm/mmzone.h>

Mike Fedyk

2004-01-05 22:53 UTC #3 lore raw

In reply to Martin J. Bligh

include/asm-i386/mmzone.h line 38

Quoted 6 lines · Martin J. Bligh

+ * for now assume that 64Gb is max amount of RAM for whole system
+ * 64Gb / 4096bytes/page = 16777216 pages
+ */
include/asm-i386/mmzone.h +3 -0
+#define MAX_NR_PAGES 16777216
+#define MAX_ELEMENTS 256
+#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)

Why not do the calculation in the define, and use PAGE_SIZE?

If PAGE_SIZE isn't 4k will it break the rest of this code, or will the calculations make sence with larger PAGE_SIZE?

Might as well make it easier to go in the direction of variable PAGE_SIZE instead of keeping the assumption.

Martin J. Bligh

2004-01-05 23:16 UTC #4 lore raw

Quoted 9 lines

include/asm-i386/mmzone.h line 38
Quoted 6 lines
+ * for now assume that 64Gb is max amount of RAM for whole system
+ * 64Gb / 4096bytes/page = 16777216 pages
+ */
include/asm-i386/mmzone.h +3 -0
+#define MAX_NR_PAGES 16777216
+#define MAX_ELEMENTS 256
+#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)
Why not do the calculation in the define, and use PAGE_SIZE?

Quoted 5 lines

If PAGE_SIZE isn't 4k will it break the rest of this code, or will the calculations make sence with larger PAGE_SIZE?
Might as well make it easier to go in the direction of variable PAGE_SIZE instead of keeping the assumption.

The patch is just moving the code, not changing it ;-) But yes, that value could probably be derived instead of hardcoded. Separate patch though.

Matthew Dobson

author

2004-03-29 15:45 UTC #6 lore raw

In reply to Martin J. Bligh

Ok... This looks good to me. It's definitely longer than my 1 line fix, but it's also more correct long-term solution.

Cheers!

-Matt

Quoted 129 lines · Martin J. Bligh

Quoted 1 line

pfn_to_nid() is defined for *almost* all configurations for i386. There is a small bug in that for CONFIG_X86_PC it is not. This just sets up a generic definition so that this function is always defined in a reasonable manner.

I'd prefer to fix the current ungodly mess in the current fashion ...
I know it's longer, but I'd prefer to clean it up more if we're going to touch it. This has been sitting in my tree for a while, waiting for the code freeze to slush out a bit.

diff -purN -X /home/mbligh/.diff.exclude 276-per_node_rss/include/asm-i386/mmzone.h 278-pfn_to_nid/include/asm-i386/mmzone.h

include/asm-i386/mmzone.h +42 ~~-39~~

@@ -10,7 +10,49 @@
 
 #ifdef CONFIG_DISCONTIGMEM
 
+#ifdef CONFIG_NUMA
+	#ifdef CONFIG_X86_NUMAQ
+		#include <asm/numaq.h>
+	#else	/* summit or generic arch */
+		#include <asm/srat.h>
+	#endif
+#else /* !CONFIG_NUMA */
+	#define get_memcfg_numa get_memcfg_numa_flat
+	#define get_zholes_size(n) (0)
+#endif /* CONFIG_NUMA */
+
 extern struct pglist_data *node_data[];
+#define NODE_DATA(nid)		(node_data[nid])
+
+/*
+ * generic node memory support, the following assumptions apply:
+ *
+ * 1) memory comes in 256Mb contigious chunks which are either present or not
+ * 2) we will not have more than 64Gb in total
+ *
+ * for now assume that 64Gb is max amount of RAM for whole system
+ *    64Gb / 4096bytes/page = 16777216 pages
+ */
+#define MAX_NR_PAGES 16777216
+#define MAX_ELEMENTS 256
+#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)
+
+extern u8 physnode_map[];
+
+static inline int pfn_to_nid(unsigned long pfn)
+{
+#ifdef CONFIG_NUMA
+	return(physnode_map[(pfn) / PAGES_PER_ELEMENT]);
+#else
+	return 0;
+#endif
+}
+
+static inline struct pglist_data *pfn_to_pgdat(unsigned long pfn)
+{
+	return(NODE_DATA(pfn_to_nid(pfn)));
+}
+
 
 /*
  * Following are macros that are specific to this numa platform.

@@ -43,11 +85,6 @@ extern struct pglist_data *node_data[];
  */
 #define kvaddr_to_nid(kaddr)	pfn_to_nid(__pa(kaddr) >> PAGE_SHIFT)
 
-/*
- * Return a pointer to the node data for node n.
- */
-#define NODE_DATA(nid)		(node_data[nid])
-
 #define node_mem_map(nid)	(NODE_DATA(nid)->node_mem_map)
 #define node_start_pfn(nid)	(NODE_DATA(nid)->node_start_pfn)
 #define node_end_pfn(nid)						\

@@ -93,40 +130,6 @@ extern struct pglist_data *node_data[];
  */ 
 #define pfn_valid(pfn)          ((pfn) < num_physpages)
 
-/*
- * generic node memory support, the following assumptions apply:
- *
- * 1) memory comes in 256Mb contigious chunks which are either present or not
- * 2) we will not have more than 64Gb in total
- *
- * for now assume that 64Gb is max amount of RAM for whole system
- *    64Gb / 4096bytes/page = 16777216 pages
- */
-#define MAX_NR_PAGES 16777216
-#define MAX_ELEMENTS 256
-#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)
-
-extern u8 physnode_map[];
-
-static inline int pfn_to_nid(unsigned long pfn)
-{
-	return(physnode_map[(pfn) / PAGES_PER_ELEMENT]);
-}
-static inline struct pglist_data *pfn_to_pgdat(unsigned long pfn)
-{
-	return(NODE_DATA(pfn_to_nid(pfn)));
-}
-
-#ifdef CONFIG_X86_NUMAQ
-#include <asm/numaq.h>
-#elif CONFIG_ACPI_SRAT
-#include <asm/srat.h>
-#elif CONFIG_X86_PC
-#define get_zholes_size(n) (0)
-#else
-#define pfn_to_nid(pfn)		(0)
-#endif /* CONFIG_X86_NUMAQ */
-
 extern int get_memcfg_numa_flat(void );
 /*
  * This allows any one NUMA architecture to be compiled

diff -purN -X /home/mbligh/.diff.exclude 276-per_node_rss/include/linux/mmzone.h 278-pfn_to_nid/include/linux/mmzone.h

include/linux/mmzone.h +1 -1

@@ -303,7 +303,7 @@ extern struct pglist_data contig_page_da
 #define NODE_DATA(nid)		(&contig_page_data)
 #define NODE_MEM_MAP(nid)	mem_map
 #define MAX_NODES_SHIFT		0
-
+#define pfn_to_nid(pfn)		(0)
 #else /* CONFIG_DISCONTIGMEM */
 
 #include <asm/mmzone.h>

Mike Fedyk

2004-03-29 15:45 UTC #7 lore raw

In reply to Martin J. Bligh

include/asm-i386/mmzone.h line 38

Quoted 6 lines · Martin J. Bligh

+ * for now assume that 64Gb is max amount of RAM for whole system
+ * 64Gb / 4096bytes/page = 16777216 pages
+ */
include/asm-i386/mmzone.h +3 -0
+#define MAX_NR_PAGES 16777216
+#define MAX_ELEMENTS 256
+#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)

Why not do the calculation in the define, and use PAGE_SIZE?

If PAGE_SIZE isn't 4k will it break the rest of this code, or will the calculations make sence with larger PAGE_SIZE?

Might as well make it easier to go in the direction of variable PAGE_SIZE instead of keeping the assumption.

Martin J. Bligh

2004-03-29 15:45 UTC #5 lore raw

In reply to Mike Fedyk

Quoted 9 lines

include/asm-i386/mmzone.h line 38
Quoted 6 lines
+ * for now assume that 64Gb is max amount of RAM for whole system
+ * 64Gb / 4096bytes/page = 16777216 pages
+ */
include/asm-i386/mmzone.h +3 -0
+#define MAX_NR_PAGES 16777216
+#define MAX_ELEMENTS 256
+#define PAGES_PER_ELEMENT (MAX_NR_PAGES/MAX_ELEMENTS)
Why not do the calculation in the define, and use PAGE_SIZE?

Quoted 5 lines

If PAGE_SIZE isn't 4k will it break the rest of this code, or will the calculations make sence with larger PAGE_SIZE?
Might as well make it easier to go in the direction of variable PAGE_SIZE instead of keeping the assumption.

The patch is just moving the code, not changing it ;-) But yes, that value could probably be derived instead of hardcoded. Separate patch though.