[PATCH] sparsemem memory model for i386

Provide the architecture specific implementation for SPARSEMEM for i386 SMP
and NUMA systems.

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Dave Hansen <haveblue@us.ibm.com>
Signed-off-by: Martin Bligh <mbligh@aracnet.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

5 files changed, 52 insertions, 36 deletions

diff --git a/arch/i386/Kconfig b/arch/i386/Kconfig
index f0064b5e3702..bfdcedef06e1 100644
--- a/arch/i386/Kconfig
+++ b/arch/i386/Kconfig
@@ -68,7 +68,6 @@ config X86_VOYAGER
 
 config X86_NUMAQ
         bool "NUMAQ (IBM/Sequent)"
-        select DISCONTIGMEM
         select NUMA
         help
           This option is used for getting Linux to run on a (IBM/Sequent) NUMA
@@ -783,11 +782,6 @@ comment "NUMA (NUMA-Q) requires SMP, 64GB highmem support"
 comment "NUMA (Summit) requires SMP, 64GB highmem support, ACPI"
         depends on X86_SUMMIT && (!HIGHMEM64G || !ACPI)
 
-config ARCH_DISCONTIGMEM_ENABLE
-        bool
-        depends on NUMA
-        default y
-
 config HAVE_ARCH_BOOTMEM_NODE
         bool
         depends on NUMA
@@ -800,7 +794,7 @@ config ARCH_HAVE_MEMORY_PRESENT
 
 config NEED_NODE_MEMMAP_SIZE
         bool
-        depends on DISCONTIGMEM
+        depends on DISCONTIGMEM || SPARSEMEM
         default y
 
 config HAVE_ARCH_ALLOC_REMAP
@@ -808,6 +802,22 @@ config HAVE_ARCH_ALLOC_REMAP
         depends on NUMA
         default y
 
+config ARCH_DISCONTIGMEM_ENABLE
+        def_bool y
+        depends on NUMA
+
+config ARCH_DISCONTIGMEM_DEFAULT
+        def_bool y
+        depends on NUMA
+
+config ARCH_SPARSEMEM_ENABLE
+        def_bool y
+        depends on NUMA
+
+config ARCH_SELECT_MEMORY_MODEL
+        def_bool y
+        depends on ARCH_SPARSEMEM_ENABLE
+
 source "mm/Kconfig"
 
 config HAVE_ARCH_EARLY_PFN_TO_NID
diff --git a/arch/i386/kernel/setup.c b/arch/i386/kernel/setup.c
index 2bfbddebdbf8..0d689335c8d6 100644
--- a/arch/i386/kernel/setup.c
+++ b/arch/i386/kernel/setup.c
@@ -25,6 +25,7 @@
 
 #include <linux/sched.h>
 #include <linux/mm.h>
+#include <linux/mmzone.h>
 #include <linux/tty.h>
 #include <linux/ioport.h>
 #include <linux/acpi.h>
@@ -1022,7 +1023,7 @@ static void __init reserve_ebda_region(void)
                 reserve_bootmem(addr, PAGE_SIZE);       
 }
 
-#ifndef CONFIG_DISCONTIGMEM
+#ifndef CONFIG_NEED_MULTIPLE_NODES
 void __init setup_bootmem_allocator(void);
 static unsigned long __init setup_memory(void)
 {
@@ -1072,9 +1073,9 @@ void __init zone_sizes_init(void)
         free_area_init(zones_size);
 }
 #else
-extern unsigned long setup_memory(void);
+extern unsigned long __init setup_memory(void);
 extern void zone_sizes_init(void);
-#endif /* !CONFIG_DISCONTIGMEM */
+#endif /* !CONFIG_NEED_MULTIPLE_NODES */
 
 void __init setup_bootmem_allocator(void)
 {
@@ -1475,6 +1476,7 @@ void __init setup_arch(char **cmdline_p)
 #endif
         paging_init();
         remapped_pgdat_init();
+        sparse_init();
         zone_sizes_init();
 
         /*
diff --git a/arch/i386/mm/Makefile b/arch/i386/mm/Makefile
index fc3272506846..80908b5aa60f 100644
--- a/arch/i386/mm/Makefile
+++ b/arch/i386/mm/Makefile
@@ -4,7 +4,7 @@
 
 obj-y   := init.o pgtable.o fault.o ioremap.o extable.o pageattr.o mmap.o
 
-obj-$(CONFIG_DISCONTIGMEM)      += discontig.o
+obj-$(CONFIG_NUMA) += discontig.o
 obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
 obj-$(CONFIG_HIGHMEM) += highmem.o
 obj-$(CONFIG_BOOT_IOREMAP) += boot_ioremap.o
diff --git a/arch/i386/mm/discontig.c b/arch/i386/mm/discontig.c
index 088ca4722183..0efeb96ba5d4 100644
--- a/arch/i386/mm/discontig.c
+++ b/arch/i386/mm/discontig.c
@@ -42,12 +42,16 @@ bootmem_data_t node0_bdata;
  *                  populated the following initialisation.
  *
  * 1) node_online_map  - the map of all nodes configured (online) in the system
- * 2) physnode_map     - the mapping between a pfn and owning node
- * 3) node_start_pfn   - the starting page frame number for a node
+ * 2) node_start_pfn   - the starting page frame number for a node
  * 3) node_end_pfn     - the ending page fram number for a node
  */
+unsigned long node_start_pfn[MAX_NUMNODES];
+unsigned long node_end_pfn[MAX_NUMNODES];
+
 
+#ifdef CONFIG_DISCONTIGMEM
 /*
+ * 4) physnode_map     - the mapping between a pfn and owning node
  * physnode_map keeps track of the physical memory layout of a generic
  * numa node on a 256Mb break (each element of the array will
  * represent 256Mb of memory and will be marked by the node id.  so,
@@ -85,9 +89,7 @@ unsigned long node_memmap_size_bytes(int nid, unsigned long start_pfn,
 
         return (nr_pages + 1) * sizeof(struct page);
 }
-
-unsigned long node_start_pfn[MAX_NUMNODES];
-unsigned long node_end_pfn[MAX_NUMNODES];
+#endif
 
 extern unsigned long find_max_low_pfn(void);
 extern void find_max_pfn(void);
@@ -390,24 +392,26 @@ void __init set_highmem_pages_init(int bad_ppro)
 {
 #ifdef CONFIG_HIGHMEM
         struct zone *zone;
+        struct page *page;
 
         for_each_zone(zone) {
-                unsigned long node_pfn, node_high_size, zone_start_pfn;
-                struct page * zone_mem_map;
-                
+                unsigned long node_pfn, zone_start_pfn, zone_end_pfn;
+
                 if (!is_highmem(zone))
                         continue;
 
-                printk("Initializing %s for node %d\n", zone->name,
-                        zone->zone_pgdat->node_id);
-
-                node_high_size = zone->spanned_pages;
-                zone_mem_map = zone->zone_mem_map;
                 zone_start_pfn = zone->zone_start_pfn;
+                zone_end_pfn = zone_start_pfn + zone->spanned_pages;
+
+                printk("Initializing %s for node %d (%08lx:%08lx)\n",
+                                zone->name, zone->zone_pgdat->node_id,
+                                zone_start_pfn, zone_end_pfn);
 
-                for (node_pfn = 0; node_pfn < node_high_size; node_pfn++) {
-                        one_highpage_init((struct page *)(zone_mem_map + node_pfn),
-                                          zone_start_pfn + node_pfn, bad_ppro);
+                for (node_pfn = zone_start_pfn; node_pfn < zone_end_pfn; node_pfn++) {
+                        if (!pfn_valid(node_pfn))
+                                continue;
+                        page = pfn_to_page(node_pfn);
+                        one_highpage_init(page, node_pfn, bad_ppro);
                 }
         }
         totalram_pages += totalhigh_pages;
diff --git a/arch/i386/mm/init.c b/arch/i386/mm/init.c
index 666ca79fb50a..48ebfab77a3c 100644
--- a/arch/i386/mm/init.c
+++ b/arch/i386/mm/init.c
@@ -276,7 +276,9 @@ void __init one_highpage_init(struct page *page, int pfn, int bad_ppro)
                 SetPageReserved(page);
 }
 
-#ifndef CONFIG_DISCONTIGMEM
+#ifdef CONFIG_NUMA
+extern void set_highmem_pages_init(int);
+#else
 static void __init set_highmem_pages_init(int bad_ppro)
 {
         int pfn;
@@ -284,9 +286,7 @@ static void __init set_highmem_pages_init(int bad_ppro)
                 one_highpage_init(pfn_to_page(pfn), pfn, bad_ppro);
         totalram_pages += totalhigh_pages;
 }
-#else
-extern void set_highmem_pages_init(int);
-#endif /* !CONFIG_DISCONTIGMEM */
+#endif /* CONFIG_FLATMEM */
 
 #else
 #define kmap_init() do { } while (0)
@@ -297,10 +297,10 @@ extern void set_highmem_pages_init(int);
 unsigned long long __PAGE_KERNEL = _PAGE_KERNEL;
 unsigned long long __PAGE_KERNEL_EXEC = _PAGE_KERNEL_EXEC;
 
-#ifndef CONFIG_DISCONTIGMEM
-#define remap_numa_kva() do {} while (0)
-#else
+#ifdef CONFIG_NUMA
 extern void __init remap_numa_kva(void);
+#else
+#define remap_numa_kva() do {} while (0)
 #endif
 
 static void __init pagetable_init (void)
@@ -525,7 +525,7 @@ static void __init set_max_mapnr_init(void)
 #else
         num_physpages = max_low_pfn;
 #endif
-#ifndef CONFIG_DISCONTIGMEM
+#ifdef CONFIG_FLATMEM
         max_mapnr = num_physpages;
 #endif
 }
@@ -539,7 +539,7 @@ void __init mem_init(void)
         int tmp;
         int bad_ppro;
 
-#ifndef CONFIG_DISCONTIGMEM
+#ifdef CONFIG_FLATMEM
         if (!mem_map)
                 BUG();
 #endif