[PATCH] free_area_init cleanup

Patch from Martin Bligh.  It should only affect machines using
discontigmem.

"This patch cleans up free_area_init stuff, and undefines mem_map and
max_mapnr for discontigmem, where they were horrible kludges anyway
...  We just use the lmem_maps instead, which makes much more sense.
It also kills pgdat->node_start_mapnr, which is tarred with the same
brush.

It breaks free_area_init_core into a couple of sections, pulls the
allocation of the lmem_map back into the next higher function, and
passes more things via the pgdat.  But that's not very interesting,
the objective was to kill mem_map for discontigmem, which seems to
attract bugs like flypaper.  This brings any misuses to obvious
compile-time errors rather than wierd oopses, which I can't help but
feel is a good thing.

It does break other discontigmem architectures, but in a very obvious
way (they won't compile) and it's easy to fix.  I think that's a small
price to pay ...  ;-) At some point soon I will follow up with a patch
to remove free_area_init_node for the contig mem case, or at the very
least rename it to something more sensible, like __free_area_init.

Christoph has grander plans to kill mem_map more extensively in
addition to the attatched, but I've heard nobody disagree that it
should die for the discontigmem case at least.

Oh, and I renamed mem_map in drivers/pcmcia/sa1100 to pc_mem_map
because my tiny little brain (and cscope) find it confusing like that.

Tested on 16-way NUMA-Q with discontigmem + NUMA support and on a
standard PC (well, boots and appears functional).  On top of
2.5.33-mm4"

arch/alpha/mm/numa.c

@@ -286,7 +286,6 @@ void __init paging_init(void)
 	for (nid = 0; nid < numnodes; nid++) {
 		unsigned long start_pfn = plat_node_bdata[nid].node_boot_start >> PAGE_SHIFT;
 		unsigned long end_pfn = plat_node_bdata[nid].node_low_pfn;
-		unsigned long lmax_mapnr;
 
 		if (dma_local_pfn >= end_pfn - start_pfn)
 			zones_size[ZONE_DMA] = end_pfn - start_pfn;
@@ -295,11 +294,6 @@ void __init paging_init(void)
 			zones_size[ZONE_NORMAL] = (end_pfn - start_pfn) - dma_local_pfn;
 		}
 		free_area_init_node(nid, NODE_DATA(nid), NULL, zones_size, start_pfn, NULL);
-		lmax_mapnr = PLAT_NODE_DATA_STARTNR(nid) + PLAT_NODE_DATA_SIZE(nid);
-		if (lmax_mapnr > max_mapnr) {
-			max_mapnr = lmax_mapnr;
-			DBGDCONT("Grow max_mapnr to %ld\n", max_mapnr);
-		}
 	}
 
 	/* Initialize the kernel's ZERO_PGE. */

arch/i386/kernel/cpu/amd.c

@@ -25,7 +25,7 @@ __asm__(".align 4\nvide: ret");
 static void __init init_amd(struct cpuinfo_x86 *c)
 {
 	u32 l, h;
-	int mbytes = max_mapnr >> (20-PAGE_SHIFT);
+	int mbytes = num_physpages >> (20-PAGE_SHIFT);
 	int r;
 
 	/*

arch/i386/kernel/i386_ksyms.c

@@ -58,7 +58,11 @@ EXPORT_SYMBOL(boot_cpu_data);
 EXPORT_SYMBOL(EISA_bus);
 #endif
 EXPORT_SYMBOL(MCA_bus);
-#ifdef CONFIG_MULTIQUAD
+#ifdef CONFIG_DISCONTIGMEM
+EXPORT_SYMBOL(node_data);
+EXPORT_SYMBOL(pfn_to_nid);
+#endif
+#ifdef CONFIG_X86_NUMAQ
 EXPORT_SYMBOL(xquad_portio);
 #endif
 EXPORT_SYMBOL(__verify_write);

arch/i386/kernel/numaq.c

@@ -82,27 +82,19 @@ static void __init smp_dump_qct(void)
  */
 int physnode_map[MAX_ELEMENTS] = { [0 ... (MAX_ELEMENTS - 1)] = -1};
 
-#define MB_TO_ELEMENT(x) (x >> ELEMENT_REPRESENTS)
-#define PA_TO_MB(pa) (pa >> 20) 	/* assumption: a physical address is in bytes */
+#define PFN_TO_ELEMENT(pfn) (pfn / PAGES_PER_ELEMENT)
+#define PA_TO_ELEMENT(pa) (PFN_TO_ELEMENT(pa >> PAGE_SHIFT))
 
-int pa_to_nid(u64 pa)
+int pfn_to_nid(unsigned long pfn)
 {
-	int nid;
-	
-	nid = physnode_map[MB_TO_ELEMENT(PA_TO_MB(pa))];
+	int nid = physnode_map[PFN_TO_ELEMENT(pfn)];
 
-	/* the physical address passed in is not in the map for the system */
 	if (nid == -1)
-		BUG();
+		BUG(); /* address is not present */
 
 	return nid;
 }
 
-int pfn_to_nid(unsigned long pfn)
-{
-	return pa_to_nid(((u64)pfn) << PAGE_SHIFT);
-}
-
 /*
  * for each node mark the regions
  *        TOPOFMEM = hi_shrd_mem_start + hi_shrd_mem_size
@@ -132,7 +124,7 @@ static void __init initialize_physnode_map(void)
 			topofmem = eq->hi_shrd_mem_start + eq->hi_shrd_mem_size;
 			while (cur < topofmem) {
 				physnode_map[cur >> 8] = nid;
-				cur += (ELEMENT_REPRESENTS - 1);
+				cur ++;
 			}
 		}
 	}

arch/i386/mm/discontig.c

@@ -275,20 +275,9 @@ void __init set_highmem_pages_init(int bad_ppro)
 void __init set_max_mapnr_init(void)
 {
 #ifdef CONFIG_HIGHMEM
-	unsigned long lmax_mapnr;
-	int nid;
-	
-	highmem_start_page = mem_map + NODE_DATA(0)->node_zones[ZONE_HIGHMEM].zone_start_mapnr;
+	highmem_start_page = NODE_DATA(0)->node_zones[ZONE_HIGHMEM].zone_mem_map;
 	num_physpages = highend_pfn;
-
-	for (nid = 0; nid < numnodes; nid++) {
-		lmax_mapnr = node_startnr(nid) + node_size(nid);
-		if (lmax_mapnr > max_mapnr) {
-			max_mapnr = lmax_mapnr;
-		}
-	}
-	
 #else
-	max_mapnr = num_physpages = max_low_pfn;
+	num_physpages = max_low_pfn;
 #endif
 }

arch/i386/mm/init.c

@@ -440,8 +440,10 @@ void __init mem_init(void)
 	int tmp;
 	int bad_ppro;
 
+#ifndef CONFIG_DISCONTIGMEM
 	if (!mem_map)
 		BUG();
+#endif
 	
 	bad_ppro = ppro_with_ram_bug();
 
@@ -471,7 +473,7 @@ void __init mem_init(void)
 
 	printk("Memory: %luk/%luk available (%dk kernel code, %dk reserved, %dk data, %dk init, %ldk highmem)\n",
 		(unsigned long) nr_free_pages() << (PAGE_SHIFT-10),
-		max_mapnr << (PAGE_SHIFT-10),
+		num_physpages << (PAGE_SHIFT-10),
 		codesize >> 10,
 		reservedpages << (PAGE_SHIFT-10),
 		datasize >> 10,

arch/i386/mm/pgtable.c

@@ -22,26 +22,29 @@
 
 void show_mem(void)
 {
-	int pfn, total = 0, reserved = 0;
+	int total = 0, reserved = 0;
 	int shared = 0, cached = 0;
 	int highmem = 0;
 	struct page *page;
+	pg_data_t *pgdat;
+	unsigned long i;
 
 	printk("Mem-info:\n");
 	show_free_areas();
 	printk("Free swap:       %6dkB\n",nr_swap_pages<<(PAGE_SHIFT-10));
-	pfn = max_mapnr;
-	while (pfn-- > 0) {
-		page = pfn_to_page(pfn);
-		total++;
-		if (PageHighMem(page))
-			highmem++;
-		if (PageReserved(page))
-			reserved++;
-		else if (PageSwapCache(page))
-			cached++;
-		else if (page_count(page))
-			shared += page_count(page) - 1;
+	for_each_pgdat(pgdat) {
+		for (i = 0; i < pgdat->node_size; ++i) {
+			page = pgdat->node_mem_map + i;
+			total++;
+			if (PageHighMem(page))
+				highmem++;
+			if (PageReserved(page))
+				reserved++;
+			else if (PageSwapCache(page))
+				cached++;
+			else if (page_count(page))
+				shared += page_count(page) - 1;
+		}
 	}
 	printk("%d pages of RAM\n", total);
 	printk("%d pages of HIGHMEM\n",highmem);

arch/mips64/sgi-ip27/ip27-memory.c

@@ -254,10 +254,6 @@ void __init paging_init(void)
 		zones_size[ZONE_DMA] = end_pfn + 1 - start_pfn;
 		free_area_init_node(node, NODE_DATA(node), 0, zones_size, 
 						start_pfn, 0);
-		if ((PLAT_NODE_DATA_STARTNR(node) + 
-					PLAT_NODE_DATA_SIZE(node)) > pagenr)
-			pagenr = PLAT_NODE_DATA_STARTNR(node) +
-					PLAT_NODE_DATA_SIZE(node);
 	}
 }
 
@@ -271,7 +267,6 @@ void __init mem_init(void)
 	unsigned long codesize, datasize, initsize;
 	int slot, numslots;
 	struct page *pg, *pslot;
-	pfn_t pgnr;
 
 	num_physpages = numpages;	/* memory already sized by szmem */
 	max_mapnr = pagenr;		/* already found during paging_init */
@@ -293,7 +288,6 @@ void __init mem_init(void)
 		 * We need to manually do the other slots.
 		 */
 		pg = NODE_DATA(nid)->node_mem_map + slot_getsize(nid, 0);
-		pgnr = PLAT_NODE_DATA_STARTNR(nid) + slot_getsize(nid, 0);
 		numslots = node_getlastslot(nid);
 		for (slot = 1; slot <= numslots; slot++) {
 			pslot = NODE_DATA(nid)->node_mem_map + 
@@ -304,7 +298,7 @@ void __init mem_init(void)
 			 * free up the pages that hold the memmap entries.
 			 */
 			while (pg < pslot) {
-				pg++; pgnr++;
+				pg++;
 			}
 
 			/*
@@ -312,8 +306,8 @@ void __init mem_init(void)
 			 */
 			pslot += slot_getsize(nid, slot);
 			while (pg < pslot) {
-				if (!page_is_ram(pgnr))
-					continue;
+				/* if (!page_is_ram(pgnr)) continue; */
+				/* commented out until page_is_ram works */
 				ClearPageReserved(pg);
 				atomic_set(&pg->count, 1);
 				__free_page(pg);

drivers/pcmcia/sa1100.h

@@ -160,7 +160,7 @@ struct sa1100_pcmcia_socket {
    */
   socket_state_t        cs_state;
   pccard_io_map         io_map[MAX_IO_WIN];
-  pccard_mem_map        mem_map[MAX_WIN];
+  pccard_mem_map        pc_mem_map[MAX_WIN];
   void                  (*handler)(void *, unsigned int);
   void                  *handler_info;
 

drivers/pcmcia/sa1100_generic.c

@@ -686,7 +686,7 @@ sa1100_pcmcia_get_mem_map(unsigned int sock, struct pccard_mem_map *map)
   DEBUG(2, "%s() for sock %u\n", __FUNCTION__, sock);
 
   if (map->map < MAX_WIN) {
-    *map = skt->mem_map[map->map];
+    *map = skt->pc_mem_map[map->map];
     ret = 0;
   }
 
@@ -754,7 +754,7 @@ sa1100_pcmcia_set_mem_map(unsigned int sock, struct pccard_mem_map *map)
   map->sys_stop += start;
   map->sys_start = start;
 
-  skt->mem_map[map->map] = *map;
+  skt->pc_mem_map[map->map] = *map;
 
   return 0;
 }  /* sa1100_pcmcia_set_mem_map() */

include/asm-alpha/mmzone.h

@@ -46,8 +46,6 @@ extern plat_pg_data_t *plat_node_data[];
 
 #define PHYSADDR_TO_NID(pa)		ALPHA_PA_TO_NID(pa)
 #define PLAT_NODE_DATA(n)		(plat_node_data[(n)])
-#define PLAT_NODE_DATA_STARTNR(n)	\
-	(PLAT_NODE_DATA(n)->gendata.node_start_mapnr)
 #define PLAT_NODE_DATA_SIZE(n)		(PLAT_NODE_DATA(n)->gendata.node_size)
 
 #if 1

include/asm-i386/mmzone.h

@@ -11,7 +11,6 @@
 #ifdef CONFIG_X86_NUMAQ
 #include <asm/numaq.h>
 #else
-#define pa_to_nid(pa)	(0)
 #define pfn_to_nid(pfn)		(0)
 #ifdef CONFIG_NUMA
 #define _cpu_to_node(cpu) 0
@@ -44,7 +43,6 @@ extern struct pglist_data *node_data[];
 #define alloc_bootmem_low_pages_node(ignore, x) \
 	__alloc_bootmem_node(NODE_DATA(0), (x), PAGE_SIZE, 0)
 
-#define node_startnr(nid)	(node_data[nid]->node_start_mapnr)
 #define node_size(nid)		(node_data[nid]->node_size)
 #define node_localnr(pfn, nid)	((pfn) - node_data[nid]->node_start_pfn)
 
@@ -55,7 +53,7 @@ extern struct pglist_data *node_data[];
 /*
  * Given a kernel address, find the home node of the underlying memory.
  */
-#define kvaddr_to_nid(kaddr)	pa_to_nid(__pa(kaddr))
+#define kvaddr_to_nid(kaddr)	pfn_to_nid(__pa(kaddr) >> PAGE_SHIFT)
 
 /*
  * Return a pointer to the node data for node n.
@@ -64,6 +62,8 @@ extern struct pglist_data *node_data[];
 
 #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)       (NODE_DATA(nid)->node_start_pfn + \
+				 NODE_DATA(nid)->node_size)
 
 #define local_mapnr(kvaddr) \
 	( (__pa(kvaddr) >> PAGE_SHIFT) - node_start_pfn(kvaddr_to_nid(kvaddr)) )
@@ -74,5 +74,13 @@ extern struct pglist_data *node_data[];
 #define pfn_to_page(pfn)	(node_mem_map(pfn_to_nid(pfn)) + node_localnr(pfn, pfn_to_nid(pfn)))
 #define page_to_pfn(page)	((page - page_zone(page)->zone_mem_map) + page_zone(page)->zone_start_pfn)
 #define pmd_page(pmd)		(pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT))
+/*
+ * pfn_valid should be made as fast as possible, and the current definition 
+ * is valid for machines that are NUMA, but still contiguous, which is what
+ * is currently supported. A more generalised, but slower definition would
+ * be something like this - mbligh:
+ * ( pfn_to_pgdat(pfn) && (pfn < node_end_pfn(pfn_to_nid(pfn))) ) 
+ */ 
+#define pfn_valid(pfn)          (pfn < num_physpages)
 #endif /* CONFIG_DISCONTIGMEM */
 #endif /* _ASM_MMZONE_H_ */

include/asm-i386/numaq.h

@@ -32,17 +32,18 @@
 
 /*
  * for now assume that 64Gb is max amount of RAM for whole system
- *    64Gb * 1024Mb/Gb = 65536 Mb
- *    65536 Mb / 256Mb = 256
+ *    64Gb / 4096bytes/page = 16777216 pages
  */
+#define MAX_NR_PAGES 16777216
 #define MAX_ELEMENTS 256
-#define ELEMENT_REPRESENTS 8 /* 256 Mb */
+#define PAGES_PER_ELEMENT (16777216/256)
 
+#define pfn_to_pgdat(pfn) NODE_DATA(pfn_to_nid(pfn))
+#define PHYSADDR_TO_NID(pa) pfn_to_nid(pa >> PAGE_SHIFT)
 #define MAX_NUMNODES		8
 #ifdef CONFIG_NUMA
 #define _cpu_to_node(cpu) (cpu_to_logical_apicid(cpu) >> 4)
 #endif /* CONFIG_NUMA */
-extern int pa_to_nid(u64);
 extern int pfn_to_nid(unsigned long);
 extern void get_memcfg_numaq(void);
 #define get_memcfg_numa() get_memcfg_numaq()

include/asm-i386/page.h

@@ -145,10 +145,10 @@ static __inline__ int get_order(unsigned long size)
 #ifndef CONFIG_DISCONTIGMEM
 #define pfn_to_page(pfn)	(mem_map + (pfn))
 #define page_to_pfn(page)	((unsigned long)((page) - mem_map))
+#define pfn_valid(pfn)		((pfn) < max_mapnr)
 #endif /* !CONFIG_DISCONTIGMEM */
 #define virt_to_page(kaddr)	pfn_to_page(__pa(kaddr) >> PAGE_SHIFT)
 
-#define pfn_valid(pfn)		((pfn) < max_mapnr)
 #define virt_addr_valid(kaddr)	pfn_valid(__pa(kaddr) >> PAGE_SHIFT)
 
 #define VM_DATA_DEFAULT_FLAGS	(VM_READ | VM_WRITE | VM_EXEC | \

include/asm-mips64/mmzone.h

@@ -24,7 +24,6 @@ extern plat_pg_data_t *plat_node_data[];
 
 #define PHYSADDR_TO_NID(pa)		NASID_TO_COMPACT_NODEID(NASID_GET(pa))
 #define PLAT_NODE_DATA(n)		(plat_node_data[n])
-#define PLAT_NODE_DATA_STARTNR(n)    (PLAT_NODE_DATA(n)->gendata.node_start_mapnr)
 #define PLAT_NODE_DATA_SIZE(n)	     (PLAT_NODE_DATA(n)->gendata.node_size)
 #define PLAT_NODE_DATA_LOCALNR(p, n) \
 		(((p) >> PAGE_SHIFT) - PLAT_NODE_DATA(n)->gendata.node_start_pfn)

include/asm-mips64/pgtable.h

@@ -373,10 +373,10 @@ extern inline void pgd_clear(pgd_t *pgdp)
 #ifndef CONFIG_DISCONTIGMEM
 #define pte_page(x)		(mem_map+(unsigned long)((pte_val(x) >> PAGE_SHIFT)))
 #else
-#define mips64_pte_pagenr(x) \
-	(PLAT_NODE_DATA_STARTNR(PHYSADDR_TO_NID(pte_val(x))) + \
-	PLAT_NODE_DATA_LOCALNR(pte_val(x), PHYSADDR_TO_NID(pte_val(x))))
-#define pte_page(x)		(mem_map+mips64_pte_pagenr(x))
+
+#define pte_page(x) ( NODE_MEM_MAP(PHYSADDR_TO_NID(pte_val(x))) +
+	PLAT_NODE_DATA_LOCALNR(pte_val(x), PHYSADDR_TO_NID(pte_val(x))) )
+				  
 #endif
 
 /*

include/linux/mm.h

@@ -15,7 +15,10 @@
 #include <linux/rbtree.h>
 #include <linux/fs.h>
 
+#ifndef CONFIG_DISCONTIGMEM          /* Don't use mapnrs, do it properly */
 extern unsigned long max_mapnr;
+#endif
+
 extern unsigned long num_physpages;
 extern void * high_memory;
 extern int page_cluster;
@@ -345,8 +348,10 @@ static inline int page_mapped(struct page *page)
 #define VM_FAULT_MINOR	1
 #define VM_FAULT_MAJOR	2
 
-/* The array of struct pages */
+#ifndef CONFIG_DISCONTIGMEM
+/* The array of struct pages - for discontigmem use pgdat->lmem_map */
 extern struct page *mem_map;
+#endif 
 
 extern void show_free_areas(void);
 

include/linux/mmzone.h

@@ -112,7 +112,6 @@ struct zone {
 	struct page		*zone_mem_map;
 	/* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
 	unsigned long		zone_start_pfn;
-	unsigned long		zone_start_mapnr;
 
 	/*
 	 * rarely used fields:
@@ -163,7 +162,6 @@ typedef struct pglist_data {
 	unsigned long *valid_addr_bitmap;
 	struct bootmem_data *bdata;
 	unsigned long node_start_pfn;
-	unsigned long node_start_mapnr;
 	unsigned long node_size;
 	int node_id;
 	struct pglist_data *pgdat_next;
@@ -187,9 +185,10 @@ memclass(struct zone *pgzone, struct zone *classzone)
  * prototypes for the discontig memory code.
  */
 struct page;
-void free_area_init_core(int nid, pg_data_t *pgdat, struct page **gmap,
-  unsigned long *zones_size, unsigned long paddr, unsigned long *zholes_size,
-  struct page *pmap);
+extern void calculate_totalpages (pg_data_t *pgdat, unsigned long *zones_size,
+		unsigned long *zholes_size);
+extern void free_area_init_core(pg_data_t *pgdat, unsigned long *zones_size,
+		unsigned long *zholes_size);
 void get_zone_counts(unsigned long *active, unsigned long *inactive);
 
 extern pg_data_t contig_page_data;

kernel/ksyms.c

@@ -115,9 +115,11 @@ EXPORT_SYMBOL(vmalloc_32);
 EXPORT_SYMBOL(vmap);
 EXPORT_SYMBOL(vunmap);
 EXPORT_SYMBOL(vmalloc_to_page);
-EXPORT_SYMBOL(mem_map);
 EXPORT_SYMBOL(remap_page_range);
+#ifndef CONFIG_DISCONTIGMEM
+EXPORT_SYMBOL(mem_map);
 EXPORT_SYMBOL(max_mapnr);
+#endif
 EXPORT_SYMBOL(high_memory);
 EXPORT_SYMBOL(vmtruncate);
 EXPORT_SYMBOL(find_vma);

kernel/suspend.c

@@ -471,10 +471,12 @@ static int count_and_copy_data_pages(struct pbe *pagedir_p)
 	int nr_copy_pages = 0;
 	int pfn;
 	struct page *page;
-	
+
+#ifndef CONFIG_DISCONTIGMEM	
 	if (max_mapnr != num_physpages)
 		panic("mapnr is not expected");
-	for (pfn = 0; pfn < max_mapnr; pfn++) {
+#endif
+	for (pfn = 0; pfn < num_physpages; pfn++) {
 		page = pfn_to_page(pfn);
 		if (PageHighMem(page))
 			panic("Swsusp not supported on highmem boxes. Send 1GB of RAM to <pavel@ucw.cz> and try again ;-).");
@@ -514,19 +516,20 @@ static int count_and_copy_data_pages(struct pbe *pagedir_p)
 
 static void free_suspend_pagedir(unsigned long this_pagedir)
 {
-	struct page *page = mem_map;
-	int i;
+	struct page *page;
+	int pfn;
 	unsigned long this_pagedir_end = this_pagedir +
 		(PAGE_SIZE << pagedir_order);
 
-	for(i=0; i < num_physpages; i++, page++) {
+	for(pfn = 0; pfn < num_physpages; pfn++) {
+		page = pfn_to_page(pfn);
 		if (!TestClearPageNosave(page))
 			continue;
 
-		if (ADDRESS(i) >= this_pagedir && ADDRESS(i) < this_pagedir_end)
+		if (ADDRESS(pfn) >= this_pagedir && ADDRESS(pfn) < this_pagedir_end)
 			continue; /* old pagedir gets freed in one */
 		
-		free_page(ADDRESS(i));
+		free_page(ADDRESS(pfn));
 	}
 	free_pages(this_pagedir, pagedir_order);
 }

mm/memory.c

@@ -53,7 +53,12 @@
 
 #include <linux/swapops.h>
 
+#ifndef CONFIG_DISCONTIGMEM
+/* use the per-pgdat data instead for discontigmem - mbligh */
 unsigned long max_mapnr;
+struct page *mem_map;
+#endif
+
 unsigned long num_physpages;
 void * high_memory;
 struct page *highmem_start_page;
@@ -72,8 +77,6 @@ static inline void copy_cow_page(struct page * from, struct page * to, unsigned
 	copy_user_highpage(to, from, address);
 }
 
-struct page *mem_map;
-
 /*
  * Note: this doesn't free the actual pages themselves. That
  * has been handled earlier when unmapping all the memory regions.

mm/numa.c

@@ -22,11 +22,21 @@ pg_data_t contig_page_data = { .bdata = &contig_bootmem_data };
  * Should be invoked with paramters (0, 0, unsigned long *[], start_paddr).
  */
 void __init free_area_init_node(int nid, pg_data_t *pgdat, struct page *pmap,
-	unsigned long *zones_size, unsigned long zone_start_pfn, 
+	unsigned long *zones_size, unsigned long node_start_pfn, 
 	unsigned long *zholes_size)
 {
-	free_area_init_core(0, &contig_page_data, &mem_map, zones_size, 
-				zone_start_pfn, zholes_size, pmap);
+	unsigned long size;
+
+	contig_page_data.node_id = 0;
+	contig_page_data.node_start_pfn = node_start_pfn;
+	calculate_totalpages (&contig_page_data, zones_size, zholes_size);
+	if (pmap == (struct page *)0) {
+		size = (pgdat->node_size + 1) * sizeof(struct page);
+		pmap = (struct page *) alloc_bootmem_node(pgdat, size);
+	}
+	contig_page_data.node_mem_map = pmap;
+	free_area_init_core(&contig_page_data, zones_size, zholes_size);
+	mem_map = contig_page_data.node_mem_map;
 }
 
 #endif /* !CONFIG_DISCONTIGMEM */
@@ -48,22 +58,26 @@ struct page * alloc_pages_node(int nid, unsigned int gfp_mask, unsigned int orde
  * Nodes can be initialized parallely, in no particular order.
  */
 void __init free_area_init_node(int nid, pg_data_t *pgdat, struct page *pmap,
-	unsigned long *zones_size, unsigned long zone_start_pfn, 
+	unsigned long *zones_size, unsigned long node_start_pfn, 
 	unsigned long *zholes_size)
 {
-	int i, size = 0;
-	struct page *discard;
-
-	if (mem_map == NULL)
-		mem_map = (struct page *)PAGE_OFFSET;
+	int i;
+	unsigned long size;
 
-	free_area_init_core(nid, pgdat, &discard, zones_size, zone_start_pfn,
-					zholes_size, pmap);
 	pgdat->node_id = nid;
+	pgdat->node_start_pfn = node_start_pfn;
+	calculate_totalpages (pgdat, zones_size, zholes_size);
+	if (pmap == (struct page *)0) {
+		size = (pgdat->node_size + 1) * sizeof(struct page); 
+		pmap = (struct page *) alloc_bootmem_node(pgdat, size);
+	}
+	pgdat->node_mem_map = pmap;
+	free_area_init_core(pgdat, zones_size, zholes_size);
 
 	/*
 	 * Get space for the valid bitmap.
 	 */
+	size = 0;
 	for (i = 0; i < MAX_NR_ZONES; i++)
 		size += zones_size[i];
 	size = LONG_ALIGN((size + 7) >> 3);

mm/page_alloc.c

@@ -724,6 +724,23 @@ static inline void build_zonelists(pg_data_t *pgdat)
 	} 
 }
 
+void __init calculate_totalpages (pg_data_t *pgdat, unsigned long *zones_size,
+	unsigned long *zholes_size)
+{
+	unsigned long realtotalpages, totalpages = 0;
+	int i;
+
+	for (i = 0; i < MAX_NR_ZONES; i++)
+		totalpages += zones_size[i];
+	pgdat->node_size = totalpages;
+
+	realtotalpages = totalpages;
+	if (zholes_size)
+		for (i = 0; i < MAX_NR_ZONES; i++)
+			realtotalpages -= zholes_size[i];
+	printk("On node %d totalpages: %lu\n", pgdat->node_id, realtotalpages);
+}
+
 /*
  * Helper functions to size the waitqueue hash table.
  * Essentially these want to choose hash table sizes sufficiently
@@ -774,46 +791,18 @@ static inline unsigned long wait_table_bits(unsigned long size)
  *   - mark all memory queues empty
  *   - clear the memory bitmaps
  */
-void __init free_area_init_core(int nid, pg_data_t *pgdat, struct page **gmap,
-	unsigned long *zones_size, unsigned long zone_start_pfn, 
-	unsigned long *zholes_size, struct page *lmem_map)
+void __init free_area_init_core(pg_data_t *pgdat,
+		unsigned long *zones_size, unsigned long *zholes_size)
 {
 	unsigned long i, j;
-	unsigned long map_size;
-	unsigned long totalpages, offset, realtotalpages;
+	unsigned long local_offset;
 	const unsigned long zone_required_alignment = 1UL << (MAX_ORDER-1);
+	int nid = pgdat->node_id;
+	struct page *lmem_map = pgdat->node_mem_map;
+	unsigned long zone_start_pfn = pgdat->node_start_pfn;
 
-	totalpages = 0;
-	for (i = 0; i < MAX_NR_ZONES; i++)
-		totalpages += zones_size[i];
-
-	realtotalpages = totalpages;
-	if (zholes_size)
-		for (i = 0; i < MAX_NR_ZONES; i++)
-			realtotalpages -= zholes_size[i];
-			
-	printk("On node %d totalpages: %lu\n", nid, realtotalpages);
-
-	/*
-	 * Some architectures (with lots of mem and discontinous memory
-	 * maps) have to search for a good mem_map area:
-	 * For discontigmem, the conceptual mem map array starts from 
-	 * PAGE_OFFSET, we need to align the actual array onto a mem map 
-	 * boundary, so that MAP_NR works.
-	 */
-	map_size = (totalpages + 1)*sizeof(struct page);
-	if (lmem_map == (struct page *)0) {
-		lmem_map = (struct page *) alloc_bootmem_node(pgdat, map_size);
-		lmem_map = (struct page *)(PAGE_OFFSET + 
-			MAP_ALIGN((unsigned long)lmem_map - PAGE_OFFSET));
-	}
-	*gmap = pgdat->node_mem_map = lmem_map;
-	pgdat->node_size = totalpages;
-	pgdat->node_start_pfn = zone_start_pfn;
-	pgdat->node_start_mapnr = (lmem_map - mem_map);
 	pgdat->nr_zones = 0;
-
-	offset = lmem_map - mem_map;	
+	local_offset = 0;                /* offset within lmem_map */
 	for (j = 0; j < MAX_NR_ZONES; j++) {
 		struct zone *zone = pgdat->node_zones + j;
 		unsigned long mask;
@@ -865,8 +854,7 @@ void __init free_area_init_core(int nid, pg_data_t *pgdat, struct page **gmap,
 		zone->pages_low = mask*2;
 		zone->pages_high = mask*3;
 
-		zone->zone_mem_map = mem_map + offset;
-		zone->zone_start_mapnr = offset;
+		zone->zone_mem_map = lmem_map + local_offset;
 		zone->zone_start_pfn = zone_start_pfn;
 
 		if ((zone_start_pfn) & (zone_required_alignment-1))
@@ -878,7 +866,7 @@ void __init free_area_init_core(int nid, pg_data_t *pgdat, struct page **gmap,
 		 * done. Non-atomic initialization, single-pass.
 		 */
 		for (i = 0; i < size; i++) {
-			struct page *page = mem_map + offset + i;
+			struct page *page = lmem_map + local_offset + i;
 			set_page_zone(page, nid * MAX_NR_ZONES + j);
 			set_page_count(page, 0);
 			SetPageReserved(page);
@@ -892,7 +880,7 @@ void __init free_area_init_core(int nid, pg_data_t *pgdat, struct page **gmap,
 			zone_start_pfn++;
 		}
 
-		offset += size;
+		local_offset += size;
 		for (i = 0; ; i++) {
 			unsigned long bitmap_size;
 
@@ -934,10 +922,13 @@ void __init free_area_init_core(int nid, pg_data_t *pgdat, struct page **gmap,
 	build_zonelists(pgdat);
 }
 
+#ifndef CONFIG_DISCONTIGMEM
 void __init free_area_init(unsigned long *zones_size)
 {
-	free_area_init_core(0, &contig_page_data, &mem_map, zones_size, 0, 0, 0);
+	free_area_init_node(0, &contig_page_data, NULL, zones_size, 0, NULL);
+	mem_map = contig_page_data.node_mem_map;
 }
+#endif
 
 static int __init setup_mem_frac(char *str)
 {