x86, NUMA: Enable emulation on 32bit too

Now that NUMA init path is unified, NUMA emulation can be enabled on
32bit.  Make numa_emluation.c safe on 32bit by doing the followings.

* Define MAX_DMA32_PFN on 32bit too.

* Include bootmem.h for max_pfn declaration.

* Use u64 explicitly and always use PFN_PHYS() when converting page
  number to address.

* Avoid __udivdi3() generation on 32bit by doing number of pages
  calculation instead in split_nodes_interleave().

And drop X86_64 dependency from Kconfig.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>

arch/x86/Kconfig

@@ -1201,7 +1201,7 @@ config NODES_SPAN_OTHER_NODES
 
 config NUMA_EMU
 	bool "NUMA emulation"
-	depends on X86_64 && NUMA
+	depends on NUMA
 	---help---
 	  Enable NUMA emulation. A flat machine will be split
 	  into virtual nodes when booted with "numa=fake=N", where N is the

arch/x86/include/asm/dma.h

@@ -72,19 +72,15 @@
 /* 16MB ISA DMA zone */
 #define MAX_DMA_PFN   ((16 * 1024 * 1024) >> PAGE_SHIFT)
 
-#ifdef CONFIG_X86_32
+/* 4GB broken PCI/AGP hardware bus master zone */
+#define MAX_DMA32_PFN ((4UL * 1024 * 1024 * 1024) >> PAGE_SHIFT)
 
+#ifdef CONFIG_X86_32
 /* The maximum address that we can perform a DMA transfer to on this platform */
 #define MAX_DMA_ADDRESS      (PAGE_OFFSET + 0x1000000)
-
 #else
-
-/* 4GB broken PCI/AGP hardware bus master zone */
-#define MAX_DMA32_PFN ((4UL * 1024 * 1024 * 1024) >> PAGE_SHIFT)
-
 /* Compat define for old dma zone */
 #define MAX_DMA_ADDRESS ((unsigned long)__va(MAX_DMA_PFN << PAGE_SHIFT))
-
 #endif
 
 /* 8237 DMA controllers */

arch/x86/mm/numa_emulation.c

@@ -5,6 +5,7 @@
 #include <linux/errno.h>
 #include <linux/topology.h>
 #include <linux/memblock.h>
+#include <linux/bootmem.h>
 #include <asm/dma.h>
 
 #include "numa_internal.h"
@@ -84,7 +85,13 @@ static int __init split_nodes_interleave(struct numa_meminfo *ei,
 		nr_nodes = MAX_NUMNODES;
 	}
 
-	size = (max_addr - addr - memblock_x86_hole_size(addr, max_addr)) / nr_nodes;
+	/*
+	 * Calculate target node size.  x86_32 freaks on __udivdi3() so do
+	 * the division in ulong number of pages and convert back.
+	 */
+	size = max_addr - addr - memblock_x86_hole_size(addr, max_addr);
+	size = PFN_PHYS((unsigned long)(size >> PAGE_SHIFT) / nr_nodes);
+
 	/*
 	 * Calculate the number of big nodes that can be allocated as a result
 	 * of consolidating the remainder.
@@ -226,7 +233,7 @@ static int __init split_nodes_size_interleave(struct numa_meminfo *ei,
 	 */
 	while (nodes_weight(physnode_mask)) {
 		for_each_node_mask(i, physnode_mask) {
-			u64 dma32_end = MAX_DMA32_PFN << PAGE_SHIFT;
+			u64 dma32_end = PFN_PHYS(MAX_DMA32_PFN);
 			u64 start, limit, end;
 			int phys_blk;
 
@@ -298,7 +305,7 @@ void __init numa_emulation(struct numa_meminfo *numa_meminfo, int numa_dist_cnt)
 {
 	static struct numa_meminfo ei __initdata;
 	static struct numa_meminfo pi __initdata;
-	const u64 max_addr = max_pfn << PAGE_SHIFT;
+	const u64 max_addr = PFN_PHYS(max_pfn);
 	u8 *phys_dist = NULL;
 	size_t phys_size = numa_dist_cnt * numa_dist_cnt * sizeof(phys_dist[0]);
 	int max_emu_nid, dfl_phys_nid;
@@ -342,8 +349,7 @@ void __init numa_emulation(struct numa_meminfo *numa_meminfo, int numa_dist_cnt)
 	if (numa_dist_cnt) {
 		u64 phys;
 
-		phys = memblock_find_in_range(0,
-					      (u64)max_pfn_mapped << PAGE_SHIFT,
+		phys = memblock_find_in_range(0, PFN_PHYS(max_pfn_mapped),
 					      phys_size, PAGE_SIZE);
 		if (phys == MEMBLOCK_ERROR) {
 			pr_warning("NUMA: Warning: can't allocate copy of distance table, disabling emulation\n");