[PATCH] Unify pxm_to_node() and node_to_pxm()

Consolidate the various arch-specific implementations of pxm_to_node() and
node_to_pxm() into a single generic version.

Signed-off-by: Yasunori Goto <[email protected]>
Cc: "Luck, Tony" <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Dave Hansen <[email protected]>
Cc: "Brown, Len" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Author: Yasunori Goto

arch/i386/Kconfig

@@ -173,6 +173,12 @@ config ACPI_SRAT
 	bool
 	default y
 	depends on NUMA && (X86_SUMMIT || X86_GENERICARCH)
+	select ACPI_NUMA
+
+config HAVE_ARCH_PARSE_SRAT
+       bool
+       default y
+       depends on ACPI_SRAT
 
 config X86_SUMMIT_NUMA
 	bool

arch/i386/kernel/srat.c

@@ -39,7 +39,6 @@
 #define NODE_ARRAY_OFFSET(x)	((x) % 8)	/* 8 bits/char */
 #define BMAP_SET(bmap, bit)	((bmap)[NODE_ARRAY_INDEX(bit)] |= 1 << NODE_ARRAY_OFFSET(bit))
 #define BMAP_TEST(bmap, bit)	((bmap)[NODE_ARRAY_INDEX(bit)] & (1 << NODE_ARRAY_OFFSET(bit)))
-#define MAX_PXM_DOMAINS		256	/* 1 byte and no promises about values */
 /* bitmap length; _PXM is at most 255 */
 #define PXM_BITMAP_LEN (MAX_PXM_DOMAINS / 8) 
 static u8 pxm_bitmap[PXM_BITMAP_LEN];	/* bitmap of proximity domains */
@@ -213,19 +212,11 @@ static __init void node_read_chunk(int nid, struct node_memory_chunk_s *memory_c
 		node_end_pfn[nid] = memory_chunk->end_pfn;
 }
 
-static u8 pxm_to_nid_map[MAX_PXM_DOMAINS];/* _PXM to logical node ID map */
-
-int pxm_to_node(int pxm)
-{
-	return pxm_to_nid_map[pxm];
-}
-
 /* Parse the ACPI Static Resource Affinity Table */
 static int __init acpi20_parse_srat(struct acpi_table_srat *sratp)
 {
 	u8 *start, *end, *p;
 	int i, j, nid;
-	u8 nid_to_pxm_map[MAX_NUMNODES];/* logical node ID to _PXM map */
 
 	start = (u8 *)(&(sratp->reserved) + 1);	/* skip header */
 	p = start;
@@ -235,10 +226,6 @@ static int __init acpi20_parse_srat(struct acpi_table_srat *sratp)
 	memset(node_memory_chunk, 0, sizeof(node_memory_chunk));
 	memset(zholes_size, 0, sizeof(zholes_size));
 
-	/* -1 in these maps means not available */
-	memset(pxm_to_nid_map, -1, sizeof(pxm_to_nid_map));
-	memset(nid_to_pxm_map, -1, sizeof(nid_to_pxm_map));
-
 	num_memory_chunks = 0;
 	while (p < end) {
 		switch (*p) {
@@ -278,17 +265,15 @@ static int __init acpi20_parse_srat(struct acpi_table_srat *sratp)
 	nodes_clear(node_online_map);
 	for (i = 0; i < MAX_PXM_DOMAINS; i++) {
 		if (BMAP_TEST(pxm_bitmap, i)) {
-			nid = num_online_nodes();
-			pxm_to_nid_map[i] = nid;
-			nid_to_pxm_map[nid] = i;
+			int nid = acpi_map_pxm_to_node(i);
 			node_set_online(nid);
 		}
 	}
 	BUG_ON(num_online_nodes() == 0);
 
 	/* set cnode id in memory chunk structure */
 	for (i = 0; i < num_memory_chunks; i++)
-		node_memory_chunk[i].nid = pxm_to_nid_map[node_memory_chunk[i].pxm];
+		node_memory_chunk[i].nid = pxm_to_node(node_memory_chunk[i].pxm);
 
 	printk("pxm bitmap: ");
 	for (i = 0; i < sizeof(pxm_bitmap); i++) {

arch/ia64/hp/common/sba_iommu.c

@@ -1958,7 +1958,7 @@ sba_map_ioc_to_node(struct ioc *ioc, acpi_handle handle)
 	if (pxm < 0)
 		return;
 
-	node = pxm_to_nid_map[pxm];
+	node = pxm_to_node(pxm);
 
 	if (node >= MAX_NUMNODES || !node_online(node))
 		return;

arch/ia64/kernel/acpi.c

@@ -415,9 +415,6 @@ static int __initdata srat_num_cpus;	/* number of cpus */
 static u32 __devinitdata pxm_flag[PXM_FLAG_LEN];
 #define pxm_bit_set(bit)	(set_bit(bit,(void *)pxm_flag))
 #define pxm_bit_test(bit)	(test_bit(bit,(void *)pxm_flag))
-/* maps to convert between proximity domain and logical node ID */
-int __devinitdata pxm_to_nid_map[MAX_PXM_DOMAINS];
-int __initdata nid_to_pxm_map[MAX_NUMNODES];
 static struct acpi_table_slit __initdata *slit_table;
 
 static int get_processor_proximity_domain(struct acpi_table_processor_affinity *pa)
@@ -533,22 +530,17 @@ void __init acpi_numa_arch_fixup(void)
 	 * MCD - This can probably be dropped now.  No need for pxm ID to node ID
 	 * mapping with sparse node numbering iff MAX_PXM_DOMAINS <= MAX_NUMNODES.
 	 */
-	/* calculate total number of nodes in system from PXM bitmap */
-	memset(pxm_to_nid_map, -1, sizeof(pxm_to_nid_map));
-	memset(nid_to_pxm_map, -1, sizeof(nid_to_pxm_map));
 	nodes_clear(node_online_map);
 	for (i = 0; i < MAX_PXM_DOMAINS; i++) {
 		if (pxm_bit_test(i)) {
-			int nid = num_online_nodes();
-			pxm_to_nid_map[i] = nid;
-			nid_to_pxm_map[nid] = i;
+			int nid = acpi_map_pxm_to_node(i);
 			node_set_online(nid);
 		}
 	}
 
 	/* set logical node id in memory chunk structure */
 	for (i = 0; i < num_node_memblks; i++)
-		node_memblk[i].nid = pxm_to_nid_map[node_memblk[i].nid];
+		node_memblk[i].nid = pxm_to_node(node_memblk[i].nid);
 
 	/* assign memory bank numbers for each chunk on each node */
 	for_each_online_node(i) {
@@ -562,7 +554,7 @@ void __init acpi_numa_arch_fixup(void)
 
 	/* set logical node id in cpu structure */
 	for (i = 0; i < srat_num_cpus; i++)
-		node_cpuid[i].nid = pxm_to_nid_map[node_cpuid[i].nid];
+		node_cpuid[i].nid = pxm_to_node(node_cpuid[i].nid);
 
 	printk(KERN_INFO "Number of logical nodes in system = %d\n",
 	       num_online_nodes());
@@ -575,11 +567,11 @@ void __init acpi_numa_arch_fixup(void)
 	for (i = 0; i < slit_table->localities; i++) {
 		if (!pxm_bit_test(i))
 			continue;
-		node_from = pxm_to_nid_map[i];
+		node_from = pxm_to_node(i);
 		for (j = 0; j < slit_table->localities; j++) {
 			if (!pxm_bit_test(j))
 				continue;
-			node_to = pxm_to_nid_map[j];
+			node_to = pxm_to_node(j);
 			node_distance(node_from, node_to) =
 			    slit_table->entry[i * slit_table->localities + j];
 		}
@@ -785,9 +777,9 @@ int acpi_map_cpu2node(acpi_handle handle, int cpu, long physid)
 
 	/*
 	 * Assuming that the container driver would have set the proximity
-	 * domain and would have initialized pxm_to_nid_map[pxm_id] && pxm_flag
+	 * domain and would have initialized pxm_to_node(pxm_id) && pxm_flag
 	 */
-	node_cpuid[cpu].nid = (pxm_id < 0) ? 0 : pxm_to_nid_map[pxm_id];
+	node_cpuid[cpu].nid = (pxm_id < 0) ? 0 : pxm_to_node(pxm_id);
 
 	node_cpuid[cpu].phys_id = physid;
 #endif
@@ -966,7 +958,7 @@ acpi_map_iosapic(acpi_handle handle, u32 depth, void *context, void **ret)
 	if (pxm < 0)
 		return AE_OK;
 
-	node = pxm_to_nid_map[pxm];
+	node = pxm_to_node(pxm);
 
 	if (node >= MAX_NUMNODES || !node_online(node) ||
 	    cpus_empty(node_to_cpumask(node)))

arch/ia64/pci/pci.c

@@ -352,7 +352,7 @@ pci_acpi_scan_root(struct acpi_device *device, int domain, int bus)
 	pxm = acpi_get_pxm(controller->acpi_handle);
 #ifdef CONFIG_NUMA
 	if (pxm >= 0)
-		controller->node = pxm_to_nid_map[pxm];
+		controller->node = pxm_to_node(pxm);
 #endif
 
 	acpi_walk_resources(device->handle, METHOD_NAME__CRS, count_window,

arch/ia64/sn/kernel/setup.c

@@ -139,7 +139,7 @@ static int __init pxm_to_nasid(int pxm)
 	int i;
 	int nid;
 
-	nid = pxm_to_nid_map[pxm];
+	nid = pxm_to_node(pxm);
 	for (i = 0; i < num_node_memblks; i++) {
 		if (node_memblk[i].nid == nid) {
 			return NASID_GET(node_memblk[i].start_paddr);
@@ -704,7 +704,7 @@ void __init build_cnode_tables(void)
 	 * cnode == node for all C & M bricks.
 	 */
 	for_each_online_node(node) {
-		nasid = pxm_to_nasid(nid_to_pxm_map[node]);
+		nasid = pxm_to_nasid(node_to_pxm(node));
 		sn_cnodeid_to_nasid[node] = nasid;
 		physical_node_map[nasid] = node;
 	}

arch/x86_64/mm/srat.c

@@ -30,41 +30,21 @@
 static struct acpi_table_slit *acpi_slit;
 
 static nodemask_t nodes_parsed __initdata;
-static nodemask_t nodes_found __initdata;
 static struct bootnode nodes[MAX_NUMNODES] __initdata;
 static struct bootnode nodes_add[MAX_NUMNODES] __initdata;
 static int found_add_area __initdata;
 int hotadd_percent __initdata = 0;
 #ifndef RESERVE_HOTADD
 #define hotadd_percent 0	/* Ignore all settings */
 #endif
-static u8 pxm2node[256] = { [0 ... 255] = 0xff };
 
 /* Too small nodes confuse the VM badly. Usually they result
    from BIOS bugs. */
 #define NODE_MIN_SIZE (4*1024*1024)
 
-static int node_to_pxm(int n);
-
-int pxm_to_node(int pxm)
-{
-	if ((unsigned)pxm >= 256)
-		return -1;
-	/* Extend 0xff to (int)-1 */
-	return (signed char)pxm2node[pxm];
-}
-
 static __init int setup_node(int pxm)
 {
-	unsigned node = pxm2node[pxm];
-	if (node == 0xff) {
-		if (nodes_weight(nodes_found) >= MAX_NUMNODES)
-			return -1;
-		node = first_unset_node(nodes_found); 
-		node_set(node, nodes_found);
-		pxm2node[pxm] = node;
-	}
-	return pxm2node[pxm];
+	return acpi_map_pxm_to_node(pxm);
 }
 
 static __init int conflicting_nodes(unsigned long start, unsigned long end)
@@ -440,17 +420,6 @@ int __init acpi_scan_nodes(unsigned long start, unsigned long end)
 	return 0;
 }
 
-static int node_to_pxm(int n)
-{
-       int i;
-       if (pxm2node[n] == n)
-               return n;
-       for (i = 0; i < 256; i++)
-               if (pxm2node[i] == n)
-                       return i;
-       return 0;
-}
-
 void __init srat_reserve_add_area(int nodeid)
 {
 	if (found_add_area && nodes_add[nodeid].end) {

drivers/acpi/Kconfig

@@ -162,7 +162,7 @@ config ACPI_THERMAL
 config ACPI_NUMA
 	bool "NUMA support"
 	depends on NUMA
-	depends on (IA64 || X86_64)
+	depends on (X86 || IA64)
 	default y if IA64_GENERIC || IA64_SGI_SN2
 
 config ACPI_ASUS

drivers/acpi/numa.c

@@ -36,12 +36,60 @@
 #define _COMPONENT	ACPI_NUMA
 ACPI_MODULE_NAME("numa")
 
+static nodemask_t nodes_found_map = NODE_MASK_NONE;
+#define PXM_INVAL	-1
+#define NID_INVAL	-1
+
+/* maps to convert between proximity domain and logical node ID */
+int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS]
+				= { [0 ... MAX_PXM_DOMAINS - 1] = NID_INVAL };
+int __cpuinitdata node_to_pxm_map[MAX_NUMNODES]
+				= { [0 ... MAX_NUMNODES - 1] = PXM_INVAL };
+
 extern int __init acpi_table_parse_madt_family(enum acpi_table_id id,
 					       unsigned long madt_size,
 					       int entry_id,
 					       acpi_madt_entry_handler handler,
 					       unsigned int max_entries);
 
+int __cpuinit pxm_to_node(int pxm)
+{
+	if (pxm < 0)
+		return NID_INVAL;
+	return pxm_to_node_map[pxm];
+}
+
+int __cpuinit node_to_pxm(int node)
+{
+	if (node < 0)
+		return PXM_INVAL;
+	return node_to_pxm_map[node];
+}
+
+int __cpuinit acpi_map_pxm_to_node(int pxm)
+{
+	int node = pxm_to_node_map[pxm];
+
+	if (node < 0){
+		if (nodes_weight(nodes_found_map) >= MAX_NUMNODES)
+			return NID_INVAL;
+		node = first_unset_node(nodes_found_map);
+		pxm_to_node_map[pxm] = node;
+		node_to_pxm_map[node] = pxm;
+		node_set(node, nodes_found_map);
+	}
+
+	return node;
+}
+
+void __cpuinit acpi_unmap_pxm_to_node(int node)
+{
+	int pxm = node_to_pxm_map[node];
+	pxm_to_node_map[pxm] = NID_INVAL;
+	node_to_pxm_map[node] = PXM_INVAL;
+	node_clear(node, nodes_found_map);
+}
+
 void __init acpi_table_print_srat_entry(acpi_table_entry_header * header)
 {
 

include/acpi/acpi_numa.h

@@ -0,0 +1,23 @@
+#ifndef __ACPI_NUMA_H
+#define __ACPI_NUMA_H
+
+#ifdef CONFIG_ACPI_NUMA
+#include <linux/kernel.h>
+
+/* Proximity bitmap length */
+#if MAX_NUMNODES > 256
+#define MAX_PXM_DOMAINS MAX_NUMNODES
+#else
+#define MAX_PXM_DOMAINS (256) /* Old pxm spec is defined 8 bit */
+#endif
+
+extern int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS];
+extern int __cpuinitdata node_to_pxm_map[MAX_NUMNODES];
+
+extern int __cpuinit pxm_to_node(int);
+extern int __cpuinit node_to_pxm(int);
+extern int __cpuinit acpi_map_pxm_to_node(int);
+extern void __cpuinit acpi_unmap_pxm_to_node(int);
+
+#endif				/* CONFIG_ACPI_NUMA */
+#endif				/* __ACP_NUMA_H */

include/asm-x86_64/numa.h

@@ -8,7 +8,6 @@ struct bootnode {
 };
 
 extern int compute_hash_shift(struct bootnode *nodes, int numnodes);
-extern int pxm_to_node(int nid);
 
 #define ZONE_ALIGN (1UL << (MAX_ORDER+PAGE_SHIFT))
 

include/linux/acpi.h

@@ -37,6 +37,7 @@
 #include <acpi/acpi.h>
 #include <acpi/acpi_bus.h>
 #include <acpi/acpi_drivers.h>
+#include <acpi/acpi_numa.h>
 #include <asm/acpi.h>
 
 
@@ -407,10 +408,18 @@ void acpi_table_print_madt_entry (acpi_table_entry_header *madt);
 void acpi_table_print_srat_entry (acpi_table_entry_header *srat);
 
 /* the following four functions are architecture-dependent */
+#ifdef CONFIG_HAVE_ARCH_PARSE_SRAT
+#define NR_NODE_MEMBLKS MAX_NUMNODES
+#define acpi_numa_slit_init(slit) do {} while (0)
+#define acpi_numa_processor_affinity_init(pa) do {} while (0)
+#define acpi_numa_memory_affinity_init(ma) do {} while (0)
+#define acpi_numa_arch_fixup() do {} while (0)
+#else
 void acpi_numa_slit_init (struct acpi_table_slit *slit);
 void acpi_numa_processor_affinity_init (struct acpi_table_processor_affinity *pa);
 void acpi_numa_memory_affinity_init (struct acpi_table_memory_affinity *ma);
 void acpi_numa_arch_fixup(void);
+#endif
 
 #ifdef CONFIG_ACPI_HOTPLUG_CPU
 /* Arch dependent functions for cpu hotplug support */