[deliverable/linux.git] / arch / x86_64 / mm / srat.c

/*
 * ACPI 3.0 based NUMA setup
 * Copyright 2004 Andi Kleen, SuSE Labs.
 *
 * Reads the ACPI SRAT table to figure out what memory belongs to which CPUs.
 *
 * Called from acpi_numa_init while reading the SRAT and SLIT tables.
 * Assumes all memory regions belonging to a single proximity domain
 * are in one chunk. Holes between them will be included in the node.
 */

#include <linux/kernel.h>
#include <linux/acpi.h>
#include <linux/mmzone.h>
#include <linux/bitmap.h>
#include <linux/module.h>
#include <linux/topology.h>
#include <asm/proto.h>
#include <asm/numa.h>

static struct acpi_table_slit *acpi_slit;

/* Internal processor count */
static unsigned int __initdata num_processors = 0;

static nodemask_t nodes_parsed __initdata;
static nodemask_t nodes_found __initdata;
static struct node nodes[MAX_NUMNODES] __initdata;
static __u8  pxm2node[256] = { [0 ... 255] = 0xff };

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];
}

static __init int conflicting_nodes(unsigned long start, unsigned long end)
{
	int i;
	for_each_online_node(i) {
		struct node *nd = &nodes[i];
		if (nd->start == nd->end)
			continue;
		if (nd->end > start && nd->start < end)
			return 1;
		if (nd->end == end && nd->start == start)
			return 1;
	}
	return -1;
}

static __init void cutoff_node(int i, unsigned long start, unsigned long end)
{
	struct node *nd = &nodes[i];
	if (nd->start < start) {
		nd->start = start;
		if (nd->end < nd->start)
			nd->start = nd->end;
	}
	if (nd->end > end) {
		if (!(end & 0xfff))
			end--;
		nd->end = end;
		if (nd->start > nd->end)
			nd->start = nd->end;
	}
}

static __init void bad_srat(void)
{
	printk(KERN_ERR "SRAT: SRAT not used.\n");
	acpi_numa = -1;
}

static __init inline int srat_disabled(void)
{
	return numa_off || acpi_numa < 0;
}

/* Callback for SLIT parsing */
void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
{
	acpi_slit = slit;
}

/* Callback for Proximity Domain -> LAPIC mapping */
void __init
acpi_numa_processor_affinity_init(struct acpi_table_processor_affinity *pa)
{
	int pxm, node;
	if (srat_disabled() || pa->flags.enabled == 0)
		return;
	pxm = pa->proximity_domain;
	node = setup_node(pxm);
	if (node < 0) {
		printk(KERN_ERR "SRAT: Too many proximity domains %x\n", pxm);
		bad_srat();
		return;
	}
	if (num_processors >= NR_CPUS) {
		printk(KERN_ERR "SRAT: Processor #%d (lapic %u) INVALID. (Max ID: %d).\n",
			num_processors, pa->apic_id, NR_CPUS);
		bad_srat();
		return;
	}
	cpu_to_node[num_processors] = node;
	acpi_numa = 1;
	printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> CPU %u -> Node %u\n",
	       pxm, pa->apic_id, num_processors, node);

	num_processors++;
}

/* Callback for parsing of the Proximity Domain <-> Memory Area mappings */
void __init
acpi_numa_memory_affinity_init(struct acpi_table_memory_affinity *ma)
{
	struct node *nd;
	unsigned long start, end;
	int node, pxm;
	int i;

	if (srat_disabled() || ma->flags.enabled == 0)
		return;
	pxm = ma->proximity_domain;
	node = setup_node(pxm);
	if (node < 0) {
		printk(KERN_ERR "SRAT: Too many proximity domains.\n");
		bad_srat();
		return;
	}
	start = ma->base_addr_lo | ((u64)ma->base_addr_hi << 32);
	end = start + (ma->length_lo | ((u64)ma->length_hi << 32));
	/* It is fine to add this area to the nodes data it will be used later*/
	if (ma->flags.hot_pluggable == 1)
		printk(KERN_INFO "SRAT: hot plug zone found %lx - %lx \n",
				start, end);
	i = conflicting_nodes(start, end);
	if (i >= 0) {
		printk(KERN_ERR
		       "SRAT: pxm %d overlap %lx-%lx with node %d(%Lx-%Lx)\n",
		       pxm, start, end, i, nodes[i].start, nodes[i].end);
		bad_srat();
		return;
	}
	nd = &nodes[node];
	if (!node_test_and_set(node, nodes_parsed)) {
		nd->start = start;
		nd->end = end;
	} else {
		if (start < nd->start)
			nd->start = start;
		if (nd->end < end)
			nd->end = end;
	}
	if (!(nd->end & 0xfff))
		nd->end--;
	printk(KERN_INFO "SRAT: Node %u PXM %u %Lx-%Lx\n", node, pxm,
	       nd->start, nd->end);
}

void __init acpi_numa_arch_fixup(void) {}

/* Use the information discovered above to actually set up the nodes. */
int __init acpi_scan_nodes(unsigned long start, unsigned long end)
{
	int i;
	if (acpi_numa <= 0)
		return -1;
	memnode_shift = compute_hash_shift(nodes, nodes_weight(nodes_parsed));
	if (memnode_shift < 0) {
		printk(KERN_ERR
		     "SRAT: No NUMA node hash function found. Contact maintainer\n");
		bad_srat();
		return -1;
	}
	for (i = 0; i < MAX_NUMNODES; i++) {
		if (!node_isset(i, nodes_parsed))
			continue;
		cutoff_node(i, start, end);
		if (nodes[i].start == nodes[i].end) { 
			node_clear(i, nodes_parsed);
			continue;
		}
		setup_node_bootmem(i, nodes[i].start, nodes[i].end);
	}
	for (i = 0; i < NR_CPUS; i++) { 
		if (cpu_to_node[i] == NUMA_NO_NODE)
			continue;
		if (!node_isset(cpu_to_node[i], nodes_parsed))
			cpu_to_node[i] = NUMA_NO_NODE; 
	}
	numa_init_array();
	return 0;
}

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;
}

int __node_distance(int a, int b)
{
	int index;

	if (!acpi_slit)
		return a == b ? 10 : 20;
	index = acpi_slit->localities * node_to_pxm(a);
	return acpi_slit->entry[index + node_to_pxm(b)];
}

EXPORT_SYMBOL(__node_distance);
Commit	Line	Data
1da177e4 LT	1	/*
	2	* ACPI 3.0 based NUMA setup
	3	* Copyright 2004 Andi Kleen, SuSE Labs.
	4	*
	5	* Reads the ACPI SRAT table to figure out what memory belongs to which CPUs.
	6	*
	7	* Called from acpi_numa_init while reading the SRAT and SLIT tables.
	8	* Assumes all memory regions belonging to a single proximity domain
	9	* are in one chunk. Holes between them will be included in the node.
	10	*/
	11
	12	#include <linux/kernel.h>
	13	#include <linux/acpi.h>
	14	#include <linux/mmzone.h>
	15	#include <linux/bitmap.h>
	16	#include <linux/module.h>
	17	#include <linux/topology.h>
	18	#include <asm/proto.h>
	19	#include <asm/numa.h>
	20
	21	static struct acpi_table_slit *acpi_slit;
	22
8c56ac3f RT	23	/* Internal processor count */
	24	static unsigned int __initdata num_processors = 0;
	25
1da177e4 LT	26	static nodemask_t nodes_parsed __initdata;
	27	static nodemask_t nodes_found __initdata;
	28	static struct node nodes[MAX_NUMNODES] __initdata;
	29	static __u8 pxm2node[256] = { [0 ... 255] = 0xff };
	30
	31	static __init int setup_node(int pxm)
	32	{
	33	unsigned node = pxm2node[pxm];
	34	if (node == 0xff) {
	35	if (nodes_weight(nodes_found) >= MAX_NUMNODES)
	36	return -1;
	37	node = first_unset_node(nodes_found);
	38	node_set(node, nodes_found);
	39	pxm2node[pxm] = node;
	40	}
	41	return pxm2node[pxm];
	42	}
	43
	44	static __init int conflicting_nodes(unsigned long start, unsigned long end)
	45	{
	46	int i;
	47	for_each_online_node(i) {
	48	struct node *nd = &nodes[i];
	49	if (nd->start == nd->end)
	50	continue;
	51	if (nd->end > start && nd->start < end)
	52	return 1;
	53	if (nd->end == end && nd->start == start)
	54	return 1;
	55	}
	56	return -1;
	57	}
	58
	59	static __init void cutoff_node(int i, unsigned long start, unsigned long end)
	60	{
	61	struct node *nd = &nodes[i];
	62	if (nd->start < start) {
	63	nd->start = start;
	64	if (nd->end < nd->start)
	65	nd->start = nd->end;
	66	}
	67	if (nd->end > end) {
	68	if (!(end & 0xfff))
	69	end--;
	70	nd->end = end;
	71	if (nd->start > nd->end)
	72	nd->start = nd->end;
	73	}
	74	}
	75
	76	static __init void bad_srat(void)
	77	{
	78	printk(KERN_ERR "SRAT: SRAT not used.\n");
	79	acpi_numa = -1;
	80	}
	81
	82	static __init inline int srat_disabled(void)
	83	{
	84	return numa_off \|\| acpi_numa < 0;
	85	}
	86
	87	/* Callback for SLIT parsing */
	88	void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
	89	{
90	acpi_slit = slit;
91	}
92
93	/* Callback for Proximity Domain -> LAPIC mapping */
94	void __init
95	acpi_numa_processor_affinity_init(struct acpi_table_processor_affinity *pa)
96	{
97	int pxm, node;
98	if (srat_disabled() \|\| pa->flags.enabled == 0)
99	return;
100	pxm = pa->proximity_domain;
101	node = setup_node(pxm);
102	if (node < 0) {
103	printk(KERN_ERR "SRAT: Too many proximity domains %x\n", pxm);
104	bad_srat();
105	return;
106	}
8c56ac3f RT	107	if (num_processors >= NR_CPUS) {
	108	printk(KERN_ERR "SRAT: Processor #%d (lapic %u) INVALID. (Max ID: %d).\n",
	109	num_processors, pa->apic_id, NR_CPUS);
1da177e4 LT	110	bad_srat();
	111	return;
	112	}
8c56ac3f	113	cpu_to_node[num_processors] = node;
1da177e4	114	acpi_numa = 1;
8c56ac3f RT	115	printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> CPU %u -> Node %u\n",
	116	pxm, pa->apic_id, num_processors, node);
	117
	118	num_processors++;
1da177e4 LT	119	}
	120
	121	/* Callback for parsing of the Proximity Domain <-> Memory Area mappings */
	122	void __init
	123	acpi_numa_memory_affinity_init(struct acpi_table_memory_affinity *ma)
	124	{
	125	struct node *nd;
	126	unsigned long start, end;
	127	int node, pxm;
	128	int i;
	129
	130	if (srat_disabled() \|\| ma->flags.enabled == 0)
	131	return;
1da177e4 LT	132	pxm = ma->proximity_domain;
	133	node = setup_node(pxm);
	134	if (node < 0) {
	135	printk(KERN_ERR "SRAT: Too many proximity domains.\n");
	136	bad_srat();
	137	return;
	138	}
	139	start = ma->base_addr_lo \| ((u64)ma->base_addr_hi << 32);
	140	end = start + (ma->length_lo \| ((u64)ma->length_hi << 32));
69cb62eb AK	141	/* It is fine to add this area to the nodes data it will be used later*/
	142	if (ma->flags.hot_pluggable == 1)
	143	printk(KERN_INFO "SRAT: hot plug zone found %lx - %lx \n",
	144	start, end);
1da177e4 LT	145	i = conflicting_nodes(start, end);
	146	if (i >= 0) {
	147	printk(KERN_ERR
	148	"SRAT: pxm %d overlap %lx-%lx with node %d(%Lx-%Lx)\n",
	149	pxm, start, end, i, nodes[i].start, nodes[i].end);
	150	bad_srat();
	151	return;
	152	}
	153	nd = &nodes[node];
	154	if (!node_test_and_set(node, nodes_parsed)) {
	155	nd->start = start;
	156	nd->end = end;
	157	} else {
	158	if (start < nd->start)
	159	nd->start = start;
	160	if (nd->end < end)
	161	nd->end = end;
	162	}
	163	if (!(nd->end & 0xfff))
	164	nd->end--;
	165	printk(KERN_INFO "SRAT: Node %u PXM %u %Lx-%Lx\n", node, pxm,
	166	nd->start, nd->end);
	167	}
	168
	169	void __init acpi_numa_arch_fixup(void) {}
	170
	171	/* Use the information discovered above to actually set up the nodes. */
	172	int __init acpi_scan_nodes(unsigned long start, unsigned long end)
	173	{
	174	int i;
	175	if (acpi_numa <= 0)
	176	return -1;
	177	memnode_shift = compute_hash_shift(nodes, nodes_weight(nodes_parsed));
	178	if (memnode_shift < 0) {
	179	printk(KERN_ERR
	180	"SRAT: No NUMA node hash function found. Contact maintainer\n");
	181	bad_srat();
	182	return -1;
	183	}
	184	for (i = 0; i < MAX_NUMNODES; i++) {
	185	if (!node_isset(i, nodes_parsed))
	186	continue;
	187	cutoff_node(i, start, end);
	188	if (nodes[i].start == nodes[i].end) {
	189	node_clear(i, nodes_parsed);
	190	continue;
	191	}
	192	setup_node_bootmem(i, nodes[i].start, nodes[i].end);
	193	}
	194	for (i = 0; i < NR_CPUS; i++) {
	195	if (cpu_to_node[i] == NUMA_NO_NODE)
	196	continue;
	197	if (!node_isset(cpu_to_node[i], nodes_parsed))
	198	cpu_to_node[i] = NUMA_NO_NODE;
	199	}
	200	numa_init_array();
	201	return 0;
	202	}
	203
	204	int node_to_pxm(int n)
	205	{
	206	int i;
	207	if (pxm2node[n] == n)
	208	return n;
209	for (i = 0; i < 256; i++)
210	if (pxm2node[i] == n)
211	return i;
212	return 0;
213	}
214
215	int __node_distance(int a, int b)
216	{
217	int index;
218
219	if (!acpi_slit)
220	return a == b ? 10 : 20;
221	index = acpi_slit->localities * node_to_pxm(a);
222	return acpi_slit->entry[index + node_to_pxm(b)];
223	}
224
225	EXPORT_SYMBOL(__node_distance);