[net-next-2.6.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>
#include <asm/e820.h>

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 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];
}

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

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

static __init void bad_srat(void)
{
	int i;
	printk(KERN_ERR "SRAT: SRAT not used.\n");
	acpi_numa = -1;
	for (i = 0; i < MAX_LOCAL_APIC; i++)
		apicid_to_node[i] = NUMA_NO_NODE;
}

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

/*
 * A lot of BIOS fill in 10 (= no distance) everywhere. This messes
 * up the NUMA heuristics which wants the local node to have a smaller
 * distance than the others.
 * Do some quick checks here and only use the SLIT if it passes.
 */
static __init int slit_valid(struct acpi_table_slit *slit)
{
	int i, j;
	int d = slit->localities;
	for (i = 0; i < d; i++) {
		for (j = 0; j < d; j++)  {
			u8 val = slit->entry[d*i + j];
			if (i == j) {
				if (val != 10)
					return 0;
			} else if (val <= 10)
				return 0;
		}
	}
	return 1;
}

/* Callback for SLIT parsing */
void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
{
	if (!slit_valid(slit)) {
		printk(KERN_INFO "ACPI: SLIT table looks invalid. Not used.\n");
		return;
	}
	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())
		return;
	if (pa->header.length != sizeof(struct acpi_table_processor_affinity)) {		bad_srat();
		return;
	}
	if (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;
	}
	apicid_to_node[pa->apic_id] = node;
	acpi_numa = 1;
	printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n",
	       pxm, pa->apic_id, node);
}

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

	if (srat_disabled())
		return;
	if (ma->header.length != sizeof(struct acpi_table_memory_affinity)) {
		bad_srat();
		return;
	}
	if (ma->flags.enabled == 0)
		return;
	start = ma->base_addr_lo | ((u64)ma->base_addr_hi << 32);
	end = start + (ma->length_lo | ((u64)ma->length_hi << 32));
	pxm = ma->proximity_domain;
	node = setup_node(pxm);
	if (node < 0) {
		printk(KERN_ERR "SRAT: Too many proximity domains.\n");
		bad_srat();
		return;
	}
	/* 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 == node) {
		printk(KERN_WARNING
		"SRAT: Warning: PXM %d (%lx-%lx) overlaps with itself (%Lx-%Lx)\n",
			pxm, start, end, nodes[i].start, nodes[i].end);
	} else if (i >= 0) {
		printk(KERN_ERR
		       "SRAT: PXM %d (%lx-%lx) overlaps with PXM %d (%Lx-%Lx)\n",
		       pxm, start, end, node_to_pxm(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;
	}
	printk(KERN_INFO "SRAT: Node %u PXM %u %Lx-%Lx\n", node, pxm,
	       nd->start, nd->end);
}

/* Sanity check to catch more bad SRATs (they are amazingly common).
   Make sure the PXMs cover all memory. */
static int nodes_cover_memory(void)
{
	int i;
	unsigned long pxmram, e820ram;

	pxmram = 0;
	for_each_node_mask(i, nodes_parsed) {
		unsigned long s = nodes[i].start >> PAGE_SHIFT;
		unsigned long e = nodes[i].end >> PAGE_SHIFT;
		pxmram += e - s;
		pxmram -= e820_hole_size(s, e);
	}

	e820ram = end_pfn - e820_hole_size(0, end_pfn);
	/* We seem to lose 3 pages somewhere. Allow a bit of slack. */
	if ((long)(e820ram - pxmram) >= 1*1024*1024) {
		printk(KERN_ERR
	"SRAT: PXMs only cover %luMB of your %luMB e820 RAM. Not used.\n",
			(pxmram << PAGE_SHIFT) >> 20,
			(e820ram << PAGE_SHIFT) >> 20);
		return 0;
	}
	return 1;
}

static void unparse_node(int node)
{
	int i;
	node_clear(node, nodes_parsed);
	for (i = 0; i < MAX_LOCAL_APIC; i++) {
		if (apicid_to_node[i] == node)
			apicid_to_node[i] = NUMA_NO_NODE;
	}
}

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;

	/* First clean up the node list */
	for (i = 0; i < MAX_NUMNODES; i++) {
		cutoff_node(i, start, end);
		if ((nodes[i].end - nodes[i].start) < NODE_MIN_SIZE)
			unparse_node(i);
	}

	if (acpi_numa <= 0)
		return -1;

	if (!nodes_cover_memory()) {
		bad_srat();
		return -1;
	}

	memnode_shift = compute_hash_shift(nodes, MAX_NUMNODES);
	if (memnode_shift < 0) {
		printk(KERN_ERR
		     "SRAT: No NUMA node hash function found. Contact maintainer\n");
		bad_srat();
		return -1;
	}

	/* Finally register nodes */
	for_each_node_mask(i, nodes_parsed)
		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))
			numa_set_node(i, NUMA_NO_NODE);
	}
	numa_init_array();
	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;
}

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>
8a6fdd3e	20	#include <asm/e820.h>
1da177e4 LT	21
	22	static struct acpi_table_slit *acpi_slit;
	23
	24	static nodemask_t nodes_parsed __initdata;
	25	static nodemask_t nodes_found __initdata;
abe059e7	26	static struct bootnode nodes[MAX_NUMNODES] __initdata;
e4e94072	27	static u8 pxm2node[256] = { [0 ... 255] = 0xff };
1da177e4	28
9391a3f9 AK	29	/* Too small nodes confuse the VM badly. Usually they result
	30	from BIOS bugs. */
	31	#define NODE_MIN_SIZE (410241024)
	32
05d1fa4b AK	33	static int node_to_pxm(int n);
05d1fa4b AK	34
69e1a33f AK	35	int pxm_to_node(int pxm)
	36	{
	37	if ((unsigned)pxm >= 256)
e4e94072 AK	38	return -1;
	39	/* Extend 0xff to (int)-1 */
	40	return (signed char)pxm2node[pxm];
69e1a33f AK	41	}
69e1a33f AK	42
1da177e4 LT	43	static __init int setup_node(int pxm)
	44	{
	45	unsigned node = pxm2node[pxm];
	46	if (node == 0xff) {
	47	if (nodes_weight(nodes_found) >= MAX_NUMNODES)
	48	return -1;
	49	node = first_unset_node(nodes_found);
	50	node_set(node, nodes_found);
	51	pxm2node[pxm] = node;
	52	}
	53	return pxm2node[pxm];
	54	}
	55
	56	static __init int conflicting_nodes(unsigned long start, unsigned long end)
	57	{
	58	int i;
4b6a455c	59	for_each_node_mask(i, nodes_parsed) {
abe059e7	60	struct bootnode *nd = &nodes[i];
1da177e4 LT	61	if (nd->start == nd->end)
	62	continue;
	63	if (nd->end > start && nd->start < end)
05d1fa4b	64	return i;
1da177e4	65	if (nd->end == end && nd->start == start)
05d1fa4b	66	return i;
1da177e4 LT	67	}
	68	return -1;
	69	}
	70
	71	static __init void cutoff_node(int i, unsigned long start, unsigned long end)
	72	{
abe059e7	73	struct bootnode *nd = &nodes[i];
1da177e4 LT	74	if (nd->start < start) {
	75	nd->start = start;
	76	if (nd->end < nd->start)
	77	nd->start = nd->end;
	78	}
	79	if (nd->end > end) {
1da177e4 LT	80	nd->end = end;
	81	if (nd->start > nd->end)
	82	nd->start = nd->end;
	83	}
	84	}
	85
	86	static __init void bad_srat(void)
	87	{
2bce2b54	88	int i;
1da177e4 LT	89	printk(KERN_ERR "SRAT: SRAT not used.\n");
1da177e4 LT	90	acpi_numa = -1;
2bce2b54 AK	91	for (i = 0; i < MAX_LOCAL_APIC; i++)
2bce2b54 AK	92	apicid_to_node[i] = NUMA_NO_NODE;
1da177e4 LT	93	}
	94
	95	static __init inline int srat_disabled(void)
	96	{
	97	return numa_off \|\| acpi_numa < 0;
	98	}
	99
1584b89c AK	100	/*
	101	* A lot of BIOS fill in 10 (= no distance) everywhere. This messes
	102	* up the NUMA heuristics which wants the local node to have a smaller
	103	* distance than the others.
	104	* Do some quick checks here and only use the SLIT if it passes.
	105	*/
	106	static __init int slit_valid(struct acpi_table_slit *slit)
	107	{
	108	int i, j;
	109	int d = slit->localities;
	110	for (i = 0; i < d; i++) {
	111	for (j = 0; j < d; j++) {
	112	u8 val = slit->entry[d*i + j];
	113	if (i == j) {
	114	if (val != 10)
	115	return 0;
	116	} else if (val <= 10)
	117	return 0;
	118	}
	119	}
	120	return 1;
	121	}
	122
1da177e4 LT	123	/* Callback for SLIT parsing */
	124	void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
	125	{
1584b89c AK	126	if (!slit_valid(slit)) {
	127	printk(KERN_INFO "ACPI: SLIT table looks invalid. Not used.\n");
	128	return;
	129	}
1da177e4 LT	130	acpi_slit = slit;
	131	}
	132
	133	/* Callback for Proximity Domain -> LAPIC mapping */
	134	void __init
	135	acpi_numa_processor_affinity_init(struct acpi_table_processor_affinity *pa)
	136	{
	137	int pxm, node;
d22fe808 AK	138	if (srat_disabled())
	139	return;
	140	if (pa->header.length != sizeof(struct acpi_table_processor_affinity)) { bad_srat();
	141	return;
	142	}
	143	if (pa->flags.enabled == 0)
1da177e4 LT	144	return;
	145	pxm = pa->proximity_domain;
	146	node = setup_node(pxm);
	147	if (node < 0) {
	148	printk(KERN_ERR "SRAT: Too many proximity domains %x\n", pxm);
	149	bad_srat();
	150	return;
	151	}
0b07e984	152	apicid_to_node[pa->apic_id] = node;
1da177e4	153	acpi_numa = 1;
0b07e984 AK	154	printk(KERN_INFO "SRAT: PXM %u -> APIC %u -> Node %u\n",
0b07e984 AK	155	pxm, pa->apic_id, node);
1da177e4 LT	156	}
	157
	158	/* Callback for parsing of the Proximity Domain <-> Memory Area mappings */
	159	void __init
	160	acpi_numa_memory_affinity_init(struct acpi_table_memory_affinity *ma)
	161	{
abe059e7	162	struct bootnode *nd;
1da177e4 LT	163	unsigned long start, end;
	164	int node, pxm;
	165	int i;
	166
d22fe808	167	if (srat_disabled())
1da177e4	168	return;
d22fe808 AK	169	if (ma->header.length != sizeof(struct acpi_table_memory_affinity)) {
	170	bad_srat();
	171	return;
	172	}
	173	if (ma->flags.enabled == 0)
	174	return;
	175	start = ma->base_addr_lo \| ((u64)ma->base_addr_hi << 32);
	176	end = start + (ma->length_lo \| ((u64)ma->length_hi << 32));
1da177e4 LT	177	pxm = ma->proximity_domain;
	178	node = setup_node(pxm);
	179	if (node < 0) {
	180	printk(KERN_ERR "SRAT: Too many proximity domains.\n");
	181	bad_srat();
	182	return;
	183	}
69cb62eb AK	184	/* It is fine to add this area to the nodes data it will be used later*/
	185	if (ma->flags.hot_pluggable == 1)
	186	printk(KERN_INFO "SRAT: hot plug zone found %lx - %lx \n",
	187	start, end);
1da177e4	188	i = conflicting_nodes(start, end);
05d1fa4b AK	189	if (i == node) {
	190	printk(KERN_WARNING
	191	"SRAT: Warning: PXM %d (%lx-%lx) overlaps with itself (%Lx-%Lx)\n",
	192	pxm, start, end, nodes[i].start, nodes[i].end);
	193	} else if (i >= 0) {
1da177e4	194	printk(KERN_ERR
05d1fa4b AK	195	"SRAT: PXM %d (%lx-%lx) overlaps with PXM %d (%Lx-%Lx)\n",
	196	pxm, start, end, node_to_pxm(i),
	197	nodes[i].start, nodes[i].end);
1da177e4 LT	198	bad_srat();
	199	return;
	200	}
	201	nd = &nodes[node];
	202	if (!node_test_and_set(node, nodes_parsed)) {
	203	nd->start = start;
	204	nd->end = end;
	205	} else {
	206	if (start < nd->start)
	207	nd->start = start;
	208	if (nd->end < end)
	209	nd->end = end;
	210	}
1da177e4 LT	211	printk(KERN_INFO "SRAT: Node %u PXM %u %Lx-%Lx\n", node, pxm,
	212	nd->start, nd->end);
	213	}
	214
8a6fdd3e AK	215	/* Sanity check to catch more bad SRATs (they are amazingly common).
	216	Make sure the PXMs cover all memory. */
	217	static int nodes_cover_memory(void)
	218	{
	219	int i;
	220	unsigned long pxmram, e820ram;
	221
	222	pxmram = 0;
	223	for_each_node_mask(i, nodes_parsed) {
	224	unsigned long s = nodes[i].start >> PAGE_SHIFT;
	225	unsigned long e = nodes[i].end >> PAGE_SHIFT;
	226	pxmram += e - s;
	227	pxmram -= e820_hole_size(s, e);
	228	}
	229
	230	e820ram = end_pfn - e820_hole_size(0, end_pfn);
fdb9df94 AK	231	/* We seem to lose 3 pages somewhere. Allow a bit of slack. */
fdb9df94 AK	232	if ((long)(e820ram - pxmram) >= 110241024) {
8a6fdd3e AK	233	printk(KERN_ERR
	234	"SRAT: PXMs only cover %luMB of your %luMB e820 RAM. Not used.\n",
	235	(pxmram << PAGE_SHIFT) >> 20,
	236	(e820ram << PAGE_SHIFT) >> 20);
	237	return 0;
	238	}
	239	return 1;
	240	}
	241
9391a3f9 AK	242	static void unparse_node(int node)
	243	{
	244	int i;
	245	node_clear(node, nodes_parsed);
	246	for (i = 0; i < MAX_LOCAL_APIC; i++) {
	247	if (apicid_to_node[i] == node)
	248	apicid_to_node[i] = NUMA_NO_NODE;
	249	}
	250	}
	251
1da177e4 LT	252	void __init acpi_numa_arch_fixup(void) {}
	253
	254	/* Use the information discovered above to actually set up the nodes. */
	255	int __init acpi_scan_nodes(unsigned long start, unsigned long end)
	256	{
	257	int i;
8a6fdd3e	258
e58e0d03	259	/* First clean up the node list */
9391a3f9	260	for (i = 0; i < MAX_NUMNODES; i++) {
e58e0d03	261	cutoff_node(i, start, end);
9391a3f9 AK	262	if ((nodes[i].end - nodes[i].start) < NODE_MIN_SIZE)
9391a3f9 AK	263	unparse_node(i);
e58e0d03 AK	264	}
e58e0d03 AK	265
9391a3f9 AK	266	if (acpi_numa <= 0)
	267	return -1;
	268
8a6fdd3e AK	269	if (!nodes_cover_memory()) {
	270	bad_srat();
	271	return -1;
	272	}
	273
2aed711a	274	memnode_shift = compute_hash_shift(nodes, MAX_NUMNODES);
1da177e4 LT	275	if (memnode_shift < 0) {
	276	printk(KERN_ERR
	277	"SRAT: No NUMA node hash function found. Contact maintainer\n");
	278	bad_srat();
	279	return -1;
	280	}
e58e0d03 AK	281
	282	/* Finally register nodes */
	283	for_each_node_mask(i, nodes_parsed)
1da177e4	284	setup_node_bootmem(i, nodes[i].start, nodes[i].end);
1da177e4 LT	285	for (i = 0; i < NR_CPUS; i++) {
	286	if (cpu_to_node[i] == NUMA_NO_NODE)
	287	continue;
	288	if (!node_isset(cpu_to_node[i], nodes_parsed))
69d81fcd	289	numa_set_node(i, NUMA_NO_NODE);
1da177e4 LT	290	}
	291	numa_init_array();
	292	return 0;
	293	}
	294
05d1fa4b	295	static int node_to_pxm(int n)
1da177e4 LT	296	{
	297	int i;
	298	if (pxm2node[n] == n)
	299	return n;
	300	for (i = 0; i < 256; i++)
	301	if (pxm2node[i] == n)
	302	return i;
	303	return 0;
	304	}
	305
	306	int __node_distance(int a, int b)
	307	{
	308	int index;
	309
	310	if (!acpi_slit)
	311	return a == b ? 10 : 20;
	312	index = acpi_slit->localities * node_to_pxm(a);
	313	return acpi_slit->entry[index + node_to_pxm(b)];
	314	}
	315
	316	EXPORT_SYMBOL(__node_distance);