numa_64.c

linux 内核源代码

/* 
 * Generic VM initialization for x86-64 NUMA setups.
 * Copyright 2002,2003 Andi Kleen, SuSE Labs.
 */ 
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <linux/init.h>
#include <linux/bootmem.h>
#include <linux/mmzone.h>
#include <linux/ctype.h>
#include <linux/module.h>
#include <linux/nodemask.h>

#include <asm/e820.h>
#include <asm/proto.h>
#include <asm/dma.h>
#include <asm/numa.h>
#include <asm/acpi.h>

#ifndef Dprintk
#define Dprintk(x...)
#endif

struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
bootmem_data_t plat_node_bdata[MAX_NUMNODES];

struct memnode memnode;

unsigned char cpu_to_node[NR_CPUS] __read_mostly = {
	[0 ... NR_CPUS-1] = NUMA_NO_NODE
};
unsigned char apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
 	[0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
};
cpumask_t node_to_cpumask[MAX_NUMNODES] __read_mostly;

int numa_off __initdata;
unsigned long __initdata nodemap_addr;
unsigned long __initdata nodemap_size;


/*
 * Given a shift value, try to populate memnodemap[]
 * Returns :
 * 1 if OK
 * 0 if memnodmap[] too small (of shift too small)
 * -1 if node overlap or lost ram (shift too big)
 */
static int __init
populate_memnodemap(const struct bootnode *nodes, int numnodes, int shift)
{
	int i; 
	int res = -1;
	unsigned long addr, end;

	memset(memnodemap, 0xff, memnodemapsize);
	for (i = 0; i < numnodes; i++) {
		addr = nodes[i].start;
		end = nodes[i].end;
		if (addr >= end)
			continue;
		if ((end >> shift) >= memnodemapsize)
			return 0;
		do {
			if (memnodemap[addr >> shift] != 0xff)
				return -1;
			memnodemap[addr >> shift] = i;
			addr += (1UL << shift);
		} while (addr < end);
		res = 1;
	} 
	return res;
}

static int __init allocate_cachealigned_memnodemap(void)
{
	unsigned long pad, pad_addr;

	memnodemap = memnode.embedded_map;
	if (memnodemapsize <= 48)
		return 0;

	pad = L1_CACHE_BYTES - 1;
	pad_addr = 0x8000;
	nodemap_size = pad + memnodemapsize;
	nodemap_addr = find_e820_area(pad_addr, end_pfn<<PAGE_SHIFT,
				      nodemap_size);
	if (nodemap_addr == -1UL) {
		printk(KERN_ERR
		       "NUMA: Unable to allocate Memory to Node hash map\n");
		nodemap_addr = nodemap_size = 0;
		return -1;
	}
	pad_addr = (nodemap_addr + pad) & ~pad;
	memnodemap = phys_to_virt(pad_addr);

	printk(KERN_DEBUG "NUMA: Allocated memnodemap from %lx - %lx\n",
	       nodemap_addr, nodemap_addr + nodemap_size);
	return 0;
}

/*
 * The LSB of all start and end addresses in the node map is the value of the
 * maximum possible shift.
 */
static int __init
extract_lsb_from_nodes (const struct bootnode *nodes, int numnodes)
{
	int i, nodes_used = 0;
	unsigned long start, end;
	unsigned long bitfield = 0, memtop = 0;

	for (i = 0; i < numnodes; i++) {
		start = nodes[i].start;
		end = nodes[i].end;
		if (start >= end)
			continue;
		bitfield |= start;
		nodes_used++;
		if (end > memtop)
			memtop = end;
	}
	if (nodes_used <= 1)
		i = 63;
	else
		i = find_first_bit(&bitfield, sizeof(unsigned long)*8);
	memnodemapsize = (memtop >> i)+1;
	return i;
}

int __init compute_hash_shift(struct bootnode *nodes, int numnodes)
{
	int shift;

	shift = extract_lsb_from_nodes(nodes, numnodes);
	if (allocate_cachealigned_memnodemap())
		return -1;
	printk(KERN_DEBUG "NUMA: Using %d for the hash shift.\n",
		shift);

	if (populate_memnodemap(nodes, numnodes, shift) != 1) {
		printk(KERN_INFO
	"Your memory is not aligned you need to rebuild your kernel "
	"with a bigger NODEMAPSIZE shift=%d\n",
			shift);
		return -1;
	}
	return shift;
}

#ifdef CONFIG_SPARSEMEM
int early_pfn_to_nid(unsigned long pfn)
{
	return phys_to_nid(pfn << PAGE_SHIFT);
}
#endif

static void * __init
early_node_mem(int nodeid, unsigned long start, unsigned long end,
	      unsigned long size)
{
	unsigned long mem = find_e820_area(start, end, size);
	void *ptr;
	if (mem != -1L)
		return __va(mem);
	ptr = __alloc_bootmem_nopanic(size,
				SMP_CACHE_BYTES, __pa(MAX_DMA_ADDRESS));
	if (ptr == NULL) {
		printk(KERN_ERR "Cannot find %lu bytes in node %d\n",
			size, nodeid);
		return NULL;
	}
	return ptr;
}

/* Initialize bootmem allocator for a node */
void __init setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
{ 
	unsigned long start_pfn, end_pfn, bootmap_pages, bootmap_size, bootmap_start; 
	unsigned long nodedata_phys;
	void *bootmap;
	const int pgdat_size = round_up(sizeof(pg_data_t), PAGE_SIZE);

	start = round_up(start, ZONE_ALIGN); 

	printk(KERN_INFO "Bootmem setup node %d %016lx-%016lx\n", nodeid, start, end);

	start_pfn = start >> PAGE_SHIFT;
	end_pfn = end >> PAGE_SHIFT;

	node_data[nodeid] = early_node_mem(nodeid, start, end, pgdat_size);
	if (node_data[nodeid] == NULL)
		return;
	nodedata_phys = __pa(node_data[nodeid]);

	memset(NODE_DATA(nodeid), 0, sizeof(pg_data_t));
	NODE_DATA(nodeid)->bdata = &plat_node_bdata[nodeid];
	NODE_DATA(nodeid)->node_start_pfn = start_pfn;
	NODE_DATA(nodeid)->node_spanned_pages = end_pfn - start_pfn;

	/* Find a place for the bootmem map */
	bootmap_pages = bootmem_bootmap_pages(end_pfn - start_pfn); 
	bootmap_start = round_up(nodedata_phys + pgdat_size, PAGE_SIZE);
	bootmap = early_node_mem(nodeid, bootmap_start, end,
					bootmap_pages<<PAGE_SHIFT);
	if (bootmap == NULL)  {
		if (nodedata_phys < start || nodedata_phys >= end)
			free_bootmem((unsigned long)node_data[nodeid],pgdat_size);
		node_data[nodeid] = NULL;
		return;
	}
	bootmap_start = __pa(bootmap);
	Dprintk("bootmap start %lu pages %lu\n", bootmap_start, bootmap_pages); 
	
	bootmap_size = init_bootmem_node(NODE_DATA(nodeid),
					 bootmap_start >> PAGE_SHIFT, 
					 start_pfn, end_pfn); 

	free_bootmem_with_active_regions(nodeid, end);

	reserve_bootmem_node(NODE_DATA(nodeid), nodedata_phys, pgdat_size); 
	reserve_bootmem_node(NODE_DATA(nodeid), bootmap_start, bootmap_pages<<PAGE_SHIFT);
#ifdef CONFIG_ACPI_NUMA
	srat_reserve_add_area(nodeid);
#endif
	node_set_online(nodeid);
} 

/* Initialize final allocator for a zone */
void __init setup_node_zones(int nodeid)
{ 
	unsigned long start_pfn, end_pfn, memmapsize, limit;

 	start_pfn = node_start_pfn(nodeid);
 	end_pfn = node_end_pfn(nodeid);

	Dprintk(KERN_INFO "Setting up memmap for node %d %lx-%lx\n",
		nodeid, start_pfn, end_pfn);

	/* Try to allocate mem_map at end to not fill up precious <4GB
	   memory. */
	memmapsize = sizeof(struct page) * (end_pfn-start_pfn);
	limit = end_pfn << PAGE_SHIFT;
#ifdef CONFIG_FLAT_NODE_MEM_MAP
	NODE_DATA(nodeid)->node_mem_map = 
		__alloc_bootmem_core(NODE_DATA(nodeid)->bdata, 
				memmapsize, SMP_CACHE_BYTES, 
				round_down(limit - memmapsize, PAGE_SIZE), 
				limit);
#endif
} 

void __init numa_init_array(void)
{
	int rr, i;
	/* There are unfortunately some poorly designed mainboards around
	   that only connect memory to a single CPU. This breaks the 1:1 cpu->node
	   mapping. To avoid this fill in the mapping for all possible
	   CPUs, as the number of CPUs is not known yet. 
	   We round robin the existing nodes. */
	rr = first_node(node_online_map);
	for (i = 0; i < NR_CPUS; i++) {
		if (cpu_to_node(i) != NUMA_NO_NODE)
			continue;
 		numa_set_node(i, rr);
		rr = next_node(rr, node_online_map);
		if (rr == MAX_NUMNODES)
			rr = first_node(node_online_map);
	}

}

#ifdef CONFIG_NUMA_EMU
/* Numa emulation */
char *cmdline __initdata;

/*
 * Setups up nid to range from addr to addr + size.  If the end boundary is
 * greater than max_addr, then max_addr is used instead.  The return value is 0
 * if there is additional memory left for allocation past addr and -1 otherwise.
 * addr is adjusted to be at the end of the node.
 */
static int __init setup_node_range(int nid, struct bootnode *nodes, u64 *addr,
				   u64 size, u64 max_addr)
{
	int ret = 0;
	nodes[nid].start = *addr;
	*addr += size;
	if (*addr >= max_addr) {
		*addr = max_addr;
		ret = -1;
	}
	nodes[nid].end = *addr;
	node_set(nid, node_possible_map);
	printk(KERN_INFO "Faking node %d at %016Lx-%016Lx (%LuMB)\n", nid,
	       nodes[nid].start, nodes[nid].end,
	       (nodes[nid].end - nodes[nid].start) >> 20);
	return ret;
}

/*
 * Splits num_nodes nodes up equally starting at node_start.  The return value
 * is the number of nodes split up and addr is adjusted to be at the end of the
 * last node allocated.
 */
static int __init split_nodes_equally(struct bootnode *nodes, u64 *addr,
				      u64 max_addr, int node_start,
				      int num_nodes)
{
	unsigned int big;
	u64 size;
	int i;

	if (num_nodes <= 0)
		return -1;
	if (num_nodes > MAX_NUMNODES)
		num_nodes = MAX_NUMNODES;
	size = (max_addr - *addr - e820_hole_size(*addr, max_addr)) /
	       num_nodes;
	/*
	 * Calculate the number of big nodes that can be allocated as a result
	 * of consolidating the leftovers.
	 */
	big = ((size & ~FAKE_NODE_MIN_HASH_MASK) * num_nodes) /