#define Dprintk(x...)
#endif
-struct pglist_data *node_data[MAX_NUMNODES];
+struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
bootmem_data_t plat_node_bdata[MAX_NUMNODES];
int memnode_shift;
u8 memnodemap[NODEMAPSIZE];
-unsigned char cpu_to_node[NR_CPUS] = { [0 ... NR_CPUS-1] = NUMA_NO_NODE };
-cpumask_t node_to_cpumask[MAX_NUMNODES];
+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;
-int __init compute_hash_shift(struct node *nodes, int numnodes)
+
+/*
+ * 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 node *nodes, int numnodes, int shift)
{
int i;
- int shift = 24;
- u64 addr;
-
- /* When in doubt use brute force. */
- while (shift < 48) {
- memset(memnodemap,0xff,sizeof(*memnodemap) * NODEMAPSIZE);
- for (i = 0; i < numnodes; i++) {
- if (nodes[i].start == nodes[i].end)
- continue;
- for (addr = nodes[i].start;
- addr < nodes[i].end;
- addr += (1UL << shift)) {
- if (memnodemap[addr >> shift] != 0xff &&
- memnodemap[addr >> shift] != i) {
- printk(KERN_INFO
- "node %d shift %d addr %Lx conflict %d\n",
- i, shift, addr, memnodemap[addr>>shift]);
- goto next;
- }
- memnodemap[addr >> shift] = i;
- }
- }
- return shift;
- next:
- shift++;
+ int res = -1;
+ unsigned long addr, end;
+
+ if (shift >= 64)
+ return -1;
+ memset(memnodemap, 0xff, sizeof(memnodemap));
+ for (i = 0; i < numnodes; i++) {
+ addr = nodes[i].start;
+ end = nodes[i].end;
+ if (addr >= end)
+ continue;
+ if ((end >> shift) >= NODEMAPSIZE)
+ return 0;
+ do {
+ if (memnodemap[addr >> shift] != 0xff)
+ return -1;
+ memnodemap[addr >> shift] = i;
+ addr += (1UL << shift);
+ } while (addr < end);
+ res = 1;
}
- memset(memnodemap,0,sizeof(*memnodemap) * NODEMAPSIZE);
- return -1;
+ return res;
+}
+
+int __init compute_hash_shift(struct node *nodes, int numnodes)
+{
+ int shift = 20;
+
+ while (populate_memnodemap(nodes, numnodes, shift + 1) >= 0)
+ shift++;
+
+ 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
+
/* Initialize bootmem allocator for a node */
void __init setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
{
start = round_up(start, ZONE_ALIGN);
- printk("Bootmem setup node %d %016lx-%016lx\n", nodeid, start, end);
+ printk(KERN_INFO "Bootmem setup node %d %016lx-%016lx\n", nodeid, start, end);
start_pfn = start >> PAGE_SHIFT;
end_pfn = end >> PAGE_SHIFT;
{
unsigned long start_pfn, end_pfn;
unsigned long zones[MAX_NR_ZONES];
- unsigned long dma_end_pfn;
+ unsigned long holes[MAX_NR_ZONES];
- memset(zones, 0, sizeof(unsigned long) * MAX_NR_ZONES);
+ start_pfn = node_start_pfn(nodeid);
+ end_pfn = node_end_pfn(nodeid);
- start_pfn = node_start_pfn(nodeid);
- end_pfn = node_end_pfn(nodeid);
+ Dprintk(KERN_INFO "Setting up node %d %lx-%lx\n",
+ nodeid, start_pfn, end_pfn);
- Dprintk(KERN_INFO "setting up node %d %lx-%lx\n", nodeid, start_pfn, end_pfn);
-
- /* All nodes > 0 have a zero length zone DMA */
- dma_end_pfn = __pa(MAX_DMA_ADDRESS) >> PAGE_SHIFT;
- if (start_pfn < dma_end_pfn) {
- zones[ZONE_DMA] = dma_end_pfn - start_pfn;
- zones[ZONE_NORMAL] = end_pfn - dma_end_pfn;
- } else {
- zones[ZONE_NORMAL] = end_pfn - start_pfn;
- }
-
+ size_zones(zones, holes, start_pfn, end_pfn);
free_area_init_node(nodeid, NODE_DATA(nodeid), zones,
- start_pfn, NULL);
+ start_pfn, holes);
}
void __init numa_init_array(void)
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 = 0;
+ 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);
- cpu_to_node[i] = rr;
- rr++;
}
- set_bit(0, &node_to_cpumask[cpu_to_node(0)]);
}
#ifdef CONFIG_NUMA_EMU
while ((x << 1) < sz)
x <<= 1;
if (x < sz/2)
- printk("Numa emulation unbalanced. Complain to maintainer\n");
+ printk(KERN_ERR "Numa emulation unbalanced. Complain to maintainer\n");
sz = x;
}
if (i == numa_fake-1)
sz = (end_pfn<<PAGE_SHIFT) - nodes[i].start;
nodes[i].end = nodes[i].start + sz;
- if (i != numa_fake-1)
- nodes[i].end--;
printk(KERN_INFO "Faking node %d at %016Lx-%016Lx (%LuMB)\n",
i,
nodes[i].start, nodes[i].end,
nodes_clear(node_online_map);
node_set_online(0);
for (i = 0; i < NR_CPUS; i++)
- cpu_to_node[i] = 0;
+ numa_set_node(i, 0);
node_to_cpumask[0] = cpumask_of_cpu(0);
setup_node_bootmem(0, start_pfn << PAGE_SHIFT, end_pfn << PAGE_SHIFT);
}
-__init void numa_add_cpu(int cpu)
+__cpuinit void numa_add_cpu(int cpu)
{
- /* BP is initialized elsewhere */
- if (cpu)
- set_bit(cpu, &node_to_cpumask[cpu_to_node(cpu)]);
+ set_bit(cpu, &node_to_cpumask[cpu_to_node(cpu)]);
}
+void __cpuinit numa_set_node(int cpu, int node)
+{
+ cpu_pda(cpu)->nodenumber = node;
+ cpu_to_node[cpu] = node;
+}
+
unsigned long __init numa_free_all_bootmem(void)
{
int i;
return pages;
}
+#ifdef CONFIG_SPARSEMEM
+static void __init arch_sparse_init(void)
+{
+ int i;
+
+ for_each_online_node(i)
+ memory_present(i, node_start_pfn(i), node_end_pfn(i));
+
+ sparse_init();
+}
+#else
+#define arch_sparse_init() do {} while (0)
+#endif
+
void __init paging_init(void)
{
int i;
+
+ arch_sparse_init();
+
for_each_online_node(i) {
setup_node_zones(i);
}
return 1;
}
+/*
+ * Setup early cpu_to_node.
+ *
+ * Populate cpu_to_node[] only if x86_cpu_to_apicid[],
+ * and apicid_to_node[] tables have valid entries for a CPU.
+ * This means we skip cpu_to_node[] initialisation for NUMA
+ * emulation and faking node case (when running a kernel compiled
+ * for NUMA on a non NUMA box), which is OK as cpu_to_node[]
+ * is already initialized in a round robin manner at numa_init_array,
+ * prior to this call, and this initialization is good enough
+ * for the fake NUMA cases.
+ */
+void __init init_cpu_to_node(void)
+{
+ int i;
+ for (i = 0; i < NR_CPUS; i++) {
+ u8 apicid = x86_cpu_to_apicid[i];
+ if (apicid == BAD_APICID)
+ continue;
+ if (apicid_to_node[apicid] == NUMA_NO_NODE)
+ continue;
+ numa_set_node(i,apicid_to_node[apicid]);
+ }
+}
+
EXPORT_SYMBOL(cpu_to_node);
EXPORT_SYMBOL(node_to_cpumask);
EXPORT_SYMBOL(memnode_shift);
EXPORT_SYMBOL(memnodemap);
EXPORT_SYMBOL(node_data);
+
+#ifdef CONFIG_DISCONTIGMEM
+/*
+ * Functions to convert PFNs from/to per node page addresses.
+ * These are out of line because they are quite big.
+ * They could be all tuned by pre caching more state.
+ * Should do that.
+ */
+
+/* Requires pfn_valid(pfn) to be true */
+struct page *pfn_to_page(unsigned long pfn)
+{
+ int nid = phys_to_nid(((unsigned long)(pfn)) << PAGE_SHIFT);
+ return (pfn - node_start_pfn(nid)) + NODE_DATA(nid)->node_mem_map;
+}
+EXPORT_SYMBOL(pfn_to_page);
+
+unsigned long page_to_pfn(struct page *page)
+{
+ return (long)(((page) - page_zone(page)->zone_mem_map) +
+ page_zone(page)->zone_start_pfn);
+}
+EXPORT_SYMBOL(page_to_pfn);
+
+int pfn_valid(unsigned long pfn)
+{
+ unsigned nid;
+ if (pfn >= num_physpages)
+ return 0;
+ nid = pfn_to_nid(pfn);
+ if (nid == 0xff)
+ return 0;
+ return pfn >= node_start_pfn(nid) && (pfn) < node_end_pfn(nid);
+}
+EXPORT_SYMBOL(pfn_valid);
+#endif
git://git.onelab.eu / linux-2.6.git / blobdiff