Merge master.kernel.org:/pub/scm/linux/kernel/git/aegl/linux-2.6

31 files changed, 1255 insertions, 544 deletions

diff --git a/arch/ia64/kernel/Makefile b/arch/ia64/kernel/Makefile
index b2e2f6509eb..e1fb68ddec2 100644
--- a/arch/ia64/kernel/Makefile
+++ b/arch/ia64/kernel/Makefile
@@ -17,6 +17,7 @@ obj-$(CONFIG_IA64_PALINFO)	+= palinfo.o
 obj-$(CONFIG_IOSAPIC)		+= iosapic.o
 obj-$(CONFIG_MODULES)		+= module.o
 obj-$(CONFIG_SMP)		+= smp.o smpboot.o domain.o
+obj-$(CONFIG_NUMA)		+= numa.o
 obj-$(CONFIG_PERFMON)		+= perfmon_default_smpl.o
 obj-$(CONFIG_IA64_CYCLONE)	+= cyclone.o
 obj-$(CONFIG_IA64_MCA_RECOVERY)	+= mca_recovery.o
diff --git a/arch/ia64/kernel/acpi.c b/arch/ia64/kernel/acpi.c
index cda06f88c66..542256e98e6 100644
--- a/arch/ia64/kernel/acpi.c
+++ b/arch/ia64/kernel/acpi.c
@@ -640,9 +640,11 @@ acpi_boot_init (void)
 			if (smp_boot_data.cpu_phys_id[cpu] != hard_smp_processor_id())
 				node_cpuid[i++].phys_id = smp_boot_data.cpu_phys_id[cpu];
 	}
-	build_cpu_to_node_map();
 # endif
 #endif
+#ifdef CONFIG_ACPI_NUMA
+	build_cpu_to_node_map();
+#endif
 	/* Make boot-up look pretty */
 	printk(KERN_INFO "%d CPUs available, %d CPUs total\n", available_cpus, total_cpus);
 	return 0;
diff --git a/arch/ia64/kernel/numa.c b/arch/ia64/kernel/numa.c
new file mode 100644
index 00000000000..a68ce667809
--- /dev/null
+++ b/arch/ia64/kernel/numa.c
@@ -0,0 +1,57 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ *
+ * ia64 kernel NUMA specific stuff
+ *
+ * Copyright (C) 2002 Erich Focht <efocht@ess.nec.de>
+ * Copyright (C) 2004 Silicon Graphics, Inc.
+ *   Jesse Barnes <jbarnes@sgi.com>
+ */
+#include <linux/config.h>
+#include <linux/topology.h>
+#include <linux/module.h>
+#include <asm/processor.h>
+#include <asm/smp.h>
+
+u8 cpu_to_node_map[NR_CPUS] __cacheline_aligned;
+EXPORT_SYMBOL(cpu_to_node_map);
+
+cpumask_t node_to_cpu_mask[MAX_NUMNODES] __cacheline_aligned;
+
+/**
+ * build_cpu_to_node_map - setup cpu to node and node to cpumask arrays
+ *
+ * Build cpu to node mapping and initialize the per node cpu masks using
+ * info from the node_cpuid array handed to us by ACPI.
+ */
+void __init build_cpu_to_node_map(void)
+{
+	int cpu, i, node;
+
+	for(node=0; node < MAX_NUMNODES; node++)
+		cpus_clear(node_to_cpu_mask[node]);
+
+	for(cpu = 0; cpu < NR_CPUS; ++cpu) {
+		node = -1;
+		for (i = 0; i < NR_CPUS; ++i)
+			if (cpu_physical_id(cpu) == node_cpuid[i].phys_id) {
+				node = node_cpuid[i].nid;
+				break;
+			}
+		cpu_to_node_map[cpu] = (node >= 0) ? node : 0;
+		if (node >= 0)
+			cpu_set(cpu, node_to_cpu_mask[node]);
+	}
+}
diff --git a/arch/ia64/kernel/signal.c b/arch/ia64/kernel/signal.c
index edd9f07860b..b8a0a7d257a 100644
--- a/arch/ia64/kernel/signal.c
+++ b/arch/ia64/kernel/signal.c
@@ -143,6 +143,7 @@ restore_sigcontext (struct sigcontext __user *sc, struct sigscratch *scr)
 
 		__copy_from_user(current->thread.fph, &sc->sc_fr[32], 96*16);
 		psr->mfh = 0;	/* drop signal handler's fph contents... */
+		preempt_disable();
 		if (psr->dfh)
 			ia64_drop_fpu(current);
 		else {
@@ -150,6 +151,7 @@ restore_sigcontext (struct sigcontext __user *sc, struct sigscratch *scr)
 			__ia64_load_fpu(current->thread.fph);
 			ia64_set_local_fpu_owner(current);
 		}
+		preempt_enable();
 	}
 	return err;
 }
diff --git a/arch/ia64/kernel/smpboot.c b/arch/ia64/kernel/smpboot.c
index 623b0a54670..7d72c0d872b 100644
--- a/arch/ia64/kernel/smpboot.c
+++ b/arch/ia64/kernel/smpboot.c
@@ -525,47 +525,6 @@ smp_build_cpu_map (void)
 	}
 }
 
-#ifdef CONFIG_NUMA
-
-/* on which node is each logical CPU (one cacheline even for 64 CPUs) */
-u8 cpu_to_node_map[NR_CPUS] __cacheline_aligned;
-EXPORT_SYMBOL(cpu_to_node_map);
-/* which logical CPUs are on which nodes */
-cpumask_t node_to_cpu_mask[MAX_NUMNODES] __cacheline_aligned;
-
-/*
- * Build cpu to node mapping and initialize the per node cpu masks.
- */
-void __init
-build_cpu_to_node_map (void)
-{
-	int cpu, i, node;
-
-	for(node=0; node<MAX_NUMNODES; node++)
-		cpus_clear(node_to_cpu_mask[node]);
-	for(cpu = 0; cpu < NR_CPUS; ++cpu) {
-		/*
-		 * All Itanium NUMA platforms I know use ACPI, so maybe we
-		 * can drop this ifdef completely.                    [EF]
-		 */
-#ifdef CONFIG_ACPI_NUMA
-		node = -1;
-		for (i = 0; i < NR_CPUS; ++i)
-			if (cpu_physical_id(cpu) == node_cpuid[i].phys_id) {
-				node = node_cpuid[i].nid;
-				break;
-			}
-#else
-#		error Fixme: Dunno how to build CPU-to-node map.
-#endif
-		cpu_to_node_map[cpu] = (node >= 0) ? node : 0;
-		if (node >= 0)
-			cpu_set(cpu, node_to_cpu_mask[node]);
-	}
-}
-
-#endif /* CONFIG_NUMA */
-
 /*
  * Cycle through the APs sending Wakeup IPIs to boot each.
  */
diff --git a/arch/ia64/kernel/traps.c b/arch/ia64/kernel/traps.c
index e7e520d90f0..4440c8343fa 100644
--- a/arch/ia64/kernel/traps.c
+++ b/arch/ia64/kernel/traps.c
@@ -90,14 +90,16 @@ die (const char *str, struct pt_regs *regs, long err)
 		.lock_owner_depth =	0
 	};
 	static int die_counter;
+	int cpu = get_cpu();
 
-	if (die.lock_owner != smp_processor_id()) {
+	if (die.lock_owner != cpu) {
 		console_verbose();
 		spin_lock_irq(&die.lock);
-		die.lock_owner = smp_processor_id();
+		die.lock_owner = cpu;
 		die.lock_owner_depth = 0;
 		bust_spinlocks(1);
 	}
+	put_cpu();
 
 	if (++die.lock_owner_depth < 3) {
 		printk("%s[%d]: %s %ld [%d]\n",
diff --git a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c
index f3fd528ead3..b5c90e54819 100644
--- a/arch/ia64/mm/discontig.c
+++ b/arch/ia64/mm/discontig.c
@@ -44,150 +44,7 @@ struct early_node_data {
 };
 
 static struct early_node_data mem_data[MAX_NUMNODES] __initdata;
-
-/**
- * reassign_cpu_only_nodes - called from find_memory to move CPU-only nodes to a memory node
- *
- * This function will move nodes with only CPUs (no memory)
- * to a node with memory which is at the minimum numa_slit distance.
- * Any reassigments will result in the compression of the nodes
- * and renumbering the nid values where appropriate.
- * The static declarations below are to avoid large stack size which
- * makes the code not re-entrant.
- */
-static void __init reassign_cpu_only_nodes(void)
-{
-	struct node_memblk_s *p;
-	int i, j, k, nnode, nid, cpu, cpunid, pxm;
-	u8 cslit, slit;
-	static DECLARE_BITMAP(nodes_with_mem, MAX_NUMNODES) __initdata;
-	static u8 numa_slit_fix[MAX_NUMNODES * MAX_NUMNODES] __initdata;
-	static int node_flip[MAX_NUMNODES] __initdata;
-	static int old_nid_map[NR_CPUS] __initdata;
-
-	for (nnode = 0, p = &node_memblk[0]; p < &node_memblk[num_node_memblks]; p++)
-		if (!test_bit(p->nid, (void *) nodes_with_mem)) {
-			set_bit(p->nid, (void *) nodes_with_mem);
-			nnode++;
-		}
-
-	/*
-	 * All nids with memory.
-	 */
-	if (nnode == num_online_nodes())
-		return;
-
-	/*
-	 * Change nids and attempt to migrate CPU-only nodes
-	 * to the best numa_slit (closest neighbor) possible.
-	 * For reassigned CPU nodes a nid can't be arrived at
-	 * until after this loop because the target nid's new
-	 * identity might not have been established yet. So
-	 * new nid values are fabricated above num_online_nodes() and
-	 * mapped back later to their true value.
-	 */
-	/* MCD - This code is a bit complicated, but may be unnecessary now.
-	 * We can now handle much more interesting node-numbering.
-	 * The old requirement that 0 <= nid <= numnodes <= MAX_NUMNODES
-	 * and that there be no holes in the numbering 0..numnodes
-	 * has become simply 0 <= nid <= MAX_NUMNODES.
-	 */
-	nid = 0;
-	for_each_online_node(i)  {
-		if (test_bit(i, (void *) nodes_with_mem)) {
-			/*
-			 * Save original nid value for numa_slit
-			 * fixup and node_cpuid reassignments.
-			 */
-			node_flip[nid] = i;
-
-			if (i == nid) {
-				nid++;
-				continue;
-			}
-
-			for (p = &node_memblk[0]; p < &node_memblk[num_node_memblks]; p++)
-				if (p->nid == i)
-					p->nid = nid;
-
-			cpunid = nid;
-			nid++;
-		} else
-			cpunid = MAX_NUMNODES;
-
-		for (cpu = 0; cpu < NR_CPUS; cpu++)
-			if (node_cpuid[cpu].nid == i) {
-				/*
-				 * For nodes not being reassigned just
-				 * fix the cpu's nid and reverse pxm map
-				 */
-				if (cpunid < MAX_NUMNODES) {
-					pxm = nid_to_pxm_map[i];
-					pxm_to_nid_map[pxm] =
-					          node_cpuid[cpu].nid = cpunid;
-					continue;
-				}
-
-				/*
-				 * For nodes being reassigned, find best node by
-				 * numa_slit information and then make a temporary
-				 * nid value based on current nid and num_online_nodes().
-				 */
-				slit = 0xff;
-				k = 2*num_online_nodes();
-				for_each_online_node(j) {
-					if (i == j)
-						continue;
-					else if (test_bit(j, (void *) nodes_with_mem)) {
-						cslit = numa_slit[i * num_online_nodes() + j];
-						if (cslit < slit) {
-							k = num_online_nodes() + j;
-							slit = cslit;
-						}
-					}
-				}
-
-				/* save old nid map so we can update the pxm */
-				old_nid_map[cpu] = node_cpuid[cpu].nid;
-				node_cpuid[cpu].nid = k;
-			}
-	}
-
-	/*
-	 * Fixup temporary nid values for CPU-only nodes.
-	 */
-	for (cpu = 0; cpu < NR_CPUS; cpu++)
-		if (node_cpuid[cpu].nid == (2*num_online_nodes())) {
-			pxm = nid_to_pxm_map[old_nid_map[cpu]];
-			pxm_to_nid_map[pxm] = node_cpuid[cpu].nid = nnode - 1;
-		} else {
-			for (i = 0; i < nnode; i++) {
-				if (node_flip[i] != (node_cpuid[cpu].nid - num_online_nodes()))
-					continue;
-
-				pxm = nid_to_pxm_map[old_nid_map[cpu]];
-				pxm_to_nid_map[pxm] = node_cpuid[cpu].nid = i;
-				break;
-			}
-		}
-
-	/*
-	 * Fix numa_slit by compressing from larger
-	 * nid array to reduced nid array.
-	 */
-	for (i = 0; i < nnode; i++)
-		for (j = 0; j < nnode; j++)
-			numa_slit_fix[i * nnode + j] =
-				numa_slit[node_flip[i] * num_online_nodes() + node_flip[j]];
-
-	memcpy(numa_slit, numa_slit_fix, sizeof (numa_slit));
-
-	nodes_clear(node_online_map);
-	for (i = 0; i < nnode; i++)
-		node_set_online(i);
-
-	return;
-}
+static nodemask_t memory_less_mask __initdata;
 
 /*
  * To prevent cache aliasing effects, align per-node structures so that they
@@ -233,44 +90,101 @@ static int __init build_node_maps(unsigned long start, unsigned long len,
 }
 
 /**
- * early_nr_phys_cpus_node - return number of physical cpus on a given node
+ * early_nr_cpus_node - return number of cpus on a given node
  * @node: node to check
  *
- * Count the number of physical cpus on @node.  These are cpus that actually
- * exist.  We can't use nr_cpus_node() yet because
+ * Count the number of cpus on @node.  We can't use nr_cpus_node() yet because
  * acpi_boot_init() (which builds the node_to_cpu_mask array) hasn't been
- * called yet.
+ * called yet.  Note that node 0 will also count all non-existent cpus.
  */
-static int early_nr_phys_cpus_node(int node)
+static int __init early_nr_cpus_node(int node)
 {
 	int cpu, n = 0;
 
 	for (cpu = 0; cpu < NR_CPUS; cpu++)
 		if (node == node_cpuid[cpu].nid)
-			if ((cpu == 0) || node_cpuid[cpu].phys_id)
-				n++;
+			n++;
 
 	return n;
 }
 
+/**
+ * compute_pernodesize - compute size of pernode data
+ * @node: the node id.
+ */
+static unsigned long __init compute_pernodesize(int node)
+{
+	unsigned long pernodesize = 0, cpus;
+
+	cpus = early_nr_cpus_node(node);
+	pernodesize += PERCPU_PAGE_SIZE * cpus;
+	pernodesize += node * L1_CACHE_BYTES;
+	pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t));
+	pernodesize += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
+	pernodesize = PAGE_ALIGN(pernodesize);
+	return pernodesize;
+}
 
 /**
- * early_nr_cpus_node - return number of cpus on a given node
- * @node: node to check
+ * per_cpu_node_setup - setup per-cpu areas on each node
+ * @cpu_data: per-cpu area on this node
+ * @node: node to setup
  *
- * Count the number of cpus on @node.  We can't use nr_cpus_node() yet because
- * acpi_boot_init() (which builds the node_to_cpu_mask array) hasn't been
- * called yet.  Note that node 0 will also count all non-existent cpus.
+ * Copy the static per-cpu data into the region we just set aside and then
+ * setup __per_cpu_offset for each CPU on this node.  Return a pointer to
+ * the end of the area.
  */
-static int early_nr_cpus_node(int node)
+static void *per_cpu_node_setup(void *cpu_data, int node)
 {
-	int cpu, n = 0;
+#ifdef CONFIG_SMP
+	int cpu;
 
-	for (cpu = 0; cpu < NR_CPUS; cpu++)
-		if (node == node_cpuid[cpu].nid)
-			n++;
+	for (cpu = 0; cpu < NR_CPUS; cpu++) {
+		if (node == node_cpuid[cpu].nid) {
+			memcpy(__va(cpu_data), __phys_per_cpu_start,
+			       __per_cpu_end - __per_cpu_start);
+			__per_cpu_offset[cpu] = (char*)__va(cpu_data) -
+				__per_cpu_start;
+			cpu_data += PERCPU_PAGE_SIZE;
+		}
+	}
+#endif
+	return cpu_data;
+}
 
-	return n;
+/**
+ * fill_pernode - initialize pernode data.
+ * @node: the node id.
+ * @pernode: physical address of pernode data
+ * @pernodesize: size of the pernode data
+ */
+static void __init fill_pernode(int node, unsigned long pernode,
+	unsigned long pernodesize)
+{
+	void *cpu_data;
+	int cpus = early_nr_cpus_node(node);
+	struct bootmem_data *bdp = &mem_data[node].bootmem_data;
+
+	mem_data[node].pernode_addr = pernode;
+	mem_data[node].pernode_size = pernodesize;
+	memset(__va(pernode), 0, pernodesize);
+
+	cpu_data = (void *)pernode;
+	pernode += PERCPU_PAGE_SIZE * cpus;
+	pernode += node * L1_CACHE_BYTES;
+
+	mem_data[node].pgdat = __va(pernode);
+	pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
+
+	mem_data[node].node_data = __va(pernode);
+	pernode += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
+
+	mem_data[node].pgdat->bdata = bdp;
+	pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
+
+	cpu_data = per_cpu_node_setup(cpu_data, node);
+
+	return;
 }
 
 /**
@@ -304,9 +218,8 @@ static int early_nr_cpus_node(int node)
 static int __init find_pernode_space(unsigned long start, unsigned long len,
 				     int node)
 {
-	unsigned long epfn, cpu, cpus, phys_cpus;
+	unsigned long epfn;
 	unsigned long pernodesize = 0, pernode, pages, mapsize;
-	void *cpu_data;
 	struct bootmem_data *bdp = &mem_data[node].bootmem_data;
 
 	epfn = (start + len) >> PAGE_SHIFT;
@@ -329,49 +242,12 @@ static int __init find_pernode_space(unsigned long start, unsigned long len,
 	 * Calculate total size needed, incl. what's necessary
 	 * for good alignment and alias prevention.
 	 */
-	cpus = early_nr_cpus_node(node);
-	phys_cpus = early_nr_phys_cpus_node(node);
-	pernodesize += PERCPU_PAGE_SIZE * cpus;
-	pernodesize += node * L1_CACHE_BYTES;
-	pernodesize += L1_CACHE_ALIGN(sizeof(pg_data_t));
-	pernodesize += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
-	pernodesize = PAGE_ALIGN(pernodesize);
+	pernodesize = compute_pernodesize(node);
 	pernode = NODEDATA_ALIGN(start, node);
 
 	/* Is this range big enough for what we want to store here? */
-	if (start + len > (pernode + pernodesize + mapsize)) {
-		mem_data[node].pernode_addr = pernode;
-		mem_data[node].pernode_size = pernodesize;
-		memset(__va(pernode), 0, pernodesize);
-
-		cpu_data = (void *)pernode;
-		pernode += PERCPU_PAGE_SIZE * cpus;
-		pernode += node * L1_CACHE_BYTES;
-
-		mem_data[node].pgdat = __va(pernode);
-		pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
-
-		mem_data[node].node_data = __va(pernode);
-		pernode += L1_CACHE_ALIGN(sizeof(struct ia64_node_data));
-
-		mem_data[node].pgdat->bdata = bdp;
-		pernode += L1_CACHE_ALIGN(sizeof(pg_data_t));
-
-		/*
-		 * Copy the static per-cpu data into the region we
-		 * just set aside and then setup __per_cpu_offset
-		 * for each CPU on this node.
-		 */
-		for (cpu = 0; cpu < NR_CPUS; cpu++) {
-			if (node == node_cpuid[cpu].nid) {
-				memcpy(__va(cpu_data), __phys_per_cpu_start,
-				       __per_cpu_end - __per_cpu_start);
-				__per_cpu_offset[cpu] = (char*)__va(cpu_data) -
-					__per_cpu_start;
-				cpu_data += PERCPU_PAGE_SIZE;
-			}
-		}
-	}
+	if (start + len > (pernode + pernodesize + mapsize))
+		fill_pernode(node, pernode, pernodesize);
 
 	return 0;
 }
@@ -411,6 +287,9 @@ static void __init reserve_pernode_space(void)
 	for_each_online_node(node) {
 		pg_data_t *pdp = mem_data[node].pgdat;
 
+		if (node_isset(node, memory_less_mask))
+			continue;
+
 		bdp = pdp->bdata;
 
 		/* First the bootmem_map itself */
@@ -436,8 +315,8 @@ static void __init reserve_pernode_space(void)
  */
 static void __init initialize_pernode_data(void)
 {
-	int cpu, node;
 	pg_data_t *pgdat_list[MAX_NUMNODES];
+	int cpu, node;
 
 	for_each_online_node(node)
 		pgdat_list[node] = mem_data[node].pgdat;
@@ -447,12 +326,99 @@ static void __init initialize_pernode_data(void)
 		memcpy(mem_data[node].node_data->pg_data_ptrs, pgdat_list,
 		       sizeof(pgdat_list));
 	}
-
+#ifdef CONFIG_SMP
 	/* Set the node_data pointer for each per-cpu struct */
 	for (cpu = 0; cpu < NR_CPUS; cpu++) {
 		node = node_cpuid[cpu].nid;
 		per_cpu(cpu_info, cpu).node_data = mem_data[node].node_data;
 	}
+#else
+	{
+		struct cpuinfo_ia64 *cpu0_cpu_info;
+		cpu = 0;
+		node = node_cpuid[cpu].nid;
+		cpu0_cpu_info = (struct cpuinfo_ia64 *)(__phys_per_cpu_start +
+			((char *)&per_cpu__cpu_info - __per_cpu_start));
+		cpu0_cpu_info->node_data = mem_data[node].node_data;
+	}
+#endif /* CONFIG_SMP */
+}
+
+/**
+ * memory_less_node_alloc - * attempt to allocate memory on the best NUMA slit
+ * 	node but fall back to any other node when __alloc_bootmem_node fails
+ *	for best.
+ * @nid: node id
+ * @pernodesize: size of this node's pernode data
+ * @align: alignment to use for this node's pernode data
+ */
+static void __init *memory_less_node_alloc(int nid, unsigned long pernodesize,
+	unsigned long align)
+{
+	void *ptr = NULL;
+	u8 best = 0xff;
+	int bestnode = -1, node;
+
+	for_each_online_node(node) {
+		if (node_isset(node, memory_less_mask))
+			continue;
+		else if (node_distance(nid, node) < best) {
+			best = node_distance(nid, node);
+			bestnode = node;
+		}
+	}
+
+	ptr = __alloc_bootmem_node(mem_data[bestnode].pgdat,
+		pernodesize, align, __pa(MAX_DMA_ADDRESS));
+
+	if (!ptr)
+		panic("NO memory for memory less node\n");
+	return ptr;
+}
+
+/**
+ * pgdat_insert - insert the pgdat into global pgdat_list
+ * @pgdat: the pgdat for a node.
+ */
+static void __init pgdat_insert(pg_data_t *pgdat)
+{
+	pg_data_t *prev = NULL, *next;
+
+	for_each_pgdat(next)
+		if (pgdat->node_id < next->node_id)
+			break;
+		else
+			prev = next;
+
+	if (prev) {
+		prev->pgdat_next = pgdat;
+		pgdat->pgdat_next = next;
+	} else {
+		pgdat->pgdat_next = pgdat_list;
+		pgdat_list = pgdat;
+	}
+
+	return;
+}
+
+/**
+ * memory_less_nodes - allocate and initialize CPU only nodes pernode
+ *	information.
+ */
+static void __init memory_less_nodes(void)
+{
+	unsigned long pernodesize;
+	void *pernode;
+	int node;
+
+	for_each_node_mask(node, memory_less_mask) {
+		pernodesize = compute_pernodesize(node);
+		pernode = memory_less_node_alloc(node, pernodesize,
+			(node) ? (node * PERCPU_PAGE_SIZE) : (1024*1024));
+		fill_pernode(node, __pa(pernode), pernodesize);
+	}
+
+	return;
 }
 
 /**
@@ -472,16 +438,19 @@ void __init find_memory(void)
 		node_set_online(0);
 	}
 
+	nodes_or(memory_less_mask, memory_less_mask, node_online_map);
 	min_low_pfn = -1;
 	max_low_pfn = 0;
 
-	if (num_online_nodes() > 1)
-		reassign_cpu_only_nodes();
-
 	/* These actually end up getting called by call_pernode_memory() */
 	efi_memmap_walk(filter_rsvd_memory, build_node_maps);
 	efi_memmap_walk(filter_rsvd_memory, find_pernode_space);
 
+	for_each_online_node(node)
+		if (mem_data[node].bootmem_data.node_low_pfn) {
+			node_clear(node, memory_less_mask);
+			mem_data[node].min_pfn = ~0UL;
+		}
 	/*
 	 * Initialize the boot memory maps in reverse order since that's
 	 * what the bootmem allocator expects
@@ -492,17 +461,14 @@ void __init find_memory(void)
 
 		if (!node_online(node))
 			continue;
+		else if (node_isset(node, memory_less_mask))
+			continue;
 
 		bdp = &mem_data[node].bootmem_data;
 		pernode = mem_data[node].pernode_addr;
 		pernodesize = mem_data[node].pernode_size;
 		map = pernode + pernodesize;
 
-		/* Sanity check... */
-		if (!pernode)
-			panic("pernode space for node %d "
-			      "could not be allocated!", node);
-
 		init_bootmem_node(mem_data[node].pgdat,
 				  map>>PAGE_SHIFT,
 				  bdp->node_boot_start>>PAGE_SHIFT,
@@ -512,6 +478,7 @@ void __init find_memory(void)
 	efi_memmap_walk(filter_rsvd_memory, free_node_bootmem);
 
 	reserve_pernode_space();
+	memory_less_nodes();
 	initialize_pernode_data();
 
 	max_pfn = max_low_pfn;
@@ -519,6 +486,7 @@ void __init find_memory(void)
 	find_initrd();
 }
 
+#ifdef CONFIG_SMP
 /**
  * per_cpu_init - setup per-cpu variables
  *
@@ -529,15 +497,15 @@ void *per_cpu_init(void)
 {
 	int cpu;
 
-	if (smp_processor_id() == 0) {
-		for (cpu = 0; cpu < NR_CPUS; cpu++) {
-			per_cpu(local_per_cpu_offset, cpu) =
-				__per_cpu_offset[cpu];
-		}
-	}
+	if (smp_processor_id() != 0)
+		return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
+
+	for (cpu = 0; cpu < NR_CPUS; cpu++)
+		per_cpu(local_per_cpu_offset, cpu) = __per_cpu_offset[cpu];
 
 	return __per_cpu_start + __per_cpu_offset[smp_processor_id()];
 }
+#endif /* CONFIG_SMP */
 
 /**
  * show_mem - give short summary of memory stats
@@ -680,12 +648,13 @@ void __init paging_init(void)
 
 	max_dma = virt_to_phys((void *) MAX_DMA_ADDRESS) >> PAGE_SHIFT;
 
-	/* so min() will work in count_node_pages */
-	for_each_online_node(node)
-		mem_data[node].min_pfn = ~0UL;
-
 	efi_memmap_walk(filter_rsvd_memory, count_node_pages);
 
+	vmalloc_end -= PAGE_ALIGN(max_low_pfn * sizeof(struct page));
+	vmem_map = (struct page *) vmalloc_end;
+	efi_memmap_walk(create_mem_map_page_table, NULL);
+	printk("Virtual mem_map starts at 0x%p\n", vmem_map);
+
 	for_each_online_node(node) {
 		memset(zones_size, 0, sizeof(zones_size));
 		memset(zholes_size, 0, sizeof(zholes_size));
@@ -719,15 +688,6 @@ void __init paging_init(void)
 				 mem_data[node].num_dma_physpages);
 		}
 
-		if (node == 0) {
-			vmalloc_end -=
-				PAGE_ALIGN(max_low_pfn * sizeof(struct page));
-			vmem_map = (struct page *) vmalloc_end;
-
-			efi_memmap_walk(create_mem_map_page_table, NULL);
-			printk("Virtual mem_map starts at 0x%p\n", vmem_map);
-		}
-
 		pfn_offset = mem_data[node].min_pfn;
 
 		NODE_DATA(node)->node_mem_map = vmem_map + pfn_offset;
@@ -735,5 +695,11 @@ void __init paging_init(void)
 				    pfn_offset, zholes_size);
 	}
 
+	/*
+	 * Make memory less nodes become a member of the known nodes.
+	 */
+	for_each_node_mask(node, memory_less_mask)
+		pgdat_insert(mem_data[node].pgdat);
+
 	zero_page_memmap_ptr = virt_to_page(ia64_imva(empty_zero_page));
 }
diff --git a/arch/ia64/mm/init.c b/arch/ia64/mm/init.c
index 4eb2f52b87a..65f9958db9f 100644
--- a/arch/ia64/mm/init.c
+++ b/arch/ia64/mm/init.c
@@ -597,7 +597,8 @@ mem_init (void)
 	kclist_add(&kcore_kernel, _stext, _end - _stext);
 
 	for_each_pgdat(pgdat)
-		totalram_pages += free_all_bootmem_node(pgdat);
+		if (pgdat->bdata->node_bootmem_map)
+			totalram_pages += free_all_bootmem_node(pgdat);
 
 	reserved_pages = 0;
 	efi_memmap_walk(count_reserved_pages, &reserved_pages);
diff --git a/arch/ia64/sn/include/xtalk/hubdev.h b/arch/ia64/sn/include/xtalk/hubdev.h
index 868e7ecae84..580a1c0403a 100644
--- a/arch/ia64/sn/include/xtalk/hubdev.h
+++ b/arch/ia64/sn/include/xtalk/hubdev.h
@@ -8,6 +8,8 @@
 #ifndef _ASM_IA64_SN_XTALK_HUBDEV_H
 #define _ASM_IA64_SN_XTALK_HUBDEV_H
 
+#include "xtalk/xwidgetdev.h"
+
 #define HUB_WIDGET_ID_MAX 0xf
 #define DEV_PER_WIDGET (2*2*8)
 #define IIO_ITTE_WIDGET_BITS    4       /* size of widget field */
diff --git a/arch/ia64/sn/kernel/io_init.c b/arch/ia64/sn/kernel/io_init.c
index 783eb432384..a67f39e448c 100644
--- a/arch/ia64/sn/kernel/io_init.c
+++ b/arch/ia64/sn/kernel/io_init.c
@@ -9,21 +9,28 @@
 #include <linux/bootmem.h>
 #include <linux/nodemask.h>
 #include <asm/sn/types.h>
-#include <asm/sn/sn_sal.h>
 #include <asm/sn/addrs.h>
-#include <asm/sn/pcibus_provider_defs.h>
-#include <asm/sn/pcidev.h>
-#include "pci/pcibr_provider.h"
-#include "xtalk/xwidgetdev.h"
 #include <asm/sn/geo.h>
-#include "xtalk/hubdev.h"
 #include <asm/sn/io.h>
+#include <asm/sn/pcibr_provider.h>
+#include <asm/sn/pcibus_provider_defs.h>
+#include <asm/sn/pcidev.h>
 #include <asm/sn/simulator.h>
+#include <asm/sn/sn_sal.h>
 #include <asm/sn/tioca_provider.h>
+#include "xtalk/hubdev.h"
+#include "xtalk/xwidgetdev.h"
 
-char master_baseio_wid;
 nasid_t master_nasid = INVALID_NASID;	/* Partition Master */
 
+static struct list_head sn_sysdata_list;
+
+/* sysdata list struct */
+struct sysdata_el {
+	struct list_head entry;
+	void *sysdata;
+};
+
 struct slab_info {
 	struct hubdev_info hubdev;
 };
@@ -138,23 +145,6 @@ sal_get_pcidev_info(u64 segment, u64 bus_number, u64 devfn, u64 pci_dev,
 }
 
 /*
- * sn_alloc_pci_sysdata() - This routine allocates a pci controller
- *	which is expected as the pci_dev and pci_bus sysdata by the Linux
- *	PCI infrastructure.
- */
-static inline struct pci_controller *sn_alloc_pci_sysdata(void)
-{
-	struct pci_controller *pci_sysdata;
-
-	pci_sysdata = kmalloc(sizeof(*pci_sysdata), GFP_KERNEL);
-	if (!pci_sysdata)
-		BUG();</