1 files changed, 71 insertions, 60 deletions

diff --git a/arch/ia64/sn/kernel/sn2/sn_hwperf.c b/arch/ia64/sn/kernel/sn2/sn_hwperf.c
index 19b54fbcd7e..b9992571c03 100644
--- a/arch/ia64/sn/kernel/sn2/sn_hwperf.c
+++ b/arch/ia64/sn/kernel/sn2/sn_hwperf.c
@@ -3,7 +3,7 @@
  * License.  See the file "COPYING" in the main directory of this archive
  * for more details.
  *
- * Copyright (C) 2004-2005 Silicon Graphics, Inc. All rights reserved.
+ * Copyright (C) 2004-2006 Silicon Graphics, Inc. All rights reserved.
  *
  * SGI Altix topology and hardware performance monitoring API.
  * Mark Goodwin <markgw@sgi.com>. 
@@ -25,17 +25,18 @@
 
 #include <linux/fs.h>
 #include <linux/slab.h>
+#include <linux/export.h>
 #include <linux/vmalloc.h>
 #include <linux/seq_file.h>
 #include <linux/miscdevice.h>
 #include <linux/utsname.h>
 #include <linux/cpumask.h>
-#include <linux/smp_lock.h>
 #include <linux/nodemask.h>
+#include <linux/smp.h>
+#include <linux/mutex.h>
+
 #include <asm/processor.h>
 #include <asm/topology.h>
-#include <asm/smp.h>
-#include <asm/semaphore.h>
 #include <asm/uaccess.h>
 #include <asm/sal.h>
 #include <asm/sn/io.h>
@@ -49,7 +50,9 @@ static void *sn_hwperf_salheap = NULL;
 static int sn_hwperf_obj_cnt = 0;
 static nasid_t sn_hwperf_master_nasid = INVALID_NASID;
 static int sn_hwperf_init(void);
-static DECLARE_MUTEX(sn_hwperf_init_mutex);
+static DEFINE_MUTEX(sn_hwperf_init_mutex);
+
+#define cnode_possible(n)	((n) < num_cnodes)
 
 static int sn_hwperf_enum_objects(int *nobj, struct sn_hwperf_object_info **ret)
 {
@@ -63,7 +66,8 @@ static int sn_hwperf_enum_objects(int *nobj, struct sn_hwperf_object_info **ret)
 	}
 
 	sz = sn_hwperf_obj_cnt * sizeof(struct sn_hwperf_object_info);
-	if ((objbuf = (struct sn_hwperf_object_info *) vmalloc(sz)) == NULL) {
+	objbuf = vmalloc(sz);
+	if (objbuf == NULL) {
 		printk("sn_hwperf_enum_objects: vmalloc(%d) failed\n", (int)sz);
 		e = -ENOMEM;
 		goto out;
@@ -110,7 +114,11 @@ static int sn_hwperf_geoid_to_cnode(char *location)
 	if (sn_hwperf_location_to_bpos(location, &rack, &bay, &slot, &slab))
 		return -1;
 
-	for_each_node(cnode) {
+	/*
+	 * FIXME: replace with cleaner for_each_XXX macro which addresses
+	 * both compute and IO nodes once ACPI3.0 is available.
+	 */
+	for (cnode = 0; cnode < num_cnodes; cnode++) {
 		geoid = cnodeid_get_geoid(cnode);
 		module_id = geo_module(geoid);
 		this_rack = MODULE_GET_RACK(module_id);
@@ -123,14 +131,14 @@ static int sn_hwperf_geoid_to_cnode(char *location)
 		}
 	}
 
-	return node_possible(cnode) ? cnode : -1;
+	return cnode_possible(cnode) ? cnode : -1;
 }
 
 static int sn_hwperf_obj_to_cnode(struct sn_hwperf_object_info * obj)
 {
 	if (!SN_HWPERF_IS_NODE(obj) && !SN_HWPERF_IS_IONODE(obj))
 		BUG();
-	if (!obj->sn_hwp_this_part)
+	if (SN_HWPERF_FOREIGN(obj))
 		return -1;
 	return sn_hwperf_geoid_to_cnode(obj->location);
 }
@@ -182,7 +190,7 @@ static void print_pci_topology(struct seq_file *s)
 	int e;
 
 	for (sz = PAGE_SIZE; sz < 16 * PAGE_SIZE; sz += PAGE_SIZE) {
-		if (!(p = (char *)kmalloc(sz, GFP_KERNEL)))
+		if (!(p = kmalloc(sz, GFP_KERNEL)))
 			break;
 		e = ia64_sn_ioif_get_pci_topology(__pa(p), sz);
 		if (e == SALRET_OK)
@@ -195,12 +203,12 @@ static void print_pci_topology(struct seq_file *s)
 
 static inline int sn_hwperf_has_cpus(cnodeid_t node)
 {
-	return node_online(node) && nr_cpus_node(node);
+	return node < MAX_NUMNODES && node_online(node) && nr_cpus_node(node);
 }
 
 static inline int sn_hwperf_has_mem(cnodeid_t node)
 {
-	return node_online(node) && NODE_DATA(node)->node_present_pages;
+	return node < MAX_NUMNODES && node_online(node) && NODE_DATA(node)->node_present_pages;
 }
 
 static struct sn_hwperf_object_info *
@@ -233,7 +241,7 @@ static int sn_hwperf_get_nearest_node_objdata(struct sn_hwperf_object_info *objb
 	int found_mem = 0;
 	int found_cpu = 0;
 
-	if (!node_possible(node))
+	if (!cnode_possible(node))
 		return -EINVAL;
 
 	if (sn_hwperf_has_cpus(node)) {
@@ -267,8 +275,7 @@ static int sn_hwperf_get_nearest_node_objdata(struct sn_hwperf_object_info *objb
 
 	/* get it's interconnect topology */
 	sz = op->ports * sizeof(struct sn_hwperf_port_info);
-	if (sz > sizeof(ptdata))
-		BUG();
+	BUG_ON(sz > sizeof(ptdata));
 	e = ia64_sn_hwperf_op(sn_hwperf_master_nasid,
 			      SN_HWPERF_ENUM_PORTS, nodeobj->id, sz,
 			      (u64)&ptdata, 0, 0, NULL);
@@ -280,6 +287,8 @@ static int sn_hwperf_get_nearest_node_objdata(struct sn_hwperf_object_info *objb
 	/* find nearest node with cpus and nearest memory */
 	for (router=NULL, j=0; j < op->ports; j++) {
 		dest = sn_hwperf_findobj_id(objbuf, nobj, ptdata[j].conn_id);
+		if (dest && SN_HWPERF_IS_ROUTER(dest))
+			router = dest;
 		if (!dest || SN_HWPERF_FOREIGN(dest) ||
 		    !SN_HWPERF_IS_NODE(dest) || SN_HWPERF_IS_IONODE(dest)) {
 			continue;
@@ -295,15 +304,12 @@ static int sn_hwperf_get_nearest_node_objdata(struct sn_hwperf_object_info *objb
 				*near_mem_node = c;
 			found_mem++;
 		}
-		if (SN_HWPERF_IS_ROUTER(dest))
-			router = dest;
 	}
 
 	if (router && (!found_cpu || !found_mem)) {
 		/* search for a node connected to the same router */
 		sz = router->ports * sizeof(struct sn_hwperf_port_info);
-		if (sz > sizeof(ptdata))
-			BUG();
+		BUG_ON(sz > sizeof(ptdata));
 		e = ia64_sn_hwperf_op(sn_hwperf_master_nasid,
 				      SN_HWPERF_ENUM_PORTS, router->id, sz,
 				      (u64)&ptdata, 0, 0, NULL);
@@ -377,7 +383,6 @@ static int sn_topology_show(struct seq_file *s, void *d)
 	int j;
 	const char *slabname;
 	int ordinal;
-	cpumask_t cpumask;
 	char slice;
 	struct cpuinfo_ia64 *c;
 	struct sn_hwperf_port_info *ptdata;
@@ -409,14 +414,14 @@ static int sn_topology_show(struct seq_file *s, void *d)
 		}
 		seq_printf(s, "partition %u %s local "
 			"shubtype %s, "
-			"nasid_mask 0x%016lx, "
+			"nasid_mask 0x%016llx, "
 			"nasid_bits %d:%d, "
 			"system_size %d, "
 			"sharing_size %d, "
 			"coherency_domain %d, "
 			"region_size %d\n",
 
-			partid, system_utsname.nodename,
+			partid, utsname()->nodename,
 			shubtype ? "shub2" : "shub1", 
 			(u64)nasid_mask << nasid_shift, nasid_msb, nasid_shift,
 			system_size, sharing_size, coher, region_size);
@@ -438,7 +443,7 @@ static int sn_topology_show(struct seq_file *s, void *d)
 	seq_printf(s, "%s %d %s %s asic %s", slabname, ordinal, obj->location,
 		obj->sn_hwp_this_part ? "local" : "shared", obj->name);
 
-	if (!SN_HWPERF_IS_NODE(obj) && !SN_HWPERF_IS_IONODE(obj))
+	if (ordinal < 0 || (!SN_HWPERF_IS_NODE(obj) && !SN_HWPERF_IS_IONODE(obj)))
 		seq_putc(s, '\n');
 	else {
 		cnodeid_t near_mem = -1;
@@ -464,20 +469,20 @@ static int sn_topology_show(struct seq_file *s, void *d)
 		/*
 		 * CPUs on this node, if any
 		 */
-		cpumask = node_to_cpumask(ordinal);
-		for_each_online_cpu(i) {
-			if (cpu_isset(i, cpumask)) {
+		if (!SN_HWPERF_IS_IONODE(obj)) {
+			for_each_cpu_and(i, cpu_online_mask,
+					 cpumask_of_node(ordinal)) {
 				slice = 'a' + cpuid_to_slice(i);
 				c = cpu_data(i);
 				seq_printf(s, "cpu %d %s%c local"
-					" freq %luMHz, arch ia64",
-					i, obj->location, slice,
-					c->proc_freq / 1000000);
+					   " freq %luMHz, arch ia64",
+					   i, obj->location, slice,
+					   c->proc_freq / 1000000);
 				for_each_online_cpu(j) {
 					seq_printf(s, j ? ":%d" : ", dist %d",
-						node_distance(
-						    cpu_to_node(i),
-						    cpu_to_node(j)));
+						   node_distance(
+						    	cpu_to_node(i),
+						    	cpu_to_node(j)));
 				}
 				seq_putc(s, '\n');
 			}
@@ -489,7 +494,7 @@ static int sn_topology_show(struct seq_file *s, void *d)
 		 * numalink ports
 		 */
 		sz = obj->ports * sizeof(struct sn_hwperf_port_info);
-		if ((ptdata = vmalloc(sz)) == NULL)
+		if ((ptdata = kmalloc(sz, GFP_KERNEL)) == NULL)
 			return -ENOMEM;
 		e = ia64_sn_hwperf_op(sn_hwperf_master_nasid,
 				      SN_HWPERF_ENUM_PORTS, obj->id, sz,
@@ -519,7 +524,7 @@ static int sn_topology_show(struct seq_file *s, void *d)
 			if (obj->sn_hwp_this_part && p->sn_hwp_this_part)
 				/* both ends local to this partition */
 				seq_puts(s, " local");
-			else if (!obj->sn_hwp_this_part && !p->sn_hwp_this_part)
+			else if (SN_HWPERF_FOREIGN(p))
 				/* both ends of the link in foreign partiton */
 				seq_puts(s, " foreign");
 			else
@@ -537,7 +542,7 @@ static int sn_topology_show(struct seq_file *s, void *d)
 				(SN_HWPERF_IS_NL3ROUTER(obj) ||
 				SN_HWPERF_IS_NL3ROUTER(p)) ?  "LLP3" : "LLP4");
 		}
-		vfree(ptdata);
+		kfree(ptdata);
 	}
 
 	return 0;
@@ -567,7 +572,7 @@ static void sn_topology_stop(struct seq_file *m, void *v)
 /*
  * /proc/sgi_sn/sn_topology, read-only using seq_file
  */
-static struct seq_operations sn_topology_seq_ops = {
+static const struct seq_operations sn_topology_seq_ops = {
 	.start = sn_topology_start,
 	.next = sn_topology_next,
 	.stop = sn_topology_stop,
@@ -605,28 +610,32 @@ static int sn_hwperf_op_cpu(struct sn_hwperf_op_info *op_info)
 	op_info->a->arg &= SN_HWPERF_ARG_OBJID_MASK;
 
 	if (cpu != SN_HWPERF_ARG_ANY_CPU) {
-		if (cpu >= num_online_cpus() || !cpu_online(cpu)) {
+		if (cpu >= nr_cpu_ids || !cpu_online(cpu)) {
 			r = -EINVAL;
 			goto out;
 		}
 	}
 
-	if (cpu == SN_HWPERF_ARG_ANY_CPU || cpu == get_cpu()) {
-		/* don't care, or already on correct cpu */
+	if (cpu == SN_HWPERF_ARG_ANY_CPU) {
+		/* don't care which cpu */
 		sn_hwperf_call_sal(op_info);
-	}
-	else {
+	} else if (cpu == get_cpu()) {
+		/* already on correct cpu */
+		sn_hwperf_call_sal(op_info);
+		put_cpu();
+	} else {
+		put_cpu();
 		if (use_ipi) {
 			/* use an interprocessor interrupt to call SAL */
 			smp_call_function_single(cpu, sn_hwperf_call_sal,
-				op_info, 1, 1);
+				op_info, 1);
 		}
 		else {
 			/* migrate the task before calling SAL */ 
 			save_allowed = current->cpus_allowed;
-			set_cpus_allowed(current, cpumask_of_cpu(cpu));
+			set_cpus_allowed_ptr(current, cpumask_of(cpu));
 			sn_hwperf_call_sal(op_info);
-			set_cpus_allowed(current, save_allowed);
+			set_cpus_allowed_ptr(current, &save_allowed);
 		}
 	}
 	r = op_info->ret;
@@ -677,8 +686,7 @@ static int sn_hwperf_map_err(int hwperf_err)
 /*
  * ioctl for "sn_hwperf" misc device
  */
-static int
-sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
+static long sn_hwperf_ioctl(struct file *fp, u32 op, unsigned long arg)
 {
 	struct sn_hwperf_ioctl_args a;
 	struct cpuinfo_ia64 *cdata;
@@ -694,8 +702,6 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 	int i;
 	int j;
 
-	unlock_kernel();
-
 	/* only user requests are allowed here */
 	if ((op & SN_HWPERF_OP_MASK) < 10) {
 		r = -EINVAL;
@@ -741,9 +747,10 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 			goto error;
 		} else
 		if ((r = sn_hwperf_enum_objects(&nobj, &objs)) == 0) {
+			int cpuobj_index = 0;
+
 			memset(p, 0, a.sz);
 			for (i = 0; i < nobj; i++) {
-				int cpuobj_index = 0;
 				if (!SN_HWPERF_IS_NODE(objs + i))
 					continue;
 				node = sn_hwperf_obj_to_cnode(objs + i);
@@ -772,7 +779,7 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 
 	case SN_HWPERF_GET_NODE_NASID:
 		if (a.sz != sizeof(u64) ||
-		   (node = a.arg) < 0 || !node_possible(node)) {
+		   (node = a.arg) < 0 || !cnode_possible(node)) {
 			r = -EINVAL;
 			goto error;
 		}
@@ -780,17 +787,18 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 		break;
 
 	case SN_HWPERF_GET_OBJ_NODE:
-		if (a.sz != sizeof(u64) || a.arg < 0) {
+		i = a.arg;
+		if (a.sz != sizeof(u64) || i < 0) {
 			r = -EINVAL;
 			goto error;
 		}
 		if ((r = sn_hwperf_enum_objects(&nobj, &objs)) == 0) {
-			if (a.arg >= nobj) {
+			if (i >= nobj) {
 				r = -EINVAL;
 				vfree(objs);
 				goto error;
 			}
-			if (objs[(i = a.arg)].id != a.arg) {
+			if (objs[i].id != a.arg) {
 				for (i = 0; i < nobj; i++) {
 					if (objs[i].id == a.arg)
 						break;
@@ -852,12 +860,12 @@ sn_hwperf_ioctl(struct inode *in, struct file *fp, u32 op, u64 arg)
 error:
 	vfree(p);
 
-	lock_kernel();
 	return r;
 }
 
-static struct file_operations sn_hwperf_fops = {
-	.ioctl = sn_hwperf_ioctl,
+static const struct file_operations sn_hwperf_fops = {
+	.unlocked_ioctl = sn_hwperf_ioctl,
+	.llseek = noop_llseek,
 };
 
 static struct miscdevice sn_hwperf_dev = {
@@ -873,10 +881,10 @@ static int sn_hwperf_init(void)
 	int e = 0;
 
 	/* single threaded, once-only initialization */
-	down(&sn_hwperf_init_mutex);
+	mutex_lock(&sn_hwperf_init_mutex);
 
 	if (sn_hwperf_salheap) {
-		up(&sn_hwperf_init_mutex);
+		mutex_unlock(&sn_hwperf_init_mutex);
 		return e;
 	}
 
@@ -925,7 +933,7 @@ out:
 		sn_hwperf_salheap = NULL;
 		sn_hwperf_obj_cnt = 0;
 	}
-	up(&sn_hwperf_init_mutex);
+	mutex_unlock(&sn_hwperf_init_mutex);
 	return e;
 }
 
@@ -969,10 +977,13 @@ int sn_hwperf_get_nearest_node(cnodeid_t node,
 	return e;
 }
 
-static int __devinit sn_hwperf_misc_register_init(void)
+static int sn_hwperf_misc_register_init(void)
 {
 	int e;
 
+	if (!ia64_platform_is("sn2"))
+		return 0;
+
 	sn_hwperf_init();
 
 	/*