aboutsummaryrefslogtreecommitdiff
path: root/arch/powerpc/platforms/cell/spu_manage.c
diff options
context:
space:
mode:
Diffstat (limited to 'arch/powerpc/platforms/cell/spu_manage.c')
-rw-r--r--arch/powerpc/platforms/cell/spu_manage.c339
1 files changed, 237 insertions, 102 deletions
diff --git a/arch/powerpc/platforms/cell/spu_manage.c b/arch/powerpc/platforms/cell/spu_manage.c
index d8b39fe39cd..c3327f3d8cf 100644
--- a/arch/powerpc/platforms/cell/spu_manage.c
+++ b/arch/powerpc/platforms/cell/spu_manage.c
@@ -21,9 +21,8 @@
#include <linux/interrupt.h>
#include <linux/list.h>
-#include <linux/module.h>
+#include <linux/export.h>
#include <linux/ptrace.h>
-#include <linux/slab.h>
#include <linux/wait.h>
#include <linux/mm.h>
#include <linux/io.h>
@@ -35,6 +34,7 @@
#include <asm/firmware.h>
#include <asm/prom.h>
+#include "spufs/spufs.h"
#include "interrupt.h"
struct device_node *spu_devnode(struct spu *spu)
@@ -48,72 +48,23 @@ static u64 __init find_spu_unit_number(struct device_node *spe)
{
const unsigned int *prop;
int proplen;
- prop = get_property(spe, "unit-id", &proplen);
+
+ /* new device trees should provide the physical-id attribute */
+ prop = of_get_property(spe, "physical-id", &proplen);
if (proplen == 4)
return (u64)*prop;
- prop = get_property(spe, "reg", &proplen);
+ /* celleb device tree provides the unit-id */
+ prop = of_get_property(spe, "unit-id", &proplen);
if (proplen == 4)
return (u64)*prop;
- return 0;
-}
-
-static int __init cell_spuprop_present(struct spu *spu, struct device_node *spe,
- const char *prop)
-{
- const struct address_prop {
- unsigned long address;
- unsigned int len;
- } __attribute__((packed)) *p;
- int proplen;
-
- unsigned long start_pfn, nr_pages;
- struct pglist_data *pgdata;
- struct zone *zone;
- int ret;
-
- p = get_property(spe, prop, &proplen);
- WARN_ON(proplen != sizeof (*p));
-
- start_pfn = p->address >> PAGE_SHIFT;
- nr_pages = ((unsigned long)p->len + PAGE_SIZE - 1) >> PAGE_SHIFT;
-
- pgdata = NODE_DATA(spu->node);
- zone = pgdata->node_zones;
-
- ret = __add_pages(zone, start_pfn, nr_pages);
-
- return ret;
-}
-
-static void __iomem * __init map_spe_prop(struct spu *spu,
- struct device_node *n, const char *name)
-{
- const struct address_prop {
- unsigned long address;
- unsigned int len;
- } __attribute__((packed)) *prop;
-
- const void *p;
- int proplen;
- void __iomem *ret = NULL;
- int err = 0;
-
- p = get_property(n, name, &proplen);
- if (proplen != sizeof (struct address_prop))
- return NULL;
-
- prop = p;
-
- err = cell_spuprop_present(spu, n, name);
- if (err && (err != -EEXIST))
- goto out;
-
- ret = ioremap(prop->address, prop->len);
+ /* legacy device trees provide the id in the reg attribute */
+ prop = of_get_property(spe, "reg", &proplen);
+ if (proplen == 4)
+ return (u64)*prop;
- out:
- return ret;
+ return 0;
}
static void spu_unmap(struct spu *spu)
@@ -133,14 +84,14 @@ static int __init spu_map_interrupts_old(struct spu *spu,
int nid;
/* Get the interrupt source unit from the device-tree */
- tmp = get_property(np, "isrc", NULL);
+ tmp = of_get_property(np, "isrc", NULL);
if (!tmp)
return -ENODEV;
isrc = tmp[0];
- tmp = get_property(np->parent->parent, "node-id", NULL);
+ tmp = of_get_property(np->parent->parent, "node-id", NULL);
if (!tmp) {
- printk(KERN_WARNING "%s: can't find node-id\n", __FUNCTION__);
+ printk(KERN_WARNING "%s: can't find node-id\n", __func__);
nid = spu->node;
} else
nid = tmp[0];
@@ -157,6 +108,23 @@ static int __init spu_map_interrupts_old(struct spu *spu,
return spu->irqs[2] == NO_IRQ ? -EINVAL : 0;
}
+static void __iomem * __init spu_map_prop_old(struct spu *spu,
+ struct device_node *n,
+ const char *name)
+{
+ const struct address_prop {
+ unsigned long address;
+ unsigned int len;
+ } __attribute__((packed)) *prop;
+ int proplen;
+
+ prop = of_get_property(n, name, &proplen);
+ if (prop == NULL || proplen != sizeof (struct address_prop))
+ return NULL;
+
+ return ioremap(prop->address, prop->len);
+}
+
static int __init spu_map_device_old(struct spu *spu)
{
struct device_node *node = spu->devnode;
@@ -164,36 +132,36 @@ static int __init spu_map_device_old(struct spu *spu)
int ret;
ret = -ENODEV;
- spu->name = get_property(node, "name", NULL);
+ spu->name = of_get_property(node, "name", NULL);
if (!spu->name)
goto out;
- prop = get_property(node, "local-store", NULL);
+ prop = of_get_property(node, "local-store", NULL);
if (!prop)
goto out;
spu->local_store_phys = *(unsigned long *)prop;
/* we use local store as ram, not io memory */
spu->local_store = (void __force *)
- map_spe_prop(spu, node, "local-store");
+ spu_map_prop_old(spu, node, "local-store");
if (!spu->local_store)
goto out;
- prop = get_property(node, "problem", NULL);
+ prop = of_get_property(node, "problem", NULL);
if (!prop)
goto out_unmap;
spu->problem_phys = *(unsigned long *)prop;
- spu->problem = map_spe_prop(spu, node, "problem");
+ spu->problem = spu_map_prop_old(spu, node, "problem");
if (!spu->problem)
goto out_unmap;
- spu->priv2 = map_spe_prop(spu, node, "priv2");
+ spu->priv2 = spu_map_prop_old(spu, node, "priv2");
if (!spu->priv2)
goto out_unmap;
if (!firmware_has_feature(FW_FEATURE_LPAR)) {
- spu->priv1 = map_spe_prop(spu, node, "priv1");
+ spu->priv1 = spu_map_prop_old(spu, node, "priv1");
if (!spu->priv1)
goto out_unmap;
}
@@ -209,21 +177,20 @@ out:
static int __init spu_map_interrupts(struct spu *spu, struct device_node *np)
{
- struct of_irq oirq;
+ struct of_phandle_args oirq;
int ret;
int i;
for (i=0; i < 3; i++) {
- ret = of_irq_map_one(np, i, &oirq);
+ ret = of_irq_parse_one(np, i, &oirq);
if (ret) {
pr_debug("spu_new: failed to get irq %d\n", i);
goto err;
}
ret = -EINVAL;
- pr_debug(" irq %d no 0x%x on %s\n", i, oirq.specifier[0],
- oirq.controller->full_name);
- spu->irqs[i] = irq_create_of_mapping(oirq.controller,
- oirq.specifier, oirq.size);
+ pr_debug(" irq %d no 0x%x on %s\n", i, oirq.args[0],
+ oirq.np->full_name);
+ spu->irqs[i] = irq_create_of_mapping(&oirq);
if (spu->irqs[i] == NO_IRQ) {
pr_debug("spu_new: failed to map it !\n");
goto err;
@@ -232,7 +199,7 @@ static int __init spu_map_interrupts(struct spu *spu, struct device_node *np)
return 0;
err:
- pr_debug("failed to map irq %x for spu %s\n", *oirq.specifier,
+ pr_debug("failed to map irq %x for spu %s\n", *oirq.args,
spu->name);
for (; i >= 0; i--) {
if (spu->irqs[i] != NO_IRQ)
@@ -245,34 +212,20 @@ static int spu_map_resource(struct spu *spu, int nr,
void __iomem** virt, unsigned long *phys)
{
struct device_node *np = spu->devnode;
- unsigned long start_pfn, nr_pages;
- struct pglist_data *pgdata;
- struct zone *zone;
struct resource resource = { };
unsigned long len;
int ret;
ret = of_address_to_resource(np, nr, &resource);
if (ret)
- goto out;
-
+ return ret;
if (phys)
*phys = resource.start;
- len = resource.end - resource.start + 1;
+ len = resource_size(&resource);
*virt = ioremap(resource.start, len);
if (!*virt)
- ret = -EINVAL;
-
- start_pfn = resource.start >> PAGE_SHIFT;
- nr_pages = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
-
- pgdata = NODE_DATA(spu->node);
- zone = pgdata->node_zones;
-
- ret = __add_pages(zone, start_pfn, nr_pages);
-
-out:
- return ret;
+ return -EINVAL;
+ return 0;
}
static int __init spu_map_device(struct spu *spu)
@@ -280,7 +233,7 @@ static int __init spu_map_device(struct spu *spu)
struct device_node *np = spu->devnode;
int ret = -ENODEV;
- spu->name = get_property(np, "name", NULL);
+ spu->name = of_get_property(np, "name", NULL);
if (!spu->name)
goto out;
@@ -333,6 +286,7 @@ static int __init of_enumerate_spus(int (*fn)(void *data))
{
int ret;
struct device_node *node;
+ unsigned int n = 0;
ret = -ENODEV;
for (node = of_find_node_by_type(NULL, "spe");
@@ -340,11 +294,12 @@ static int __init of_enumerate_spus(int (*fn)(void *data))
ret = fn(node);
if (ret) {
printk(KERN_WARNING "%s: Error initializing %s\n",
- __FUNCTION__, node->name);
+ __func__, node->name);
break;
}
+ n++;
}
- return ret;
+ return ret ? ret : n;
}
static int __init of_create_spu(struct spu *spu, void *data)
@@ -370,7 +325,7 @@ static int __init of_create_spu(struct spu *spu, void *data)
if (!legacy_map) {
legacy_map = 1;
printk(KERN_WARNING "%s: Legacy device tree found, "
- "trying to map old style\n", __FUNCTION__);
+ "trying to map old style\n", __func__);
}
ret = spu_map_device_old(spu);
if (ret) {
@@ -385,11 +340,11 @@ static int __init of_create_spu(struct spu *spu, void *data)
if (!legacy_irq) {
legacy_irq = 1;
printk(KERN_WARNING "%s: Legacy device tree found, "
- "trying old style irq\n", __FUNCTION__);
+ "trying old style irq\n", __func__);
}
ret = spu_map_interrupts_old(spu, spe);
if (ret) {
- printk(KERN_ERR "%s: could not map interrupts",
+ printk(KERN_ERR "%s: could not map interrupts\n",
spu->name);
goto out_unmap;
}
@@ -413,8 +368,188 @@ static int of_destroy_spu(struct spu *spu)
return 0;
}
+static void enable_spu_by_master_run(struct spu_context *ctx)
+{
+ ctx->ops->master_start(ctx);
+}
+
+static void disable_spu_by_master_run(struct spu_context *ctx)
+{
+ ctx->ops->master_stop(ctx);
+}
+
+/* Hardcoded affinity idxs for qs20 */
+#define QS20_SPES_PER_BE 8
+static int qs20_reg_idxs[QS20_SPES_PER_BE] = { 0, 2, 4, 6, 7, 5, 3, 1 };
+static int qs20_reg_memory[QS20_SPES_PER_BE] = { 1, 1, 0, 0, 0, 0, 0, 0 };
+
+static struct spu *spu_lookup_reg(int node, u32 reg)
+{
+ struct spu *spu;
+ const u32 *spu_reg;
+
+ list_for_each_entry(spu, &cbe_spu_info[node].spus, cbe_list) {
+ spu_reg = of_get_property(spu_devnode(spu), "reg", NULL);
+ if (*spu_reg == reg)
+ return spu;
+ }
+ return NULL;
+}
+
+static void init_affinity_qs20_harcoded(void)
+{
+ int node, i;
+ struct spu *last_spu, *spu;
+ u32 reg;
+
+ for (node = 0; node < MAX_NUMNODES; node++) {
+ last_spu = NULL;
+ for (i = 0; i < QS20_SPES_PER_BE; i++) {
+ reg = qs20_reg_idxs[i];
+ spu = spu_lookup_reg(node, reg);
+ if (!spu)
+ continue;
+ spu->has_mem_affinity = qs20_reg_memory[reg];
+ if (last_spu)
+ list_add_tail(&spu->aff_list,
+ &last_spu->aff_list);
+ last_spu = spu;
+ }
+ }
+}
+
+static int of_has_vicinity(void)
+{
+ struct device_node *dn;
+
+ for_each_node_by_type(dn, "spe") {
+ if (of_find_property(dn, "vicinity", NULL)) {
+ of_node_put(dn);
+ return 1;
+ }
+ }
+ return 0;
+}
+
+static struct spu *devnode_spu(int cbe, struct device_node *dn)
+{
+ struct spu *spu;
+
+ list_for_each_entry(spu, &cbe_spu_info[cbe].spus, cbe_list)
+ if (spu_devnode(spu) == dn)
+ return spu;
+ return NULL;
+}
+
+static struct spu *
+neighbour_spu(int cbe, struct device_node *target, struct device_node *avoid)
+{
+ struct spu *spu;
+ struct device_node *spu_dn;
+ const phandle *vic_handles;
+ int lenp, i;
+
+ list_for_each_entry(spu, &cbe_spu_info[cbe].spus, cbe_list) {
+ spu_dn = spu_devnode(spu);
+ if (spu_dn == avoid)
+ continue;
+ vic_handles = of_get_property(spu_dn, "vicinity", &lenp);
+ for (i=0; i < (lenp / sizeof(phandle)); i++) {
+ if (vic_handles[i] == target->phandle)
+ return spu;
+ }
+ }
+ return NULL;
+}
+
+static void init_affinity_node(int cbe)
+{
+ struct spu *spu, *last_spu;
+ struct device_node *vic_dn, *last_spu_dn;
+ phandle avoid_ph;
+ const phandle *vic_handles;
+ const char *name;
+ int lenp, i, added;
+
+ last_spu = list_first_entry(&cbe_spu_info[cbe].spus, struct spu,
+ cbe_list);
+ avoid_ph = 0;
+ for (added = 1; added < cbe_spu_info[cbe].n_spus; added++) {
+ last_spu_dn = spu_devnode(last_spu);
+ vic_handles = of_get_property(last_spu_dn, "vicinity", &lenp);
+
+ /*
+ * Walk through each phandle in vicinity property of the spu
+ * (tipically two vicinity phandles per spe node)
+ */
+ for (i = 0; i < (lenp / sizeof(phandle)); i++) {
+ if (vic_handles[i] == avoid_ph)
+ continue;
+
+ vic_dn = of_find_node_by_phandle(vic_handles[i]);
+ if (!vic_dn)
+ continue;
+
+ /* a neighbour might be spe, mic-tm, or bif0 */
+ name = of_get_property(vic_dn, "name", NULL);
+ if (!name)
+ continue;
+
+ if (strcmp(name, "spe") == 0) {
+ spu = devnode_spu(cbe, vic_dn);
+ avoid_ph = last_spu_dn->phandle;
+ } else {
+ /*
+ * "mic-tm" and "bif0" nodes do not have
+ * vicinity property. So we need to find the
+ * spe which has vic_dn as neighbour, but
+ * skipping the one we came from (last_spu_dn)
+ */
+ spu = neighbour_spu(cbe, vic_dn, last_spu_dn);
+ if (!spu)
+ continue;
+ if (!strcmp(name, "mic-tm")) {
+ last_spu->has_mem_affinity = 1;
+ spu->has_mem_affinity = 1;
+ }
+ avoid_ph = vic_dn->phandle;
+ }
+
+ list_add_tail(&spu->aff_list, &last_spu->aff_list);
+ last_spu = spu;
+ break;
+ }
+ }
+}
+
+static void init_affinity_fw(void)
+{
+ int cbe;
+
+ for (cbe = 0; cbe < MAX_NUMNODES; cbe++)
+ init_affinity_node(cbe);
+}
+
+static int __init init_affinity(void)
+{
+ if (of_has_vicinity()) {
+ init_affinity_fw();
+ } else {
+ long root = of_get_flat_dt_root();
+ if (of_flat_dt_is_compatible(root, "IBM,CPBW-1.0"))
+ init_affinity_qs20_harcoded();
+ else
+ printk("No affinity configuration found\n");
+ }
+
+ return 0;
+}
+
const struct spu_management_ops spu_management_of_ops = {
.enumerate_spus = of_enumerate_spus,
.create_spu = of_create_spu,
.destroy_spu = of_destroy_spu,
+ .enable_spu = enable_spu_by_master_run,
+ .disable_spu = disable_spu_by_master_run,
+ .init_affinity = init_affinity,
};