[V2 1/8] drm/amd/pm: introduce a new set of OD interfaces
Evan Quan
evan.quan at amd.com
Wed Aug 23 06:33:59 UTC 2023
There will be multiple interfaces(sysfs files) exposed with each representing
a single OD functionality. And all those interface will be arranged in a tree
liked hierarchy with the top dir as "gpu_od". Meanwhile all functionalities
for the same component will be arranged under the same directory.
Signed-off-by: Evan Quan <evan.quan at amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +
drivers/gpu/drm/amd/pm/amdgpu_pm.c | 264 ++++++++++++++++++++-
drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h | 2 +
3 files changed, 266 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 533daba2accb..7aae2801b36e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -3623,6 +3623,8 @@ int amdgpu_device_init(struct amdgpu_device *adev,
INIT_LIST_HEAD(&adev->ras_list);
+ INIT_LIST_HEAD(&adev->pm.od_kobj_list);
+
INIT_DELAYED_WORK(&adev->delayed_init_work,
amdgpu_device_delayed_init_work_handler);
INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work,
diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c
index b0ef3d8b0ecf..beb3303fc832 100644
--- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c
+++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c
@@ -35,6 +35,44 @@
#include <linux/pm_runtime.h>
#include <asm/processor.h>
+#define MAX_NUM_OF_FEATURES_PER_SUBSET 8
+#define MAX_NUM_OF_SUBSETS 8
+
+struct od_attribute {
+ struct kobj_attribute attribute;
+ struct list_head entry;
+};
+
+struct od_kobj {
+ struct kobject kobj;
+ struct list_head entry;
+ struct list_head attribute;
+ void *priv;
+};
+
+struct od_feature_ops {
+ umode_t (*is_visible)(struct amdgpu_device *adev);
+ ssize_t (*show)(struct kobject *kobj, struct kobj_attribute *attr,
+ char *buf);
+ ssize_t (*store)(struct kobject *kobj, struct kobj_attribute *attr,
+ const char *buf, size_t count);
+};
+
+struct od_feature_item {
+ const char *name;
+ struct od_feature_ops ops;
+};
+
+struct od_feature_container {
+ char *name;
+ struct od_feature_ops ops;
+ struct od_feature_item sub_feature[MAX_NUM_OF_FEATURES_PER_SUBSET];
+};
+
+struct od_feature_set {
+ struct od_feature_container containers[MAX_NUM_OF_SUBSETS];
+};
+
static const struct hwmon_temp_label {
enum PP_HWMON_TEMP channel;
const char *label;
@@ -3345,10 +3383,216 @@ static const struct attribute_group *hwmon_groups[] = {
NULL
};
-int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
+static struct od_feature_set amdgpu_od_set;
+
+static void od_kobj_release(struct kobject *kobj)
+{
+ struct od_kobj *od_kobj = container_of(kobj, struct od_kobj, kobj);
+
+ kfree(od_kobj);
+}
+
+static const struct kobj_type od_ktype = {
+ .release = od_kobj_release,
+ .sysfs_ops = &kobj_sysfs_ops,
+};
+
+static void amdgpu_od_set_fini(struct amdgpu_device *adev)
+{
+ struct od_kobj *container, *container_next;
+ struct od_attribute *attribute, *attribute_next;
+
+ if (list_empty(&adev->pm.od_kobj_list))
+ return;
+
+ list_for_each_entry_safe(container, container_next,
+ &adev->pm.od_kobj_list, entry) {
+ list_del(&container->entry);
+
+ list_for_each_entry_safe(attribute, attribute_next,
+ &container->attribute, entry) {
+ list_del(&attribute->entry);
+ sysfs_remove_file(&container->kobj,
+ &attribute->attribute.attr);
+ kfree(attribute);
+ }
+
+ kobject_put(&container->kobj);
+ }
+}
+
+static bool amdgpu_is_od_feature_supported(struct amdgpu_device *adev,
+ struct od_feature_ops *feature_ops)
+{
+ umode_t mode;
+
+ if (!feature_ops->is_visible)
+ return false;
+
+ /*
+ * If the feature has no user read and write mode set,
+ * we can assume the feature is actually not supported.(?)
+ * And the revelant sysfs interface should not be exposed.
+ */
+ mode = feature_ops->is_visible(adev);
+ if (mode & (S_IRUSR | S_IWUSR))
+ return true;
+
+ return false;
+}
+
+static bool amdgpu_od_is_self_contained(struct amdgpu_device *adev,
+ struct od_feature_container *container)
+{
+ int i;
+
+ /*
+ * If there is no valid entry within the container, the container
+ * is recognized as a self contained container. And the valid entry
+ * here means it has a valid naming and it is visible/supported by
+ * the ASIC.
+ */
+ for (i = 0; i < ARRAY_SIZE(container->sub_feature); i++) {
+ if (container->sub_feature[i].name &&
+ amdgpu_is_od_feature_supported(adev,
+ &container->sub_feature[i].ops))
+ return false;
+ }
+
+ return true;
+}
+
+static int amdgpu_od_set_init(struct amdgpu_device *adev)
{
+ struct od_kobj *top_set, *sub_set;
+ struct od_attribute *attribute;
+ struct od_feature_container *container;
+ struct od_feature_item *feature;
+ int i, j;
int ret;
+
+ /* Setup the top `gpu_od` directory which holds all other OD interfaces */
+ top_set = kzalloc(sizeof(*top_set), GFP_KERNEL);
+ if (!top_set)
+ return -ENOMEM;
+ list_add(&top_set->entry, &adev->pm.od_kobj_list);
+
+ ret = kobject_init_and_add(&top_set->kobj,
+ &od_ktype,
+ &adev->dev->kobj,
+ "%s",
+ "gpu_od");
+ if (ret)
+ goto err_out;
+ INIT_LIST_HEAD(&top_set->attribute);
+ top_set->priv = adev;
+
+ for (i = 0; i < ARRAY_SIZE(amdgpu_od_set.containers); i++) {
+ container = &amdgpu_od_set.containers[i];
+
+ if (!container->name)
+ continue;
+
+ /*
+ * If there is valid entries within the container, the container
+ * will be presented as a sub directory and all its holding entries
+ * will be presented as plain files under it.
+ * While if there is no valid entry within the container, the container
+ * itself will be presented as a plain file under top `gpu_od` directory.
+ */
+ if (amdgpu_od_is_self_contained(adev, container)) {
+ if (!amdgpu_is_od_feature_supported(adev,
+ &container->ops))
+ continue;
+
+ /*
+ * The container is presented as a plain file under top `gpu_od`
+ * directory.
+ */
+ attribute = kzalloc(sizeof(*attribute), GFP_KERNEL);
+ if (!attribute) {
+ ret = -ENOMEM;
+ goto err_out;
+ }
+ list_add(&attribute->entry, &top_set->attribute);
+
+ attribute->attribute.attr.mode =
+ container->ops.is_visible(adev);
+ attribute->attribute.attr.name = container->name;
+ attribute->attribute.show =
+ container->ops.show;
+ attribute->attribute.store =
+ container->ops.store;
+ ret = sysfs_create_file(&top_set->kobj,
+ &attribute->attribute.attr);
+ if (ret)
+ goto err_out;
+ } else {
+ /* The container is presented as a sub directory. */
+ sub_set = kzalloc(sizeof(*sub_set), GFP_KERNEL);
+ if (!sub_set) {
+ ret = -ENOMEM;
+ goto err_out;
+ }
+ list_add(&sub_set->entry, &adev->pm.od_kobj_list);
+
+ ret = kobject_init_and_add(&sub_set->kobj,
+ &od_ktype,
+ &top_set->kobj,
+ "%s",
+ container->name);
+ if (ret)
+ goto err_out;
+ INIT_LIST_HEAD(&sub_set->attribute);
+ sub_set->priv = adev;
+
+ for (j = 0; j < ARRAY_SIZE(container->sub_feature); j++) {
+ feature = &container->sub_feature[j];
+ if (!feature->name)
+ continue;
+
+ if (!amdgpu_is_od_feature_supported(adev,
+ &feature->ops))
+ continue;
+
+ /*
+ * With the container presented as a sub directory, the entry within
+ * it is presented as a plain file under the sub directory.
+ */
+ attribute = kzalloc(sizeof(*attribute), GFP_KERNEL);
+ if (!attribute) {
+ ret = -ENOMEM;
+ goto err_out;
+ }
+ list_add(&attribute->entry, &sub_set->attribute);
+
+ attribute->attribute.attr.mode =
+ feature->ops.is_visible(adev);
+ attribute->attribute.attr.name = feature->name;
+ attribute->attribute.show =
+ feature->ops.show;
+ attribute->attribute.store =
+ feature->ops.store;
+ ret = sysfs_create_file(&sub_set->kobj,
+ &attribute->attribute.attr);
+ if (ret)
+ goto err_out;
+ }
+ }
+ }
+
+ return 0;
+
+err_out:
+ amdgpu_od_set_fini(adev);
+
+ return ret;
+}
+
+int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
+{
uint32_t mask = 0;
+ int ret;
if (adev->pm.sysfs_initialized)
return 0;
@@ -3387,15 +3631,31 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev)
mask,
&adev->pm.pm_attr_list);
if (ret)
- return ret;
+ goto err_out0;
+
+ if (amdgpu_dpm_is_overdrive_supported(adev)) {
+ ret = amdgpu_od_set_init(adev);
+ if (ret)
+ goto err_out1;
+ }
adev->pm.sysfs_initialized = true;
return 0;
+
+err_out1:
+ amdgpu_device_attr_remove_groups(adev, &adev->pm.pm_attr_list);
+err_out0:
+ if (adev->pm.int_hwmon_dev)
+ hwmon_device_unregister(adev->pm.int_hwmon_dev);
+
+ return ret;
}
void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev)
{
+ amdgpu_od_set_fini(adev);
+
if (adev->pm.int_hwmon_dev)
hwmon_device_unregister(adev->pm.int_hwmon_dev);
diff --git a/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h b/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h
index 42172b00be66..4cab6a2efb63 100644
--- a/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h
+++ b/drivers/gpu/drm/amd/pm/inc/amdgpu_dpm.h
@@ -366,6 +366,8 @@ struct amdgpu_pm {
struct config_table_setting config_table;
/* runtime mode */
enum amdgpu_runpm_mode rpm_mode;
+
+ struct list_head od_kobj_list;
};
int amdgpu_dpm_read_sensor(struct amdgpu_device *adev, enum amd_pp_sensors sensor,
--
2.34.1
More information about the amd-gfx
mailing list