static DEFINE_MUTEX(xgmi_mutex);
-#define AMDGPU_MAX_XGMI_HIVE 8
#define AMDGPU_MAX_XGMI_DEVICE_PER_HIVE 4
-static struct amdgpu_hive_info xgmi_hives[AMDGPU_MAX_XGMI_HIVE];
-static unsigned hive_count = 0;
+static LIST_HEAD(xgmi_hive_list);
static const int xgmi_pcs_err_status_reg_vg20[] = {
smnXGMI0_PCS_GOPX16_PCS_ERROR_STATUS,
*
*/
+static struct attribute amdgpu_xgmi_hive_id = {
+ .name = "xgmi_hive_id",
+ .mode = S_IRUGO
+};
-static ssize_t amdgpu_xgmi_show_hive_id(struct device *dev,
- struct device_attribute *attr, char *buf)
-{
- struct amdgpu_hive_info *hive =
- container_of(attr, struct amdgpu_hive_info, dev_attr);
-
- return snprintf(buf, PAGE_SIZE, "%llu\n", hive->hive_id);
-}
+static struct attribute *amdgpu_xgmi_hive_attrs[] = {
+ &amdgpu_xgmi_hive_id,
+ NULL
+};
-static int amdgpu_xgmi_sysfs_create(struct amdgpu_device *adev,
- struct amdgpu_hive_info *hive)
+static ssize_t amdgpu_xgmi_show_attrs(struct kobject *kobj,
+ struct attribute *attr, char *buf)
{
- int ret = 0;
-
- if (WARN_ON(hive->kobj))
- return -EINVAL;
-
- hive->kobj = kobject_create_and_add("xgmi_hive_info", &adev->dev->kobj);
- if (!hive->kobj) {
- dev_err(adev->dev, "XGMI: Failed to allocate sysfs entry!\n");
- return -EINVAL;
- }
-
- hive->dev_attr = (struct device_attribute) {
- .attr = {
- .name = "xgmi_hive_id",
- .mode = S_IRUGO,
+ struct amdgpu_hive_info *hive = container_of(
+ kobj, struct amdgpu_hive_info, kobj);
- },
- .show = amdgpu_xgmi_show_hive_id,
- };
-
- ret = sysfs_create_file(hive->kobj, &hive->dev_attr.attr);
- if (ret) {
- dev_err(adev->dev, "XGMI: Failed to create device file xgmi_hive_id\n");
- kobject_del(hive->kobj);
- kobject_put(hive->kobj);
- hive->kobj = NULL;
- }
+ if (attr == &amdgpu_xgmi_hive_id)
+ return snprintf(buf, PAGE_SIZE, "%llu\n", hive->hive_id);
- return ret;
+ return 0;
}
-static void amdgpu_xgmi_sysfs_destroy(struct amdgpu_device *adev,
- struct amdgpu_hive_info *hive)
+static void amdgpu_xgmi_hive_release(struct kobject *kobj)
{
- sysfs_remove_file(hive->kobj, &hive->dev_attr.attr);
- kobject_del(hive->kobj);
- kobject_put(hive->kobj);
- hive->kobj = NULL;
+ struct amdgpu_hive_info *hive = container_of(
+ kobj, struct amdgpu_hive_info, kobj);
+
+ mutex_destroy(&hive->hive_lock);
+ kfree(hive);
}
+static const struct sysfs_ops amdgpu_xgmi_hive_ops = {
+ .show = amdgpu_xgmi_show_attrs,
+};
+
+struct kobj_type amdgpu_xgmi_hive_type = {
+ .release = amdgpu_xgmi_hive_release,
+ .sysfs_ops = &amdgpu_xgmi_hive_ops,
+ .default_attrs = amdgpu_xgmi_hive_attrs,
+};
+
static ssize_t amdgpu_xgmi_show_device_id(struct device *dev,
struct device_attribute *attr,
char *buf)
/* Create sysfs link to hive info folder on the first device */
- if (adev != hive->adev) {
- ret = sysfs_create_link(&adev->dev->kobj, hive->kobj,
+ if (hive->kobj.parent != (&adev->dev->kobj)) {
+ ret = sysfs_create_link(&adev->dev->kobj, &hive->kobj,
"xgmi_hive_info");
if (ret) {
dev_err(adev->dev, "XGMI: Failed to create link to hive info");
}
}
- sprintf(node, "node%d", hive->number_devices);
+ sprintf(node, "node%d", atomic_read(&hive->number_devices));
/* Create sysfs link form the hive folder to yourself */
- ret = sysfs_create_link(hive->kobj, &adev->dev->kobj, node);
+ ret = sysfs_create_link(&hive->kobj, &adev->dev->kobj, node);
if (ret) {
dev_err(adev->dev, "XGMI: Failed to create link from hive info");
goto remove_link;
device_remove_file(adev->dev, &dev_attr_xgmi_device_id);
device_remove_file(adev->dev, &dev_attr_xgmi_error);
- if (adev != hive->adev)
+ if (hive->kobj.parent != (&adev->dev->kobj))
sysfs_remove_link(&adev->dev->kobj,"xgmi_hive_info");
- sprintf(node, "node%d", hive->number_devices);
- sysfs_remove_link(hive->kobj, node);
+ sprintf(node, "node%d", atomic_read(&hive->number_devices));
+ sysfs_remove_link(&hive->kobj, node);
}
-struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev, int lock)
+struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev)
{
- int i;
- struct amdgpu_hive_info *tmp;
+ struct amdgpu_hive_info *hive = NULL, *tmp = NULL;
+ int ret;
if (!adev->gmc.xgmi.hive_id)
return NULL;
+ if (adev->hive) {
+ kobject_get(&adev->hive->kobj);
+ return adev->hive;
+ }
+
mutex_lock(&xgmi_mutex);
- for (i = 0 ; i < hive_count; ++i) {
- tmp = &xgmi_hives[i];
- if (tmp->hive_id == adev->gmc.xgmi.hive_id) {
- if (lock)
- mutex_lock(&tmp->hive_lock);
- mutex_unlock(&xgmi_mutex);
- return tmp;
+ if (!list_empty(&xgmi_hive_list)) {
+ list_for_each_entry_safe(hive, tmp, &xgmi_hive_list, node) {
+ if (hive->hive_id == adev->gmc.xgmi.hive_id)
+ goto pro_end;
}
}
- if (i >= AMDGPU_MAX_XGMI_HIVE) {
- mutex_unlock(&xgmi_mutex);
- return NULL;
+
+ hive = kzalloc(sizeof(*hive), GFP_KERNEL);
+ if (!hive) {
+ dev_err(adev->dev, "XGMI: allocation failed\n");
+ hive = NULL;
+ goto pro_end;
}
/* initialize new hive if not exist */
- tmp = &xgmi_hives[hive_count++];
-
- if (amdgpu_xgmi_sysfs_create(adev, tmp)) {
- mutex_unlock(&xgmi_mutex);
- return NULL;
+ ret = kobject_init_and_add(&hive->kobj,
+ &amdgpu_xgmi_hive_type,
+ &adev->dev->kobj,
+ "%s", "xgmi_hive_info");
+ if (ret) {
+ dev_err(adev->dev, "XGMI: failed initializing kobject for xgmi hive\n");
+ kfree(hive);
+ hive = NULL;
+ goto pro_end;
}
- tmp->adev = adev;
- tmp->hive_id = adev->gmc.xgmi.hive_id;
- INIT_LIST_HEAD(&tmp->device_list);
- mutex_init(&tmp->hive_lock);
- atomic_set(&tmp->in_reset, 0);
- task_barrier_init(&tmp->tb);
-
- if (lock)
- mutex_lock(&tmp->hive_lock);
- tmp->pstate = AMDGPU_XGMI_PSTATE_UNKNOWN;
- tmp->hi_req_gpu = NULL;
+ hive->hive_id = adev->gmc.xgmi.hive_id;
+ INIT_LIST_HEAD(&hive->device_list);
+ INIT_LIST_HEAD(&hive->node);
+ mutex_init(&hive->hive_lock);
+ atomic_set(&hive->in_reset, 0);
+ atomic_set(&hive->number_devices, 0);
+ task_barrier_init(&hive->tb);
+ hive->pstate = AMDGPU_XGMI_PSTATE_UNKNOWN;
+ hive->hi_req_gpu = NULL;
/*
* hive pstate on boot is high in vega20 so we have to go to low
* pstate on after boot.
*/
- tmp->hi_req_count = AMDGPU_MAX_XGMI_DEVICE_PER_HIVE;
+ hive->hi_req_count = AMDGPU_MAX_XGMI_DEVICE_PER_HIVE;
+ list_add_tail(&hive->node, &xgmi_hive_list);
+
+pro_end:
+ if (hive)
+ kobject_get(&hive->kobj);
mutex_unlock(&xgmi_mutex);
+ return hive;
+}
- return tmp;
+void amdgpu_put_xgmi_hive(struct amdgpu_hive_info *hive)
+{
+ if (hive)
+ kobject_put(&hive->kobj);
}
int amdgpu_xgmi_set_pstate(struct amdgpu_device *adev, int pstate)
{
int ret = 0;
- struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev, 0);
+ struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
struct amdgpu_device *request_adev = hive->hi_req_gpu ?
hive->hi_req_gpu : adev;
bool is_hi_req = pstate == AMDGPU_XGMI_PSTATE_MAX_VEGA20;
bool init_low = hive->pstate == AMDGPU_XGMI_PSTATE_UNKNOWN;
+ amdgpu_put_xgmi_hive(hive);
/* fw bug so temporarily disable pstate switching */
return 0;
/* Each psp need to set the latest topology */
ret = psp_xgmi_set_topology_info(&adev->psp,
- hive->number_devices,
+ atomic_read(&hive->number_devices),
&adev->psp.xgmi_context.top_info);
if (ret)
dev_err(adev->dev,
adev->gmc.xgmi.node_id = adev->gmc.xgmi.physical_node_id + 16;
}
- hive = amdgpu_get_xgmi_hive(adev, 1);
+ hive = amdgpu_get_xgmi_hive(adev);
if (!hive) {
ret = -EINVAL;
dev_err(adev->dev,
adev->gmc.xgmi.node_id, adev->gmc.xgmi.hive_id);
goto exit;
}
+ mutex_lock(&hive->hive_lock);
top_info = &adev->psp.xgmi_context.top_info;
list_for_each_entry(entry, &hive->device_list, head)
top_info->nodes[count++].node_id = entry->node_id;
top_info->num_nodes = count;
- hive->number_devices = count;
+ atomic_set(&hive->number_devices, count);
task_barrier_add_task(&hive->tb);
exit_unlock:
mutex_unlock(&hive->hive_lock);
exit:
- if (!ret)
+ if (!ret) {
+ adev->hive = hive;
dev_info(adev->dev, "XGMI: Add node %d, hive 0x%llx.\n",
adev->gmc.xgmi.physical_node_id, adev->gmc.xgmi.hive_id);
- else
+ } else {
+ amdgpu_put_xgmi_hive(hive);
dev_err(adev->dev, "XGMI: Failed to add node %d, hive 0x%llx ret: %d\n",
adev->gmc.xgmi.physical_node_id, adev->gmc.xgmi.hive_id,
ret);
+ }
return ret;
}
int amdgpu_xgmi_remove_device(struct amdgpu_device *adev)
{
- struct amdgpu_hive_info *hive;
+ struct amdgpu_hive_info *hive = adev->hive;
if (!adev->gmc.xgmi.supported)
return -EINVAL;
- hive = amdgpu_get_xgmi_hive(adev, 1);
if (!hive)
return -EINVAL;
+ mutex_lock(&hive->hive_lock);
task_barrier_rem_task(&hive->tb);
amdgpu_xgmi_sysfs_rem_dev_info(adev, hive);
+ if (hive->hi_req_gpu == adev)
+ hive->hi_req_gpu = NULL;
+ list_del(&adev->gmc.xgmi.head);
mutex_unlock(&hive->hive_lock);
- if(!(--hive->number_devices)){
- amdgpu_xgmi_sysfs_destroy(adev, hive);
- mutex_destroy(&hive->hive_lock);
+ amdgpu_put_xgmi_hive(hive);
+ adev->hive = NULL;
+
+ if (atomic_dec_return(&hive->number_devices) == 0) {
+ /* Remove the hive from global hive list */
+ mutex_lock(&xgmi_mutex);
+ list_del(&hive->node);
+ mutex_unlock(&xgmi_mutex);
+
+ amdgpu_put_xgmi_hive(hive);
}
return psp_xgmi_terminate(&adev->psp);
projects / linux.git / commitdiff