drm/amdkfd: Add thermal throttling SMI event

Add support for reporting thermal throttling events through SMI.
Also, add a counter to count the number of throttling interrupts
observed and report the count in the SMI event message.

Signed-off-by: Mukul Joshi <mukul.joshi@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

1 files changed, 50 insertions, 17 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index 7b348bf9df21..86c2c3e97944 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
@@ -24,6 +24,7 @@
 #include <linux/wait.h>
 #include <linux/anon_inodes.h>
 #include <uapi/linux/kfd_ioctl.h>
+#include "amdgpu.h"
 #include "amdgpu_vm.h"
 #include "kfd_priv.h"
 #include "kfd_smi_events.h"
@@ -148,6 +149,54 @@ static int kfd_smi_ev_release(struct inode *inode, struct file *filep)
 	return 0;
 }
 
+static void add_event_to_kfifo(struct kfd_dev *dev, unsigned long long smi_event,
+			      char *event_msg, int len)
+{
+	struct kfd_smi_client *client;
+
+	rcu_read_lock();
+
+	list_for_each_entry_rcu(client, &dev->smi_clients, list) {
+		if (!(READ_ONCE(client->events) & smi_event))
+			continue;
+		spin_lock(&client->lock);
+		if (kfifo_avail(&client->fifo) >= len) {
+			kfifo_in(&client->fifo, event_msg, len);
+			wake_up_all(&client->wait_queue);
+		} else {
+			pr_debug("smi_event(EventID: %llu): no space left\n",
+					smi_event);
+		}
+		spin_unlock(&client->lock);
+	}
+
+	rcu_read_unlock();
+}
+
+void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev,
+					     uint32_t throttle_bitmask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)dev->kgd;
+	/*
+	 * ThermalThrottle msg = throttle_bitmask(8):
+	 * 			 thermal_interrupt_count(16):
+	 * 16 bytes event + 1 byte space + 8 byte throttle_bitmask +
+	 * 1 byte : + 16 byte thermal_interupt_counter + 1 byte \n +
+	 * 1 byte \0 = 44
+	 */
+	char fifo_in[44];
+	int len;
+
+	if (list_empty(&dev->smi_clients))
+		return;
+
+	len = snprintf(fifo_in, 44, "%x %x:%llx\n",
+		       KFD_SMI_EVENT_THERMAL_THROTTLE, throttle_bitmask,
+		       atomic64_read(&adev->smu.throttle_int_counter));
+
+	add_event_to_kfifo(dev, KFD_SMI_EVENT_THERMAL_THROTTLE, fifo_in, len);
+}
+
 void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)dev->kgd;
@@ -156,7 +205,6 @@ void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid)
 	/* 16 bytes event + 1 byte space + 25 bytes msg + 1 byte \n = 43
 	 */
 	char fifo_in[43];
-	struct kfd_smi_client *client;
 	int len;
 
 	if (list_empty(&dev->smi_clients))
@@ -171,22 +219,7 @@ void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid)
 	len = snprintf(fifo_in, 43, "%x %x:%s\n", KFD_SMI_EVENT_VMFAULT,
 		task_info.pid, task_info.task_name);
 
-	rcu_read_lock();
-
-	list_for_each_entry_rcu(client, &dev->smi_clients, list) {
-		if (!(READ_ONCE(client->events) & KFD_SMI_EVENT_VMFAULT))
-			continue;
-		spin_lock(&client->lock);
-		if (kfifo_avail(&client->fifo) >= len) {
-			kfifo_in(&client->fifo, fifo_in, len);
-			wake_up_all(&client->wait_queue);
-		}
-		else
-			pr_debug("smi_event(vmfault): no space left\n");
-		spin_unlock(&client->lock);
-	}
-
-	rcu_read_unlock();
+	add_event_to_kfifo(dev, KFD_SMI_EVENT_VMFAULT, fifo_in, len);
 }
 
 int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd)