From 71cd434f4f03f82f729ce39c7a429337607e6c10 Mon Sep 17 00:00:00 2001
From: Rajkumar Kasirajan <rkasirajan@nvidia.com>
Date: Sun, 14 Nov 2021 23:14:31 +0800
Subject: [PATCH] gpu: nvgpu: add thermal cooling support

Add devfeq thermal cooling device for GPU software thermal throttling.

Bug 3287074

Change-Id: Ib0b53a58177964dfda3c8993da9c4835e2cb8a6e
Signed-off-by: Rajkumar Kasirajan <rkasirajan@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/c/linux-nvgpu/+/2625659
Reviewed-by: svcacv <svcacv@nvidia.com>
Reviewed-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-by: Srikar Srimath Tirumala <srikars@nvidia.com>
Reviewed-by: Bibek Basu <bbasu@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: Seshendra Gadagottu <sgadagottu@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
GVS: Gerrit_Virtual_Submit
---
 drivers/gpu/nvgpu/os/linux/os_linux.h |  3 +++
 drivers/gpu/nvgpu/os/linux/scale.c    | 21 +++++++++++++++++++++
 2 files changed, 24 insertions(+)

diff --git a/drivers/gpu/nvgpu/os/linux/os_linux.h b/drivers/gpu/nvgpu/os/linux/os_linux.h
index 902db354d..a7ef5319b 100644
--- a/drivers/gpu/nvgpu/os/linux/os_linux.h
+++ b/drivers/gpu/nvgpu/os/linux/os_linux.h
@@ -74,6 +74,9 @@ struct nvgpu_os_linux {
 	struct device *dev;
 	struct dgpu_thermal_alert thermal_alert;
 	struct nvgpu_interrupts interrupts;
+#ifdef CONFIG_DEVFREQ_THERMAL
+	struct thermal_cooling_device *cooling;
+#endif
 
 	struct nvgpu_list_node class_list_head;
 	struct nvgpu_list_node cdev_list_head;
diff --git a/drivers/gpu/nvgpu/os/linux/scale.c b/drivers/gpu/nvgpu/os/linux/scale.c
index bb812d2e4..b89928553 100644
--- a/drivers/gpu/nvgpu/os/linux/scale.c
+++ b/drivers/gpu/nvgpu/os/linux/scale.c
@@ -17,6 +17,9 @@
  */
 
 #include <linux/devfreq.h>
+#ifdef CONFIG_DEVFREQ_THERMAL
+#include <linux/devfreq_cooling.h>
+#endif
 #include <linux/export.h>
 #include <linux/pm_qos.h>
 #include <linux/version.h>
@@ -384,6 +387,9 @@ void gk20a_scale_init(struct device *dev)
 	struct gk20a *g = platform->g;
 	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
 	struct gk20a_scale_profile *profile;
+#ifdef CONFIG_DEVFREQ_THERMAL
+	struct thermal_cooling_device *cooling;
+#endif
 	int err;
 
 	if (g->scale_profile)
@@ -434,6 +440,14 @@ void gk20a_scale_init(struct device *dev)
 
 		l->devfreq = devfreq;
 
+#ifdef CONFIG_DEVFREQ_THERMAL
+		cooling = of_devfreq_cooling_register(dev->of_node, devfreq);
+		if (IS_ERR(cooling))
+			dev_info(dev, "Failed to register cooling device\n");
+		else
+			l->cooling = cooling;
+#endif
+
 		/* create symlink /sys/devices/gpu.0/devfreq_dev */
 		if (devfreq != NULL) {
 			error = sysfs_create_link(&dev->kobj,
@@ -485,6 +499,13 @@ void gk20a_scale_exit(struct device *dev)
 	}
 #endif
 
+#ifdef CONFIG_DEVFREQ_THERMAL
+	if (l->cooling) {
+		devfreq_cooling_unregister(l->cooling);
+		l->cooling = NULL;
+	}
+#endif
+
 	if (platform->devfreq_governor) {
 		sysfs_remove_link(&dev->kobj, "devfreq_dev");