[PATCH 8/8] drm/msm/gpu: Add devfreq support for the GPU

[Jordan Crouse <jcrouse@codeaurora.org>]

Add support for devfreq to dynamically control the GPU frequency.
By default try to use the 'simple_ondemand' governor which can
adjust the frequency based on GPU load.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
---
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c   | 12 +++++
 drivers/gpu/drm/msm/adreno/adreno_gpu.c |  1 -
 drivers/gpu/drm/msm/msm_gpu.c           | 90 +++++++++++++++++++++++++++++++++
 drivers/gpu/drm/msm/msm_gpu.h           |  7 +++
 4 files changed, 109 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
index 56c2c44..7e09d44 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
@@ -600,6 +600,9 @@ static int a5xx_hw_init(struct msm_gpu *gpu)
 	/* Select CP0 to always count cycles */
 	gpu_write(gpu, REG_A5XX_CP_PERFCTR_CP_SEL_0, PERF_CP_ALWAYS_COUNT);
 
+	/* Select RBBM0 to countable 6 to get the busy status for devfreq */
+	gpu_write(gpu, REG_A5XX_RBBM_PERFCTR_RBBM_SEL_0, 6);
+
 	/* Increase VFD cache access so LRZ and other data gets evicted less */
 	gpu_write(gpu, REG_A5XX_UCHE_CACHE_WAYS, 0x02);
 
@@ -1170,6 +1173,14 @@ static struct msm_ringbuffer *a5xx_active_ring(struct msm_gpu *gpu)
 	return a5xx_gpu->cur_ring;
 }
 
+static int a5xx_gpu_busy(struct msm_gpu *gpu, uint64_t *value)
+{
+	*value = gpu_read64(gpu, REG_A5XX_RBBM_PERFCTR_RBBM_0_LO,
+		REG_A5XX_RBBM_PERFCTR_RBBM_0_HI);
+
+	return 0;
+}
+
 static const struct adreno_gpu_funcs funcs = {
 	.base = {
 		.get_param = adreno_get_param,
@@ -1185,6 +1196,7 @@ static struct msm_ringbuffer *a5xx_active_ring(struct msm_gpu *gpu)
 #ifdef CONFIG_DEBUG_FS
 		.show = a5xx_show,
 #endif
+		.gpu_busy = a5xx_gpu_busy,
 	},
 	.get_timestamp = a5xx_get_timestamp,
 };
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
index b4bac84..de63ff2 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
@@ -22,7 +22,6 @@
 #include "msm_gem.h"
 #include "msm_mmu.h"
 
-
 int adreno_get_param(struct msm_gpu *gpu, uint32_t param, uint64_t *value)
 {
 	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
diff --git a/drivers/gpu/drm/msm/msm_gpu.c b/drivers/gpu/drm/msm/msm_gpu.c
index 3d00e7a..243e06e 100644
--- a/drivers/gpu/drm/msm/msm_gpu.c
+++ b/drivers/gpu/drm/msm/msm_gpu.c
@@ -21,12 +21,90 @@
 #include "msm_fence.h"
 
 #include <linux/string_helpers.h>
+#include <linux/pm_opp.h>
+#include <linux/devfreq.h>
 
 
 /*
  * Power Management:
  */
 
+static int msm_devfreq_target(struct device *dev, unsigned long *freq,
+		u32 flags)
+{
+	struct msm_gpu *gpu = platform_get_drvdata(to_platform_device(dev));
+	struct dev_pm_opp *opp;
+
+	opp = dev_pm_opp_find_freq_ceil(dev, freq);
+
+	if (!IS_ERR(opp)) {
+		clk_set_rate(gpu->core_clk, *freq);
+		dev_pm_opp_put(opp);
+	}
+
+	return 0;
+}
+
+static int msm_devfreq_get_dev_status(struct device *dev,
+		struct devfreq_dev_status *status)
+{
+	struct msm_gpu *gpu = platform_get_drvdata(to_platform_device(dev));
+	u64 cycles;
+	ktime_t time;
+
+	status->current_frequency = (unsigned long) clk_get_rate(gpu->core_clk);
+	gpu->funcs->gpu_busy(gpu, &cycles);
+
+	status->busy_time = (cycles - gpu->devfreq.busy_cycles) /
+		(status->current_frequency / 1000000);
+
+	gpu->devfreq.busy_cycles = cycles;
+
+	time = ktime_get();
+	status->total_time = ktime_us_delta(time, gpu->devfreq.time);
+	gpu->devfreq.time = time;
+
+	return 0;
+}
+
+static int msm_devfreq_get_cur_freq(struct device *dev, unsigned long *freq)
+{
+	struct msm_gpu *gpu = platform_get_drvdata(to_platform_device(dev));
+
+	*freq = (unsigned long) clk_get_rate(gpu->core_clk);
+
+	return 0;
+}
+
+static struct devfreq_dev_profile msm_devfreq_profile = {
+	.polling_ms = 10,
+	.target = msm_devfreq_target,
+	.get_dev_status = msm_devfreq_get_dev_status,
+	.get_cur_freq = msm_devfreq_get_cur_freq,
+};
+
+static void msm_devfreq_init(struct msm_gpu *gpu)
+{
+	/* We need target support to do devfreq */
+	if (!gpu->funcs->gpu_busy)
+		return;
+
+	msm_devfreq_profile.initial_freq = gpu->fast_rate;
+
+	/*
+	 * Don't set the freq_table or max_state and let devfreq build the table
+	 * from OPP
+	 */
+
+	gpu->devfreq.devfreq = devm_devfreq_add_device(&gpu->pdev->dev,
+			&msm_devfreq_profile, "simple_ondemand", NULL);
+
+	if (IS_ERR(gpu->devfreq.devfreq)) {
+		dev_err(&gpu->pdev->dev, "Couldn't initialize GPU devfreq\n");
+		gpu->devfreq.devfreq = NULL;
+	}
+}
+
 static int enable_pwrrail(struct msm_gpu *gpu)
 {
 	struct drm_device *dev = gpu->dev;
@@ -140,6 +218,13 @@ int msm_gpu_pm_resume(struct msm_gpu *gpu)
 	if (ret)
 		return ret;
 
+	if (gpu->devfreq.devfreq) {
+		gpu->devfreq.busy_cycles = 0;
+		gpu->devfreq.time = ktime_get();
+
+		devfreq_resume_device(gpu->devfreq.devfreq);
+	}
+
 	gpu->needs_hw_init = true;
 
 	return 0;
@@ -151,6 +236,9 @@ int msm_gpu_pm_suspend(struct msm_gpu *gpu)
 
 	DBG("%s", gpu->name);
 
+	if (gpu->devfreq.devfreq)
+		devfreq_suspend_device(gpu->devfreq.devfreq);
+
 	ret = disable_axi(gpu);
 	if (ret)
 		return ret;
@@ -721,6 +809,8 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	gpu->pdev = pdev;
 	platform_set_drvdata(pdev, gpu);
 
+	msm_devfreq_init(gpu);
+
 	gpu->aspace = msm_gpu_create_address_space(gpu, pdev,
 		config->va_start, config->va_end);
 
diff --git a/drivers/gpu/drm/msm/msm_gpu.h b/drivers/gpu/drm/msm/msm_gpu.h
index 0de26b6..fccfccd 100644
--- a/drivers/gpu/drm/msm/msm_gpu.h
+++ b/drivers/gpu/drm/msm/msm_gpu.h
@@ -66,6 +66,7 @@ struct msm_gpu_funcs {
 	/* show GPU status in debugfs: */
 	void (*show)(struct msm_gpu *gpu, struct seq_file *m);
 #endif
+	int (*gpu_busy)(struct msm_gpu *gpu, uint64_t *value);
 };
 
 struct msm_gpu {
@@ -120,6 +121,12 @@ struct msm_gpu {
 	struct work_struct recover_work;
 
 	struct drm_gem_object *memptrs_bo;
+
+	struct {
+		struct devfreq *devfreq;
+		u64 busy_cycles;
+		ktime_t time;
+	} devfreq;
 };
 
 /* It turns out that all targets use the same ringbuffer size */
-- 
1.9.1