diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
index c424e9a37669..3dcec7acb384 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
@@ -1666,18 +1666,10 @@ static u64 a5xx_gpu_busy(struct msm_gpu *gpu, unsigned long *out_sample_rate)
 {
 	u64 busy_cycles;
 
-	/* Only read the gpu busy if the hardware is already active */
-	if (pm_runtime_get_if_in_use(&gpu->pdev->dev) == 0) {
-		*out_sample_rate = 1;
-		return 0;
-	}
-
 	busy_cycles = gpu_read64(gpu, REG_A5XX_RBBM_PERFCTR_RBBM_0_LO,
 			REG_A5XX_RBBM_PERFCTR_RBBM_0_HI);
 	*out_sample_rate = clk_get_rate(gpu->core_clk);
 
-	pm_runtime_put(&gpu->pdev->dev);
-
 	return busy_cycles;
 }
 
diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
index 9f76f5b15759..dc715d88ff21 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
@@ -102,7 +102,8 @@ bool a6xx_gmu_gx_is_on(struct a6xx_gmu *gmu)
 		A6XX_GMU_SPTPRAC_PWR_CLK_STATUS_GX_HM_CLK_OFF));
 }
 
-void a6xx_gmu_set_freq(struct msm_gpu *gpu, struct dev_pm_opp *opp)
+void a6xx_gmu_set_freq(struct msm_gpu *gpu, struct dev_pm_opp *opp,
+		       bool suspended)
 {
 	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
 	struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
@@ -127,15 +128,16 @@ void a6xx_gmu_set_freq(struct msm_gpu *gpu, struct dev_pm_opp *opp)
 
 	/*
 	 * This can get called from devfreq while the hardware is idle. Don't
-	 * bring up the power if it isn't already active
+	 * bring up the power if it isn't already active. All we're doing here
+	 * is updating the frequency so that when we come back online we're at
+	 * the right rate.
 	 */
-	if (pm_runtime_get_if_in_use(gmu->dev) == 0)
+	if (suspended)
 		return;
 
 	if (!gmu->legacy) {
 		a6xx_hfi_set_freq(gmu, perf_index);
 		dev_pm_opp_set_opp(&gpu->pdev->dev, opp);
-		pm_runtime_put(gmu->dev);
 		return;
 	}
 
@@ -159,7 +161,6 @@ void a6xx_gmu_set_freq(struct msm_gpu *gpu, struct dev_pm_opp *opp)
 		dev_err(gmu->dev, "GMU set GPU frequency error: %d\n", ret);
 
 	dev_pm_opp_set_opp(&gpu->pdev->dev, opp);
-	pm_runtime_put(gmu->dev);
 }
 
 unsigned long a6xx_gmu_get_freq(struct msm_gpu *gpu)
@@ -895,7 +896,7 @@ static void a6xx_gmu_set_initial_freq(struct msm_gpu *gpu, struct a6xx_gmu *gmu)
 		return;
 
 	gmu->freq = 0; /* so a6xx_gmu_set_freq() doesn't exit early */
-	a6xx_gmu_set_freq(gpu, gpu_opp);
+	a6xx_gmu_set_freq(gpu, gpu_opp, false);
 	dev_pm_opp_put(gpu_opp);
 }
 
diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
index 42ed9a3c4905..8c02a67f29f2 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
@@ -1658,27 +1658,21 @@ static u64 a6xx_gpu_busy(struct msm_gpu *gpu, unsigned long *out_sample_rate)
 	/* 19.2MHz */
 	*out_sample_rate = 19200000;
 
-	/* Only read the gpu busy if the hardware is already active */
-	if (pm_runtime_get_if_in_use(a6xx_gpu->gmu.dev) == 0)
-		return 0;
-
 	busy_cycles = gmu_read64(&a6xx_gpu->gmu,
 			REG_A6XX_GMU_CX_GMU_POWER_COUNTER_XOCLK_0_L,
 			REG_A6XX_GMU_CX_GMU_POWER_COUNTER_XOCLK_0_H);
 
-
-	pm_runtime_put(a6xx_gpu->gmu.dev);
-
 	return busy_cycles;
 }
 
-static void a6xx_gpu_set_freq(struct msm_gpu *gpu, struct dev_pm_opp *opp)
+static void a6xx_gpu_set_freq(struct msm_gpu *gpu, struct dev_pm_opp *opp,
+			      bool suspended)
 {
 	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
 	struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
 
 	mutex_lock(&a6xx_gpu->gmu.lock);
-	a6xx_gmu_set_freq(gpu, opp);
+	a6xx_gmu_set_freq(gpu, opp, suspended);
 	mutex_unlock(&a6xx_gpu->gmu.lock);
 }
 
diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.h b/drivers/gpu/drm/msm/adreno/a6xx_gpu.h
index 86e0a7c3fe6d..ab853f61db63 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.h
@@ -77,7 +77,8 @@ void a6xx_gmu_clear_oob(struct a6xx_gmu *gmu, enum a6xx_gmu_oob_state state);
 int a6xx_gmu_init(struct a6xx_gpu *a6xx_gpu, struct device_node *node);
 void a6xx_gmu_remove(struct a6xx_gpu *a6xx_gpu);
 
-void a6xx_gmu_set_freq(struct msm_gpu *gpu, struct dev_pm_opp *opp);
+void a6xx_gmu_set_freq(struct msm_gpu *gpu, struct dev_pm_opp *opp,
+		       bool suspended);
 unsigned long a6xx_gmu_get_freq(struct msm_gpu *gpu);
 
 void a6xx_show(struct msm_gpu *gpu, struct msm_gpu_state *state,
diff --git a/drivers/gpu/drm/msm/msm_gpu.h b/drivers/gpu/drm/msm/msm_gpu.h
index 6def00883046..31269c1c896b 100644
--- a/drivers/gpu/drm/msm/msm_gpu.h
+++ b/drivers/gpu/drm/msm/msm_gpu.h
@@ -64,11 +64,14 @@ struct msm_gpu_funcs {
 	/* for generation specific debugfs: */
 	void (*debugfs_init)(struct msm_gpu *gpu, struct drm_minor *minor);
 #endif
+	/* note: gpu_busy() can assume that we have been pm_resumed */
 	u64 (*gpu_busy)(struct msm_gpu *gpu, unsigned long *out_sample_rate);
 	struct msm_gpu_state *(*gpu_state_get)(struct msm_gpu *gpu);
 	int (*gpu_state_put)(struct msm_gpu_state *state);
 	unsigned long (*gpu_get_freq)(struct msm_gpu *gpu);
-	void (*gpu_set_freq)(struct msm_gpu *gpu, struct dev_pm_opp *opp);
+	/* note: gpu_set_freq() can assume that we have been pm_resumed */
+	void (*gpu_set_freq)(struct msm_gpu *gpu, struct dev_pm_opp *opp,
+			     bool suspended);
 	struct msm_gem_address_space *(*create_address_space)
 		(struct msm_gpu *gpu, struct platform_device *pdev);
 	struct msm_gem_address_space *(*create_private_address_space)
@@ -92,6 +95,9 @@ struct msm_gpu_devfreq {
 	/** devfreq: devfreq instance */
 	struct devfreq *devfreq;
 
+	/** lock: lock for "suspended", "busy_cycles", and "time" */
+	struct mutex lock;
+
 	/**
 	 * idle_constraint:
 	 *
@@ -135,6 +141,9 @@ struct msm_gpu_devfreq {
 	 * elapsed
 	 */
 	struct msm_hrtimer_work boost_work;
+
+	/** suspended: tracks if we're suspended */
+	bool suspended;
 };
 
 struct msm_gpu {
diff --git a/drivers/gpu/drm/msm/msm_gpu_devfreq.c b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
index d2539ca78c29..ea94bc18e72e 100644
--- a/drivers/gpu/drm/msm/msm_gpu_devfreq.c
+++ b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
@@ -20,6 +20,7 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
 		u32 flags)
 {
 	struct msm_gpu *gpu = dev_to_gpu(dev);
+	struct msm_gpu_devfreq *df = &gpu->devfreq;
 	struct dev_pm_opp *opp;
 
 	/*
@@ -32,10 +33,13 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
 
 	trace_msm_gpu_freq_change(dev_pm_opp_get_freq(opp));
 
-	if (gpu->funcs->gpu_set_freq)
-		gpu->funcs->gpu_set_freq(gpu, opp);
-	else
+	if (gpu->funcs->gpu_set_freq) {
+		mutex_lock(&df->lock);
+		gpu->funcs->gpu_set_freq(gpu, opp, df->suspended);
+		mutex_unlock(&df->lock);
+	} else {
 		clk_set_rate(gpu->core_clk, *freq);
+	}
 
 	dev_pm_opp_put(opp);
 
@@ -58,16 +62,25 @@ static void get_raw_dev_status(struct msm_gpu *gpu,
 	unsigned long sample_rate;
 	ktime_t time;
 
+	mutex_lock(&df->lock);
+
 	status->current_frequency = get_freq(gpu);
-	busy_cycles = gpu->funcs->gpu_busy(gpu, &sample_rate);
 	time = ktime_get();
-
-	busy_time = busy_cycles - df->busy_cycles;
 	status->total_time = ktime_us_delta(time, df->time);
-
-	df->busy_cycles = busy_cycles;
 	df->time = time;
 
+	if (df->suspended) {
+		mutex_unlock(&df->lock);
+		status->busy_time = 0;
+		return;
+	}
+
+	busy_cycles = gpu->funcs->gpu_busy(gpu, &sample_rate);
+	busy_time = busy_cycles - df->busy_cycles;
+	df->busy_cycles = busy_cycles;
+
+	mutex_unlock(&df->lock);
+
 	busy_time *= USEC_PER_SEC;
 	do_div(busy_time, sample_rate);
 	if (WARN_ON(busy_time > ~0LU))
@@ -175,6 +188,8 @@ void msm_devfreq_init(struct msm_gpu *gpu)
 	if (!gpu->funcs->gpu_busy)
 		return;
 
+	mutex_init(&df->lock);
+
 	dev_pm_qos_add_request(&gpu->pdev->dev, &df->idle_freq,
 			       DEV_PM_QOS_MAX_FREQUENCY,
 			       PM_QOS_MAX_FREQUENCY_DEFAULT_VALUE);
@@ -244,12 +259,16 @@ void msm_devfreq_cleanup(struct msm_gpu *gpu)
 void msm_devfreq_resume(struct msm_gpu *gpu)
 {
 	struct msm_gpu_devfreq *df = &gpu->devfreq;
+	unsigned long sample_rate;
 
 	if (!has_devfreq(gpu))
 		return;
 
-	df->busy_cycles = 0;
+	mutex_lock(&df->lock);
+	df->busy_cycles = gpu->funcs->gpu_busy(gpu, &sample_rate);
 	df->time = ktime_get();
+	df->suspended = false;
+	mutex_unlock(&df->lock);
 
 	devfreq_resume_device(df->devfreq);
 }
@@ -261,6 +280,10 @@ void msm_devfreq_suspend(struct msm_gpu *gpu)
 	if (!has_devfreq(gpu))
 		return;
 
+	mutex_lock(&df->lock);
+	df->suspended = true;
+	mutex_unlock(&df->lock);
+
 	devfreq_suspend_device(df->devfreq);
 
 	cancel_idle_work(df);