[PATCH 0/8] msm/gpu: devfreq support

[PATCH 0/8] msm/gpu: devfreq support

[Jordan Crouse]

Here is a stack of patches to enable devfreq support for Adreno GPUs.
This stack adds support for a5xx only, but it should be trivial to
add support for the other targets.  I'll send the DT changes 
separately.

Jordan Crouse (8):
  drm/msm/adreno: Call dev_pm_opp_put()
  drm/msm/adreno: Remove a useless call to dev_pm_opp_get_freq()
  drm/msm/gpu: Remove unused bus scaling code
  drm/msm/adreno: Cleanup chipid parsing
  drm/msm/adreno: Move clock parsing to adreno_gpu_init()
  drm/msm/adreno: Read the speed bins for a5xx targets
  drm/msm/adreno: a5xx: Explicitly program the CP0 performance counter
  drm/msm/gpu: Add devfreq support for the GPU

 drivers/gpu/drm/msm/adreno/a5xx_gpu.c      |  38 ++++++++++
 drivers/gpu/drm/msm/adreno/a5xx_power.c    |   8 ++-
 drivers/gpu/drm/msm/adreno/adreno_device.c | 112 ++++++-----------------------
 drivers/gpu/drm/msm/adreno/adreno_gpu.c    |  83 ++++++++++++++++++---
 drivers/gpu/drm/msm/adreno/adreno_gpu.h    |   4 --
 drivers/gpu/drm/msm/msm_gpu.c              | 105 ++++++++++++++++++++-------
 drivers/gpu/drm/msm/msm_gpu.h              |  14 ++--
 7 files changed, 225 insertions(+), 139 deletions(-)

-- 
1.9.1

[PATCH 1/8] drm/msm/adreno: Call dev_pm_opp_put()

[Jordan Crouse]

We need to call dev_pm_opp_put() to put back the reference
for the OPP struct after calling the various dev_pm_opp_get_*
functions.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
---
 drivers/gpu/drm/msm/adreno/a5xx_power.c    | 8 +++++++-
 drivers/gpu/drm/msm/adreno/adreno_device.c | 4 +++-
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/msm/adreno/a5xx_power.c b/drivers/gpu/drm/msm/adreno/a5xx_power.c
index e5700bb..4e4d965 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_power.c
+++ b/drivers/gpu/drm/msm/adreno/a5xx_power.c
@@ -103,10 +103,16 @@ static inline uint32_t _get_mvolts(struct msm_gpu *gpu, uint32_t freq)
 	struct msm_drm_private *priv = dev->dev_private;
 	struct platform_device *pdev = priv->gpu_pdev;
 	struct dev_pm_opp *opp;
+	u32 ret = 0;
 
 	opp = dev_pm_opp_find_freq_exact(&pdev->dev, freq, true);
 
-	return (!IS_ERR(opp)) ? dev_pm_opp_get_voltage(opp) / 1000 : 0;
+	if (!IS_ERR(opp)) {
+		ret = dev_pm_opp_get_voltage(opp) / 1000;
+		dev_pm_opp_put(opp);
+	}
+
+	return ret;
 }
 
 /* Setup thermal limit management */
diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index 05022ea..e1eb7e1 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -239,8 +239,10 @@ static int adreno_get_pwrlevels(struct device *dev,
 
 	/* Find the fastest defined rate */
 	opp = dev_pm_opp_find_freq_floor(dev, &freq);
-	if (!IS_ERR(opp))
+	if (!IS_ERR(opp)) {
 		config->fast_rate = dev_pm_opp_get_freq(opp);
+		dev_pm_opp_put(opp);
+	}
 
 	if (!config->fast_rate) {
 		DRM_DEV_INFO(dev,
-- 
1.9.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 2/8] drm/msm/adreno: Remove a useless call to dev_pm_opp_get_freq()

[Jordan Crouse]

Calling dev_pm_opp_find_freq_floor() returns the matched frequency
in 'freq'.  We don't need to call dev_pm_opp_get_freq() again
to get the frequency value.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
---
 drivers/gpu/drm/msm/adreno/adreno_device.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index e1eb7e1..e85a1cc 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -240,7 +240,7 @@ static int adreno_get_pwrlevels(struct device *dev,
 	/* Find the fastest defined rate */
 	opp = dev_pm_opp_find_freq_floor(dev, &freq);
 	if (!IS_ERR(opp)) {
-		config->fast_rate = dev_pm_opp_get_freq(opp);
+		config->fast_rate = freq;
 		dev_pm_opp_put(opp);
 	}
 
-- 
1.9.1

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 3/8] drm/msm/gpu: Remove unused bus scaling code

[Jordan Crouse]

Remove the downstream bus scaling code. It isn't needed for for
compatibility with a downstream or vendor kernel. Get it out of the
way to clear space for devfreq support.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
---
 drivers/gpu/drm/msm/adreno/adreno_gpu.c |  7 +-----
 drivers/gpu/drm/msm/adreno/adreno_gpu.h |  5 +----
 drivers/gpu/drm/msm/msm_gpu.c           | 39 ---------------------------------
 drivers/gpu/drm/msm/msm_gpu.h           |  7 +-----
 4 files changed, 3 insertions(+), 55 deletions(-)

diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
index 2f0610f..61e3091 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
@@ -480,13 +480,8 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	adreno_gpu->rev = config->rev;
 
 	gpu->fast_rate = config->fast_rate;
-	gpu->bus_freq  = config->bus_freq;
-#ifdef DOWNSTREAM_CONFIG_MSM_BUS_SCALING
-	gpu->bus_scale_table = config->bus_scale_table;
-#endif
 
-	DBG("fast_rate=%u, slow_rate=27000000, bus_freq=%u",
-			gpu->fast_rate, gpu->bus_freq);
+	DBG("fast_rate=%u, slow_rate=27000000", gpu->fast_rate);
 
 	adreno_gpu_config.ioname = "kgsl_3d0_reg_memory";
 	adreno_gpu_config.irqname = "kgsl_3d0_irq";
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
index 28e3de6..88d1bdf 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
@@ -129,10 +129,7 @@ struct adreno_gpu {
 /* platform config data (ie. from DT, or pdata) */
 struct adreno_platform_config {
 	struct adreno_rev rev;
-	uint32_t fast_rate, bus_freq;
-#ifdef DOWNSTREAM_CONFIG_MSM_BUS_SCALING
-	struct msm_bus_scale_pdata *bus_scale_table;
-#endif
+	uint32_t fast_rate;
 };
 
 #define ADRENO_IDLE_TIMEOUT msecs_to_jiffies(1000)
diff --git a/drivers/gpu/drm/msm/msm_gpu.c b/drivers/gpu/drm/msm/msm_gpu.c
index fb5a1e0..3d00e7a 100644
--- a/drivers/gpu/drm/msm/msm_gpu.c
+++ b/drivers/gpu/drm/msm/msm_gpu.c
@@ -27,37 +27,6 @@
  * Power Management:
  */
 
-#ifdef DOWNSTREAM_CONFIG_MSM_BUS_SCALING
-#include <mach/board.h>
-static void bs_init(struct msm_gpu *gpu)
-{
-	if (gpu->bus_scale_table) {
-		gpu->bsc = msm_bus_scale_register_client(gpu->bus_scale_table);
-		DBG("bus scale client: %08x", gpu->bsc);
-	}
-}
-
-static void bs_fini(struct msm_gpu *gpu)
-{
-	if (gpu->bsc) {
-		msm_bus_scale_unregister_client(gpu->bsc);
-		gpu->bsc = 0;
-	}
-}
-
-static void bs_set(struct msm_gpu *gpu, int idx)
-{
-	if (gpu->bsc) {
-		DBG("set bus scaling: %d", idx);
-		msm_bus_scale_client_update_request(gpu->bsc, idx);
-	}
-}
-#else
-static void bs_init(struct msm_gpu *gpu) {}
-static void bs_fini(struct msm_gpu *gpu) {}
-static void bs_set(struct msm_gpu *gpu, int idx) {}
-#endif
-
 static int enable_pwrrail(struct msm_gpu *gpu)
 {
 	struct drm_device *dev = gpu->dev;
@@ -143,8 +112,6 @@ static int enable_axi(struct msm_gpu *gpu)
 {
 	if (gpu->ebi1_clk)
 		clk_prepare_enable(gpu->ebi1_clk);
-	if (gpu->bus_freq)
-		bs_set(gpu, gpu->bus_freq);
 	return 0;
 }
 
@@ -152,8 +119,6 @@ static int disable_axi(struct msm_gpu *gpu)
 {
 	if (gpu->ebi1_clk)
 		clk_disable_unprepare(gpu->ebi1_clk);
-	if (gpu->bus_freq)
-		bs_set(gpu, 0);
 	return 0;
 }
 
@@ -756,8 +721,6 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	gpu->pdev = pdev;
 	platform_set_drvdata(pdev, gpu);
 
-	bs_init(gpu);
-
 	gpu->aspace = msm_gpu_create_address_space(gpu, pdev,
 		config->va_start, config->va_end);
 
@@ -827,8 +790,6 @@ void msm_gpu_cleanup(struct msm_gpu *gpu)
 
 	WARN_ON(!list_empty(&gpu->active_list));
 
-	bs_fini(gpu);
-
 	for (i = 0; i < ARRAY_SIZE(gpu->rb); i++) {
 		msm_ringbuffer_destroy(gpu->rb[i]);
 		gpu->rb[i] = NULL;
diff --git a/drivers/gpu/drm/msm/msm_gpu.h b/drivers/gpu/drm/msm/msm_gpu.h
index e113d64..0de26b6 100644
--- a/drivers/gpu/drm/msm/msm_gpu.h
+++ b/drivers/gpu/drm/msm/msm_gpu.h
@@ -108,12 +108,7 @@ struct msm_gpu {
 	struct clk **grp_clks;
 	int nr_clocks;
 	struct clk *ebi1_clk, *core_clk, *rbbmtimer_clk;
-	uint32_t fast_rate, bus_freq;
-
-#ifdef DOWNSTREAM_CONFIG_MSM_BUS_SCALING
-	struct msm_bus_scale_pdata *bus_scale_table;
-	uint32_t bsc;
-#endif
+	uint32_t fast_rate;
 
 	/* Hang and Inactivity Detection:
 	 */
-- 
1.9.1

_______________________________________________
Freedreno mailing list
Freedreno@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/freedreno

[PATCH 4/8] drm/msm/adreno: Cleanup chipid parsing

[Jordan Crouse]

We don't need to convert the chipid to an intermediate value and
then back again into a struct adreno_rev.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
---
 drivers/gpu/drm/msm/adreno/adreno_device.c | 44 +++++++++++++++---------------
 1 file changed, 22 insertions(+), 22 deletions(-)

diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index e85a1cc..80d26b9 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -153,39 +153,45 @@ static void set_gpu_pdev(struct drm_device *dev,
 	priv->gpu_pdev = pdev;
 }
 
-static int find_chipid(struct device *dev, u32 *chipid)
+static int find_chipid(struct device *dev, struct adreno_rev *rev)
 {
 	struct device_node *node = dev->of_node;
 	const char *compat;
 	int ret;
+	u32 chipid;
 
 	/* first search the compat strings for qcom,adreno-XYZ.W: */
 	ret = of_property_read_string_index(node, "compatible", 0, &compat);
 	if (ret == 0) {
-		unsigned rev, patch;
+		unsigned int r, patch;
 
-		if (sscanf(compat, "qcom,adreno-%u.%u", &rev, &patch) == 2) {
-			*chipid = 0;
-			*chipid |= (rev / 100) << 24;  /* core */
-			rev %= 100;
-			*chipid |= (rev / 10) << 16;   /* major */
-			rev %= 10;
-			*chipid |= rev << 8;           /* minor */
-			*chipid |= patch;
+		if (sscanf(compat, "qcom,adreno-%u.%u", &r, &patch) == 2) {
+			rev->core = r / 100;
+			r %= 100;
+			rev->major = r / 10;
+			r %= 10;
+			rev->minor = r;
+			rev->patchid = patch;
 
 			return 0;
 		}
 	}
 
 	/* and if that fails, fall back to legacy "qcom,chipid" property: */
-	ret = of_property_read_u32(node, "qcom,chipid", chipid);
-	if (ret)
+	ret = of_property_read_u32(node, "qcom,chipid", &chipid);
+	if (ret) {
+		dev_err(dev, "could not parse qcom,chipid: %d\n", ret);
 		return ret;
+	}
+
+	rev->core = (chipid >> 24) & 0xff;
+	rev->major = (chipid >> 16) & 0xff;
+	rev->minor = (chipid >> 8) & 0xff;
+	rev->patchid = (chipid & 0xff);
 
 	dev_warn(dev, "Using legacy qcom,chipid binding!\n");
 	dev_warn(dev, "Use compatible qcom,adreno-%u%u%u.%u instead.\n",
-			(*chipid >> 24) & 0xff, (*chipid >> 16) & 0xff,
-			(*chipid >> 8) & 0xff, *chipid & 0xff);
+		rev->core, rev->major, rev->minor, rev->patchid);
 
 	return 0;
 }
@@ -260,17 +266,11 @@ static int adreno_bind(struct device *dev, struct device *master, void *data)
 	const struct adreno_info *info;
 	struct drm_device *drm = dev_get_drvdata(master);
 	struct msm_gpu *gpu;
-	u32 val;
 	int ret;
 
-	ret = find_chipid(dev, &val);
-	if (ret) {
-		dev_err(dev, "could not find chipid: %d\n", ret);
+	ret = find_chipid(dev, &config.rev);
+	if (ret)
 		return ret;
-	}
-
-	config.rev = ADRENO_REV((val >> 24) & 0xff,
-			(val >> 16) & 0xff, (val >> 8) & 0xff, val & 0xff);
 
 	/* find clock rates: */
 	config.fast_rate = 0;
-- 
1.9.1

[PATCH 5/8] drm/msm/adreno: Move clock parsing to adreno_gpu_init()

[Jordan Crouse]

Move the clock parsing to adreno_gpu_init() to allow for target
specific probing and manipulation of the clock tables.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
---
 drivers/gpu/drm/msm/adreno/adreno_device.c | 72 ----------------------------
 drivers/gpu/drm/msm/adreno/adreno_gpu.c    | 77 ++++++++++++++++++++++++++++--
 drivers/gpu/drm/msm/adreno/adreno_gpu.h    |  1 -
 3 files changed, 73 insertions(+), 77 deletions(-)

diff --git a/drivers/gpu/drm/msm/adreno/adreno_device.c b/drivers/gpu/drm/msm/adreno/adreno_device.c
index 80d26b9..250fa1e 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_device.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_device.c
@@ -17,7 +17,6 @@
  * this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include <linux/pm_opp.h>
 #include "adreno_gpu.h"
 
 #define ANY_ID 0xff
@@ -196,70 +195,6 @@ static int find_chipid(struct device *dev, struct adreno_rev *rev)
 	return 0;
 }
 
-/* Get legacy powerlevels from qcom,gpu-pwrlevels and populate the opp table */
-static int adreno_get_legacy_pwrlevels(struct device *dev)
-{
-	struct device_node *child, *node;
-	int ret;
-
-	node = of_find_compatible_node(dev->of_node, NULL,
-		"qcom,gpu-pwrlevels");
-	if (!node) {
-		dev_err(dev, "Could not find the GPU powerlevels\n");
-		return -ENXIO;
-	}
-
-	for_each_child_of_node(node, child) {
-		unsigned int val;
-
-		ret = of_property_read_u32(child, "qcom,gpu-freq", &val);
-		if (ret)
-			continue;
-
-		/*
-		 * Skip the intentionally bogus clock value found at the bottom
-		 * of most legacy frequency tables
-		 */
-		if (val != 27000000)
-			dev_pm_opp_add(dev, val, 0);
-	}
-
-	return 0;
-}
-
-static int adreno_get_pwrlevels(struct device *dev,
-		struct adreno_platform_config *config)
-{
-	unsigned long freq = ULONG_MAX;
-	struct dev_pm_opp *opp;
-	int ret;
-
-	/* You down with OPP? */
-	if (!of_find_property(dev->of_node, "operating-points-v2", NULL))
-		ret = adreno_get_legacy_pwrlevels(dev);
-	else
-		ret = dev_pm_opp_of_add_table(dev);
-
-	if (ret)
-		return ret;
-
-	/* Find the fastest defined rate */
-	opp = dev_pm_opp_find_freq_floor(dev, &freq);
-	if (!IS_ERR(opp)) {
-		config->fast_rate = freq;
-		dev_pm_opp_put(opp);
-	}
-
-	if (!config->fast_rate) {
-		DRM_DEV_INFO(dev,
-			"Could not find clock rate. Using default\n");
-		/* Pick a suitably safe clock speed for any target */
-		config->fast_rate = 200000000;
-	}
-
-	return 0;
-}
-
 static int adreno_bind(struct device *dev, struct device *master, void *data)
 {
 	static struct adreno_platform_config config = {};
@@ -272,13 +207,6 @@ static int adreno_bind(struct device *dev, struct device *master, void *data)
 	if (ret)
 		return ret;
 
-	/* find clock rates: */
-	config.fast_rate = 0;
-
-	ret = adreno_get_pwrlevels(dev, &config);
-	if (ret)
-		return ret;
-
 	dev->platform_data = &config;
 	set_gpu_pdev(drm, to_platform_device(dev));
 
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
index 61e3091..b4bac84 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
@@ -17,6 +17,7 @@
  * this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
+#include <linux/pm_opp.h>
 #include "adreno_gpu.h"
 #include "msm_gem.h"
 #include "msm_mmu.h"
@@ -465,6 +466,76 @@ void adreno_wait_ring(struct msm_ringbuffer *ring, uint32_t ndwords)
 			ring->id);
 }
 
+/* Get legacy powerlevels from qcom,gpu-pwrlevels and populate the opp table */
+static int adreno_get_legacy_pwrlevels(struct device *dev)
+{
+	struct device_node *child, *node;
+	int ret;
+
+	node = of_find_compatible_node(dev->of_node, NULL,
+		"qcom,gpu-pwrlevels");
+	if (!node) {
+		dev_err(dev, "Could not find the GPU powerlevels\n");
+		return -ENXIO;
+	}
+
+	for_each_child_of_node(node, child) {
+		unsigned int val;
+
+		ret = of_property_read_u32(child, "qcom,gpu-freq", &val);
+		if (ret)
+			continue;
+
+		/*
+		 * Skip the intentionally bogus clock value found at the bottom
+		 * of most legacy frequency tables
+		 */
+		if (val != 27000000)
+			dev_pm_opp_add(dev, val, 0);
+	}
+
+	return 0;
+}
+
+static int adreno_get_pwrlevels(struct device *dev,
+		struct msm_gpu *gpu)
+{
+	unsigned long freq = ULONG_MAX;
+	struct dev_pm_opp *opp;
+	int ret;
+
+	gpu->fast_rate = 0;
+
+	/* You down with OPP? */
+	if (!of_find_property(dev->of_node, "operating-points-v2", NULL))
+		ret = adreno_get_legacy_pwrlevels(dev);
+	else {
+		ret = dev_pm_opp_of_add_table(dev);
+		if (ret)
+			dev_err(dev, "Unable to set the OPP table\n");
+	}
+
+	if (!ret) {
+		/* Find the fastest defined rate */
+		opp = dev_pm_opp_find_freq_floor(dev, &freq);
+		if (!IS_ERR(opp)) {
+			gpu->fast_rate = freq;
+			dev_pm_opp_put(opp);
+		}
+	}
+
+	if (!gpu->fast_rate) {
+		dev_warn(dev,
+			"Could not find a clock rate. Using a reasonable default\n");
+		/* Pick a suitably safe clock speed for any target */
+		gpu->fast_rate = 200000000;
+	}
+
+	DBG("fast_rate=%u, slow_rate=27000000", gpu->fast_rate);
+
+	return 0;
+}
+
 int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 		struct adreno_gpu *adreno_gpu,
 		const struct adreno_gpu_funcs *funcs, int nr_rings)
@@ -479,10 +550,6 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	adreno_gpu->revn = adreno_gpu->info->revn;
 	adreno_gpu->rev = config->rev;
 
-	gpu->fast_rate = config->fast_rate;
-
-	DBG("fast_rate=%u, slow_rate=27000000", gpu->fast_rate);
-
 	adreno_gpu_config.ioname = "kgsl_3d0_reg_memory";
 	adreno_gpu_config.irqname = "kgsl_3d0_irq";
 
@@ -491,6 +558,8 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 
 	adreno_gpu_config.nr_rings = nr_rings;
 
+	adreno_get_pwrlevels(&pdev->dev, gpu);
+
 	pm_runtime_set_autosuspend_delay(&pdev->dev, DRM_MSM_INACTIVE_PERIOD);
 	pm_runtime_use_autosuspend(&pdev->dev);
 	pm_runtime_enable(&pdev->dev);
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
index 88d1bdf..8d3d0a9 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
@@ -129,7 +129,6 @@ struct adreno_gpu {
 /* platform config data (ie. from DT, or pdata) */
 struct adreno_platform_config {
 	struct adreno_rev rev;
-	uint32_t fast_rate;
 };
 
 #define ADRENO_IDLE_TIMEOUT msecs_to_jiffies(1000)
-- 
1.9.1

_______________________________________________
Freedreno mailing list
Freedreno@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/freedreno

[PATCH 6/8] drm/msm/adreno: Read the speed bins for a5xx targets

[Jordan Crouse]

Some 5xx based chipsets have different bins for GPU clock speeds.
Read the fuses (if applicable) and set the appropriate OPP table.
This will only work with OPP v2 tables - the bin will be ignored
for legacy pwrlevel tables.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
---
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
index a1f4eee..fdbe9e9 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
@@ -17,6 +17,8 @@
 #include <linux/dma-mapping.h>
 #include <linux/of_address.h>
 #include <linux/soc/qcom/mdt_loader.h>
+#include <linux/pm_opp.h>
+#include <linux/nvmem-consumer.h>
 #include "msm_gem.h"
 #include "msm_mmu.h"
 #include "a5xx_gpu.h"
@@ -1184,6 +1186,25 @@ static struct msm_ringbuffer *a5xx_active_ring(struct msm_gpu *gpu)
 	.get_timestamp = a5xx_get_timestamp,
 };
 
+static void check_speed_bin(struct device *dev)
+{
+	struct nvmem_cell *cell;
+	u32 bin, val;
+
+	cell = nvmem_cell_get(dev, "speed_bin");
+
+	/* If a nvmem cell isn't defined, nothing to do */
+	if (IS_ERR(cell))
+		return;
+
+	bin = *((u32 *) nvmem_cell_read(cell, NULL));
+	nvmem_cell_put(cell);
+
+	val = (1 << bin);
+
+	dev_pm_opp_set_supported_hw(dev, &val, 1);
+}
+
 struct msm_gpu *a5xx_gpu_init(struct drm_device *dev)
 {
 	struct msm_drm_private *priv = dev->dev_private;
@@ -1210,6 +1231,8 @@ struct msm_gpu *a5xx_gpu_init(struct drm_device *dev)
 
 	a5xx_gpu->lm_leakage = 0x4E001A;
 
+	check_speed_bin(&pdev->dev);
+
 	ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs, 4);
 	if (ret) {
 		a5xx_destroy(&(a5xx_gpu->base.base));
-- 
1.9.1

[PATCH 7/8] drm/msm/adreno: a5xx: Explicitly program the CP0 performance counter

[Jordan Crouse]

Even though the default countable for CP0 is CP_ALWAYS_COUNT (0),
program the selector during HW initialization in an effort to be
up front about which counters are programmed and why.

Signed-off-by: Jordan Crouse <jcrouse@codeaurora.org>
---
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
index fdbe9e9..56c2c44 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
@@ -597,6 +597,9 @@ static int a5xx_hw_init(struct msm_gpu *gpu)
 	/* Turn on performance counters */
 	gpu_write(gpu, REG_A5XX_RBBM_PERFCTR_CNTL, 0x01);
 
+	/* Select CP0 to always count cycles */
+	gpu_write(gpu, REG_A5XX_CP_PERFCTR_CP_SEL_0, PERF_CP_ALWAYS_COUNT);
+
 	/* Increase VFD cache access so LRZ and other data gets evicted less */
 	gpu_write(gpu, REG_A5XX_UCHE_CACHE_WAYS, 0x02);
 
-- 
1.9.1

_______________________________________________
Freedreno mailing list
Freedreno@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/freedreno

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

[Jordan Crouse]

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

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

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 1131 bytes --]

Hi Jordan,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on robclark/msm-next]
[also build test ERROR on next-20171122]
[cannot apply to v4.14]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Jordan-Crouse/msm-gpu-devfreq-support/20171124-022911
base:   git://people.freedesktop.org/~robclark/linux msm-next
config: arm-multi_v7_defconfig (attached as .config)
compiler: arm-linux-gnueabi-gcc (Debian 7.2.0-11) 7.2.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        make.cross ARCH=arm 

All errors (new ones prefixed by >>):

   drivers/gpu/drm/msm/msm_gpu.o: In function `msm_devfreq_get_dev_status':
>> msm_gpu.c:(.text+0x558): undefined reference to `__aeabi_uldivmod'

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 42342 bytes --]

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

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 1048 bytes --]

Hi Jordan,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on robclark/msm-next]
[also build test ERROR on next-20171122]
[cannot apply to v4.14]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url:    https://github.com/0day-ci/linux/commits/Jordan-Crouse/msm-gpu-devfreq-support/20171124-022911
base:   git://people.freedesktop.org/~robclark/linux msm-next
config: arm-allmodconfig (attached as .config)
compiler: arm-linux-gnueabi-gcc (Debian 7.2.0-11) 7.2.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # save the attached .config to linux build tree
        make.cross ARCH=arm 

All errors (new ones prefixed by >>):

>> ERROR: "__aeabi_uldivmod" [drivers/gpu/drm/msm/msm.ko] undefined!

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 64188 bytes --]