Skip to contentSkip to conversation

ARM: davinci: complete the conversion to using the reset framework

PATCH Applied Bartosz Golaszewski 3 participants 16 messages Mar 30, 2018 – Apr 6, 2018

Subsystems (7)
Device tree bindings Arm (32-bit) Arm sub-architectures Common clk +3 more subsystems

v5 of 7 Applied Latest is v7 →

lore
Share & tools
lore

Bartosz Golaszewski

author
#0 lore raw

v5 cover ARM: davinci: complete the conversion to using the reset framework

From: Bartosz Golaszewski <[email protected]>

This series converts the only user of the handcoded, mach-specific reset routines in the davinci platform to using the reset framework.

Patches 1-3 add necessary lookups/DT-properties.

Patches 4-6 fix issues found in the remoteproc davinci driver.

Patch 7 converts the davinci-rproc driver to the reset framework.

Patch 8 removes now dead code.

Tested both in DT and legacy modes by booting the examples from ti-ipc-rtos recipe in meta-ti.

This series applies on top of David Lechner's common-clk-v9 branch[1] with Philipp Zabel's reset/next branch[2] pulled in.

It can be found in my github tree as well[3].

[1] git://github.com/dlech/ev3dev-kernel.git common-clk-v9
[2] git://git.pengutronix.de/git/pza/linux reset/next
[3] git at github.com:brgl/linux.git topic/davinci-reset

v1 -> v2:
- fixed the device tree patches the descriptions of which were mixed up
- return -EPROBE_DEFER from davinci-rproc's probe() if we can't get the reset provider, since it's possible that the lookup table was not yet registered
- made the local variable naming consistent in the davinci-rproc driver
- fixed a typo in PATCH 5/8

v2 -> v3:
- modify PATCH 1/8: drop the provider argument from the function adding lookup entries and instead pass the provider name to the RESET_LOOKUP macro, return -EPROBE_DEFER if we locate a correct lookup entry but cannot get the corresponding reset controller
- modify the reset lookup entry in psc-da850
- don't manually return -EPROBE_DEFER from davinci-rproc, instead don't emit an error message if devm_reset_control_get_exclusive() returns this error code

v3 -> v4:
- make index the second parameter in RESET_LOOKUP() (right after the provider name)

v4 -> v5:
- fix a bug where the dsp_reset object correctly stored in drproc struct

Bartosz Golaszewski (8):
ARM: davinci: dts: make psc0 a reset provider
ARM: davinci: dts: add a reset control to the dsp node
clk: davinci: add a reset lookup table for psc0 remoteproc/davinci: add the missing retval check for clk_enable() remoteproc/davinci: prepare and unprepare the clock where needed remoteproc/davinci: use octal permissions for module_param() remoteproc/davinci: use the reset framework
clk: davinci: kill davinci_clk_reset_assert/deassert()

 arch/arm/boot/dts/da850.dtsi               |  2 ++
 arch/arm/mach-davinci/include/mach/clock.h | 21 --------------
 drivers/clk/davinci/psc-da850.c            |  7 +++++
 drivers/clk/davinci/psc.c                  | 19 +-----------
 drivers/remoteproc/da8xx_remoteproc.c      | 46 ++++++++++++++++++++++++------
 5 files changed, 48 insertions(+), 47 deletions(-)
 delete mode 100644 arch/arm/mach-davinci/include/mach/clock.h
Signature · 1 line
2.16.3
author
#1 lore raw

PATCH v5 1/8 ARM: davinci: dts: make psc0 a reset provider

From: Bartosz Golaszewski <[email protected]>

The psc driver registers with the reset framework as a provider. Add the #reset-cells property to the psc0 node.

Signed-off-by Bartosz Golaszewski <[email protected]>
---
 arch/arm/boot/dts/da850.dtsi | 1 +
 1 file changed, 1 insertion(+)
arch/arm/boot/dts/da850.dtsi +1 -0
index a3b9a99b43ca..dad64dbf1f23 100644
@@ -123,6 +123,7 @@
 			compatible = "ti,da850-psc0";
 			reg = <0x10000 0x1000>;
 			#clock-cells = <1>;
+			#reset-cells = <1>;
 			#power-domain-cells = <1>;
 			clocks = <&pll0_sysclk 1>, <&pll0_sysclk 2>,
 				 <&pll0_sysclk 4>, <&pll0_sysclk 6>,
Signature · 1 line
2.16.3
maintainer Reviewed-by on v5 1/8
#2 lore raw

Re: [PATCH v5 1/8] ARM: davinci: dts: make psc0 a reset provider

Quoted 16 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The psc driver registers with the reset framework as a provider. Add the #reset-cells property to the psc0 node.

Signed-off-by Bartosz Golaszewski <[email protected]>
---
  arch/arm/boot/dts/da850.dtsi | 1 +
  1 file changed, 1 insertion(+)
arch/arm/boot/dts/da850.dtsi +0 -0
index a3b9a99b43ca..dad64dbf1f23 100644

arch/arm/boot/dts/da850.dtsi line 129

Quoted 7 lines
arch/arm/boot/dts/da850.dtsi +1 -0
@@ -123,6 +123,7 @@
  			compatible = "ti,da850-psc0";
  			reg = <0x10000 0x1000>;
  			#clock-cells = <1>;
+			#reset-cells = <1>;
  			#power-domain-cells = <1>;
  			clocks = <&pll0_sysclk 1>, <&pll0_sysclk 2>,
  				 <&pll0_sysclk 4>, <&pll0_sysclk 6>,
Reviewed-by David Lechner <[email protected]>
author
#3 lore raw

PATCH v5 2/8 ARM: davinci: dts: add a reset control to the dsp node

In reply to Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The davinci-rproc driver will soon use the reset framework. Add the resets property to the dsp node in da850.dtsi.

Signed-off-by Bartosz Golaszewski <[email protected]>
---
 arch/arm/boot/dts/da850.dtsi | 1 +
 1 file changed, 1 insertion(+)
arch/arm/boot/dts/da850.dtsi +1 -0
index dad64dbf1f23..5b485e44b83e 100644
@@ -109,6 +109,7 @@
 		interrupt-parent = <&intc>;
 		interrupts = <28>;
 		clocks = <&psc0 15>;
+		resets = <&psc0 15>;
 		status = "disabled";
 	};
 	soc at 1c00000 {
Signature · 1 line
2.16.3
maintainer Reviewed-by on v5 2/8
#4 lore raw

Re: [PATCH v5 2/8] ARM: davinci: dts: add a reset control to the dsp node

Quoted 16 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The davinci-rproc driver will soon use the reset framework. Add the resets property to the dsp node in da850.dtsi.

Signed-off-by Bartosz Golaszewski <[email protected]>
---
  arch/arm/boot/dts/da850.dtsi | 1 +
  1 file changed, 1 insertion(+)
arch/arm/boot/dts/da850.dtsi +0 -0
index dad64dbf1f23..5b485e44b83e 100644

arch/arm/boot/dts/da850.dtsi line 115

Quoted 7 lines
arch/arm/boot/dts/da850.dtsi +1 -0
@@ -109,6 +109,7 @@
  		interrupt-parent = <&intc>;
  		interrupts = <28>;
  		clocks = <&psc0 15>;
+		resets = <&psc0 15>;
  		status = "disabled";
  	};
  	soc at 1c00000 {
Reviewed-by David Lechner <[email protected]>
author
#5 lore raw

PATCH v5 3/8 clk: davinci: add a reset lookup table for psc0

In reply to Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

In order to be able to use the reset framework in legacy boot mode as well, add the reset lookup table to the psc driver for da850 variant.

Signed-off-by Bartosz Golaszewski <[email protected]>
---
 drivers/clk/davinci/psc-da850.c | 7 +++++++
 drivers/clk/davinci/psc.c       | 1 +
 2 files changed, 8 insertions(+)
drivers/clk/davinci/psc-da850.c +7 -0
index ccc7eb17bf3a..d196dcbed560 100644
@@ -6,6 +6,7 @@
  */
 
 #include <linux/clk-provider.h>
+#include <linux/reset-controller.h>
 #include <linux/clk.h>
 #include <linux/clkdev.h>
 #include <linux/init.h>

@@ -66,8 +67,14 @@ LPSC_CLKDEV3(ecap_clkdev,	"fck",	"ecap.0",
 				"fck",	"ecap.1",
 				"fck",	"ecap.2");
 
+static struct reset_control_lookup da850_psc0_reset_lookup_table[] = {
+	RESET_LOOKUP("da850-psc0", 15, "davinci-rproc.0", NULL),
+};
+
 static int da850_psc0_init(struct device *dev, void __iomem *base)
 {
+	reset_controller_add_lookup(da850_psc0_reset_lookup_table,
+				    ARRAY_SIZE(da850_psc0_reset_lookup_table));
 	return davinci_psc_register_clocks(dev, da850_psc0_info, 16, base);
 }
drivers/clk/davinci/psc.c +1 -0
index 3b0e59dfbdd7..063df62381ea 100644
@@ -425,6 +425,7 @@ __davinci_psc_register_clocks(struct device *dev,
 
 	psc->rcdev.ops = &davinci_psc_reset_ops;
 	psc->rcdev.owner = THIS_MODULE;
+	psc->rcdev.dev = dev;
 	psc->rcdev.of_node = dev->of_node;
 	psc->rcdev.of_reset_n_cells = 1;
 	psc->rcdev.of_xlate = davinci_psc_reset_of_xlate;
Signature · 1 line
2.16.3
author
#8 lore raw

PATCH v5 4/8 remoteproc/davinci: add the missing retval check for clk_enable()

In reply to Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The davinci platform is being switched to using the common clock framework, where clk_enable() can fail. Add the return value check.

Signed-off-by Bartosz Golaszewski <[email protected]>
Acked-by Suman Anna <[email protected]>
---
 drivers/remoteproc/da8xx_remoteproc.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)
drivers/remoteproc/da8xx_remoteproc.c +7 -1
index bf3b9034c319..2b24291337b7 100644
@@ -138,6 +138,7 @@ static int da8xx_rproc_start(struct rproc *rproc)
 	struct device *dev = rproc->dev.parent;
 	struct da8xx_rproc *drproc = (struct da8xx_rproc *)rproc->priv;
 	struct clk *dsp_clk = drproc->dsp_clk;
+	int ret;
 
 	/* hw requires the start (boot) address be on 1KB boundary */
 	if (rproc->bootaddr & 0x3ff) {

@@ -148,7 +149,12 @@ static int da8xx_rproc_start(struct rproc *rproc)
 
 	writel(rproc->bootaddr, drproc->bootreg);
 
-	clk_enable(dsp_clk);
+	ret = clk_enable(dsp_clk);
+	if (ret) {
+		dev_err(dev, "clk_enable() failed: %d\n", ret);
+		return ret;
+	}
+
 	davinci_clk_reset_deassert(dsp_clk);
 
 	return 0;
Signature · 1 line
2.16.3
maintainer Reviewed-by on v5 4/8
#9 lore raw

Re: [PATCH v5 4/8] remoteproc/davinci: add the missing retval check for clk_enable()

Quoted 8 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

The davinci platform is being switched to using the common clock framework, where clk_enable() can fail. Add the return value check.

Signed-off-by Bartosz Golaszewski <[email protected]>
Acked-by Suman Anna <[email protected]>

---

Reviewed-by David Lechner <[email protected]>
author
#10 lore raw

PATCH v5 5/8 remoteproc/davinci: prepare and unprepare the clock where needed

In reply to Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

We're currently switching the platform to using the common clock framework. We need to explicitly prepare and unprepare the rproc clock.

Signed-off-by Bartosz Golaszewski <[email protected]>
Acked-by Suman Anna <[email protected]>
---
 drivers/remoteproc/da8xx_remoteproc.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/remoteproc/da8xx_remoteproc.c +3 -3
index 2b24291337b7..f134192922e0 100644
@@ -149,9 +149,9 @@ static int da8xx_rproc_start(struct rproc *rproc)
 
 	writel(rproc->bootaddr, drproc->bootreg);
 
-	ret = clk_enable(dsp_clk);
+	ret = clk_prepare_enable(dsp_clk);
 	if (ret) {
-		dev_err(dev, "clk_enable() failed: %d\n", ret);
+		dev_err(dev, "clk_prepare_enable() failed: %d\n", ret);
 		return ret;
 	}
 

@@ -165,7 +165,7 @@ static int da8xx_rproc_stop(struct rproc *rproc)
 	struct da8xx_rproc *drproc = rproc->priv;
 
 	davinci_clk_reset_assert(drproc->dsp_clk);
-	clk_disable(drproc->dsp_clk);
+	clk_disable_unprepare(drproc->dsp_clk);
 
 	return 0;
 }
Signature · 1 line
2.16.3
maintainer Reviewed-by on v5 5/8
#11 lore raw

Re: [PATCH v5 5/8] remoteproc/davinci: prepare and unprepare the clock where needed

Quoted 9 lines · Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

We're currently switching the platform to using the common clock framework. We need to explicitly prepare and unprepare the rproc clock.

Signed-off-by Bartosz Golaszewski <[email protected]>
Acked-by Suman Anna <[email protected]>

---

Reviewed-by David Lechner <[email protected]>
author
#12 lore raw

PATCH v5 6/8 remoteproc/davinci: use octal permissions for module_param()

In reply to Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

Checkpatch recommends to use octal perms instead of S_IRUGO.

Signed-off-by Bartosz Golaszewski <[email protected]>
---
 drivers/remoteproc/da8xx_remoteproc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/remoteproc/da8xx_remoteproc.c +1 -1
index f134192922e0..b668e32996e2 100644
@@ -25,7 +25,7 @@
 #include "remoteproc_internal.h"
 
 static char *da8xx_fw_name;
-module_param(da8xx_fw_name, charp, S_IRUGO);
+module_param(da8xx_fw_name, charp, 0444);
 MODULE_PARM_DESC(da8xx_fw_name,
 		 "Name of DSP firmware file in /lib/firmware (if not specified defaults to 'rproc-dsp-fw')");
Signature · 1 line
2.16.3
author
#13 lore raw

PATCH v5 7/8 remoteproc/davinci: use the reset framework

In reply to Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

Switch to using the reset framework instead of handcoded reset routines we used so far.

Signed-off-by Bartosz Golaszewski <[email protected]>
---
 drivers/remoteproc/da8xx_remoteproc.c | 34 +++++++++++++++++++++++++++++-----
 1 file changed, 29 insertions(+), 5 deletions(-)
drivers/remoteproc/da8xx_remoteproc.c +29 -5
index b668e32996e2..76c06b70a1c6 100644
@@ -10,6 +10,7 @@
 
 #include <linux/bitops.h>
 #include <linux/clk.h>
+#include <linux/reset.h>
 #include <linux/err.h>
 #include <linux/interrupt.h>
 #include <linux/io.h>

@@ -20,8 +21,6 @@
 #include <linux/platform_device.h>
 #include <linux/remoteproc.h>
 
-#include <mach/clock.h>   /* for davinci_clk_reset_assert/deassert() */
-
 #include "remoteproc_internal.h"
 
 static char *da8xx_fw_name;

@@ -72,6 +71,7 @@ struct da8xx_rproc {
 	struct da8xx_rproc_mem *mem;
 	int num_mems;
 	struct clk *dsp_clk;
+	struct reset_control *dsp_reset;
 	void (*ack_fxn)(struct irq_data *data);
 	struct irq_data *irq_data;
 	void __iomem *chipsig;

@@ -138,6 +138,7 @@ static int da8xx_rproc_start(struct rproc *rproc)
 	struct device *dev = rproc->dev.parent;
 	struct da8xx_rproc *drproc = (struct da8xx_rproc *)rproc->priv;
 	struct clk *dsp_clk = drproc->dsp_clk;
+	struct reset_control *dsp_reset = drproc->dsp_reset;
 	int ret;
 
 	/* hw requires the start (boot) address be on 1KB boundary */

@@ -155,7 +156,12 @@ static int da8xx_rproc_start(struct rproc *rproc)
 		return ret;
 	}
 
-	davinci_clk_reset_deassert(dsp_clk);
+	ret = reset_control_deassert(dsp_reset);
+	if (ret) {
+		dev_err(dev, "reset_control_deassert() failed: %d\n", ret);
+		clk_disable_unprepare(dsp_clk);
+		return ret;
+	}
 
 	return 0;
 }

@@ -163,8 +169,15 @@ static int da8xx_rproc_start(struct rproc *rproc)
 static int da8xx_rproc_stop(struct rproc *rproc)
 {
 	struct da8xx_rproc *drproc = rproc->priv;
+	struct device *dev = rproc->dev.parent;
+	int ret;
+
+	ret = reset_control_assert(drproc->dsp_reset);
+	if (ret) {
+		dev_err(dev, "reset_control_assert() failed: %d\n", ret);
+		return ret;
+	}
 
-	davinci_clk_reset_assert(drproc->dsp_clk);
 	clk_disable_unprepare(drproc->dsp_clk);
 
 	return 0;

@@ -232,6 +245,7 @@ static int da8xx_rproc_probe(struct platform_device *pdev)
 	struct resource *bootreg_res;
 	struct resource *chipsig_res;
 	struct clk *dsp_clk;
+	struct reset_control *dsp_reset;
 	void __iomem *chipsig;
 	void __iomem *bootreg;
 	int irq;

@@ -268,6 +282,15 @@ static int da8xx_rproc_probe(struct platform_device *pdev)
 		return PTR_ERR(dsp_clk);
 	}
 
+	dsp_reset = devm_reset_control_get_exclusive(dev, NULL);
+	if (IS_ERR(dsp_reset)) {
+		if (PTR_ERR(dsp_reset) != -EPROBE_DEFER)
+			dev_err(dev, "unable to get reset control: %ld\n",
+				PTR_ERR(dsp_reset));
+
+		return PTR_ERR(dsp_reset);
+	}
+
 	if (dev->of_node) {
 		ret = of_reserved_mem_device_init(dev);
 		if (ret) {

@@ -287,6 +310,7 @@ static int da8xx_rproc_probe(struct platform_device *pdev)
 	drproc = rproc->priv;
 	drproc->rproc = rproc;
 	drproc->dsp_clk = dsp_clk;
+	drproc->dsp_reset = dsp_reset;
 	rproc->has_iommu = false;
 
 	ret = da8xx_rproc_get_internal_memories(pdev, drproc);

@@ -309,7 +333,7 @@ static int da8xx_rproc_probe(struct platform_device *pdev)
 	 * *not* in reset, but da8xx_rproc_start() needs the DSP to be
 	 * held in reset at the time it is called.
 	 */
-	ret = davinci_clk_reset_assert(drproc->dsp_clk);
+	ret = reset_control_assert(dsp_reset);
 	if (ret)
 		goto free_rproc;
Signature · 1 line
2.16.3
author
#14 lore raw

PATCH v5 8/8 clk: davinci: kill davinci_clk_reset_assert/deassert()

In reply to Bartosz Golaszewski

From: Bartosz Golaszewski <[email protected]>

This code is no longer used. Remove it.

Signed-off-by Bartosz Golaszewski <[email protected]>
Reviewed-by David Lechner <[email protected]>
---
 arch/arm/mach-davinci/include/mach/clock.h | 21 ---------------------
 drivers/clk/davinci/psc.c                  | 18 ------------------
 2 files changed, 39 deletions(-)
 delete mode 100644 arch/arm/mach-davinci/include/mach/clock.h
arch/arm/mach-davinci/include/mach/clock.h deleted +0 -21
deleted file mode 100644
index 42ed4f2f5ce4..000000000000
@@ -1,21 +0,0 @@
-/*
- * arch/arm/mach-davinci/include/mach/clock.h
- *
- * Clock control driver for DaVinci - header file
- *
- * Authors: Vladimir Barinov <[email protected]>
- *
- * 2007 (c) MontaVista Software, Inc. This file is licensed under
- * the terms of the GNU General Public License version 2. This program
- * is licensed "as is" without any warranty of any kind, whether express
- * or implied.
- */
-#ifndef __ASM_ARCH_DAVINCI_CLOCK_H
-#define __ASM_ARCH_DAVINCI_CLOCK_H
-
-struct clk;
-
-int davinci_clk_reset_assert(struct clk *c);
-int davinci_clk_reset_deassert(struct clk *c);
-
-#endif
drivers/clk/davinci/psc.c +0 -18
index 063df62381ea..75c4bc6f23e3 100644
@@ -295,24 +295,6 @@ static int davinci_lpsc_clk_reset(struct clk *clk, bool reset)
 	return 0;
 }
 
-/*
- * REVISIT: These exported functions can be removed after a non-DT lookup is
- * added to the reset controller framework and the davinci-rproc driver is
- * updated to use the generic reset controller framework.
- */
-
-int davinci_clk_reset_assert(struct clk *clk)
-{
-	return davinci_lpsc_clk_reset(clk, true);
-}
-EXPORT_SYMBOL(davinci_clk_reset_assert);
-
-int davinci_clk_reset_deassert(struct clk *clk)
-{
-	return davinci_lpsc_clk_reset(clk, false);
-}
-EXPORT_SYMBOL(davinci_clk_reset_deassert);
-
 static int davinci_psc_reset_assert(struct reset_controller_dev *rcdev,
 				    unsigned long id)
 {
Signature · 1 line
2.16.3
maintainer Acked-by on v5 8/8
#15 lore raw

Re: [PATCH v5 8/8] clk: davinci: kill davinci_clk_reset_assert/deassert()

Quoting Bartosz Golaszewski (2018-03-30 08:28:56)

Quoted 7 lines

From: Bartosz Golaszewski <[email protected]>

This code is no longer used. Remove it.

Signed-off-by Bartosz Golaszewski <[email protected]>
Reviewed-by David Lechner <[email protected]>

---

I'll try to remember, but please send this again if we need to apply it to the clk tree for 4.18.

Acked-by Stephen Boyd <[email protected]>