Skip to contentSkip to conversation

clk: add LP8788 clock driver

PATCH Superseded Kim, Milo 2 participants 4 messages Mar 21, 2013 – Apr 2, 2013

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Kim, Milo

author
#0 lore raw

PATCH 1/1 clk: add LP8788 clock driver

LP8788 PMU consists of regulator, battery charger, RTC, ADC, backlight driver and current sinks.
Here is additional driver, clock.

* Clock source
LP8788 clock is generated by two clock source.
One is internal oscillator. The other is attached external crystal oscillator. LP8788 provides automatic clock source selection through the I2C register. This operation is executed in 'prepare' of common clock driver architecture.

* Clock rate
LP8788 generates a fixed 32768Hz clock which is used for the system.

* Supported operation
.prepare: Before the clock is enabled, automatic clock source selection is done
by register access.
.unprepare: No register for disable or remove the clock source, so do nothing. .is_enabled
.recalc_rate: Fixed clock rate, 32.768KHz.

* clk_register_fixed_rate() vs devm_clk_register() and clkdev_add() Fixed clock driver can be created by common clock helper function but but LP8788 should be built as a module.
Using devm_clk_register() and clkdev_add() is more appropriate method to implement LP8788 clock driver.

---
 drivers/clk/Kconfig        |    7 ++
 drivers/clk/Makefile       |    1 +
 drivers/clk/clk-lp8788.c   |  186 ++++++++++++++++++++++++++++++++++++++++++++
 drivers/mfd/lp8788.c       |    2 +
 include/linux/mfd/lp8788.h |    1 +
 5 files changed, 197 insertions(+)
 create mode 100644 drivers/clk/clk-lp8788.c
drivers/clk/Kconfig +7 -0
index a64caef..4b5109c 100644
@@ -71,6 +71,13 @@ config COMMON_CLK_AXI_CLKGEN
 	  Support for the Analog Devices axi-clkgen pcore clock generator for Xilinx
 	  FPGAs. It is commonly used in Analog Devices' reference designs.
 
+config CLK_LP8788
+	tristate "Clock driver for TI LP8788 PMU"
+	depends on MFD_LP8788
+	---help---
+	  Supports the clock subsystem of TI LP8788. It generates the 32KHz
+	  clock output.
+
 endmenu
 
 source "drivers/clk/mvebu/Kconfig"
drivers/clk/Makefile +1 -0
index 1c22f9d..06d0763 100644
@@ -35,3 +35,4 @@ obj-$(CONFIG_COMMON_CLK_AXI_CLKGEN) += clk-axi-clkgen.o
 obj-$(CONFIG_COMMON_CLK_WM831X) += clk-wm831x.o
 obj-$(CONFIG_COMMON_CLK_MAX77686) += clk-max77686.o
 obj-$(CONFIG_CLK_TWL6040)	+= clk-twl6040.o
+obj-$(CONFIG_CLK_LP8788)	+= clk-lp8788.o
drivers/clk/clk-lp8788.c new file +186 -0

186 lines not on this page · raw message

drivers/mfd/lp8788.c +2 -0
index c3d3c9b..9f5596e 100644
@@ -72,6 +72,8 @@ static struct resource rtc_irqs[] = {
 };
 
 static struct mfd_cell lp8788_devs[] = {
+	MFD_DEV_SIMPLE(CLK),
+
 	/* 4 bucks */
 	MFD_DEV_WITH_ID(BUCK, 1),
 	MFD_DEV_WITH_ID(BUCK, 2),
include/linux/mfd/lp8788.h +1 -0
index 786bf66..5f33be8 100644
@@ -28,6 +28,7 @@
 #define LP8788_DEV_VIBRATOR	"lp8788-vibrator"
 #define LP8788_DEV_KEYLED	"lp8788-keyled"
 #define LP8788_DEV_ADC		"lp8788-adc"
+#define LP8788_DEV_CLK		"lp8788-clk"
 
 #define LP8788_NUM_BUCKS	4
 #define LP8788_NUM_DLDOS	12
Signature · 5 lines
1.7.9.5


Best Regards,
Milo
#1 lore raw

PATCH 1/1 clk: add LP8788 clock driver

Quoting Kim, Milo (2013-03-20 17:37:14)

Quoted 102 lines

LP8788 PMU consists of regulator, battery charger, RTC, ADC, backlight driver and current sinks.
Here is additional driver, clock.

* Clock source
LP8788 clock is generated by two clock source.
One is internal oscillator. The other is attached external crystal oscillator. LP8788 provides automatic clock source selection through the I2C register. This operation is executed in 'prepare' of common clock driver architecture.

* Clock rate
LP8788 generates a fixed 32768Hz clock which is used for the system.

* Supported operation
.prepare: Before the clock is enabled, automatic clock source selection is done
by register access.
.unprepare: No register for disable or remove the clock source, so do nothing. .is_enabled
.recalc_rate: Fixed clock rate, 32.768KHz.

* clk_register_fixed_rate() vs devm_clk_register() and clkdev_add() Fixed clock driver can be created by common clock helper function but but LP8788 should be built as a module.
Using devm_clk_register() and clkdev_add() is more appropriate method to implement LP8788 clock driver.

---
 drivers/clk/Kconfig        |    7 ++
 drivers/clk/Makefile       |    1 +
 drivers/clk/clk-lp8788.c   |  186 ++++++++++++++++++++++++++++++++++++++++++++
 drivers/mfd/lp8788.c       |    2 +
 include/linux/mfd/lp8788.h |    1 +
 5 files changed, 197 insertions(+)
 create mode 100644 drivers/clk/clk-lp8788.c
drivers/clk/Kconfig +7 -0
index a64caef..4b5109c 100644
@@ -71,6 +71,13 @@ config COMMON_CLK_AXI_CLKGEN
          Support for the Analog Devices axi-clkgen pcore clock generator for Xilinx
          FPGAs. It is commonly used in Analog Devices' reference designs.
 
+config CLK_LP8788
+       tristate "Clock driver for TI LP8788 PMU"
+       depends on MFD_LP8788
+       ---help---
+         Supports the clock subsystem of TI LP8788. It generates the 32KHz
+         clock output.
+
 endmenu
 
 source "drivers/clk/mvebu/Kconfig"
drivers/clk/Makefile +1 -0
index 1c22f9d..06d0763 100644
@@ -35,3 +35,4 @@ obj-$(CONFIG_COMMON_CLK_AXI_CLKGEN) += clk-axi-clkgen.o
 obj-$(CONFIG_COMMON_CLK_WM831X) += clk-wm831x.o
 obj-$(CONFIG_COMMON_CLK_MAX77686) += clk-max77686.o
 obj-$(CONFIG_CLK_TWL6040)      += clk-twl6040.o
+obj-$(CONFIG_CLK_LP8788)       += clk-lp8788.o
drivers/clk/clk-lp8788.c new file +33 -0
new file mode 100644
index 0000000..de56887
@@ -0,0 +1,186 @@
+/*
+ * TI LP8788 Clock Driver
+ *
+ * Copyright 2013 Texas Instruments
+ *
+ * Author: Milo(Woogyom) Kim <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <linux/clk.h>
+#include <linux/clkdev.h>
+#include <linux/clk-provider.h>
+#include <linux/mfd/lp8788.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+
+/* Registers */
+#define LP8788_REG_STATUS              0x06
+#define LP8788_OSC_WORKING             0x10
+
+#define LP8788_REG_CLKCTRL             0xA2
+#define LP8788_CLKMODE_MASK            0x02
+#define LP8788_CLKMODE_AUTO            0X02
+
+#define CLKOUT_32KHZ           32768
+
+struct lp8788_clk {
+       struct lp8788 *lp;

drivers/clk/clk-lp8788.c line 43

Quoted 10 lines
drivers/clk/clk-lp8788.c new file +10 -0
+       struct clk *clk;
+       struct clk_hw hw;
+       struct clk_lookup *lookup;
+       bool enabled;
+};
+
+static struct lp8788_clk *to_lp8788_clk(struct clk_hw *hw)
+{
+       return container_of(hw, struct lp8788_clk, hw);
+}

Hi Milo,

Just FYI most folks are simply using a macro to do the above. E.g:

#define to_lp8788_clk(_hw) container_of(_hw, struct lp8788_clk, hw)

Quoted 14 lines
drivers/clk/clk-lp8788.c new file +14 -0
+
+static int lp8788_clk_prepare(struct clk_hw *hw)
+{
+       struct lp8788_clk *pclk = to_lp8788_clk(hw);
+
+       /* Automatic oscillator source selection */
+       return lp8788_update_bits(pclk->lp, LP8788_REG_CLKCTRL,
+                               LP8788_CLKMODE_MASK, LP8788_CLKMODE_AUTO);
+}
+
+static void lp8788_clk_unprepare(struct clk_hw *hw)
+{
+       /* Do nothing */
+       return;

drivers/clk/clk-lp8788.c line 67

Quoted 10 lines
drivers/clk/clk-lp8788.c new file +10 -0
+}
+
+static int lp8788_clk_is_enabled(struct clk_hw *hw)
+{
+       struct lp8788_clk *pclk = to_lp8788_clk(hw);
+
+       /*
+        * Do not use the LP8788 register access here, unsafe locking problem
+        * may occur during loading the driver. So stored varible is prefered.
+        */

What unsafe locking problem?

Also have you looked into using the new .is_prepare callback? See:
https://git.linaro.org/gitweb?p=people/mturquette/linux.git;a=commitdiff;h=3d6ee287a3e341c88eafd0b4620b12d640b3736b;hp=30ee400614385ac49f4c9b4bc03d77ff8f07a61e

drivers/clk/clk-lp8788.c line 69

Quoted 2 lines
drivers/clk/clk-lp8788.c new file +2 -0
+
+       return pclk->enabled;

Why provide a .is_enabled callback when there are no .enable or .disable callbacks?

Regards,
Mike

author
#2 lore raw
Quoted 3 lines

Just FYI most folks are simply using a macro to do the above. E.g:

#define to_lp8788_clk(_hw) container_of(_hw, struct lp8788_clk, hw)

OK, thanks! Will be fixed in the 2nd patch.

Quoted 8 lines

drivers/clk/clk-lp8788.c

Quoted 7 lines
drivers/clk/clk-lp8788.c new file +7 -0
+
+static int lp8788_clk_is_enabled(struct clk_hw *hw)
+{
+       struct lp8788_clk *pclk = to_lp8788_clk(hw);
+
+       /*
+        * Do not use the LP8788 register access here, unsafe locking

problem

Quoted 5 lines
Quoted 1 line

+ * may occur during loading the driver. So stored varible is

prefered.

Quoted 1 line

+ */

What unsafe locking problem?

If I try to get LP8788 clock status via the remap-i2c here, then locking problem occurs.
The 'enable_lock' is already locked and the MUTEX of REGMAP is used on reading a register.

[    2.692443] -------------------------------------------------------
[    2.699066] swapper/0/1 is trying to acquire lock:
[    2.704132]  (&map->mutex){+.+...}, at: [<c0374ecc>] regmap_read+0x34/0x5c
[    2.711486] 
[    2.711486] but task is already holding lock:
[    2.717681]  (enable_lock){......}, at: [<c047a6c8>] clk_disable_unused_subtr
ee+0x3c/0xb4
[    2.726379] 
[    2.726379] which lock already depends on the new lock.

That's why lp8788_get_clk_status() is used on _probe().
- Read a register and store the value into local status variable. Then local variable is used in .is_enabled().

Quoted 4 lines

Also have you looked into using the new .is_prepare callback? See:
https://git.linaro.org/gitweb?p=people/mturquette/linux.git;a=commitdif
f;h=3d6ee287a3e341c88eafd0b4620b12d640b3736b;hp=30ee400614385ac49f4c9b4 bc03d77ff8f07a61e

Not checked yet, but it looks no caller for this callback at this moment. Will this be added inside the clk_prepare()?

Quoted 6 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+
+       return pclk->enabled;

Why provide a .is_enabled callback when there are no .enable or .disable
callbacks?

LP8788 Clock is always enabled, so it needs to show the clock status. However, I think no effect on working the clock operation without is_enabled(). Can I remove it?

Thanks,
Milo

#3 lore raw

PATCH 1/1 clk: add LP8788 clock driver

Quoting Kim, Milo (2013-04-01 19:18:32)

Quoted 35 lines
Quoted 3 lines

Just FYI most folks are simply using a macro to do the above. E.g:

#define to_lp8788_clk(_hw) container_of(_hw, struct lp8788_clk, hw)

OK, thanks! Will be fixed in the 2nd patch.

Quoted 8 lines

drivers/clk/clk-lp8788.c

Quoted 7 lines
drivers/clk/clk-lp8788.c new file +7 -0
+
+static int lp8788_clk_is_enabled(struct clk_hw *hw)
+{
+       struct lp8788_clk *pclk = to_lp8788_clk(hw);
+
+       /*
+        * Do not use the LP8788 register access here, unsafe locking

problem

Quoted 5 lines
Quoted 1 line

+ * may occur during loading the driver. So stored varible is

prefered.

Quoted 1 line

+ */

What unsafe locking problem?

If I try to get LP8788 clock status via the remap-i2c here, then locking problem occurs.
The 'enable_lock' is already locked and the MUTEX of REGMAP is used on reading a register.

[    2.692443] -------------------------------------------------------
[    2.699066] swapper/0/1 is trying to acquire lock:
[    2.704132]  (&map->mutex){+.+...}, at: [<c0374ecc>] regmap_read+0x34/0x5c
[    2.711486] 
[    2.711486] but task is already holding lock:
[    2.717681]  (enable_lock){......}, at: [<c047a6c8>] clk_disable_unused_subtr
ee+0x3c/0xb4
[    2.726379] 
[    2.726379] which lock already depends on the new lock.
Quoted 4 lines

That's why lp8788_get_clk_status() is used on _probe().
- Read a register and store the value into local status variable. Then local variable is used in .is_enabled().

I think that the new clk reentrancy patch should fix this for you. Can you rebase your patch onto the latest clk-next and try using remap-i2c again? You can that branch here:
git://git.linaro.org/people/mturquette/linux.git clk-next

Quoted 11 lines
Quoted 4 lines

Also have you looked into using the new .is_prepare callback? See:
https://git.linaro.org/gitweb?p=people/mturquette/linux.git;a=commitdif
f;h=3d6ee287a3e341c88eafd0b4620b12d640b3736b;hp=30ee400614385ac49f4c9b4 bc03d77ff8f07a61e

Not checked yet, but it looks no caller for this callback at this moment. Will this be added inside the clk_prepare()?

Quoted 3 lines
Quoted 2 lines
(quoted diff) +2 -0
+
+       return pclk->enabled;
Quoted 8 lines
Quoted 3 lines

Why provide a .is_enabled callback when there are no .enable or .disable
callbacks?

LP8788 Clock is always enabled, so it needs to show the clock status. However, I think no effect on working the clock operation without is_enabled(). Can I remove it?

Yes you can remove it. Since you do not provide a .disabled callback then clk_disable_unused() will not try to disable your clock. Do any drivers call clk_prepare_enable on this clock?

It is also probably a good idea to set the CLK_IGNORE_UNUSED flag on this clock (in addition to the CLK_IS_ROOT flag that you have already set). This way if your .unprepare callback ever does something some day then you won't get a nasty surprise.

Regards,
Mike

Quoted 2 lines

Thanks,
Milo