On Fri, Nov 09, 2012 at 13:41:04, Thierry Reding wrote:
> On Thu, Nov 08, 2012 at 01:23:15PM +0530, Philip, Avinash wrote:
...
> > + /* Some platforms require explicit tbclk gating */
> > + if (of_property_read_bool(pdev->dev.of_node, "tbclkgating")) {
>
> Is it really necessary to have an extra boolean property for this?
> Couldn't this just be handled by not defining a clock for the tbclk
> consumer in board setup/DT

If no clk node defined, driver can still continue expecting this platform
the requirement is not there. So I will check for tbclk with devm_clk_get()
and continue by removing Boolean property.

>
> > + pc->tbclk = clk_get(&pdev->dev, "tbclk");
>
> You should be using devm_clk_get() or add a matching clk_put() in
> .remove().

Ok

Thanks
Avinash

>
> Thierry
>

2012-11-09 11:01:09

by Philip, Avinash

[permalink] [raw]

Subject: RE: [PATCH v2 04/10] pwm: pwm-tiecap: Add device-tree binding support for APWM driver

On Fri, Nov 09, 2012 at 13:22:19, Thierry Reding wrote:
> On Thu, Nov 08, 2012 at 01:23:11PM +0530, Philip, Avinash wrote:
> > This patch
> > 1. Add support for device-tree binding for ECAP APWM driver.
> > 2. Set size of pwm-cells set to 3 to support PWM channel number, PWM
> > period & polarity configuration from device tree.
> > 3. Add enable/disable clock gating in PWM subsystem common config space.
> > 4. When here set .owner member in platform_driver structure to
> > THIS_MODULE.
> >
> > Signed-off-by: Philip, Avinash <[email protected]>
> > Cc: Grant Likely <[email protected]>
> > Cc: Rob Herring <[email protected]>
> > Cc: Rob Landley <[email protected]>
> > ---
> > Changes since v1:
> > - Add separate patch for pinctrl support
> > - Add conditional check for PWM subsystem clock enable.
> > - Combined with HWMOD changes & DT bindings.
> > - Remove the custom of xlate support.
> >
> > :000000 100644 0000000... fe24cac... A Documentation/devicetree/bindings/pwm/pwm-tiecap.txt
> > :100644 100644 d6d4cf0... 0d43266... M drivers/pwm/pwm-tiecap.c
> > .../devicetree/bindings/pwm/pwm-tiecap.txt | 22 +++++++++
> > drivers/pwm/pwm-tiecap.c | 48 +++++++++++++++++++-
> > 2 files changed, 69 insertions(+), 1 deletions(-)
> >
> > diff --git a/Documentation/devicetree/bindings/pwm/pwm-tiecap.txt b/Documentation/devicetree/bindings/pwm/pwm-tiecap.txt
> > new file mode 100644
> > index 0000000..fe24cac
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/pwm/pwm-tiecap.txt
> > @@ -0,0 +1,22 @@
> > +TI SOC ECAP based APWM controller
> > +
> > +Required properties:
> > +- compatible: Must be "ti,am33xx-ecap"
> > +- #pwm-cells: Should be 3. Number of cells being used to specify PWM property.
> > + First cell specifies the per-chip index of the PWM to use, the second
> > + cell is the period cycle in nanoseconds and bit 0 in the third cell is
>
> I think this should be "period in nanoseconds". I haven't heard "period
> cycle" before.

Ok

>
> > + used to encode the polarity of PWM output.
>
> Maybe you should explicitly say how this is encoded.

Ok I will add details

>
...
> >
> > +#define ECAPCLK_EN BIT(0)
> > +#define ECAPCLK_STOP_REQ BIT(1)
>
> This one doesn't seem to align with the rest. Also, why is bit 0 called
> _EN and bit 1 _STOP_REQ? Couldn't they be made more consistent, i.e.
> _START and _STOP? Or _ENABLE and _DISABLE?

Ok I will change to PWMSS_ECAPCLK_EN & PWMSS_ECAPCLK_STPO_REQ

>
> > +
> > +#define ECAPCLK_EN_ACK BIT(0)
> > +
> > +#define PWM_CELL_SIZE 3
>
> You don't need a define for this.

I remove.

>
> > +
> > struct ecap_pwm_chip {
> > struct pwm_chip chip;
> > unsigned int clk_rate;
> > @@ -184,6 +194,16 @@ static const struct pwm_ops ecap_pwm_ops = {
> > .owner = THIS_MODULE,
> > };
> >
> > +#ifdef CONFIG_OF
> > +static const struct of_device_id ecap_of_match[] = {
> > + {
> > + .compatible = "ti,am33xx-ecap",
> > + },
> > + {},
> > +};
> > +MODULE_DEVICE_TABLE(of, ecap_of_match);
> > +#endif
> > +
>
> I'm not sure if I remember correctly, but wasn't AM33xx support supposed
> to be DT only? In that case you don't need the CONFIG_OF guards.

I will remove

>
...
> > pm_runtime_enable(&pdev->dev);
> > + pm_runtime_get_sync(&pdev->dev);
>
> Maybe put a blank line after this for readability.

Ok

>
> > + if (!(pwmss_submodule_state_change(pdev->dev.parent, ECAPCLK_EN) &
> > + ECAPCLK_EN_ACK)) {
>
> This is very hard to read, can you split this up into something like the
> following please?
>
> status = pwmss_submodule_state_change(pdev->dev.parent, ECAPCLK_EN);
> if (!(status & ECAPCLK_EN_ACK)) {
> ...
> }
>

Ok I will correct it.

> > + dev_err(&pdev->dev, "PWMSS config space clock enable failure\n");
> > + ret = -EINVAL;
> > + goto pwmss_clk_failure;
> > + }
> > + pm_runtime_put_sync(&pdev->dev);
>
> Another blank line between the two above would be good.

Ok

>
...
> > + .owner = THIS_MODULE,
> > + .of_match_table = of_match_ptr(ecap_of_match),
>
> Here as well, if AM33xx is DT-only, then the of_match_ptr() can be
> dropped.

Ok I will drop.

Thanks
Avinash
>
> > },
> > .probe = ecap_pwm_probe,
> > .remove = __devexit_p(ecap_pwm_remove),
>
> No __devexit_p() please.
>
> Thierry
>

2012-11-09 11:01:46

by Philip, Avinash

[permalink] [raw]

Subject: RE: [PATCH v2 01/10] PWMSS: Add PWM Subsystem driver for parent<->child relationship

On Fri, Nov 09, 2012 at 12:59:57, Thierry Reding wrote:
> On Thu, Nov 08, 2012 at 01:23:08PM +0530, Philip, Avinash wrote:
> > diff --git a/Documentation/devicetree/bindings/pwm/tipwmss.txt b/Documentation/devicetree/bindings/pwm/tipwmss.txt
> > new file mode 100644
> > index 0000000..b6c2814
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/pwm/tipwmss.txt
> > @@ -0,0 +1,30 @@
> [...]
> > +Also child nodes should also populated under PWMSS DT node.
> > +Example:
>
> Maybe put an blank line between these two lines for readability?

Ok I will add.

>
> > +pwmss0: pwmss@48300000 {
> > + compatible = "ti,am33xx-pwmss";
> > + reg = <0x48300000 0x10
> > + 0x48300100 0x80
> > + 0x48300180 0x80
> > + 0x48300200 0x80>;
>
> I don't think you should list the register spaces of the children here.
> From what I understand, all regions listed in the reg property are
> supposed to be requested by the corresponding driver and therefore
> cannot be used by any other device.

I will check & correct it.

>
> > + ti,hwmods = "epwmss0";
> > + #address-cells = <1>;
> > + #size-cells = <1>;
> > + status = "disabled";
> > + ranges;
>
> I think to represent which memory regions go to the children, you should
> put them in this ranges property, which would then look like this:
>
> ranges = <0x48300100 0x48300100 0x80 /* ECAP */
> 0x48300180 0x48300180 0x80 /* EQEP */
> 0x48300200 0x48300200 0x80>; /* EHRPWM */

I will correct it.

>
> > +
> > + /* child nodes go here */
> > +};
>
> Maybe you should actually list a full set of children here?

Ok.

>
> > diff --git a/drivers/pwm/Kconfig b/drivers/pwm/Kconfig
> > index 6e556c7..384a346 100644
> > --- a/drivers/pwm/Kconfig
> > +++ b/drivers/pwm/Kconfig
> > @@ -136,6 +136,7 @@ config PWM_TEGRA
> > config PWM_TIECAP
> > tristate "ECAP PWM support"
> > depends on SOC_AM33XX
> > + select PWM_TIPWMSS
> > help
> > PWM driver support for the ECAP APWM controller found on AM33XX
> > TI SOC
> > @@ -146,6 +147,7 @@ config PWM_TIECAP
> > config PWM_TIEHRPWM
> > tristate "EHRPWM PWM support"
> > depends on SOC_AM33XX
> > + select PWM_TIPWMSS
> > help
> > PWM driver support for the EHRPWM controller found on AM33XX
> > TI SOC
> > @@ -153,6 +155,15 @@ config PWM_TIEHRPWM
> > To compile this driver as a module, choose M here: the module
> > will be called pwm-tiehrpwm.
> >
> > +config PWM_TIPWMSS
> > + tristate "TI PWM Subsytem parent support"
> > + depends on SOC_AM33XX && (PWM_TIEHRPWM || PWM_TIECAP)
>
> Since you select the symbol from the PWM_TIECAP and PWM_TIEHRPWM symbols
> there is no need to depend on them, right? Oh, but maybe that's to make
> sure the symbol is deselected automatically if neither user is selected.

I want to make sure that, symbol selected only if either are selected.
>
> Perhaps this should actually be a hidden symbol (i.e. leave away the
> prompt string in the tristate option) since it's purely a dependency and
> not useful of its own.

Ok I will take care.

>
> > + help
> > + PWM Subsystem driver support for AM33xx SOC.
> > +
> > + PWM submodules require PWM config space access from submodule
> > + drivers and require common parent driver support.
> > +
> > config PWM_TWL6030
> > tristate "TWL6030 PWM support"
> > depends on TWL4030_CORE
> > diff --git a/drivers/pwm/Makefile b/drivers/pwm/Makefile
> > index 3b3f4c9..55f6fb2 100644
> > --- a/drivers/pwm/Makefile
> > +++ b/drivers/pwm/Makefile
> > @@ -12,5 +12,6 @@ obj-$(CONFIG_PWM_SPEAR) += pwm-spear.o
> > obj-$(CONFIG_PWM_TEGRA) += pwm-tegra.o
> > obj-$(CONFIG_PWM_TIECAP) += pwm-tiecap.o
> > obj-$(CONFIG_PWM_TIEHRPWM) += pwm-tiehrpwm.o
> > +obj-$(CONFIG_PWM_TIPWMSS) += tipwmss.o
>
> This should have a pwm- prefix as well.

Ok I will add.

>
> > obj-$(CONFIG_PWM_TWL6030) += pwm-twl6030.o
> > obj-$(CONFIG_PWM_VT8500) += pwm-vt8500.o
> > diff --git a/drivers/pwm/tipwmss.c b/drivers/pwm/tipwmss.c
> > new file mode 100644
> > index 0000000..c188348
> > --- /dev/null
> > +++ b/drivers/pwm/tipwmss.c
> > @@ -0,0 +1,142 @@
> [...]
> > +#include "tipwmss.h"
> > +
> > +#define PWMSS_CLKCONFIG 0x8 /* Clock gaitng reg, for PWM submodules */
>
> "gating"

Ok

>
> > +#define PWMSS_CLKSTATUS 0xc /* Clock gating status reg */
> > +
> > +struct pwmss_info {
> > + void __iomem *mmio_base;
> > + struct mutex pwmss_lock;
> > + u16 pwmss_clkconfig;
>
> The indentation looks weird on this last field.

Ok I will correct it.

>
> > +};
> > +
> > +u16 pwmss_submodule_state_change(struct device *dev, int set)
> > +{
> > + struct pwmss_info *info = dev_get_drvdata(dev);
> > + u16 val;
> > +
> > + val = readw(info->mmio_base + PWMSS_CLKCONFIG);
> > + val |= set;
> > + mutex_lock(&info->pwmss_lock);
> > + writew(val , info->mmio_base + PWMSS_CLKCONFIG);
> > + mutex_unlock(&info->pwmss_lock);
>
> The mutex needs to span the whole read-modify-write sequence, not just
> the write.

I missed. I will correct it.

>
> Also, how do you clear this state?

Here is the hardware trick comes.
For enabling should write enable bit.
For disabling stop_req bit should set & it will clear enable bit also.

>
> > + return readw(info->mmio_base + PWMSS_CLKSTATUS);
> > +}
> > +EXPORT_SYMBOL(pwmss_submodule_state_change);
> > +
> > +static const struct of_device_id pwmss_of_match[] = {
> > + {
> > + .compatible = "ti,am33xx-pwmss",
> > + },
> > + {},
> > +};
> > +MODULE_DEVICE_TABLE(of, pwmss_of_match);
> > +
> > +static int __devinit pwmss_probe(struct platform_device *pdev)
>
> __dev* annotation usage is deprecated, you should drop it.

Ok

>
> > +{
> > + int ret;
> > + struct resource *r;
> > + struct pwmss_info *info;
> > + struct device_node *node = pdev->dev.of_node;
> > +
> > + info = devm_kzalloc(&pdev->dev, sizeof(*info), GFP_KERNEL);
> > + if (!info) {
> > + dev_err(&pdev->dev, "failed to allocate memory\n");
> > + return -ENOMEM;
> > + }
> > +
> > + mutex_init(&info->pwmss_lock);
> > + r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>
> Blank line between those two lines?

Ok

>
> > + if (!r) {
> > + dev_err(&pdev->dev, "no memory resource defined\n");
> > + return -ENODEV;
> > + }
> > +
> > + info->mmio_base = devm_request_and_ioremap(&pdev->dev, r);
> > + if (!info->mmio_base)
> > + return -EADDRNOTAVAIL;
> > +
> > + pm_runtime_enable(&pdev->dev);
> > + pm_runtime_get_sync(&pdev->dev);
> > + platform_set_drvdata(pdev, info);
> > +
> > + /* Populate all the child nodes here... */
> > + ret = of_platform_populate(node, NULL, NULL, &pdev->dev);
> > + if (ret)
> > + dev_warn(&pdev->dev, "Doesn't have any child node\n");
>
> This reads oddly compared to the other error messages, maybe something
> like "no children found" or similar would be more consistent.

Ok

>
> > +
> > + return ret;
>
> Then again, since you return of_platform_populate()'s error code here,
> you may just want to skip the above warning since the driver probe won't
> succeed anyway. Or if you really want to give a hint as to why loading
> failed, maybe it would be better to make it an error message instead.

Ok I will change to error message.

>
> There is one little problem with registering the children here, which is
> that if you build the drivers as modules, because once the pwm-tipwmss
> module is unloaded, reloading it will fail since it will try to create
> the children again.
>
> AFAICT there are two solutions to this: a) do not allow the pwm-tipwmss
> code to be built as a module and b) have of_platform_populate() called
> by the architecture initialization code. Both are relatively easy to
> implement. a) can be done by making the PWM_TIPWMSS symbol bool instead
> of tristate, and b) can be done by adding "simple-bus" to the end of the
> compatible list in the DT.

Ok I will go for option a.

>
> > +}
> > +
> > +static int __devexit pwmss_remove(struct platform_device *pdev)
>
> Again, no need for __devexit anymore.

I will remove

>
...
> > +
> > +static const struct dev_pm_ops pwmss_pm_ops = {
> > + .suspend = pwmss_suspend,
> > + .resume = pwmss_resume,
> > +};
>
> Shouldn't these functions be conditionalized on CONFIG_PM_SLEEP? And
> maybe you want to use the SIMPLE_DEV_PM_OPS macro here.

I will check and correct it.

>
> > +
> > +static struct platform_driver pwmss_driver = {
> > + .driver = {
> > + .name = "pwmss",
> > + .owner = THIS_MODULE,
> > + .pm = &pwmss_pm_ops,
> > + .of_match_table = of_match_ptr(pwmss_of_match),
>
> You already define the pwmss_of_match table unconditionally, so you
> don't need the of_match_ptr() either.

Will remove

>
> > + },
> > + .probe = pwmss_probe,
> > + .remove = __devexit_p(pwmss_remove),
>
> __devexit_p() can go away.

Will remove

>
> > +};
> > +
> > +module_platform_driver(pwmss_driver);
> > +
> > +MODULE_DESCRIPTION("pwmss driver");
>
> This description could be better.

Will add.
>
> > +MODULE_AUTHOR("Texas Instruments");
> > +MODULE_LICENSE("GPL");
> > diff --git a/drivers/pwm/tipwmss.h b/drivers/pwm/tipwmss.h
> > new file mode 100644
> > index 0000000..f9cb2e2
> > --- /dev/null
> > +++ b/drivers/pwm/tipwmss.h
>
> I think this should also get the pwm- prefix for consistency with the
> source file.

Will add the prefix.

>
> > @@ -0,0 +1,30 @@
> > +/*
> > + * TI PWM Subsystem driver
> > + *
> > + * Copyright (C) 2012 Texas Instruments Incorporated - http://www.ti.com/
> > + *
> > + * This program is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License as published by
> > + * the Free Software Foundation; either version 2 of the License, or
> > + * (at your option) any later version.
> > + *
> > + * This program is distributed in the hope that it will be useful,
> > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> > + * GNU General Public License for more details.
> > + *
> > + */
> > +
> > +#ifndef __TIPWMSS_H
> > +#define __TIPWMSS_H
> > +
> > +#ifdef CONFIG_PWM_TIPWMSS
> > +extern u16 pwmss_submodule_state_change(struct device *dev, int set);
> > +#else
> > +static inline u16 pwmss_submodule_state_change(struct device *dev, int set)
> > +{
> > + /* return success status value */
> > + return 0xFFFF;
> > +}
> > +#endif
> > +#endif /* __TIPWMSS_H */
>
> Is it really necessary to provide a !PWM_TIPWMSS version of this
> function? All users that want to use it can select it and get the
> correct version, right?

Added the !PWM_TIPWMSS part to support platforms that didn't
have PWM Subsytem support.

I will send next version on nov 19th as I was on leave on next week.

Thanks
Avinash
>
> Thierry
>

2012-11-09 11:19:37

by Thierry Reding

[permalink] [raw]

Subject: Re: [PATCH v2 04/10] pwm: pwm-tiecap: Add device-tree binding support for APWM driver

On Fri, Nov 09, 2012 at 10:59:30AM +0000, Philip, Avinash wrote:
> On Fri, Nov 09, 2012 at 13:22:19, Thierry Reding wrote:
> > On Thu, Nov 08, 2012 at 01:23:11PM +0530, Philip, Avinash wrote:
> > > +#define ECAPCLK_EN BIT(0)
> > > +#define ECAPCLK_STOP_REQ BIT(1)
> >
> > This one doesn't seem to align with the rest. Also, why is bit 0 called
> > _EN and bit 1 _STOP_REQ? Couldn't they be made more consistent, i.e.
> > _START and _STOP? Or _ENABLE and _DISABLE?
>
> Ok I will change to PWMSS_ECAPCLK_EN & PWMSS_ECAPCLK_STPO_REQ

While at it, maybe move these defines to the pwm-tipwmss.h header file?
After all that's the module that accesses the register that contains
these bits.

Attachments:

(No filename) (689.00 B)
(No filename) (836.00 B)
Download all attachments