Hi Thomas,
thanks for your review:
Am Montag, 27. August 2012, 06:20:49 schrieb Thomas Abraham:
Hi Heiko,
On 26 August 2012 03:23, Heiko St?bner [off-list ref] wrote:
quoted
Until now the Exynos-SoC was the only Samsung-SoC supporting the GPIOs
via the device tree. This patch implements dt-support for the
s3c24xx arches.
The controllers contain only 3 cells, as the underlying gpio controller
does not support controlling the drive strength on a gpio level.
Tested with the gpio-keys driver on a s3c2416 based machine.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
---
.../devicetree/bindings/gpio/gpio-samsung.txt | 38 ++++++++++++
drivers/gpio/gpio-samsung.c | 63
++++++++++++++++++++ 2 files changed, 101 insertions(+), 0 deletions(-)
@@ -39,3 +39,41 @@ Example: #gpio-cells = <4>; gpio-controller; };+++Samsung S3C24XX GPIO Controller++Required properties:+- compatible: Compatible property value should be
"samsung,s3c24xx-gpio".
This is debatable, but I would choose to be specific here and use
"samsing,s3c2416-gpio". For instance, SoC's in s3c24xx family have
differing value for pull none/down/up.
Hmmm, but the logic to drive the gpio controller is the same for all arches.
The only difference is the number to use for the different pull up/down
settings. So I think the "samsung,s3c24xx-gpio" is ok but the documentation
should simply reflect the different pull settings.
quoted
+
+- reg: Physical base address of the controller and length of memory
mapped + region.
+
+- #gpio-cells: Should be 3. The syntax of the gpio specifier used by
client nodes + should be the following with values derived from the SoC
user manual. + <[phandle of the gpio controller node]
+ [pin number within the gpio controller]
+ [mux function]
+ [flags and pull up/down]
+
+ Values for gpio specifier:
+ - Pin number: depending on the controller a number from 0 up to 15.
+ - Flags and Pull Up/Down: 0 - Pull Up/Down Disabled.
+ 1 - Pull Down Enabled.
+ 3 - Pull Up Enabled.
As per s3c2416 user manual, 2 is used for Pull Up and 3 is listed as
reserved value. Any particular reason to use 3 here for pull up and
not 2 as per the user manual?
Sorry, copy and paste error :-)
quoted
+ Bit 16 (0x00010000) - Input is active low.
+
+- gpio-controller: Specifies that the node is a gpio controller.
+- #address-cells: should be 1.
+- #size-cells: should be 1.
It would be informative to add information about the 'mux function'
cell here as well. Specifically, on how to handle the banks that have
an extended GPxSEL register that allow additional pin function
selection.
Will add a mux function description.
Until now I've never realised the existence of the GPxSEL registers, so thanks
for the pointer :-).
But is this used in the driver at all? All the setcfg/getcfg functions in
gpio-samsung.c only handle the GPxCON registers - I haven't found code to
handle the GPxSEL registers at all. So my guess is that this was never
implemented - or that I'm blind ;-) .
Hi,
On Mon, Aug 27, 2012 at 11:01 AM, Heiko St?bner [off-list ref] wrote:
Hi Thomas,
thanks for your review:
Hmmm, but the logic to drive the gpio controller is the same for all arches.
The only difference is the number to use for the different pull up/down
settings. So I think the "samsung,s3c24xx-gpio" is ok but the documentation
should simply reflect the different pull settings.
Also s3c2410 and s3c2440 has different count of GPIOs. And that could
be a problem on machines
with GPIO-expanders (h1940 machine)
quoted
It would be informative to add information about the 'mux function'
cell here as well. Specifically, on how to handle the banks that have
an extended GPxSEL register that allow additional pin function
selection.
Will add a mux function description.
Until now I've never realised the existence of the GPxSEL registers, so thanks
for the pointer :-).
But is this used in the driver at all? All the setcfg/getcfg functions in
gpio-samsung.c only handle the GPxCON registers - I haven't found code to
handle the GPxSEL registers at all. So my guess is that this was never
implemented - or that I'm blind ;-) .
There're no GPxSEL registers on s3c2410/s3c2440/s3c2442. Is it
something s3c2412/sec2416 specific?
Regards
Vasily
Hi Vasily,
Am Montag, 27. August 2012, 10:57:04 schrieb Vasily Khoruzhick:
Hi,
On Mon, Aug 27, 2012 at 11:01 AM, Heiko St?bner [off-list ref] wrote:
quoted
Hi Thomas,
thanks for your review:
Hmmm, but the logic to drive the gpio controller is the same for all
arches. The only difference is the number to use for the different pull
up/down settings. So I think the "samsung,s3c24xx-gpio" is ok but the
documentation should simply reflect the different pull settings.
Also s3c2410 and s3c2440 has different count of GPIOs. And that could
be a problem on machines
with GPIO-expanders (h1940 machine)
I'm not sure I follow :-) .
The compatible property here only sets the mechanism on how to handle the
gpios defined in the devicetree - here to use the s3c24xx-style. As you can
see in gpio-samsung.c the handling is already unified for all the s3c24xx
architectures.
The definition of what gpio banks exist is then done in the respective
devicetree file for the individual SoC. And of course here one would have
individual definitions, depending on the banks present.
For reference my quite empty s3c2416.dtsi file currently looks like:
/include/ "skeleton.dtsi"
/ {
compatible = "samsung,s3c2416";
cpus {
cpu at 0 {
compatible = "arm,arm926ejs";
};
};
gpio-controllers {
#address-cells = <1>;
#size-cells = <1>;
gpio-controller;
ranges;
gpa: gpio-controller at 56000000 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000000 0x10>;
#gpio-cells = <3>;
};
gpb: gpio-controller at 56000010 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000010 0x10>;
#gpio-cells = <3>;
};
gpc: gpio-controller at 56000020 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000020 0x10>;
#gpio-cells = <3>;
};
gpd: gpio-controller at 56000030 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000030 0x10>;
#gpio-cells = <3>;
};
gpe: gpio-controller at 56000040 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000040 0x10>;
#gpio-cells = <3>;
};
gpf: gpio-controller at 56000050 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000050 0x10>;
#gpio-cells = <3>;
};
gpg: gpio-controller at 56000060 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000060 0x10>;
#gpio-cells = <3>;
};
gph: gpio-controller at 56000070 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000070 0x10>;
#gpio-cells = <3>;
};
/* s3c2443 and later */
gpj: gpio-controller at 560000D0 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x560000D0 0x10>;
#gpio-cells = <3>;
};
gpk: gpio-controller at 560000E0 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x560000E0 0x10>;
#gpio-cells = <3>;
};
gpl: gpio-controller at 560000F0 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x560000F0 0x10>;
#gpio-cells = <3>;
};
gpm: gpio-controller at 56000100 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000100 0x10>;
#gpio-cells = <3>;
};
};
};
Other s3c24xx SoCs would of course need to define their own.
quoted
quoted
It would be informative to add information about the 'mux function'
cell here as well. Specifically, on how to handle the banks that have
an extended GPxSEL register that allow additional pin function
selection.
Will add a mux function description.
Until now I've never realised the existence of the GPxSEL registers, so
thanks for the pointer :-).
But is this used in the driver at all? All the setcfg/getcfg functions in
gpio-samsung.c only handle the GPxCON registers - I haven't found code to
handle the GPxSEL registers at all. So my guess is that this was never
implemented - or that I'm blind ;-) .
There're no GPxSEL registers on s3c2410/s3c2440/s3c2442. Is it
something s3c2412/sec2416 specific?
s3c2412 also does not have it ... it seems entirely specific to
s3c2416/s3c2450 and as written also seems to be unused.
So I would tend to ignore it for now :-) .
Heiko
On Mon, Aug 27, 2012 at 12:44 PM, Heiko St?bner [off-list ref] wrote:
Hi Vasily,
I'm not sure I follow :-) .
The compatible property here only sets the mechanism on how to handle the
gpios defined in the devicetree - here to use the s3c24xx-style. As you can
see in gpio-samsung.c the handling is already unified for all the s3c24xx
architectures.
The definition of what gpio banks exist is then done in the respective
devicetree file for the individual SoC. And of course here one would have
individual definitions, depending on the banks present.
For reference my quite empty s3c2416.dtsi file currently looks like:
/include/ "skeleton.dtsi"
/ {
compatible = "samsung,s3c2416";
cpus {
cpu at 0 {
compatible = "arm,arm926ejs";
};
};
gpio-controllers {
#address-cells = <1>;
#size-cells = <1>;
gpio-controller;
ranges;
gpa: gpio-controller at 56000000 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000000 0x10>;
#gpio-cells = <3>;
};
gpb: gpio-controller at 56000010 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000010 0x10>;
#gpio-cells = <3>;
};
gpc: gpio-controller at 56000020 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000020 0x10>;
#gpio-cells = <3>;
};
gpd: gpio-controller at 56000030 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000030 0x10>;
#gpio-cells = <3>;
};
gpe: gpio-controller at 56000040 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000040 0x10>;
#gpio-cells = <3>;
};
gpf: gpio-controller at 56000050 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000050 0x10>;
#gpio-cells = <3>;
};
gpg: gpio-controller at 56000060 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000060 0x10>;
#gpio-cells = <3>;
};
gph: gpio-controller at 56000070 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000070 0x10>;
#gpio-cells = <3>;
};
/* s3c2443 and later */
gpj: gpio-controller at 560000D0 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x560000D0 0x10>;
#gpio-cells = <3>;
};
gpk: gpio-controller at 560000E0 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x560000E0 0x10>;
#gpio-cells = <3>;
};
gpl: gpio-controller at 560000F0 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x560000F0 0x10>;
#gpio-cells = <3>;
};
gpm: gpio-controller at 56000100 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000100 0x10>;
#gpio-cells = <3>;
};
};
};
Other s3c24xx SoCs would of course need to define their own.
I see. But how does it handle GPA bank (which is output-only, and
GPACON differs a bit from GP{B-J}CON? And some banks has lower number
(not 16) of GPIOs, like GPH.
Also, what would be value for S3C_GPIO_END?
Regards
Vasily
Am Montag, 27. August 2012, 11:51:27 schrieb Vasily Khoruzhick:
On Mon, Aug 27, 2012 at 12:44 PM, Heiko St?bner [off-list ref] wrote:
quoted
Hi Vasily,
I'm not sure I follow :-) .
The compatible property here only sets the mechanism on how to handle the
gpios defined in the devicetree - here to use the s3c24xx-style. As you
can see in gpio-samsung.c the handling is already unified for all the
s3c24xx architectures.
The definition of what gpio banks exist is then done in the respective
devicetree file for the individual SoC. And of course here one would have
individual definitions, depending on the banks present.
For reference my quite empty s3c2416.dtsi file currently looks like:
/include/ "skeleton.dtsi"
/ {
compatible = "samsung,s3c2416";
cpus {
cpu at 0 {
compatible = "arm,arm926ejs";
};
};
gpio-controllers {
#address-cells = <1>;
#size-cells = <1>;
gpio-controller;
ranges;
gpa: gpio-controller at 56000000 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000000 0x10>;
#gpio-cells = <3>;
};
gpb: gpio-controller at 56000010 {
compatible = "samsung,s3c24xx-gpio";
reg = <0x56000010 0x10>;
#gpio-cells = <3>;
};
[ ... ]
quoted
};
};
Other s3c24xx SoCs would of course need to define their own.
I see. But how does it handle GPA bank (which is output-only, and
GPACON differs a bit from GP{B-J}CON? And some banks has lower number
(not 16) of GPIOs, like GPH.
Also, what would be value for S3C_GPIO_END?
The creation of the banks is still done in gpio-samsung.c and the special
handling of bank-a is still sitting in the s3c24xx_gpios[] array there.
In general the gpios are still created completely like before and the
devicetree information is only attached to them so that dt-devices can
reference them.
As you can see in the original patch, after the respective chip got created in
s3c24xx_gpiolib_add_chips, s3c24xx_gpiolib_attach_ofnode searches for the
representation of it in the device tree and attaches this to the chip.
So nothing really changed, except that dt-devices can now reference gpios and
if they do s3c24xx_gpio_xlate will get called to configure them with the
target values from the dt.
Currently I'm also only using it in a mixed mode configuration [1],
supplementing a normal machine file with some devices from the devicetree -
especially as I still need to figure out, what to do about the interrupt
controller.
Heiko
[1] https://github.com/mmind/linux-es600/blob/topic/es600-
devel/arch/arm/boot/dts/s3c24xx-sg06.dts
From: Thomas Abraham <hidden> Date: 2012-08-28 04:44:02
On 27 August 2012 13:31, Heiko St?bner [off-list ref] wrote:
Hi Thomas,
thanks for your review:
Am Montag, 27. August 2012, 06:20:49 schrieb Thomas Abraham:
quoted
Hi Heiko,
On 26 August 2012 03:23, Heiko St?bner [off-list ref] wrote:
quoted
Until now the Exynos-SoC was the only Samsung-SoC supporting the GPIOs
via the device tree. This patch implements dt-support for the
s3c24xx arches.
The controllers contain only 3 cells, as the underlying gpio controller
does not support controlling the drive strength on a gpio level.
Tested with the gpio-keys driver on a s3c2416 based machine.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
---
.../devicetree/bindings/gpio/gpio-samsung.txt | 38 ++++++++++++
drivers/gpio/gpio-samsung.c | 63
++++++++++++++++++++ 2 files changed, 101 insertions(+), 0 deletions(-)
@@ -39,3 +39,41 @@ Example: #gpio-cells = <4>; gpio-controller; };+++Samsung S3C24XX GPIO Controller++Required properties:+- compatible: Compatible property value should be
"samsung,s3c24xx-gpio".
This is debatable, but I would choose to be specific here and use
"samsing,s3c2416-gpio". For instance, SoC's in s3c24xx family have
differing value for pull none/down/up.
Hmmm, but the logic to drive the gpio controller is the same for all arches.
The only difference is the number to use for the different pull up/down
settings. So I think the "samsung,s3c24xx-gpio" is ok but the documentation
should simply reflect the different pull settings.
Ok.
quoted
quoted
+
+- reg: Physical base address of the controller and length of memory
mapped + region.
+
+- #gpio-cells: Should be 3. The syntax of the gpio specifier used by
client nodes + should be the following with values derived from the SoC
user manual. + <[phandle of the gpio controller node]
+ [pin number within the gpio controller]
+ [mux function]
+ [flags and pull up/down]
+
+ Values for gpio specifier:
+ - Pin number: depending on the controller a number from 0 up to 15.
+ - Flags and Pull Up/Down: 0 - Pull Up/Down Disabled.
+ 1 - Pull Down Enabled.
+ 3 - Pull Up Enabled.
As per s3c2416 user manual, 2 is used for Pull Up and 3 is listed as
reserved value. Any particular reason to use 3 here for pull up and
not 2 as per the user manual?
Sorry, copy and paste error :-)
Ok.
quoted
quoted
+ Bit 16 (0x00010000) - Input is active low.
+
+- gpio-controller: Specifies that the node is a gpio controller.
+- #address-cells: should be 1.
+- #size-cells: should be 1.
It would be informative to add information about the 'mux function'
cell here as well. Specifically, on how to handle the banks that have
an extended GPxSEL register that allow additional pin function
selection.
Will add a mux function description.
Until now I've never realised the existence of the GPxSEL registers, so thanks
for the pointer :-).
But is this used in the driver at all? All the setcfg/getcfg functions in
gpio-samsung.c only handle the GPxCON registers - I haven't found code to
handle the GPxSEL registers at all. So my guess is that this was never
implemented - or that I'm blind ;-) .
You are right, there is no code that is handling the GPxCON registers.
Thanks,
Thomas.