A GPIO driver for the AMD Geode CS5535/5536 Companion Devices.
It uses the gpio framework and the gpio api as defined in
arch/x86/kernel/geode_32.c
Patch is against 2.6.28-rc9.
Signed-off-by: Alessandro Zummo <[email protected]>
---
arch/x86/include/asm/geode.h | 18 ++++
drivers/gpio/Kconfig | 14 +++
drivers/gpio/Makefile | 1
drivers/gpio/cs553x-gpio.c | 172 +++++++++++++++++++++++++++++++++++++++++++
4 files changed, 205 insertions(+)
--- linux-tuxrouter-2.6.28-rc9.orig/drivers/gpio/Kconfig 2008-12-24 23:34:02.000000000 +0100
+++ linux-tuxrouter-2.6.28-rc9/drivers/gpio/Kconfig 2008-12-24 23:34:44.000000000 +0100
@@ -175,4 +175,18 @@ config GPIO_MCP23S08
SPI driver for Microchip MCP23S08 I/O expander. This provides
a GPIO interface supporting inputs and outputs.
+comment "Other GPIO expanders:"
+
+config GPIO_CS553X
+ tristate "AMD CS5535/CS5536 Geode Companion Devices"
+ depends on MGEODE_LX && !CS5535_GPIO
+ help
+ Say yes here to provide support for the 28 GPIO pins on
+ the AMD CS5535 and CS5536 Geode Companion Devices.
+
+ Your board setup code will have to declare an appropriate
+ platform device for this driver to work.
+
+ If unsure, say N.
+
endif
--- linux-tuxrouter-2.6.28-rc9.orig/drivers/gpio/Makefile 2008-12-24 23:34:02.000000000 +0100
+++ linux-tuxrouter-2.6.28-rc9/drivers/gpio/Makefile 2008-12-24 23:34:44.000000000 +0100
@@ -4,6 +4,7 @@ ccflags-$(CONFIG_DEBUG_GPIO) += -DDEBUG
obj-$(CONFIG_GPIOLIB) += gpiolib.o
+obj-$(CONFIG_GPIO_CS553X) += cs553x-gpio.o
obj-$(CONFIG_GPIO_MAX7301) += max7301.o
obj-$(CONFIG_GPIO_MAX732X) += max732x.o
obj-$(CONFIG_GPIO_MCP23S08) += mcp23s08.o
--- linux-tuxrouter-2.6.28-rc9.orig/arch/x86/include/asm/geode.h 2008-12-24 23:34:02.000000000 +0100
+++ linux-tuxrouter-2.6.28-rc9/arch/x86/include/asm/geode.h 2008-12-24 23:34:44.000000000 +0100
@@ -12,6 +12,7 @@
#include <asm/processor.h>
#include <linux/io.h>
+#include <linux/platform_device.h>
/* Generic southbridge functions */
@@ -165,6 +166,23 @@ static inline void geode_gpio_event_pme(
geode_gpio_setup_event(gpio, pair, 1);
}
+struct cs553x_gpio_platform_data {
+
+ unsigned gpio_base; /* number of the first GPIO */
+
+ resource_size_t io_base;
+
+ void *context; /* param to setup/teardown */
+
+ int (*setup)(struct platform_device *pdev,
+ unsigned base, unsigned ngpio,
+ void *context);
+
+ int (*teardown)(struct platform_device *pdev,
+ unsigned base, unsigned ngpio,
+ void *context);
+};
+
/* Specific geode tests */
static inline int is_geode_gx(void)
--- /dev/null 1970-01-01 00:00:00.000000000 +0000
+++ linux-tuxrouter-2.6.28-rc9/drivers/gpio/cs553x-gpio.c 2008-12-24 23:35:41.000000000 +0100
@@ -0,0 +1,172 @@
+/*
+ * AMD CS5535/CS5536 Geode Companion Devices GPIO driver
+ *
+ * Copyright (C) 2008 Tower Technologies
+ * Written by Alessandro Zummo <[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/module.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/ioport.h>
+#include <linux/gpio.h>
+#include <linux/platform_device.h>
+
+#include <asm/geode.h>
+
+#define DRIVER_NAME "cs553x-gpio"
+#define DRIVER_VERSION "1.0"
+
+static void cs553x_gpio_set(struct gpio_chip *chip, unsigned offset, int value)
+{
+ if (value)
+ geode_gpio_set(geode_gpio(offset), GPIO_OUTPUT_VAL);
+ else
+ geode_gpio_clear(geode_gpio(offset), GPIO_OUTPUT_VAL);
+}
+
+static int cs553x_gpio_get(struct gpio_chip *chip, unsigned offset)
+{
+ return geode_gpio_isset(geode_gpio(offset), GPIO_READ_BACK);
+}
+
+static int cs553x_direction_input(struct gpio_chip *chip, unsigned offset)
+{
+ geode_gpio_clear(geode_gpio(offset), GPIO_OUTPUT_ENABLE);
+ geode_gpio_set(geode_gpio(offset), GPIO_INPUT_ENABLE);
+
+ return 0;
+}
+
+static int cs553x_direction_output(struct gpio_chip *chip, unsigned offset,
+ int value)
+{
+ cs553x_gpio_set(chip, offset, value);
+
+ geode_gpio_set(geode_gpio(offset), GPIO_OUTPUT_ENABLE);
+ geode_gpio_clear(geode_gpio(offset), GPIO_INPUT_ENABLE);
+
+ return 0;
+}
+
+static struct gpio_chip cs553x = {
+ .owner = THIS_MODULE,
+ .label = DRIVER_NAME,
+ .ngpio = 28,
+
+ .get = cs553x_gpio_get,
+ .set = cs553x_gpio_set,
+
+ .direction_input = cs553x_direction_input,
+ .direction_output = cs553x_direction_output,
+};
+
+
+static int __init cs553x_gpio_probe(struct platform_device *pdev)
+{
+ int rc;
+ struct cs553x_gpio_platform_data *pdata = pdev->dev.platform_data;
+
+ if (pdata == NULL) {
+ dev_err(&pdev->dev, "missing CS553X platform data\n");
+ return -ENODEV;
+ }
+
+ if (!pdata->io_base) {
+ dev_err(&pdev->dev, "platform data has no GPIO base region\n");
+ return -ENODEV;
+ }
+
+ if (!request_region(pdata->io_base, LBAR_GPIO_SIZE, DRIVER_NAME)) {
+ dev_err(&pdev->dev, "cannot allocate I/O region at %x\n",
+ pdata->io_base);
+ return -ENODEV;
+ }
+
+ cs553x.dev = &pdev->dev;
+ cs553x.base = pdata->gpio_base;
+
+ rc = gpiochip_add(&cs553x);
+ if (rc < 0) {
+ dev_err(&pdev->dev, "cannot add GPIO\n");
+ goto fail_release;
+ }
+
+ dev_info(&pdev->dev, "registered at 0x%x, GPIO base %d\n",
+ pdata->io_base, cs553x.base);
+
+ if (pdata->setup) {
+ rc = pdata->setup(pdev, cs553x.base,
+ cs553x.ngpio, pdata->context);
+
+ if (rc < 0) {
+ dev_err(&pdev->dev, "setup failed, %d\n", rc);
+ goto fail_remove;
+ }
+ }
+
+ return 0;
+
+fail_remove:
+ gpiochip_remove(&cs553x);
+
+fail_release:
+ release_region(pdata->io_base, LBAR_GPIO_SIZE);
+
+ return rc;
+}
+
+static int __exit cs553x_gpio_remove(struct platform_device *pdev)
+{
+ struct cs553x_gpio_platform_data *pdata = pdev->dev.platform_data;
+
+ if (pdata->teardown) {
+ int rc = pdata->teardown(pdev, cs553x.base,
+ cs553x.ngpio, pdata->context);
+ if (rc < 0) {
+ dev_err(&pdev->dev, "teardown failed, %d\n", rc);
+ return rc;
+ }
+ }
+
+ gpiochip_remove(&cs553x);
+ release_region(pdata->io_base, LBAR_GPIO_SIZE);
+
+ return 0;
+}
+
+static struct platform_driver cs553x_gpio_driver = {
+ .driver = {
+ .name = DRIVER_NAME,
+ .owner = THIS_MODULE,
+ },
+ .remove = __exit_p(cs553x_gpio_remove),
+};
+
+static int __init cs553x_gpio_init(void)
+{
+ if (!is_geode())
+ return -ENODEV;
+
+ return platform_driver_probe(&cs553x_gpio_driver, cs553x_gpio_probe);
+}
+
+static void __exit cs553x_gpio_exit(void)
+{
+ platform_driver_unregister(&cs553x_gpio_driver);
+}
+
+module_init(cs553x_gpio_init);
+module_exit(cs553x_gpio_exit);
+
+
+MODULE_AUTHOR("Alessandro Zummo <[email protected]>");
+MODULE_DESCRIPTION("AMD CS5535/CS5536 GPIO driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRIVER_VERSION);
+MODULE_ALIAS("platform:cs553x-gpio");
--
Best regards,
Alessandro Zummo,
Tower Technologies - Torino, Italy
http://www.towertech.it
On Wednesday 24 December 2008, Alessandro Zummo wrote:
>
> A GPIO driver for the AMD Geode CS5535/5536 Companion Devices.
Great -- embedded x86 looking more like embedded rest-of-Linux! ;)
> It uses the gpio framework and the gpio api as defined in
> arch/x86/kernel/geode_32.c
Eventually I'd hope to see those geode_32.c calls just vanish.
In fact, the "normal" way to package these GPIOs would be to
always provide them through the standard API, in arch/... code,
with no Kconfig option. Any reason you shouldn't do that in
this patch?
> Patch is against 2.6.28-rc9.
>
> Signed-off-by: Alessandro Zummo <[email protected]>
>
>
> ---
> arch/x86/include/asm/geode.h | 18 ++++
> drivers/gpio/Kconfig | 14 +++
> drivers/gpio/Makefile | 1
> drivers/gpio/cs553x-gpio.c | 172 +++++++++++++++++++++++++++++++++++++++++++
> 4 files changed, 205 insertions(+)
>
> --- linux-tuxrouter-2.6.28-rc9.orig/drivers/gpio/Kconfig 2008-12-24 23:34:02.000000000 +0100
> +++ linux-tuxrouter-2.6.28-rc9/drivers/gpio/Kconfig 2008-12-24 23:34:44.000000000 +0100
> @@ -175,4 +175,18 @@ config GPIO_MCP23S08
> SPI driver for Microchip MCP23S08 I/O expander. This provides
> a GPIO interface supporting inputs and outputs.
>
> +comment "Other GPIO expanders:"
This counts as "memory mapped" I'd say. Doesn't need
a new category, even if this does need to live outside
the relevant arch/... files.
> +
> +config GPIO_CS553X
> + tristate "AMD CS5535/CS5536 Geode Companion Devices"
> + depends on MGEODE_LX && !CS5535_GPIO
What's this CS5535_GPIO stuff? And why should it affect
whether this can be configured? (It's not in mainline...)
> + help
> + Say yes here to provide support for the 28 GPIO pins on
> + the AMD CS5535 and CS5536 Geode Companion Devices.
> +
> + Your board setup code will have to declare an appropriate
> + platform device for this driver to work.
> +
> + If unsure, say N.
> +
> endif
> --- linux-tuxrouter-2.6.28-rc9.orig/drivers/gpio/Makefile 2008-12-24 23:34:02.000000000 +0100
> +++ linux-tuxrouter-2.6.28-rc9/drivers/gpio/Makefile 2008-12-24 23:34:44.000000000 +0100
> @@ -4,6 +4,7 @@ ccflags-$(CONFIG_DEBUG_GPIO) += -DDEBUG
>
> obj-$(CONFIG_GPIOLIB) += gpiolib.o
>
> +obj-$(CONFIG_GPIO_CS553X) += cs553x-gpio.o
> obj-$(CONFIG_GPIO_MAX7301) += max7301.o
> obj-$(CONFIG_GPIO_MAX732X) += max732x.o
> obj-$(CONFIG_GPIO_MCP23S08) += mcp23s08.o
> --- linux-tuxrouter-2.6.28-rc9.orig/arch/x86/include/asm/geode.h 2008-12-24 23:34:02.000000000 +0100
> +++ linux-tuxrouter-2.6.28-rc9/arch/x86/include/asm/geode.h 2008-12-24 23:34:44.000000000 +0100
> @@ -12,6 +12,7 @@
>
> #include <asm/processor.h>
> #include <linux/io.h>
> +#include <linux/platform_device.h>
>
> /* Generic southbridge functions */
>
> @@ -165,6 +166,23 @@ static inline void geode_gpio_event_pme(
> geode_gpio_setup_event(gpio, pair, 1);
> }
>
> +struct cs553x_gpio_platform_data {
> +
> + unsigned gpio_base; /* number of the first GPIO */
> +
> + resource_size_t io_base;
Platform devices should use platform_get_resource() and
friends instead of passing resources through platform data.
> +
> + void *context; /* param to setup/teardown */
> +
> + int (*setup)(struct platform_device *pdev,
> + unsigned base, unsigned ngpio,
> + void *context);
> +
> + int (*teardown)(struct platform_device *pdev,
> + unsigned base, unsigned ngpio,
> + void *context);
> +};
> +
> /* Specific geode tests */
>
> static inline int is_geode_gx(void)
... other than those points, this seems like a simple and
straightforward GPIO driver. Typical of what arch/* holds
in such cases. ;)
- Dave
> --- /dev/null 1970-01-01 00:00:00.000000000 +0000
> +++ linux-tuxrouter-2.6.28-rc9/drivers/gpio/cs553x-gpio.c 2008-12-24 23:35:41.000000000 +0100
> @@ -0,0 +1,172 @@
> +/*
> + * AMD CS5535/CS5536 Geode Companion Devices GPIO driver
> + *
> + * Copyright (C) 2008 Tower Technologies
> + * Written by Alessandro Zummo <[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/module.h>
> +#include <linux/errno.h>
> +#include <linux/kernel.h>
> +#include <linux/init.h>
> +#include <linux/ioport.h>
> +#include <linux/gpio.h>
> +#include <linux/platform_device.h>
> +
> +#include <asm/geode.h>
> +
> +#define DRIVER_NAME "cs553x-gpio"
> +#define DRIVER_VERSION "1.0"
> +
> +static void cs553x_gpio_set(struct gpio_chip *chip, unsigned offset, int value)
> +{
> + if (value)
> + geode_gpio_set(geode_gpio(offset), GPIO_OUTPUT_VAL);
> + else
> + geode_gpio_clear(geode_gpio(offset), GPIO_OUTPUT_VAL);
> +}
> +
> +static int cs553x_gpio_get(struct gpio_chip *chip, unsigned offset)
> +{
> + return geode_gpio_isset(geode_gpio(offset), GPIO_READ_BACK);
> +}
> +
> +static int cs553x_direction_input(struct gpio_chip *chip, unsigned offset)
> +{
> + geode_gpio_clear(geode_gpio(offset), GPIO_OUTPUT_ENABLE);
> + geode_gpio_set(geode_gpio(offset), GPIO_INPUT_ENABLE);
> +
> + return 0;
> +}
> +
> +static int cs553x_direction_output(struct gpio_chip *chip, unsigned offset,
> + int value)
> +{
> + cs553x_gpio_set(chip, offset, value);
> +
> + geode_gpio_set(geode_gpio(offset), GPIO_OUTPUT_ENABLE);
> + geode_gpio_clear(geode_gpio(offset), GPIO_INPUT_ENABLE);
> +
> + return 0;
> +}
> +
> +static struct gpio_chip cs553x = {
> + .owner = THIS_MODULE,
> + .label = DRIVER_NAME,
> + .ngpio = 28,
> +
> + .get = cs553x_gpio_get,
> + .set = cs553x_gpio_set,
> +
> + .direction_input = cs553x_direction_input,
> + .direction_output = cs553x_direction_output,
> +};
> +
> +
> +static int __init cs553x_gpio_probe(struct platform_device *pdev)
> +{
> + int rc;
> + struct cs553x_gpio_platform_data *pdata = pdev->dev.platform_data;
> +
> + if (pdata == NULL) {
> + dev_err(&pdev->dev, "missing CS553X platform data\n");
> + return -ENODEV;
> + }
> +
> + if (!pdata->io_base) {
> + dev_err(&pdev->dev, "platform data has no GPIO base region\n");
> + return -ENODEV;
> + }
> +
> + if (!request_region(pdata->io_base, LBAR_GPIO_SIZE, DRIVER_NAME)) {
> + dev_err(&pdev->dev, "cannot allocate I/O region at %x\n",
> + pdata->io_base);
> + return -ENODEV;
> + }
> +
> + cs553x.dev = &pdev->dev;
> + cs553x.base = pdata->gpio_base;
> +
> + rc = gpiochip_add(&cs553x);
> + if (rc < 0) {
> + dev_err(&pdev->dev, "cannot add GPIO\n");
> + goto fail_release;
> + }
> +
> + dev_info(&pdev->dev, "registered at 0x%x, GPIO base %d\n",
> + pdata->io_base, cs553x.base);
> +
> + if (pdata->setup) {
> + rc = pdata->setup(pdev, cs553x.base,
> + cs553x.ngpio, pdata->context);
> +
> + if (rc < 0) {
> + dev_err(&pdev->dev, "setup failed, %d\n", rc);
> + goto fail_remove;
> + }
> + }
> +
> + return 0;
> +
> +fail_remove:
> + gpiochip_remove(&cs553x);
> +
> +fail_release:
> + release_region(pdata->io_base, LBAR_GPIO_SIZE);
> +
> + return rc;
> +}
> +
> +static int __exit cs553x_gpio_remove(struct platform_device *pdev)
> +{
> + struct cs553x_gpio_platform_data *pdata = pdev->dev.platform_data;
> +
> + if (pdata->teardown) {
> + int rc = pdata->teardown(pdev, cs553x.base,
> + cs553x.ngpio, pdata->context);
> + if (rc < 0) {
> + dev_err(&pdev->dev, "teardown failed, %d\n", rc);
> + return rc;
> + }
> + }
> +
> + gpiochip_remove(&cs553x);
> + release_region(pdata->io_base, LBAR_GPIO_SIZE);
> +
> + return 0;
> +}
> +
> +static struct platform_driver cs553x_gpio_driver = {
> + .driver = {
> + .name = DRIVER_NAME,
> + .owner = THIS_MODULE,
> + },
> + .remove = __exit_p(cs553x_gpio_remove),
> +};
> +
> +static int __init cs553x_gpio_init(void)
> +{
> + if (!is_geode())
> + return -ENODEV;
> +
> + return platform_driver_probe(&cs553x_gpio_driver, cs553x_gpio_probe);
> +}
> +
> +static void __exit cs553x_gpio_exit(void)
> +{
> + platform_driver_unregister(&cs553x_gpio_driver);
> +}
> +
> +module_init(cs553x_gpio_init);
> +module_exit(cs553x_gpio_exit);
> +
> +
> +MODULE_AUTHOR("Alessandro Zummo <[email protected]>");
> +MODULE_DESCRIPTION("AMD CS5535/CS5536 GPIO driver");
> +MODULE_LICENSE("GPL");
> +MODULE_VERSION(DRIVER_VERSION);
> +MODULE_ALIAS("platform:cs553x-gpio");
>
> --
>
> Best regards,
>
> Alessandro Zummo,
> Tower Technologies - Torino, Italy
>
> http://www.towertech.it
>
>
On Fri, 26 Dec 2008 10:24:07 -0800
David Brownell <[email protected]> wrote:
> > It uses the gpio framework and the gpio api as defined in
> > arch/x86/kernel/geode_32.c
>
> Eventually I'd hope to see those geode_32.c calls just vanish.
>
> In fact, the "normal" way to package these GPIOs would be to
> always provide them through the standard API, in arch/... code,
> with no Kconfig option. Any reason you shouldn't do that in
> this patch?
I didn't want to mess with something I did not wrote. Maybe a two steps
approach can convince people to move on ;) (you remember what happened
with people holding their old rtc code tight :) )
> > +comment "Other GPIO expanders:"
>
> This counts as "memory mapped" I'd say. Doesn't need
> a new category, even if this does need to live outside
> the relevant arch/... files.
ok
>
> > +
> > +config GPIO_CS553X
> > + tristate "AMD CS5535/CS5536 Geode Companion Devices"
> > + depends on MGEODE_LX && !CS5535_GPIO
>
> What's this CS5535_GPIO stuff? And why should it affect
> whether this can be configured? (It's not in mainline...)
drivers/char/cs5535_gpio.c (mainline)
> > +struct cs553x_gpio_platform_data {
> > +
> > + unsigned gpio_base; /* number of the first GPIO */
> > +
> > + resource_size_t io_base;
>
> Platform devices should use platform_get_resource() and
> friends instead of passing resources through platform data.
ack.
>
> ... other than those points, this seems like a simple and
> straightforward GPIO driver. Typical of what arch/* holds
> in such cases. ;)
:)
will change it a little bit and resubmit
--
Best regards,
Alessandro Zummo,
Tower Technologies - Torino, Italy
http://www.towertech.it
On Friday 26 December 2008, Alessandro Zummo wrote:
> > > ?It uses the gpio framework and the gpio api as defined in
> > > ?arch/x86/kernel/geode_32.c
> >
> > Eventually I'd hope to see those geode_32.c calls just vanish.
> >
> > In fact, the "normal" way to package these GPIOs would be to
> > always provide them through the standard API, in arch/... code,
> > with no Kconfig option. ?Any reason you shouldn't do that in
> > this patch?
>
> ?I didn't want to mess with something I did not wrote.
Such messing is part'n'parcel of updating the kernel though. :)
> ?Maybe a two steps
> ?approach can convince people to move on ;) (you remember what happened
> ?with people holding their old rtc code tight :) )
The RTC case was different ... driver structure needed to
either use the new framework, or the old. Except for that
drivers/rtc/rtc-ppc.c migration aid, letting some of them
coexist until the older ppc_md stuff gets updated. That is,
switching from one full-fledged driver to another one needs
some kind of scheduled flag day. (Plus completing kernel
infrastructure, of which NTP support is probably the most
significant missing part right now.)
If you add this to the geode_32.c file it'll just provide a
standard interfaces, which can coexist with the older stuff
until it's removed. Several of the ARM platforms have
exactly that kind of coexistence going on. (Example, OMAP;
which in 2.6.29 will stops using its legacy GPIO calls.)
- Dave
On Fri, 26 Dec 2008 11:20:28 -0800
David Brownell <[email protected]> wrote:
> some kind of scheduled flag day. (Plus completing kernel
> infrastructure, of which NTP support is probably the most
> significant missing part right now.)
almost completed it, need a few tests. suddenly all the people that
were interested disappeared :)
> If you add this to the geode_32.c file it'll just provide a
> standard interfaces, which can coexist with the older stuff
> until it's removed. Several of the ARM platforms have
> exactly that kind of coexistence going on. (Example, OMAP;
> which in 2.6.29 will stops using its legacy GPIO calls.)
ok, I'll seek some advice from the geode people.
--
Best regards,
Alessandro Zummo,
Tower Technologies - Torino, Italy
http://www.towertech.it