On 12.4.2016 03:25, Peng Fan wrote:
Hi Michal,
On Mon, Apr 11, 2016 at 02:40:06PM +0200, Michal Simek wrote:
On 11.4.2016 14:09, Michal Simek wrote:
On 11.4.2016 07:47, Peng Fan wrote:
On Sat, Apr 09, 2016 at 12:33:34PM -0600, Simon Glass wrote:
On 18 March 2016 at 03:54, Peng Fan van.freenix@gmail.com wrote:
Introduce a new driver that supports driver model for pca953x. The pca953x chips are used as I2C I/O expanders. This driver is designed to support the following chips: " 4 bits: pca9536, pca9537 8 bits: max7310, max7315, pca6107, pca9534, pca9538, pca9554, pca9556, pca9557, pca9574, tca6408, xra1202 16 bits: max7312, max7313, pca9535, pca9539, pca9555, pca9575, tca6416 24 bits: tca6424 40 bits: pca9505, pca9698 " But for now this driver only supports max 24 bits and pca953x compatible chips. pca957x compatible chips are not supported now. These can be addressed when we need to add such support for the different chips. This driver has been tested on i.MX6 SoloX Sabreauto board with max7310 i2c expander using gpio command as following:
=>gpio status -a Bank gpio@48: gpio@480: input: 1 [ ] => gpio clear gpio@480 gpio: pin gpio@480 (gpio 224) value is 0 => gpio status -a Bank gpio@48: gpio@480: output: 0 [ ]
Don't you think 480 is confusing? Perhaps you should have gpio@48_ as the bank name? Also I think you should support a gpio-bank-name property in the node, to allow a sensible name to be provided.
480 is added by gpio uclass driver I think. The dts is copied from Linux side. I'd not change the dts, will try to see how to introudce a sensible name here.
What's the binding you are using?
The part of this patch should be DT binding.
This is my node. tca6416_u61: gpio@21 { compatible = "ti,tca6416"; reg = <0x21>; gpio-controller; #gpio-cells = <2>; };
Ok. I found where the problem is.
i2c bus has #address-cells = <1>; #size-cells = <0>;
And then OF_CHECK_COUNTS thinks that it is incorrect setting for bus but it should be valid for i2c where only address without size is used. When I apply patch below driver is probed
I did not met this issue.
That's interesting. Can you please add
Can you please apply this and look at na and ns values? If you are on imx6q then you should reach the same problem as I if this code is called that's why I would like to confirm this.
diff --git a/common/fdt_support.c b/common/fdt_support.c index ced119e70d9f..9c18b312d647 100644 --- a/common/fdt_support.c +++ b/common/fdt_support.c @@ -1109,6 +1109,7 @@ static u64 __of_translate_address(void *blob, int node_offset, const fdt32_t *in
/* Cound address cells & copy address locally */ bus->count_cells(blob, parent, &na, &ns); + printf("addr cells %d, size cells %d\n", na, ns); if (!OF_CHECK_COUNTS(na, ns)) { printf("%s: Bad cell count for %s\n", __FUNCTION__, fdt_get_name(blob, node_offset, NULL));
diff --git a/common/fdt_support.c b/common/fdt_support.c index ced119e70d9f..5f5b49c6210b 100644 --- a/common/fdt_support.c +++ b/common/fdt_support.c @@ -941,7 +941,7 @@ void fdt_del_node_and_alias(void *blob, const char *alias) #define OF_MAX_ADDR_CELLS 4 #define OF_BAD_ADDR FDT_ADDR_T_NONE #define OF_CHECK_COUNTS(na, ns) ((na) > 0 && (na) <= OF_MAX_ADDR_CELLS && \
(ns) > 0)
(ns) >= 0)/* Debug utility */ #ifdef DEBUG
Regarding numbers. It is just hex to int conversion + gpio number. It means gpio@20 is 0x20 which is 32. Using hex everywhere make sense and this is the fix. Maybe better to also use underscore.
diff --git a/drivers/gpio/pca953x_gpio.c b/drivers/gpio/pca953x_gpio.c index 6b67b079a17b..1eeb4c8f199a 100644 --- a/drivers/gpio/pca953x_gpio.c +++ b/drivers/gpio/pca953x_gpio.c @@ -292,7 +292,7 @@ static int pca953x_probe(struct udevice *dev) return ret; }
snprintf(name, sizeof(name), "gpio@%u", info->addr);
snprintf(name, sizeof(name), "gpio@%x_", info->addr); str = strdup(name); if (!str) return -ENOMEM;Thanks for this. Also Would you mind I add your Tested-By in V2?
I will when this is fixed and we know more about that issue above.
Thanks, Michal