[PATCH v2] spi/imx: Fix spi-imx when the hardware SPI chipselects are used

[Grant Likely]

On Thu, Sep 15, 2011 at 03:28:57PM -0300, Fabio Estevam wrote:
> commit 22a85e4cd51 (spi/imx: add device tree probe support) broke spi-imx usage
> when the SPI chipselect is the one internal to the controller.
> 
> On a mx31pdk board the following error is seen:
> 
> Registering mxc_nand as whole device
> ------------[ cut here ]------------
> WARNING: at drivers/gpio/gpiolib.c:101 gpio_ensure_requested+0x4c/0xf4()
> autorequest GPIO-0
> Modules linked in:
> [<c0014410>] (unwind_backtrace+0x0/0xf4) from [<c0025754>] (warn_slowpath_common+0x4c/0x64)
> [<c0025754>] (warn_slowpath_common+0x4c/0x64) from [<c0025800>] (warn_slowpath_fmt+0x30/0x40)
> [<c0025800>] (warn_slowpath_fmt+0x30/0x40) from [<c0198688>] (gpio_ensure_requested+0x4c/0xf4)
> [<c0198688>] (gpio_ensure_requested+0x4c/0xf4) from [<c01988c8>] (gpio_direction_output+0xa0/0x138)
> [<c01988c8>] (gpio_direction_output+0xa0/0x138) from [<c01ed198>] (spi_imx_setup+0x38/0x4c)
> [<c01ed198>] (spi_imx_setup+0x38/0x4c) from [<c01eb5d0>] (spi_setup+0x38/0x50)
> [<c01eb5d0>] (spi_setup+0x38/0x50) from [<c01eb85c>] (spi_add_device+0x94/0x124)
> [<c01eb85c>] (spi_add_device+0x94/0x124) from [<c01eb960>] (spi_new_device+0x74/0xac)
> [<c01eb960>] (spi_new_device+0x74/0xac) from [<c01eb9b8>] (spi_match_master_to_boardinfo+0x20/0x40)
> [<c01eb9b8>] (spi_match_master_to_boardinfo+0x20/0x40) from [<c01eba88>] (spi_register_master+0xb0/0x104)
> [<c01eba88>] (spi_register_master+0xb0/0x104) from [<c01ec0b4>] (spi_bitbang_start+0x104/0x17c)
> [<c01ec0b4>] (spi_bitbang_start+0x104/0x17c) from [<c02c2c4c>] (spi_imx_probe+0x2fc/0x404)
> [<c02c2c4c>] (spi_imx_probe+0x2fc/0x404) from [<c01c2498>] (platform_drv_probe+0x18/0x1c)
> [<c01c2498>] (platform_drv_probe+0x18/0x1c) from [<c01c1058>] (driver_probe_device+0x78/0x174)
> [<c01c1058>] (driver_probe_device+0x78/0x174) from [<c01c11e0>] (__driver_attach+0x8c/0x90)
> [<c01c11e0>] (__driver_attach+0x8c/0x90) from [<c01c0860>] (bus_for_each_dev+0x60/0x8c)
> [<c01c0860>] (bus_for_each_dev+0x60/0x8c) from [<c01c0088>] (bus_add_driver+0xa0/0x288)
> [<c01c0088>] (bus_add_driver+0xa0/0x288) from [<c01c179c>] (driver_register+0x78/0x18c)
> [<c01c179c>] (driver_register+0x78/0x18c) from [<c0008490>] (do_one_initcall+0x34/0x178)
> [<c0008490>] (do_one_initcall+0x34/0x178) from [<c03a5204>] (kernel_init+0x74/0x118)
> [<c03a5204>] (kernel_init+0x74/0x118) from [<c000f65c>] (kernel_thread_exit+0x0/0x8)
> ---[ end trace 759f924b30fd5a44 ]---
> 
> Fix this issue by using the original chip select logic and make spi-imx to work again.
> 
> Tested on a mx31pdk that uses the hardware SPI chipselect pins and also
> on a mx27pdk that uses GPIO as SPI chipselect.
> 
> Signed-off-by: Fabio Estevam <fabio.estevam at freescale.com>
> ---
> Changes since v1:
> - Fix the logic for the internal chip select case and
> keep the usage of of_get_named_gpio for getting the cs_gpio
> 
>  drivers/spi/spi-imx.c |    6 ++++--
>  1 files changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/spi/spi-imx.c b/drivers/spi/spi-imx.c
> index 8ac6542..d917fa3 100644
> --- a/drivers/spi/spi-imx.c
> +++ b/drivers/spi/spi-imx.c
> @@ -784,11 +784,13 @@ static int __devinit spi_imx_probe(struct platform_device *pdev)
>  
>  	for (i = 0; i < master->num_chipselect; i++) {
>  		int cs_gpio = of_get_named_gpio(np, "cs-gpios", i);
> -		if (cs_gpio < 0)
> +		if (cs_gpio < 0) {
>  			cs_gpio = mxc_platform_info->chipselect[i];
> +			spi_imx->chipselect[i] = cs_gpio;
> +		}
>  		if (cs_gpio < 0)
>  			continue;
> -		spi_imx->chipselect[i] = cs_gpio;
> +

It looks like the removal of this line will break DT users.  Perhaps
the spi_imx->chipselect[i] line should just be moved before the
'if (cp_gpio < 0)' line.

g.


>  		ret = gpio_request(spi_imx->chipselect[i], DRIVER_NAME);
>  		if (ret) {
>  			while (i > 0) {
> -- 
> 1.7.1
> 
>