Re: [PATCH v2 09/10] MIPS: i8042: Probe this device only if it exists

[Dmitry Torokhov]

On Wed, Jun 28, 2017 at 05:47:02PM +0200, Aleksandar Markovic wrote:
> From: Miodrag Dinic <miodrag.dinic@xxxxxxxxxx>
> 
> ARCH_MIGHT_HAVE_PC_SERIO is selected by default for MIPS platforms.
> As a consequence SERIO_I8042 would be automatically selected for any
> MIPS board which wants to enable input support like keyboard
> (INPUT_KEYBOARD) regardless of i8042 controller existence.
> 
> The dependency is as follows :
> 
> config ARCH_MIGHT_HAVE_PC_SERIO [=y]
>     Defined at drivers/input/serio/Kconfig:19
>     Depends on: !UML
>     Selected by: MIPS [=y]
> 
> config SERIO
>     Defined at drivers/input/serio/Kconfig:4
>     default y
>     Depends on: !UML
>     Selected by: KEYBOARD_ATKBD [=y] && !UML && INPUT [=y] &&
>                  INPUT_KEYBOARD [=y]
> 
> config SERIO_I8042
>     Defined at drivers/input/serio/Kconfig:28
>     tristate "i8042 PC Keyboard controller"
>     default y
>     Depends on: !UML && SERIO [=y] && ARCH_MIGHT_HAVE_PC_SERIO [=y]
>     Selected by: KEYBOARD_ATKBD [=y] && !UML && INPUT [=y] &&
>                  INPUT_KEYBOARD [=y] && ARCH_MIGHT_HAVE_PC_SERIO [=y]
> 
> If this driver probes the I8042_DATA_REG not knowing if the device
> exists it can cause a kernel to crash. Using check_legacy_ioport()
> interface we can selectively enable this driver only for the MIPS
> boards which actually have the i8042 controller.
> 
> New "Ranchu" virtual platform does not support i8042 controller
> so it's added to the blacklist match table.
> 
> Each MIPS machine should update this table with it's compatible strings
> if it does not support i8042 controller.
> 
> In order to utilize this mechanism, each MIPS machine that do not
> have i8042 controller should update the blacklist table with its
> compatible strings.
> 
> Signed-off-by: Miodrag Dinic <miodrag.dinic@xxxxxxxxxx>
> Signed-off-by: Goran Ferenc <goran.ferenc@xxxxxxxxxx>
> Signed-off-by: Aleksandar Markovic <aleksandar.markovic@xxxxxxxxxx>
> ---
>  arch/mips/kernel/setup.c       | 16 ++++++++++++++++
>  drivers/input/serio/i8042-io.h |  2 +-
>  2 files changed, 17 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/mips/kernel/setup.c b/arch/mips/kernel/setup.c
> index c22cde8..c3e0d2b 100644
> --- a/arch/mips/kernel/setup.c
> +++ b/arch/mips/kernel/setup.c
> @@ -79,6 +79,15 @@ const unsigned long mips_io_port_base = -1;
>  EXPORT_SYMBOL(mips_io_port_base);
>  
>  /*
> + * Here we blacklist all MIPS boards which do not have i8042 controller
> + */
> +static const struct of_device_id i8042_blacklist_of_match[] = {
> +	{ .compatible = "mti,ranchu", },
> +	{},
> +};
> +#define I8042_DATA_REG	0x60
> +
> +/*
>   * Check for existence of legacy devices
>   *
>   * Some drivers may try to probe some I/O ports which can lead to
> @@ -90,9 +99,16 @@ EXPORT_SYMBOL(mips_io_port_base);
>   */
>  int check_legacy_ioport(unsigned long base_port)
>  {
> +	struct device_node *np;
>  	int ret = 0;
>  
>  	switch (base_port) {
> +	case I8042_DATA_REG:
> +		np = of_find_matching_node(NULL, i8042_blacklist_of_match);
> +		if (np)
> +			ret = -ENODEV;
> +		of_node_put(np);
> +		break;

Can you simply mark 8042 region as busy when you are setting up the
board, so when i8042 tries to requets it it will fail?

>  	default:
>  		/* We will assume that the I/O device port exists if
>  		 * not explicitly added to the blacklist match table
> diff --git a/drivers/input/serio/i8042-io.h b/drivers/input/serio/i8042-io.h
> index 34da81c..ec5fe9e 100644
> --- a/drivers/input/serio/i8042-io.h
> +++ b/drivers/input/serio/i8042-io.h
> @@ -72,7 +72,7 @@ static inline int i8042_platform_init(void)
>   * On some platforms touching the i8042 data register region can do really
>   * bad things. Because of this the region is always reserved on such boxes.
>   */
> -#if defined(CONFIG_PPC)
> +#if defined(CONFIG_PPC) || defined(CONFIG_MIPS)
>  	if (check_legacy_ioport(I8042_DATA_REG))
>  		return -ENODEV;
>  #endif
> -- 
> 2.7.4
> 

Thanks.

-- 
Dmitry