RE: Sample DTS entries and sample user space code

RE: Sample DTS entries and sample user space code

[Lindsey Ings -X (ingsl - Yoh Services LLC at Cisco)]

Resend of message in plain text form.


I understand the concepts involved in this project but that doesn't help me actually make use of it.

Is there a sample of how to put the mux and (if feasible) I2C clients in a DTS?

Also is there a user application sample of how to access the devices downstream from the mux?  Are the mux's created as a device so that the devices downstream are accessed as non-mux'd devices?  i.e. is dev D in segment 2 (from the diagram on the wiki) accessed as so:

int fd;
fd = open( "/dev/i2c-2", O_RDWR );

#define ADDRESS 0x38 // address of dev D
ioctl( fd, I2C_SLAVE, ADDRESS )

// Now r/w dev D

// Write 0xAA to dev D
i2c_smbus_write_byte( fd, 0xAA );
// etc.

Re: Sample DTS entries and sample user space code

[Jean Delvare]

On Fri, 10 Jul 2009 17:09:47 -0400, Lindsey Ings -X (ingsl - Yoh Services LLC at Cisco) wrote:
> 
> 
> Resend of message in plain text form.
> 
> 
> I understand the concepts involved in this project but that doesn't help me actually make use of it.

For reference for other readers who have no clue what you're talking
about: Lindsey is referring to Rodolfo's I2C multiplexing support
project, described at:
http://i2c.wiki.kernel.org/index.php/I2C_bus_multiplexing

> Is there a sample of how to put the mux and (if feasible) I2C clients in a DTS?

Not yet. Rodolfo's code is not even in mainline. But once
implementation is complete, then yes it should be possible to declare
I2C multiplexers and other I2C devices in a DTS.

> Also is there a user application sample of how to access the devices downstream from the mux?

This wouldn't make sense. Multiplexing is handled in the kernel and
user-space does not even know it is happening.

> Are the mux's created as a device so that the devices downstream are accessed as non-mux'd devices? i.e. is dev D in segment 2 (from the diagram on the wiki) accessed as so:
> 
> int fd;
> fd = open( "/dev/i2c-2", O_RDWR );
> 
> #define ADDRESS 0x38 // address of dev D
> ioctl( fd, I2C_SLAVE, ADDRESS )
> 
> // Now r/w dev D
> 
> // Write 0xAA to dev D
> i2c_smbus_write_byte( fd, 0xAA );
> // etc.

Yes, this is how the code would look like.

Note that I am still waiting for Rodolfo to update his patch set so
that it applies again on top of Linus' latest kernel.

-- 
Jean Delvare

RE: Sample DTS entries and sample user space code

[Lindsey Ings -X (ingsl - Yoh Services LLC at Cisco)]

Thanks for the reply.

Sorry for not being clear that it was for the multiplexing of I2C.  My
link into the mailing list implied that it was solely for the later and
not for all I2C in general.

 
Regards,
 
Lindsey
 
-----Original Message-----
From: Jean Delvare [mailto:khali-PUYAD+kWke1g9hUCZPvPmw@public.gmane.org] 
Sent: Tuesday, July 14, 2009 3:57 AM
To: Lindsey Ings -X (ingsl - Yoh Services LLC at Cisco)
Cc: linux-i2c-u79uwXL29TY76Z2rM5mHXA@public.gmane.org; giometti-AVVDYK/kqiJWk0Htik3J/w@public.gmane.org
Subject: Re: Sample DTS entries and sample user space code

On Fri, 10 Jul 2009 17:09:47 -0400, Lindsey Ings -X (ingsl - Yoh
Services LLC at Cisco) wrote:
> 
> 
> Resend of message in plain text form.
> 
> 
> I understand the concepts involved in this project but that doesn't
help me actually make use of it.

For reference for other readers who have no clue what you're talking
about: Lindsey is referring to Rodolfo's I2C multiplexing support
project, described at:
http://i2c.wiki.kernel.org/index.php/I2C_bus_multiplexing

> Is there a sample of how to put the mux and (if feasible) I2C clients
in a DTS?

Not yet. Rodolfo's code is not even in mainline. But once
implementation is complete, then yes it should be possible to declare
I2C multiplexers and other I2C devices in a DTS.

> Also is there a user application sample of how to access the devices
downstream from the mux?

This wouldn't make sense. Multiplexing is handled in the kernel and
user-space does not even know it is happening.

> Are the mux's created as a device so that the devices downstream are
accessed as non-mux'd devices? i.e. is dev D in segment 2 (from the
diagram on the wiki) accessed as so:
> 
> int fd;
> fd = open( "/dev/i2c-2", O_RDWR );
> 
> #define ADDRESS 0x38 // address of dev D
> ioctl( fd, I2C_SLAVE, ADDRESS )
> 
> // Now r/w dev D
> 
> // Write 0xAA to dev D
> i2c_smbus_write_byte( fd, 0xAA );
> // etc.

Yes, this is how the code would look like.

Note that I am still waiting for Rodolfo to update his patch set so
that it applies again on top of Linus' latest kernel.

-- 
Jean Delvare

RE: Sample DTS entries and sample user space code

[Lindsey Ings -X (ingsl - Yoh Services LLC at Cisco)]

Since I can't use the DTS I'm now trying to instantiate the mux through
other means.  I put the code below in
".../arch/powerpc/platforms/44x/myBoard.c".  Is this the intended
instantiation?  I have 2 i2c busses on my SOC and I have 2 muxes (with 4
ports each) on the second SOC bus.  Therefore, I'd want the first mux to
have busses 2,3,4,5 and the second mux to have 6,7,8,9:

static struct pca954x_platform_mode pca954x_modes_mux0[] = {
	    {
	    	.adap_id        = 2,
	    },
	    {
	    	.adap_id        = 3,
	    },
	    {
	    	.adap_id        = 4,
	    },
	    {
	    	.adap_id        = 5,
	    },
};

static struct pca954x_platform_mode pca954x_modes_mux1[] = {
	    {
	    	.adap_id        = 6,
	    },
	    {
	    	.adap_id        = 7,
	    },
	    {
	    	.adap_id        = 8,
	    },
	    {
	    	.adap_id        = 9,
	    },
};

static struct pca954x_platform_data pca9544_data_mux0 = {
    	.modes          = pca954x_modes_mux0,
        .num_modes      = ARRAY_SIZE(pca954x_modes_mux0),
};

static struct pca954x_platform_data pca9544_data_mux1 = {
    	.modes          = pca954x_modes_mux1,
        .num_modes      = ARRAY_SIZE(pca954x_modes_mux1),
};

static struct i2c_board_info __initdata i2c_bus1_devices[] = {
    {
        /* I2C mux */
        I2C_BOARD_INFO("pca9544", 0x70),
        .platform_data = &pca9544_data_mux0,
    },
    {
        /* I2C mux */
        I2C_BOARD_INFO("pca9544", 0x71),
        .platform_data = &pca9544_data_mux1,
    },
};

static int __init board_i2c_init(void)
{
    /* Register board i2c devices */   
    i2c_register_board_info(0, i2c_bus1_devices,
                ARRAY_SIZE(i2c_bus1_devices));

    return 0;
}
arch_initcall(board_i2c_init);

 
Regards,
 
Lindsey
 
-----Original Message-----
From: Jean Delvare [mailto:khali-PUYAD+kWke1g9hUCZPvPmw@public.gmane.org] 
Sent: Tuesday, July 14, 2009 3:57 AM
To: Lindsey Ings -X (ingsl - Yoh Services LLC at Cisco)
Cc: linux-i2c-u79uwXL29TY76Z2rM5mHXA@public.gmane.org; giometti-AVVDYK/kqiJWk0Htik3J/w@public.gmane.org
Subject: Re: Sample DTS entries and sample user space code

On Fri, 10 Jul 2009 17:09:47 -0400, Lindsey Ings -X (ingsl - Yoh
Services LLC at Cisco) wrote:
> 
> 
> Resend of message in plain text form.
> 
> 
> I understand the concepts involved in this project but that doesn't
help me actually make use of it.

For reference for other readers who have no clue what you're talking
about: Lindsey is referring to Rodolfo's I2C multiplexing support
project, described at:
http://i2c.wiki.kernel.org/index.php/I2C_bus_multiplexing

> Is there a sample of how to put the mux and (if feasible) I2C clients
in a DTS?

Not yet. Rodolfo's code is not even in mainline. But once
implementation is complete, then yes it should be possible to declare
I2C multiplexers and other I2C devices in a DTS.

> Also is there a user application sample of how to access the devices
downstream from the mux?

This wouldn't make sense. Multiplexing is handled in the kernel and
user-space does not even know it is happening.

> Are the mux's created as a device so that the devices downstream are
accessed as non-mux'd devices? i.e. is dev D in segment 2 (from the
diagram on the wiki) accessed as so:
> 
> int fd;
> fd = open( "/dev/i2c-2", O_RDWR );
> 
> #define ADDRESS 0x38 // address of dev D
> ioctl( fd, I2C_SLAVE, ADDRESS )
> 
> // Now r/w dev D
> 
> // Write 0xAA to dev D
> i2c_smbus_write_byte( fd, 0xAA );
> // etc.

Yes, this is how the code would look like.

Note that I am still waiting for Rodolfo to update his patch set so
that it applies again on top of Linus' latest kernel.

-- 
Jean Delvare

Re: Sample DTS entries and sample user space code

[Jean Delvare]

Hi Lindsey,

On Tue, 28 Jul 2009 16:38:24 -0400, Lindsey Ings -X (ingsl - Yoh Services LLC at Cisco) wrote:
> 
> Since I can't use the DTS I'm now trying to instantiate the mux through
> other means.  I put the code below in
> ".../arch/powerpc/platforms/44x/myBoard.c".  Is this the intended
> instantiation?  I have 2 i2c busses on my SOC and I have 2 muxes (with 4
> ports each) on the second SOC bus.  Therefore, I'd want the first mux to
> have busses 2,3,4,5 and the second mux to have 6,7,8,9:
> 
> static struct pca954x_platform_mode pca954x_modes_mux0[] = {
> 	    {
> 	    	.adap_id        = 2,
> 	    },
> 	    {
> 	    	.adap_id        = 3,
> 	    },
> 	    {
> 	    	.adap_id        = 4,
> 	    },
> 	    {
> 	    	.adap_id        = 5,
> 	    },
> };
> 
> static struct pca954x_platform_mode pca954x_modes_mux1[] = {
> 	    {
> 	    	.adap_id        = 6,
> 	    },
> 	    {
> 	    	.adap_id        = 7,
> 	    },
> 	    {
> 	    	.adap_id        = 8,
> 	    },
> 	    {
> 	    	.adap_id        = 9,
> 	    },
> };
> 
> static struct pca954x_platform_data pca9544_data_mux0 = {
>     	.modes          = pca954x_modes_mux0,
>         .num_modes      = ARRAY_SIZE(pca954x_modes_mux0),
> };
> 
> static struct pca954x_platform_data pca9544_data_mux1 = {
>     	.modes          = pca954x_modes_mux1,
>         .num_modes      = ARRAY_SIZE(pca954x_modes_mux1),
> };
> 
> static struct i2c_board_info __initdata i2c_bus1_devices[] = {
>     {
>         /* I2C mux */
>         I2C_BOARD_INFO("pca9544", 0x70),
>         .platform_data = &pca9544_data_mux0,
>     },
>     {
>         /* I2C mux */
>         I2C_BOARD_INFO("pca9544", 0x71),
>         .platform_data = &pca9544_data_mux1,
>     },
> };

I don't know the pca954x driver (not yet in mainline), Rodolfo would be
a better reviewer, but the above at least looks reasonable to me. I am
a little surprised that you are able to force the adapters IDs for the
segments behind the muxes though, if you do not declare any device on
them, i2c-core will not reserve the IDs for static numbering.

> static int __init board_i2c_init(void)
> {
>     /* Register board i2c devices */   
>     i2c_register_board_info(0, i2c_bus1_devices,

This, OTOH, looks wrong. If the muxes are on your second SOC I2C bus,
the bus number would be 1, not 0.

>                 ARRAY_SIZE(i2c_bus1_devices));
> 
>     return 0;
> }
> arch_initcall(board_i2c_init);

-- 
Jean Delvare