
On 07/21/2016 10:28 AM, Stefano Babic wrote:
Hi Hannes,
Hi Stefano,
On 21/07/2016 08:10, Hannes Schmelzer wrote:
Just looked around a bit more about this. Root cause for failing this patch is, that many boards do not use the 'IOMUX_PADS' macro, instead they just directly use the definition out of "mx6dl_pins.h" for example.
Both are allowed. IOMUX_PADS *must* be used in case the board supports multiple variant of the processor (DL, Quad,..). If the board has just one variant, the MX6 defines from the corresponding header can be used.
So we get in trouble there if we change the MX6_PAD_DECLARE macro for having MX6DL pads instead MX6 pads.
I am not getting where is the trouble, because there are already a lot of boards using it. Let's see....
At one point of view it would make sense to me changing all boards to use the IOMUX_PADS macro for accessing pads register, because afterwards the real accessed register would be fully in dependence of CONFIG_MX6nn. On the other hand i cannot fully predict every case could happen if we simply change that with search/replace.
So it would be OK for me to drop this patch and i will use on my board:
MX6DL_PAD_ENET_TX_EN__I2C4_SCL MX6DL_PAD_ENET_TXD1__I2C4_SDA
Now I get the point - and yes, there is an exception for I2C in the pinmux. This was discussed at the beginning when IOMUX_PADS was introduced and how to support the different layout of the SOC variants.
We agreed to tread differently I2C. This means that a i2c_pads_info structure must be set for each variant of the SOC that board supports. With help of the is_cpu_type() macro (or one of this family), the correct structure is selected and the pinmux can be set.
The right way to do is:
static struct i2c_pads_info i2c_pad = { .scl = { .i2c_mode = MX6DL_PAD_ENET_TX_EN__I2C4_SCL | <pull up>, .gpio_mode = MX6DL_PAD_ENET_TX_EN__GPIO1_IO28 | <..>, .gp = IMX_GPIO_NR(1, 28) }, .sda = { .i2c_mode = MX6DL_PAD_ENET_TXD1__I2C4_SDA | <pull >, .gpio_mode = MX6DL_PAD_ENET_TXD1__GPIO1_IO29 | <pull>, .gp = IMX_GPIO_NR(1, 29) } };
and then you call setup_i2c() with the structure.
Yeah! Now i understand the thinkings behind/around that. Many thanks for this, i will implement this for my board.
Best regards, Stefano Babic
cheers, Hannes