Dear Dong Aisheng,

On Wed, Mar 14, 2012 at 08:09:22AM +0100, Marek Vasut wrote:

quoted

Dear Dong Aisheng,

quoted

On Wed, Mar 14, 2012 at 01:58:25PM +0800, Marek Vasut wrote:

quoted

Dear Dong Aisheng,

quoted

Signed-off-by: Dong Aisheng <redacted>

........

quoted

quoted

+static const struct of_device_id mxs_mmc_dt_ids[] = {
+	{ .compatible = "fsl,imx23-mmc", .data = NULL, },
+	{ .compatible = "fsl,imx28-mmc", .data = NULL, },

Do you really need two distinct ones here?

Hmm, my original purpose is to put soc difference data in .data
to remove cpu_is_* function calls in the driver later.
Do you think if any issue?

Well, what's the difference between the interfaces on mx233 and mx28? Is
it something that can't be encoded otherwise? I think they're not so
different.

Not much difference except the one register offset and ip version.
See:
#define SSP_VERSION_LATEST      4
#define ssp_is_old()            (host->version < SSP_VERSION_LATEST)
..
#define HW_SSP_VERSION (cpu_is_mx23() ? 0x110 : 0x130)
The ip version can be handled in driver, but for offset...
it depends on cpu_is_* macro.
Putting the HW_SSP_VERSION offset difference in .data can eliminate the
need of cpu_is_*.

Despite of that, since they're two devices,
i guess it's ok to put two compatible string there, right?
Or you thought just put one as below?
{ .compatible = "fsl,mxs-mmc", .data = NULL, },

No, I understand now /wrt the register layout.

Best regards,
Marek Vasut