On 2016-12-08 18:47, Linus Walleij wrote:
> The compatible string for an EEPROM in the device tree is currently
> completely ignored by the kernel, simply stated it will not make the
> corresponding AT24 EEPROM driver probe properly. It is instead still
> relying on the DT node name to be set to one of the I2C device IDs
> which works due to a side effect in the I2C DT parsing code.
>
> Fix this up by making the DT probe mechanism a bit more elaborate:
> actually match on the compatible strings defined in the device
> tree bindings in Documentation/devicetree/bindings/eeprom/eeprom.txt:
> map these to the corresponding I2C IDs by name and look up the
> magic flags from the I2C ID before proceeding, also make the DT
> compatible string take precedence.
>
> Keep the second DT parsing callback that sets up per-chip flags as
> this needs to happen after mangling the magic flags passed from the
> I2C ID table.
>
> All vendor compatible strings listed in the binding document are
> added to the driver.
>
> After this it is possible to name the device tree node for the EEPROM
> whatever you actually like to call it, and the probing will be done
> from the compatible string.
>
> Before this patch, the following device tree node does not probe,
> which might be considered a bug:
>
> eeprom@52 {
> compatible = "atmel,at24c128";
The way I read it, that should be "atmel,24c128", i.e. w/o the "at" prefix.
> reg = <0x52>;
> pagesize = <64>;
> };
>
> After this patch, the driver probes fine from this node.
The bindings says:
Required properties:
- compatible should be "<manufacturer>,<type>", like these:
and then lists the compatibles you have added to the match table (but you
have this extra "at" prefix for the atmel manufacturer).
The way I read the above, you are free to use any manufacturer and still
have it work, and indeed, I have success with this node:
eeprom@50 {
compatible = "nxp,24c02";
reg = <0x50>;
pagesize = <16>;
};
I fear that your patch will regress this matching on the wildcard
manufacturer. I haven't tested that though, but there are enough
question marks anyway...
Cheers,
Peter
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
