On Tuesday, November 04, 2014 03:04:54 PM Grant Likely wrote:

On Tue, Nov 4, 2014 at 2:38 PM, Mika Westerberg
[off-list ref] wrote:

quoted

On Tue, Nov 04, 2014 at 02:18:26PM +0000, Grant Likely wrote:

quoted

quoted

-   strncpy(chip->name, np->name, sizeof(chip->name));
+   strncpy(chip->name, "at25", sizeof(chip->name));

This line changes behaviour of the driver. It's possibly not a problem,
but it should be commented on and whether any due diligance has been
done to make sure it there isn't anything that depends on it.

I found only one user for "atmel,at25" in any of the DT sample files.

arch/arm/boot/dts/phy3250.dts:

        eeprom: at25@0 {
                ...
                at25,byte-len = <0x8000>;
                at25,addr-mode = <2>;
                at25,page-size = <64>;

                compatible = "atmel,at25";
                reg = <0>;
                spi-max-frequency = <5000000>;
        };

I think np->name is "at25" in this case? The binding file
Documentation/devicetree/bindings/misc/at25.txt also has the same name.

Are you OK, if we add something like below to the changelog?

 In addition we hard-code the name of the chip to be "at25" for the
 reason that there is no common mechanism to fetch name of the firmware
 node. The only existing user (arch/arm/boot/dts/phy3250.dts) uses the
 same name so it should continue to work.

Yes. If somebody complains, then we can reinstate the previous
behaviour, but assume it isn't necessary for now.

OK, done in my tree.  I've added the ACK too, but that still is in my
bleeding-edge branch for now until it gets build tested and I can
pull http://marc.info/?l=linux-kernel&m=141505792126265&w=4 from
somewhere. :-)

Rafael