On Mon, May 18, 2020 at 1:55 PM Sudeep Holla [off-list ref] wrote:

On Mon, May 18, 2020 at 11:30:21AM +0200, Arnd Bergmann wrote:

quoted

On Mon, May 18, 2020 at 11:12 AM Sudeep Holla [off-list ref] wrote:

quoted

+static ssize_t
+jep106_cont_bank_code_show(struct device *dev, struct device_attribute *attr,
+                          char *buf)
+{
+       return sprintf(buf, "0x%02x\n", JEP106_BANK_CONT_CODE(soc_id_version));
+}
+
+static DEVICE_ATTR_RO(jep106_cont_bank_code);
+
+static ssize_t
+jep106_identification_code_show(struct device *dev,
+                               struct device_attribute *attr, char *buf)
+{
+       return sprintf(buf, "0x%02x\n", JEP106_ID_CODE(soc_id_version));
+}

I think we should try hard to avoid nonstandard attributes for the soc device.

I agree with that in general but this is bit different for below mentioned
reason.

quoted

Did you run into a problem with finding one of the existing attributes
that can be used to hold the fields?

Not really! The 2 JEP106 codes can be used to derive the manufacturer which
could match one of the existing attributes. However doing so might require
importing the huge JEP106 list as it needs to be maintained and updated
in the kernel. Also that approach will have the compatibility issue and
that is the reason for introducing these attributes representing raw
values for userspace.

I was thinking they codes could just be part of the normal strings rather
than get translated. Can you give an example what they would look like
with your current code?

If  you think they should be standard attributes, how about adding them
to the default list, and hardcoding them in the other soc device drivers
based on the information we have available there?

      Arnd

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel