Re: nvmem: add driver for Microchip 24AA025E48 I2C eeprom / nodeID chip

[Bartosz Golaszewski]

2018-01-25 15:37 GMT+01:00 Felix Brack <fb@xxxxxxx>:
>
> On 24.01.2018 17:02, Bartosz Golaszewski wrote:
>> 2018-01-24 16:18 GMT+01:00 Felix Brack <fb@xxxxxxx>:
>>>
>>> On 24.01.2018 16:08, Andy Shevchenko wrote:
>>>> On Wed, Jan 24, 2018 at 4:23 PM, Felix Brack <fb@xxxxxxx> wrote:
>>>>> Hello,
>>>>>
>>>>> About three years ago I wrote a driver for Microchip's 24AA025E48 I2C
>>>>> eeprom/nodeID chip. At that time I placed the source files in
>>>>> drivers/misc/eeprom. I never posted the code.
>>>>> I now plan to rewrite the driver from scratch. Is it correct to place
>>>>> the source code in drivers/nvmem and is there a special mailing list to
>>>>> which the patch should be posted when ready?
>>>>>
>>>>> many thanks and kind regards, Felix
>>>>
>>>> Does the existing driver [1] work for you (if you add ID there)?
>>>>
>>>> [1]: at24
>>>>
>>> Yes it does. Actually the driver I wrote 3 years ago is based on the
>>> at24 driver. Lot's of code in my driver originates directly from the
>>> at24 driver.
>>>
>>> --
>>> regards Felix
>>
>> Just from looking at the doc - it seems that it's a variant of
>> 24mac402. You should be able to access the memory block by
>> instantiating an 'atmel,24c02' device. As for the EUI-48 block -
>> current at24 driver will not work as the MAC is located at a different
>> offset in this chip. We need to figure out a portable way to specify
>> the addresses of such special blocks (same with the serial number) in
>> the at24 driver.
>>
>> Anyways - don't write your own driver for that, just make sure at24 works.
>>
> If no one offends against blowing up this driver by some more code, then
> that's fine with me. What about a patch that does the following:
>
> 1. Extend the DT bindings by a new optional property block_offset
>    denoting where to start reading/writing from/to this device.
> 2. Add a member block_offset to struct at24_platform_data{} to store
>    the device read/write offset returned from the DT (similar to the
>    already existing page_size member).
> 3. Complement at24_get_pdata() to read block_offset from the DT and
>    store it.
> 4. Make at24_eeprom_write_i2c() and at24_eeprom_read_serial() (and
>    eventually more?) respect the block_offset value.
>

These routines no longer exist, at24 uses regmap now. Please see
current next or wait for 4.16-rc1.

> Initializing the new block_offset member of struct at24_platform_data{}
> to 0 should leave the code semantically unchanged.
>
> Once this patch is in place I could use the 24mac402 device type with
> block_offset set to 0xfa (and read-only set to true) to read the EUI-48
> node address from Microchip's 24AA025E48. Probably the later addition of
> a device named 24eui48 would also make sense (if not redundant ?).
>
> Any comments are very appreciated, Felix

We should probably have a default value defined in chip data too,
otherwise we'd break the existing models from atmel containing the EUI
block (at24mac402/602) the offset of which is different than the one
you're using.

Best regards,
Bartosz Golaszewski