Questions about OTP driver

[Stefan Wahren]

Hi Christian,

Am 28.10.2014 um 12:07 schrieb Christian Riesch:
> Hi Stefan,
> I am not an expert in the mtd field, but in the last years I added OTP
> support for the drivers/mtd/chips/cfi_cmdset_0002.c driver and did a
> little cleanup.
>
> On Sun, Oct 26, 2014 at 9:01 PM, Stefan Wahren <stefan.wahren at i2se.com> wrote:
>> Hi,
>>
>> currently i try to port a driver for One Time Programmable memory pages on
>> Freescale MX23/MX28 (ARM9) SoC.
>> This driver gives readonly access to these memory pages via Sysfs.
>>
>> As i created a patch RFC [1], Arnd Bergmann suggested me to move the driver to
>> MTD.
>>
>> Now i have some questions:
>>
>> 1. What exact directory is suggested for OTP on SoC driver?
> drivers/mtd/devices?
>
>> 2. What is a good example implementaton of an OTP driver?
> For drivers that support OTP see git grep _read_user_prot_reg. It's
> the cfi_cmdset_0001.c (both from drivers/mtd/chips), the
> cfi_cmdset_0002.c, the drivers/mtd/devices/mtd_dataflash.c, and the
> drivers/mtd/onenand/onenand_base.c. All these devices are memory chips
> that have a little OTP memory in addition to the regular flash memory.
> I am not aware of OTP drivers for SoCs.
>
> The most simple implementation of OTP is probably the one in
> drivers/mtd/devices/mtd_dataflash.c (see the dataflash_get_otp_info
> and dataflash_read_{user,fact}_otp functions). The others are more
> complex since they support several flash devices concatenated, which
> means that there may be several OTP memory blocks. The driver must
> then walk through these blocks.
>
> Regarding your question on the arm kernel mailing list [1]:
>> Does MTD drivers have a readonly text (non binary) user interface?
> No, you offer a /dev/mtdX interface, with an ioctl you switch it to
> OTP operation (see flash_otp_dump from mtd-utils[2]). Then the OTP
> memory is offered in binary and can be read using read() etc.
>
> Regards, Christian
>
> [1] http://marc.info/?l=linux-arm-kernel&m=141381928515849&w=2
> [2] http://git.infradead.org/mtd-utils.git/blob/HEAD:/flash_otp_dump.c

thanks a lot for your explanations, but it confirms me that's not the
right way to merge all memory blocks in a single device.

I think the term OTP is misleading and i should better talk about fuses.

BR Stefan