Re: [PATCH] ASoC: pcm512x: Mend accesses to the I2S_1 and I2S_2 registers

From: Péter Ujfalusi
Date: Tue Sep 21 2021 - 00:20:50 EST


This is a multi-part message in MIME format.Hi Peter.

On 20/09/2021 22:37, Peter Rosin wrote:

> Nope, MODE1/2 are wired for I2C -> FMT is just the last I2C address
> bit. So, nothing to do with I2S. And I can't see how that would explain
> the same problem with the I2S_2 register?

true, but worth the question ;)

>>> This fix is not 100% correct. The datasheet of at least the pcm5142
>>> states that four bits (0xcc) in the I2S_1 register are "RSV"
>>> ("Reserved. Do not access.") and no hint is given as to what theHi
>>> initial values are supposed to be. So, specifying defaults for
>>> these bits is wrong. But perhaps better than a broken driver?
>>
>> The default of 0x02 (AFMT = 00b - I2S, ALEN = 10b - 24bits) is correct
>> for I2S_1 and the 0 is the default of I2S_2.
>>
>> The failure happens when updating the AFMT (bit4-5) and when regmap is
>> doing a i2c read since the default is not specified.
>>
>> It would be interesting to see what it is reading... Out of interest:
>> can you mar the I2S_1 and I2S_2 as volatile and read / print the value
>> just before the afmt and alen update?
>
> My first attempt was to mark the register volatile, and then read and
> compare if the update was needed at all. But marking volatile wasn't
> enough.

If the value is not provided in the defaults then the first read is reading out to the chip anyways.

> I also tried to set both a default and mark as volatile,

Volatile always skips the cache, default would be ignored.

> but it seems every read fails with -16 (EBUSY). I don't get why, to me
> it almost feels like a regmap issue of some sort (probably the regmap
> config is bad in some way), but I'm not fluent in regmap...

Or most likely the chip is not powered at pcm512x_set_fmt() time?

> So, since I can't read, I can't get to the initial values of the four
> reserved bits. So, I winged them as zero.

The value of the reserved bits are don't care.

Can you try the attached patch w/o without the defaults for i2s_1/2?
Not even compile tested...