Re: [PATCH] i2c: designware: Round down ACPI provided clk to nearest supported clk

[Phil Reid <preid@electromag.com.au>]

On 30/08/2017 05:00, Wolfram Sang wrote:
> 
>> The speed comes from an ACPI entry describing an i2c client,
>> any compliant i2c client must at least support 100KHz, right ?
> 
> Well, due to flaky board design, you may not be able to utilize the max
> speed because of interferences etc even if the devices would support it.
> Such knowledge of flaky boards could be encoded in the ACPI tables, or?
> But probably not in the client device, hmmm...
> 
>> Alternatively I could wrap the entire round-down for loop in
>> an if (acpi_speed) {} block.
> 
> I don't know enough about real-world ACPI tables to suggest a best
> practice here. I just wanted to add that busses < 100 kHz are legal from
> how I read the specs.
> 
> Oh well, Jarkko liked the patch, so let's all sleep over this patch and
> if nothing else comes up, I'll apply it tomorrow or so...
> 

My understanding is 100k is what the client must support.
But sometimes buses need to be run slower.
Particularly when using range extenders.

eg: I have an i2c bus running over a 10m cable that needs to run at about 40k
to be reliable.


-- 
Regards
Phil Reid