Re: [RFC 01/11] i2c: add quirk structure to describe adapter flaws

Hi,

This can describe the behavior of our current upstream driver[1], which
only support combine write-then-read.

After checking with Xudong & HW guys, it seems our HW can do more. 
On MT8135, it can support at most 2 messages, no matter read or write,
with the limitation that the length of the second message must <=
31bytes.

So this RFC is enough for our driver, but it would be better if we could
also support other case.

Hmm, I think we can convert max_comb_{read|write}_len to
max_comb_{1st|2nd}_msg_len or similar.

I'll check but it will probably not before next week.

Thanks for the input!