On Sun, Jan 18, 2026 at 10:08:12AM +0000, Russell King (Oracle) wrote:

On Sun, Jan 18, 2026 at 10:43:12AM +0100, Jonas Jelonek wrote:

quoted

Looking at the SFP MSA [1], some sentences sound like one could assume
byte access is needed at least for SFP. In Section B4, there are statements
like:
- "The memories are organized as a series of 8-bit data words that can be
    addressed individually..."
- "...provides sequential or random access to 8 bit parameters..."
- "The protocol ... sequentially transmits one or more 8-bit bytes..."

But that may be too vague and I can't judge if that's a valid argument to not
care about word-only here.

There's a whole bunch of documents. You also need to look at SFF-8472.
This contains the following paragraph:

 To guarantee coherency of the diagnostic monitoring data, the host is
 required to retrieve any multi-byte fields from the diagnostic
 monitoring data structure (e.g. Rx Power MSB - byte 104 in A2h, Rx
 Power LSB - byte 105 in A2h) by the use of a single two-byte read
 sequence across the 2-wire interface.

Hence why we don't allow hwmon when only byte accesses are available.

I would also add, SFP vendors like to ignore the MSA and do random
things. Look at the number of quirks we have for dealing with SFPs
which break what these documents say.

      Andrew