Re: [Intel-wired-lan] [RFC PATCH] ethtool: add FEC bins histogramm report

[Vadim Fedorenko <vadim.fedorenko@linux.dev>]

On 30/07/2025 02:45, Jakub Kicinski wrote:
> On Tue, 29 Jul 2025 03:23:54 -0700 Vadim Fedorenko wrote:
>> diff --git a/Documentation/netlink/specs/ethtool.yaml b/Documentation/netlink/specs/ethtool.yaml
>> index 1063d5d32fea2..3781ced722fee 100644
>> --- a/Documentation/netlink/specs/ethtool.yaml
>> +++ b/Documentation/netlink/specs/ethtool.yaml
>> @@ -1239,6 +1239,30 @@ attribute-sets:
>>           name: corr-bits
>>           type: binary
>>           sub-type: u64
>> +      -
>> +        name: hist
>> +        type: nest
>> +        multi-attr: True
>> +        nested-attributes: fec-hist
>> +      -
>> +        name: fec-hist-bin-low
>> +        type: s32
>> +      -
>> +        name: fec-hist-bin-high
>> +        type: s32
>> +      -
>> +        name: fec-hist-bin-val
>> +        type: u64
>> +  -
>> +    name: fec-hist
>> +    subset-of: fec-stat
> 
> no need to make this a subset, better to make it its own attr set

like a set for general histogram? or still fec-specific?

> 
>> +    attributes:
>> +      -
>> +        name: fec-hist-bin-low
>> +      -
>> +        name: fec-hist-bin-high
>> +      -
>> +        name: fec-hist-bin-val
>>     -
>>       name: fec
>>       attr-cnt-name: __ethtool-a-fec-cnt
> 
>> +static const struct ethtool_fec_hist_range netdevsim_fec_ranges[] = {
>> +	{  0,  0},
>> +	{  1,  3},
>> +	{  4,  7},
>> +	{ -1, -1}
>> +};
> 
> Let's add a define for the terminating element?

I believe it's about (-1, -1) case. If we end up using (0, 0) then there
is no need to define anything, right?

> 
>> +/**
>> + * struct ethtool_fec_hist_range - byte range for FEC bins histogram statistics
> 
> byte range? thought these are bit errors..
> 
>> + * sentinel value of { -1, -1 } is used as marker for end of bins array
>> + * @low: low bound of the bin (inclusive)
>> + * @high: high bound of the bin (inclusive)
>> + */
> 
>> +		len += nla_total_size_64bit(sizeof(u64) * ETHTOOL_FEC_HIST_MAX);
> 
> I don't think it's right, each attr is its own nla_total_size().
> Add a nla_total_size(8) to the calculation below

got it

> 
>> +		/* add FEC bins information */
>> +		len += (nla_total_size(0) +  /* _A_FEC_HIST */
>> +			nla_total_size(4) +  /* _A_FEC_HIST_BIN_LOW */
>> +			nla_total_size(4)) * /* _A_FEC_HIST_BIN_HI */
>> +			ETHTOOL_FEC_HIST_MAX;