Re: [PATCH net-next 7/9] mlxsw: spectrum_ptp: Add implementation for physical hardware clock operations

[Shalom Toledo <shalomt@mellanox.com>]

On 04/06/2019 17:28, Richard Cochran wrote:
> On Mon, Jun 03, 2019 at 03:12:42PM +0300, Ido Schimmel wrote:
> 
>> +static int
>> +mlxsw_sp1_ptp_update_phc_settime(struct mlxsw_sp_ptp_clock *clock, u64 nsec)
> 
> Six words ^^^
> 
> What is wrong with "mlxsw_phc_settime" ?

I can drop the "update". But as Jiri mentioned, it is aligned with the rest of
mlxsw code.

> 
>> +{
>> +	struct mlxsw_core *mlxsw_core = clock->core;
>> +	char mtutc_pl[MLXSW_REG_MTUTC_LEN];
>> +	char mtpps_pl[MLXSW_REG_MTPPS_LEN];
>> +	u64 next_sec_in_nsec, cycles;
>> +	u32 next_sec;
>> +	int err;
>> +
>> +	next_sec = nsec / NSEC_PER_SEC + 1;
>> +	next_sec_in_nsec = next_sec * NSEC_PER_SEC;
>> +
>> +	spin_lock(&clock->lock);
>> +	cycles = mlxsw_sp1_ptp_ns2cycles(&clock->tc, next_sec_in_nsec);
>> +	spin_unlock(&clock->lock);
>> +
>> +	mlxsw_reg_mtpps_vpin_pack(mtpps_pl, cycles);
>> +	err = mlxsw_reg_write(mlxsw_core, MLXSW_REG(mtpps), mtpps_pl);
>> +	if (err)
>> +		return err;
>> +
>> +	mlxsw_reg_mtutc_pack(mtutc_pl,
>> +			     MLXSW_REG_MTUTC_OPERATION_SET_TIME_AT_NEXT_SEC,
>> +			     0, next_sec);
>> +	return mlxsw_reg_write(mlxsw_core, MLXSW_REG(mtutc), mtutc_pl);
>> +}
>> +
>> +static int mlxsw_sp1_ptp_adjfine(struct ptp_clock_info *ptp, long scaled_ppm)
>> +{
>> +	struct mlxsw_sp_ptp_clock *clock =
>> +		container_of(ptp, struct mlxsw_sp_ptp_clock, ptp_info);
>> +	int neg_adj = 0;
>> +	u32 diff;
>> +	u64 adj;
>> +	s32 ppb;
>> +
>> +	ppb = ptp_clock_scaled_ppm_to_ppb(scaled_ppm);
> 
> Now I see why you did this.  Nice try.

I didn't try anything.

The reason is that the hardware units is in ppb and not in scaled_ppm(or ppm),
so I just converted to ppb in order to set the hardware.

But I got your point, I will change my calculation to use scaled_ppm (to get a
more finer resolution) and not ppb, and convert to ppb just before setting the
hardware. Is that make sense?

But I'm still need to expose scaled_ppm_to_ppb.

> 
> The 'scaled_ppm' has a finer resolution than ppb.  Please make use of
> the finer resolution in your calculation.  It does make a difference.

Will change, thanks for that!

> 
>> +
>> +	if (ppb < 0) {
>> +		neg_adj = 1;
>> +		ppb = -ppb;
>> +	}
>> +
>> +	adj = clock->nominal_c_mult;
>> +	adj *= ppb;
>> +	diff = div_u64(adj, NSEC_PER_SEC);
>> +
>> +	spin_lock(&clock->lock);
>> +	timecounter_read(&clock->tc);
>> +	clock->cycles.mult = neg_adj ? clock->nominal_c_mult - diff :
>> +				       clock->nominal_c_mult + diff;
>> +	spin_unlock(&clock->lock);
>> +
>> +	return mlxsw_sp1_ptp_update_phc_adjfreq(clock, neg_adj ? -ppb : ppb);
>> +}
> 
> Thanks,
> Richard
>