Re: [RFC] net: dsa: mv88e6xxx: ptp: improve phc2sys precision for mv88e6xxx switch in combination with imx6-fec

[Richard Cochran]

On Fri, Aug 02, 2019 at 06:32:48PM +0200, Hubert Feurstein wrote:

> diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
> index 2f6057e7335d..20f589dc5b8b 100644
> --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -1814,11 +1814,25 @@ static int fec_enet_mdio_write(struct mii_bus *bus, int mii_id, int regnum,
>  
>  	reinit_completion(&fep->mdio_done);
>  
> -	/* start a write op */
> -	writel(FEC_MMFR_ST | FEC_MMFR_OP_WRITE |
> -		FEC_MMFR_PA(mii_id) | FEC_MMFR_RA(regnum) |
> -		FEC_MMFR_TA | FEC_MMFR_DATA(value),
> -		fep->hwp + FEC_MII_DATA);
> +	if (bus->ptp_sts) {
> +		unsigned long flags = 0;
> +		local_irq_save(flags);
> +		__iowmb();
> +		/* >Take the timestamp *after* the memory barrier */
> +		ptp_read_system_prets(bus->ptp_sts);
> +		writel_relaxed(FEC_MMFR_ST | FEC_MMFR_OP_WRITE |
> +			FEC_MMFR_PA(mii_id) | FEC_MMFR_RA(regnum) |
> +			FEC_MMFR_TA | FEC_MMFR_DATA(value),
> +			fep->hwp + FEC_MII_DATA);
> +		ptp_read_system_postts(bus->ptp_sts);

Regarding generic support of this, see if you can't place the
ptp_read_system_prets/postts() calls at the mii_bus layer.  This could
mean, for example, offering a two-part write API, to split this write
operation from...

> +		local_irq_restore(flags);
> +	} else {
> +		/* start a write op */
> +		writel(FEC_MMFR_ST | FEC_MMFR_OP_WRITE |
> +			FEC_MMFR_PA(mii_id) | FEC_MMFR_RA(regnum) |
> +			FEC_MMFR_TA | FEC_MMFR_DATA(value),
> +			fep->hwp + FEC_MII_DATA);
> +	}
>
>  	/* wait for end of transfer */
>  	time_left = wait_for_completion_timeout(&fep->mdio_done,

...this kind of thing ^^^

Thanks,
Richard