Re: [PATCH net-next v5 4/5] net: phy: mediatek: Extend 1G TX/RX link pulse time

[SkyLake Huang (黃啟澤)]

On Thu, 2024-05-30 at 17:10 +0100, Russell King (Oracle) wrote:
>  	 
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
>  On Thu, May 30, 2024 at 04:01:08PM +0000, SkyLake Huang (黃啟澤) wrote:
> > I'm not going to handle timeout case here. If we can't detect
> > MTK_PHY_FINAL_SPEED_1000 in 1 second, let it go and we'll detect it
> > next round.
> 
> With this waiting up to one second for MTK_PHY_FINAL_SPEED_1000 to be
> set...
> 
> > > > +int mtk_gphy_cl22_read_status(struct phy_device *phydev)
> > > > +{
> > > > +int ret;
> > > > +
> > > > +ret = genphy_read_status(phydev);
> > > > +if (ret)
> > > > +return ret;
> > > > +
> > > > +if (phydev->autoneg == AUTONEG_ENABLE && !phydev-
> > > >autoneg_complete) {
> 
> Are you sure you want this condition like this? When the link is
> down,
> and 1G speeds are being advertised, it means that you'll call
> extend_an_new_lp_cnt_limit(). If MTK_PHY_FINAL_SPEED_1000 doesn't get
> set, that'll take one second each and every time we poll the PHY for
> its status - which will be done while holding phydev->lock.
> 
> This doesn't sound very good.
> 
> -- 
> RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
> FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!

I add another condition to make sure we enter
extend_an_new_lp_cnt_limit() only in first few seconds when we plug in
cable.

It will look like this:
===============================================================
#define MTK_PHY_AUX_CTRL_AND_STATUS		0x14
#define   MTK_PHY_LP_DETECTED_MASK		GENMASK(7, 6)

if (phydev->autoneg == AUTONEG_ENABLE && !phydev->autoneg_complete) {
	phy_select_page(phydev, MTK_PHY_PAGE_EXTENDED_1);
	ret = __phy_read(phydev, MTK_PHY_AUX_CTRL_AND_STATUS);
	phy_restore_page(phydev, MTK_PHY_PAGE_STANDARD, 0);

/* Once LP_DETECTED is set, it means that"ability_match" in IEEE 802.3
 * Figure 28-18 is set. This happens after we plug in cable. Also,
LP_DETECTED
 * will be cleared after AN complete.
 */
	if (!FIELD_GET(MTK_PHY_LP_DETECTED_MASK, ret))
		return 0;

	ret = phy_read(phydev, MII_CTRL1000);
	if (ret & (ADVERTISE_1000FULL | ADVERTISE_1000HALF)) {
		ret = extend_an_new_lp_cnt_limit(phydev);
		if (ret < 0)
			return ret;
	}
}

return 0;
===============================================================
This is tested OK on my mt7988 platform.

Sky