linux-kernel - Re: [PATCH net-next 1/4] net: pcs: xpcs: add CL37 1000BASE-X AN support

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <YmJgqSdF7LMxoSXv@shell.armlinux.org.uk>
Date:   Fri, 22 Apr 2022 09:00:41 +0100
From:   "Russell King (Oracle)" <linux@...linux.org.uk>
To:     Ong Boon Leong <boon.leong.ong@...el.com>
Cc:     Alexandre Torgue <alexandre.torgue@...com>,
        Jose Abreu <joabreu@...opsys.com>,
        Andrew Lunn <andrew@...n.ch>,
        Heiner Kallweit <hkallweit1@...il.com>,
        Paolo Abeni <pabeni@...hat.com>,
        "David S . Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Maxime Coquelin <mcoquelin.stm32@...il.com>,
        Alexandre Torgue <alexandre.torgue@...s.st.com>,
        Giuseppe Cavallaro <peppe.cavallaro@...com>,
        netdev@...r.kernel.org, linux-stm32@...md-mailman.stormreply.com,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH net-next 1/4] net: pcs: xpcs: add CL37 1000BASE-X AN
 support

Hi,

On Fri, Apr 22, 2022 at 03:35:02PM +0800, Ong Boon Leong wrote:
> @@ -774,6 +788,58 @@ static int xpcs_config_aneg_c37_sgmii(struct dw_xpcs *xpcs, unsigned int mode)
>  	return ret;
>  }
>  
> +static int xpcs_config_aneg_c37_1000basex(struct dw_xpcs *xpcs, unsigned int mode,
> +					  const unsigned long *advertising)
> +{
> +	int ret, mdio_ctrl;
> +
> +	/* For AN for 1000BASE-X mode, the settings are :-
> +	 * 1) VR_MII_MMD_CTRL Bit(12) [AN_ENABLE] = 0b (Disable C37 AN in case
> +	 *    it is already enabled)
> +	 * 2) VR_MII_AN_CTRL Bit(2:1)[PCS_MODE] = 00b (1000BASE-X C37)
> +	 * 3) SR_MII_AN_ADV Bit(6)[FD] = 1b (Full Duplex)
> +	 *    Note: Half Duplex is rarely used, so don't advertise.
> +	 * 4) VR_MII_MMD_CTRL Bit(12) [AN_ENABLE] = 1b (Enable C37 AN)

So if this function gets called to update the advertisement - even if
there is no actual change - we go through a AN-disable..AN-enable
dance and cause the link to re-negotiate. That doesn't sound like nice
behaviour.

> +	 */
> +	mdio_ctrl = xpcs_read(xpcs, MDIO_MMD_VEND2, DW_VR_MII_MMD_CTRL);
> +	if (mdio_ctrl < 0)
> +		return mdio_ctrl;
> +
> +	if (mdio_ctrl & AN_CL37_EN) {
> +		ret = xpcs_write(xpcs, MDIO_MMD_VEND2, DW_VR_MII_MMD_CTRL,
> +				 mdio_ctrl & ~AN_CL37_EN);
> +		if (ret < 0)
> +			return ret;
> +	}
> +
> +	ret = xpcs_read(xpcs, MDIO_MMD_VEND2, DW_VR_MII_AN_CTRL);
> +	if (ret < 0)
> +		return ret;
> +
> +	ret &= ~DW_VR_MII_PCS_MODE_MASK;
> +	ret = xpcs_write(xpcs, MDIO_MMD_VEND2, DW_VR_MII_AN_CTRL, ret);
> +	if (ret < 0)
> +		return ret;
> +
> +	ret = xpcs_read(xpcs, MDIO_MMD_VEND2, MII_ADVERTISE);
> +	ret |= ADVERTISE_1000XFULL;
> +	ret = xpcs_write(xpcs, MDIO_MMD_VEND2, MII_ADVERTISE, ret);

What if other bits are already set in the MII_ADVERTISE register?
Maybe consider using phylink_mii_c22_pcs_encode_advertisement()?

The pcs_config() method is also supposed to return either a negative
error, 0 for no advertisement change, or positive for an advertisement
change, in which case phylink will trigger a call to pcs_an_restart().

> +static int xpcs_get_state_c37_1000basex(struct dw_xpcs *xpcs,
> +					struct phylink_link_state *state)
> +{
> +	int lpa, adv;
> +	int ret;
> +
> +	ret = xpcs_read(xpcs, MDIO_MMD_VEND2, DW_VR_MII_MMD_CTRL);
> +	if (ret < 0)
> +		return ret;
> +
> +	if (ret & AN_CL37_EN) {
> +		/* Reset link_state */
> +		state->link = false;
> +		state->speed = SPEED_UNKNOWN;
> +		state->duplex = DUPLEX_UNKNOWN;
> +		state->pause = 0;

Phylink guarantees that speed, duplex and pause are set to something
sensible - please remove these. The only one you probably need here
is state->link.

> +
> +		lpa = xpcs_read(xpcs, MDIO_MMD_VEND2, MII_LPA);
> +		if (lpa < 0 || lpa & LPA_RFAULT)
> +			return false;

This function does not return a boolean. Returning "false" is the same
as returning 0, which means "no error" but an error has occurred.

> +
> +		adv = xpcs_read(xpcs, MDIO_MMD_VEND2, MII_ADVERTISE);
> +		if (adv < 0)
> +			return false;

Ditto.

> +
> +		if (lpa & ADVERTISE_1000XFULL &&
> +		    adv & ADVERTISE_1000XFULL) {
> +			state->speed = SPEED_1000;
> +			state->duplex = DUPLEX_FULL;
> +			state->link = true;
> +		}
> +
> +		/* Clear CL37 AN complete status */
> +		ret = xpcs_write(xpcs, MDIO_MMD_VEND2, DW_VR_MII_AN_INTR_STS, 0);
> +	} else {
> +		state->link = true;
> +		state->speed = SPEED_1000;
> +		state->duplex = DUPLEX_FULL;
> +		state->pause = 0;

If we're in AN-disabled mode, phylink will set state->speed and
state->duplex according to the user's parameters, so there should be no
need to do it here.

> @@ -994,9 +1143,21 @@ void xpcs_link_up(struct phylink_pcs *pcs, unsigned int mode,
>  		return xpcs_config_usxgmii(xpcs, speed);
>  	if (interface == PHY_INTERFACE_MODE_SGMII)
>  		return xpcs_link_up_sgmii(xpcs, mode, speed, duplex);
> +	if (interface == PHY_INTERFACE_MODE_1000BASEX)
> +		return xpcs_link_up_1000basex(xpcs, speed, duplex);
>  }
>  EXPORT_SYMBOL_GPL(xpcs_link_up);
>  
> +static void xpcs_an_restart(struct phylink_pcs *pcs)
> +{
> +	struct dw_xpcs *xpcs = phylink_pcs_to_xpcs(pcs);
> +	int ret;
> +
> +	ret = xpcs_read(xpcs, MDIO_MMD_VEND2, MDIO_CTRL1);
> +	ret |= BMCR_ANRESTART;
> +	ret = xpcs_write(xpcs, MDIO_MMD_VEND2, MDIO_CTRL1, ret);

If xpcs_read() returns an error, we try to write the error back to
the control register? Is that a good idea/

Thanks.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!