[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20230727104603.mqbnfrhri6riesrx@skbuf>
Date: Thu, 27 Jul 2023 13:46:03 +0300
From: Vladimir Oltean <olteanv@...il.com>
To: Ante Knezic <ante.knezic@...mholz.de>
Cc: netdev@...r.kernel.org, andrew@...n.ch, f.fainelli@...il.com,
davem@...emloft.net, edumazet@...gle.com, kuba@...nel.org,
pabeni@...hat.com, linux-kernel@...r.kernel.org,
Russell King <linux@...linux.org.uk>
Subject: Re: [PATCH net-next v4] net: dsa: mv88e6xxx: Add erratum 3.14 for
88E6390X and 88E6190X
On Thu, Jul 27, 2023 at 10:25:50AM +0200, Ante Knezic wrote:
> Fixes XAUI/RXAUI lane alignment errors.
> Issue causes dropped packets when trying to communicate over
> fiber via SERDES lanes of port 9 and 10.
> Errata document applies only to 88E6190X and 88E6390X devices.
> Requires poking in undocumented registers.
>
> Signed-off-by: Ante Knezic <ante.knezic@...mholz.de>
> ---
> V4 : Rework as suggested by Vladimir Oltean <olteanv@...il.com>
> and Russell King <linux@...linux.org.uk>
> * print error in case of failure to apply erratum
> * use mdiobus_c45_write instead of mdiodev_c45_write
> * use bool variable instead of embedding a chip pointer inside
> pcs struct.
> V3 : Rework to fit the new phylink_pcs infrastructure
> V2 : Rework as suggested by Andrew Lunn <andrew@....ch>
> * make int lanes[] const
> * reorder prod_nums
> * update commit message to indicate we are dealing with
> undocumented Marvell registers and magic values
> ---
> drivers/net/dsa/mv88e6xxx/pcs-639x.c | 45 ++++++++++++++++++++++++++++++++++++
> 1 file changed, 45 insertions(+)
>
> diff --git a/drivers/net/dsa/mv88e6xxx/pcs-639x.c b/drivers/net/dsa/mv88e6xxx/pcs-639x.c
> index 98dd49dac421..ba373656bfe1 100644
> --- a/drivers/net/dsa/mv88e6xxx/pcs-639x.c
> +++ b/drivers/net/dsa/mv88e6xxx/pcs-639x.c
> @@ -20,6 +20,7 @@ struct mv88e639x_pcs {
> struct mdio_device mdio;
> struct phylink_pcs sgmii_pcs;
> struct phylink_pcs xg_pcs;
> + bool erratum_3_14;
> bool supports_5g;
> phy_interface_t interface;
> unsigned int irq;
> @@ -205,13 +206,53 @@ static void mv88e639x_sgmii_pcs_pre_config(struct phylink_pcs *pcs,
> mv88e639x_sgmii_pcs_control_pwr(mpcs, false);
> }
>
> +static int mv88e6390_erratum_3_14(struct mv88e639x_pcs *mpcs)
> +{
> + const int lanes[] = { MV88E6390_PORT9_LANE0, MV88E6390_PORT9_LANE1,
> + MV88E6390_PORT9_LANE2, MV88E6390_PORT9_LANE3,
> + MV88E6390_PORT10_LANE0, MV88E6390_PORT10_LANE1,
> + MV88E6390_PORT10_LANE2, MV88E6390_PORT10_LANE3 };
> + int err, i;
> +
> + /* 88e6190x and 88e6390x errata 3.14:
> + * After chip reset, SERDES reconfiguration or SERDES core
> + * Software Reset, the SERDES lanes may not be properly aligned
> + * resulting in CRC errors
> + */
> +
> + for (i = 0; i < ARRAY_SIZE(lanes); i++) {
> + err = mdiobus_c45_write(mpcs->mdio.bus, lanes[i],
> + MDIO_MMD_PHYXS,
> + 0xf054, 0x400C);
> + if (err)
> + return err;
> +
> + err = mdiobus_c45_write(mpcs->mdio.bus, lanes[i],
> + MDIO_MMD_PHYXS,
> + 0xf054, 0x4000);
> + if (err)
> + return err;
> + }
> +
> + return 0;
> +}
> +
> static int mv88e639x_sgmii_pcs_post_config(struct phylink_pcs *pcs,
> phy_interface_t interface)
> {
> struct mv88e639x_pcs *mpcs = sgmii_pcs_to_mv88e639x_pcs(pcs);
> + int err;
>
> mv88e639x_sgmii_pcs_control_pwr(mpcs, true);
>
> + if (mpcs->erratum_3_14) {
> + err = mv88e6390_erratum_3_14(mpcs);
> + if (err)
> + dev_err(mpcs->mdio.dev.parent,
> + "failed to apply erratum 3.14: %pe\n",
> + ERR_PTR(err));
> + }
> +
> return 0;
> }
>
> @@ -524,6 +565,10 @@ static int mv88e6390_pcs_init(struct mv88e6xxx_chip *chip, int port)
> mpcs->xg_pcs.ops = &mv88e6390_xg_pcs_ops;
> mpcs->xg_pcs.neg_mode = true;
>
> + if (chip->info->prod_num == MV88E6XXX_PORT_SWITCH_ID_PROD_6190X ||
> + chip->info->prod_num == MV88E6XXX_PORT_SWITCH_ID_PROD_6390X)
> + mpcs->erratum_3_14 = true;
> +
> err = mv88e639x_pcs_setup_irq(mpcs, chip, port);
> if (err)
> goto err_free;
> --
> 2.11.0
>
It's good practice that, if you had a conversation with Russell on previous
versions of the patch, to keep him CCed on newer versions of it.
Powered by blists - more mailing lists