[PATCH net-next v2 0/2] amd-xgbe: add support for 2.5GbE and rx-adaptation

[PATCH net-next v2 0/2] amd-xgbe: add support for 2.5GbE and rx-adaptation

[Raju Rangoju]

This patch series adds support for 2.GbE in 10GBaseT mode and
rx-adaptation support for Yellow Carp devices.

1) Support for 2.5GbE:
   Add the necessary changes to the driver to fully recognize and enable
   2.5GbE speed in 10GBaseT mode.

2) Support for rx-adaptation:
   In order to support the 10G backplane mode without Auto-negotiation
   and to support the longer-length DAC cables, it requires PHY to
   perform RX Adaptation sequence as mentioned in the Synopsys databook.
   Add the necessary changes to Yellow Carp devices to ensure seamless
   RX Adaptation for 10G-SFI (LONG DAC), and 10G-KR modes without
   Auto-Negotiation (CL72 not present)

Changes since v1:
 - remove the unneeded inline usage
 - re-ordered the code to quash forward declarations; however there are
   couple of them since there is a circular dependency. Below is link
   for the same:
   https://lore.kernel.org/netdev/20230131111155.5cb607a6@kernel.org/

Raju Rangoju (2):
  amd-xgbe: add 2.5GbE support to 10G BaseT mode
  amd-xgbe: add support for rx-adaptation

 drivers/net/ethernet/amd/xgbe/xgbe-common.h |  38 ++++
 drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c | 187 +++++++++++++++++++-
 drivers/net/ethernet/amd/xgbe/xgbe.h        |   5 +
 3 files changed, 224 insertions(+), 6 deletions(-)

-- 
2.25.1

[PATCH net-next v2 1/2] amd-xgbe: add 2.5GbE support to 10G BaseT mode

[Raju Rangoju]

Add support to the driver to fully recognize and enable 2.5GbE speed in
10GBaseT mode.

Acked-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
---
 drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c b/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c
index f4683d53e58c..eea06e07b4a0 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c
@@ -1882,6 +1882,9 @@ static void xgbe_phy_an_advertising(struct xgbe_prv_data *pdata,
 		if (phy_data->phydev &&
 		    (phy_data->phydev->speed == SPEED_10000))
 			XGBE_SET_ADV(dlks, 10000baseKR_Full);
+		else if (phy_data->phydev &&
+			 (phy_data->phydev->speed == SPEED_2500))
+			XGBE_SET_ADV(dlks, 2500baseX_Full);
 		else
 			XGBE_SET_ADV(dlks, 1000baseKX_Full);
 		break;
@@ -2282,9 +2285,11 @@ static enum xgbe_mode xgbe_phy_switch_baset_mode(struct xgbe_prv_data *pdata)
 	case XGBE_MODE_SGMII_100:
 	case XGBE_MODE_SGMII_1000:
 		return XGBE_MODE_KR;
+	case XGBE_MODE_KX_2500:
+		return XGBE_MODE_SGMII_1000;
 	case XGBE_MODE_KR:
 	default:
-		return XGBE_MODE_SGMII_1000;
+		return XGBE_MODE_KX_2500;
 	}
 }
 
@@ -2644,7 +2649,8 @@ static bool xgbe_phy_valid_speed_baset_mode(struct xgbe_prv_data *pdata,
 	case SPEED_1000:
 		return true;
 	case SPEED_2500:
-		return (phy_data->port_mode == XGBE_PORT_MODE_NBASE_T);
+		return ((phy_data->port_mode == XGBE_PORT_MODE_10GBASE_T) ||
+			(phy_data->port_mode == XGBE_PORT_MODE_NBASE_T));
 	case SPEED_10000:
 		return (phy_data->port_mode == XGBE_PORT_MODE_10GBASE_T);
 	default:
@@ -3024,6 +3030,7 @@ static bool xgbe_phy_port_mode_mismatch(struct xgbe_prv_data *pdata)
 		if ((phy_data->port_speeds & XGBE_PHY_PORT_SPEED_10) ||
 		    (phy_data->port_speeds & XGBE_PHY_PORT_SPEED_100) ||
 		    (phy_data->port_speeds & XGBE_PHY_PORT_SPEED_1000) ||
+		    (phy_data->port_speeds & XGBE_PHY_PORT_SPEED_2500) ||
 		    (phy_data->port_speeds & XGBE_PHY_PORT_SPEED_10000))
 			return false;
 		break;
@@ -3474,6 +3481,10 @@ static int xgbe_phy_init(struct xgbe_prv_data *pdata)
 			XGBE_SET_SUP(lks, 1000baseT_Full);
 			phy_data->start_mode = XGBE_MODE_SGMII_1000;
 		}
+		if (phy_data->port_speeds & XGBE_PHY_PORT_SPEED_2500) {
+			XGBE_SET_SUP(lks, 2500baseT_Full);
+			phy_data->start_mode = XGBE_MODE_KX_2500;
+		}
 		if (phy_data->port_speeds & XGBE_PHY_PORT_SPEED_10000) {
 			XGBE_SET_SUP(lks, 10000baseT_Full);
 			phy_data->start_mode = XGBE_MODE_KR;
-- 
2.25.1

[PATCH net-next v2 2/2] amd-xgbe: add support for rx-adaptation

[Raju Rangoju]

The existing implementation for non-Autonegotiation 10G speed modes does
not enable RX adaptation in the Driver and FW. The RX Equalization
settings (AFE settings alone) are manually configured and the existing
link-up sequence in the driver does not perform rx adaptation process as
mentioned in the Synopsys databook. There's a customer request for 10G
backplane mode without Auto-negotiation and for the DAC cables of more
significant length that follow the non-Autonegotiation mode. These modes
require PHY to perform RX Adaptation.

The proposed logic adds the necessary changes to Yellow Carp devices to
ensure seamless RX Adaptation for 10G-SFI (LONG DAC) and 10G-KR without
AN (CL72 not present). The RX adaptation core algorithm is executed by
firmware, however, to achieve that a new mailbox sub-command is required
to be sent by the driver.

Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
---
 drivers/net/ethernet/amd/xgbe/xgbe-common.h |  38 +++++
 drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c | 172 +++++++++++++++++++-
 drivers/net/ethernet/amd/xgbe/xgbe.h        |   5 +
 3 files changed, 211 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-common.h b/drivers/net/ethernet/amd/xgbe/xgbe-common.h
index 3fd9728f817f..3b70f6737633 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-common.h
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-common.h
@@ -1285,6 +1285,22 @@
 #define MDIO_PMA_RX_CTRL1		0x8051
 #endif
 
+#ifndef MDIO_PMA_RX_LSTS
+#define MDIO_PMA_RX_LSTS		0x018020
+#endif
+
+#ifndef MDIO_PMA_RX_EQ_CTRL4
+#define MDIO_PMA_RX_EQ_CTRL4		0x0001805C
+#endif
+
+#ifndef MDIO_PMA_MP_MISC_STS
+#define MDIO_PMA_MP_MISC_STS		0x0078
+#endif
+
+#ifndef MDIO_PMA_PHY_RX_EQ_CEU
+#define MDIO_PMA_PHY_RX_EQ_CEU		0x1800E
+#endif
+
 #ifndef MDIO_PCS_DIG_CTRL
 #define MDIO_PCS_DIG_CTRL		0x8000
 #endif
@@ -1395,6 +1411,28 @@
 #define XGBE_PMA_RX_RST_0_RESET_ON	0x10
 #define XGBE_PMA_RX_RST_0_RESET_OFF	0x00
 
+#define XGBE_PMA_RX_SIG_DET_0_MASK	BIT(4)
+#define XGBE_PMA_RX_SIG_DET_0_ENABLE	BIT(4)
+#define XGBE_PMA_RX_SIG_DET_0_DISABLE	0x0000
+
+#define XGBE_PMA_RX_VALID_0_MASK	BIT(12)
+#define XGBE_PMA_RX_VALID_0_ENABLE	BIT(12)
+#define XGBE_PMA_RX_VALID_0_DISABLE	0x0000
+
+#define XGBE_PMA_RX_AD_REQ_MASK		BIT(12)
+#define XGBE_PMA_RX_AD_REQ_ENABLE	BIT(12)
+#define XGBE_PMA_RX_AD_REQ_DISABLE	0x0000
+
+#define XGBE_PMA_RX_ADPT_ACK_MASK	BIT(12)
+#define XGBE_PMA_RX_ADPT_ACK		BIT(12)
+
+#define XGBE_PMA_CFF_UPDTM1_VLD		BIT(8)
+#define XGBE_PMA_CFF_UPDT0_VLD		BIT(9)
+#define XGBE_PMA_CFF_UPDT1_VLD		BIT(10)
+#define XGBE_PMA_CFF_UPDT_MASK		(XGBE_PMA_CFF_UPDTM1_VLD |\
+					 XGBE_PMA_CFF_UPDT0_VLD | \
+					 XGBE_PMA_CFF_UPDT1_VLD)
+
 #define XGBE_PMA_PLL_CTRL_MASK		BIT(15)
 #define XGBE_PMA_PLL_CTRL_ENABLE	BIT(15)
 #define XGBE_PMA_PLL_CTRL_DISABLE	0x0000
diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c b/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c
index eea06e07b4a0..7d88caa4e623 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c
+++ b/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c
@@ -388,6 +388,9 @@ struct xgbe_phy_data {
 static DEFINE_MUTEX(xgbe_phy_comm_lock);
 
 static enum xgbe_an_mode xgbe_phy_an_mode(struct xgbe_prv_data *pdata);
+static void xgbe_phy_rrc(struct xgbe_prv_data *pdata);
+static void xgbe_phy_perform_ratechange(struct xgbe_prv_data *pdata,
+					unsigned int cmd, unsigned int sub_cmd);
 
 static int xgbe_phy_i2c_xfer(struct xgbe_prv_data *pdata,
 			     struct xgbe_i2c_op *i2c_op)
@@ -2038,6 +2041,93 @@ static void xgbe_phy_set_redrv_mode(struct xgbe_prv_data *pdata)
 	xgbe_phy_put_comm_ownership(pdata);
 }
 
+#define MAX_RX_ADAPT_RETRIES		1
+#define XGBE_PMA_RX_VAL_SIG_MASK	(XGBE_PMA_RX_SIG_DET_0_MASK | \
+					 XGBE_PMA_RX_VALID_0_MASK)
+
+static void xgbe_set_rx_adap_mode(struct xgbe_prv_data *pdata,
+				  enum xgbe_mode mode)
+{
+	if (pdata->rx_adapt_retries++ >= MAX_RX_ADAPT_RETRIES) {
+		pdata->rx_adapt_retries = 0;
+		return;
+	}
+
+	xgbe_phy_perform_ratechange(pdata,
+				    mode == XGBE_MODE_KR ?
+				    XGBE_MB_CMD_SET_10G_KR :
+				    XGBE_MB_CMD_SET_10G_SFI,
+				    XGBE_MB_SUBCMD_RX_ADAP);
+}
+
+static void xgbe_rx_adaptation(struct xgbe_prv_data *pdata)
+{
+	struct xgbe_phy_data *phy_data = pdata->phy_data;
+	unsigned int reg;
+
+	/* step 2: force PCS to send RX_ADAPT Req to PHY */
+	XMDIO_WRITE_BITS(pdata, MDIO_MMD_PMAPMD, MDIO_PMA_RX_EQ_CTRL4,
+			 XGBE_PMA_RX_AD_REQ_MASK, XGBE_PMA_RX_AD_REQ_ENABLE);
+
+	/* Step 3: Wait for RX_ADAPT ACK from the PHY */
+	msleep(200);
+
+	/* Software polls for coefficient update command (given by local PHY) */
+	reg = XMDIO_READ(pdata, MDIO_MMD_PMAPMD, MDIO_PMA_PHY_RX_EQ_CEU);
+
+	/* Clear the RX_AD_REQ bit */
+	XMDIO_WRITE_BITS(pdata, MDIO_MMD_PMAPMD, MDIO_PMA_RX_EQ_CTRL4,
+			 XGBE_PMA_RX_AD_REQ_MASK, XGBE_PMA_RX_AD_REQ_DISABLE);
+
+	/* Check if coefficient update command is set */
+	if ((reg & XGBE_PMA_CFF_UPDT_MASK) != XGBE_PMA_CFF_UPDT_MASK)
+		goto set_mode;
+
+	/* Step 4: Check for Block lock */
+
+	/* Link status is latched low, so read once to clear
+	 * and then read again to get current state
+	 */
+	reg = XMDIO_READ(pdata, MDIO_MMD_PCS, MDIO_STAT1);
+	reg = XMDIO_READ(pdata, MDIO_MMD_PCS, MDIO_STAT1);
+	if (reg & MDIO_STAT1_LSTATUS) {
+		/* If the block lock is found, update the helpers
+		 * and declare the link up
+		 */
+		netif_dbg(pdata, link, pdata->netdev, "Block_lock done");
+		pdata->rx_adapt_done = true;
+		pdata->mode_set = false;
+		return;
+	}
+
+set_mode:
+	xgbe_set_rx_adap_mode(pdata, phy_data->cur_mode);
+}
+
+static void xgbe_phy_rx_adaptation(struct xgbe_prv_data *pdata)
+{
+	unsigned int reg;
+
+rx_adapt_reinit:
+	reg = XMDIO_READ_BITS(pdata, MDIO_MMD_PMAPMD, MDIO_PMA_RX_LSTS,
+			      XGBE_PMA_RX_VAL_SIG_MASK);
+
+	/* step 1: Check for RX_VALID && LF_SIGDET */
+	if ((reg & XGBE_PMA_RX_VAL_SIG_MASK) != XGBE_PMA_RX_VAL_SIG_MASK) {
+		netif_dbg(pdata, link, pdata->netdev,
+			  "RX_VALID or LF_SIGDET is unset, issue rrc");
+		xgbe_phy_rrc(pdata);
+		if (pdata->rx_adapt_retries++ >= MAX_RX_ADAPT_RETRIES) {
+			pdata->rx_adapt_retries = 0;
+			return;
+		}
+		goto rx_adapt_reinit;
+	}
+
+	/* perform rx adaptation */
+	xgbe_rx_adaptation(pdata);
+}
+
 static void xgbe_phy_rx_reset(struct xgbe_prv_data *pdata)
 {
 	int reg;
@@ -2103,7 +2193,7 @@ static void xgbe_phy_perform_ratechange(struct xgbe_prv_data *pdata,
 	wait = XGBE_RATECHANGE_COUNT;
 	while (wait--) {
 		if (!XP_IOREAD_BITS(pdata, XP_DRIVER_INT_RO, STATUS))
-			goto reenable_pll;
+			goto do_rx_adaptation;
 
 		usleep_range(1000, 2000);
 	}
@@ -2113,6 +2203,20 @@ static void xgbe_phy_perform_ratechange(struct xgbe_prv_data *pdata,
 
 	/* Reset on error */
 	xgbe_phy_rx_reset(pdata);
+	goto reenable_pll;
+
+do_rx_adaptation:
+	if (pdata->en_rx_adap && sub_cmd == XGBE_MB_SUBCMD_RX_ADAP &&
+	    (cmd == XGBE_MB_CMD_SET_10G_KR || cmd == XGBE_MB_CMD_SET_10G_SFI)) {
+		netif_dbg(pdata, link, pdata->netdev,
+			  "Enabling RX adaptation\n");
+		pdata->mode_set = true;
+		xgbe_phy_rx_adaptation(pdata);
+		/* return from here to avoid enabling PLL ctrl
+		 * during adaptation phase
+		 */
+		return;
+	}
 
 reenable_pll:
 	/* Enable PLL re-initialization, not needed for PHY Power Off and RRC cmds */
@@ -2141,6 +2245,31 @@ static void xgbe_phy_power_off(struct xgbe_prv_data *pdata)
 	netif_dbg(pdata, link, pdata->netdev, "phy powered off\n");
 }
 
+static bool enable_rx_adap(struct xgbe_prv_data *pdata, enum xgbe_mode mode)
+{
+	struct xgbe_phy_data *phy_data = pdata->phy_data;
+	unsigned int ver;
+
+	/* Rx-Adaptation is not supported on older platforms(< 0x30H) */
+	ver = XGMAC_GET_BITS(pdata->hw_feat.version, MAC_VR, SNPSVER);
+	if (ver < 0x30)
+		return false;
+
+	/* Re-driver models 4223 && 4227 do not support Rx-Adaptation */
+	if (phy_data->redrv &&
+	    (phy_data->redrv_model == XGBE_PHY_REDRV_MODEL_4223 ||
+	     phy_data->redrv_model == XGBE_PHY_REDRV_MODEL_4227))
+		return false;
+
+	/* 10G KR mode with AN does not support Rx-Adaptation */
+	if (mode == XGBE_MODE_KR &&
+	    phy_data->port_mode != XGBE_PORT_MODE_BACKPLANE_NO_AUTONEG)
+		return false;
+
+	pdata->en_rx_adap = 1;
+	return true;
+}
+
 static void xgbe_phy_sfi_mode(struct xgbe_prv_data *pdata)
 {
 	struct xgbe_phy_data *phy_data = pdata->phy_data;
@@ -2149,7 +2278,12 @@ static void xgbe_phy_sfi_mode(struct xgbe_prv_data *pdata)
 
 	/* 10G/SFI */
 	if (phy_data->sfp_cable != XGBE_SFP_CABLE_PASSIVE) {
+		pdata->en_rx_adap = 0;
 		xgbe_phy_perform_ratechange(pdata, XGBE_MB_CMD_SET_10G_SFI, XGBE_MB_SUBCMD_ACTIVE);
+	} else if ((phy_data->sfp_cable == XGBE_SFP_CABLE_PASSIVE) &&
+		   (enable_rx_adap(pdata, XGBE_MODE_SFI))) {
+		xgbe_phy_perform_ratechange(pdata, XGBE_MB_CMD_SET_10G_SFI,
+					    XGBE_MB_SUBCMD_RX_ADAP);
 	} else {
 		if (phy_data->sfp_cable_len <= 1)
 			xgbe_phy_perform_ratechange(pdata, XGBE_MB_CMD_SET_10G_SFI,
@@ -2230,7 +2364,12 @@ static void xgbe_phy_kr_mode(struct xgbe_prv_data *pdata)
 	xgbe_phy_set_redrv_mode(pdata);
 
 	/* 10G/KR */
-	xgbe_phy_perform_ratechange(pdata, XGBE_MB_CMD_SET_10G_KR, XGBE_MB_SUBCMD_NONE);
+	if (enable_rx_adap(pdata, XGBE_MODE_KR))
+		xgbe_phy_perform_ratechange(pdata, XGBE_MB_CMD_SET_10G_KR,
+					    XGBE_MB_SUBCMD_RX_ADAP);
+	else
+		xgbe_phy_perform_ratechange(pdata, XGBE_MB_CMD_SET_10G_KR,
+					    XGBE_MB_SUBCMD_NONE);
 
 	phy_data->cur_mode = XGBE_MODE_KR;
 
@@ -2743,8 +2882,11 @@ static int xgbe_phy_link_status(struct xgbe_prv_data *pdata, int *an_restart)
 			return 0;
 		}
 
-		if (phy_data->sfp_mod_absent || phy_data->sfp_rx_los)
+		if (phy_data->sfp_mod_absent || phy_data->sfp_rx_los) {
+			if (pdata->en_rx_adap)
+				pdata->rx_adapt_done = false;
 			return 0;
+		}
 	}
 
 	if (phy_data->phydev) {
@@ -2766,7 +2908,29 @@ static int xgbe_phy_link_status(struct xgbe_prv_data *pdata, int *an_restart)
 	 */
 	reg = XMDIO_READ(pdata, MDIO_MMD_PCS, MDIO_STAT1);
 	reg = XMDIO_READ(pdata, MDIO_MMD_PCS, MDIO_STAT1);
-	if (reg & MDIO_STAT1_LSTATUS)
+
+	if (pdata->en_rx_adap) {
+		/* if the link is available and adaptation is done,
+		 * declare link up
+		 */
+		if ((reg & MDIO_STAT1_LSTATUS) && pdata->rx_adapt_done)
+			return 1;
+		/* If either link is not available or adaptation is not done,
+		 * retrigger the adaptation logic. (if the mode is not set,
+		 * then issue mailbox command first)
+		 */
+		if (pdata->mode_set) {
+			xgbe_phy_rx_adaptation(pdata);
+		} else {
+			pdata->rx_adapt_done = false;
+			xgbe_phy_set_mode(pdata, phy_data->cur_mode);
+		}
+
+		/* check again for the link and adaptation status */
+		reg = XMDIO_READ(pdata, MDIO_MMD_PCS, MDIO_STAT1);
+		if ((reg & MDIO_STAT1_LSTATUS) && pdata->rx_adapt_done)
+			return 1;
+	} else if (reg & MDIO_STAT1_LSTATUS)
 		return 1;
 
 	if (pdata->phy.autoneg == AUTONEG_ENABLE &&
diff --git a/drivers/net/ethernet/amd/xgbe/xgbe.h b/drivers/net/ethernet/amd/xgbe/xgbe.h
index 16e73df3e9b9..ad136ed493ed 100644
--- a/drivers/net/ethernet/amd/xgbe/xgbe.h
+++ b/drivers/net/ethernet/amd/xgbe/xgbe.h
@@ -625,6 +625,7 @@ enum xgbe_mb_cmd {
 
 enum xgbe_mb_subcmd {
 	XGBE_MB_SUBCMD_NONE = 0,
+	XGBE_MB_SUBCMD_RX_ADAP,
 
 	/* 10GbE SFP subcommands */
 	XGBE_MB_SUBCMD_ACTIVE = 0,
@@ -1316,6 +1317,10 @@ struct xgbe_prv_data {
 
 	bool debugfs_an_cdr_workaround;
 	bool debugfs_an_cdr_track_early;
+	bool en_rx_adap;
+	int rx_adapt_retries;
+	bool rx_adapt_done;
+	bool mode_set;
 };
 
 /* Function prototypes*/
-- 
2.25.1

Re: [PATCH net-next v2 2/2] amd-xgbe: add support for rx-adaptation

[Simon Horman]

On Wed, Feb 01, 2023 at 11:19:32AM +0530, Raju Rangoju wrote:
> The existing implementation for non-Autonegotiation 10G speed modes does
> not enable RX adaptation in the Driver and FW. The RX Equalization
> settings (AFE settings alone) are manually configured and the existing
> link-up sequence in the driver does not perform rx adaptation process as
> mentioned in the Synopsys databook. There's a customer request for 10G
> backplane mode without Auto-negotiation and for the DAC cables of more
> significant length that follow the non-Autonegotiation mode. These modes
> require PHY to perform RX Adaptation.
> 
> The proposed logic adds the necessary changes to Yellow Carp devices to
> ensure seamless RX Adaptation for 10G-SFI (LONG DAC) and 10G-KR without
> AN (CL72 not present). The RX adaptation core algorithm is executed by
> firmware, however, to achieve that a new mailbox sub-command is required
> to be sent by the driver.
> 
> Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>

...

> diff --git a/drivers/net/ethernet/amd/xgbe/xgbe.h b/drivers/net/ethernet/amd/xgbe/xgbe.h
> index 16e73df3e9b9..ad136ed493ed 100644
> --- a/drivers/net/ethernet/amd/xgbe/xgbe.h
> +++ b/drivers/net/ethernet/amd/xgbe/xgbe.h
> @@ -625,6 +625,7 @@ enum xgbe_mb_cmd {
>  
>  enum xgbe_mb_subcmd {
>  	XGBE_MB_SUBCMD_NONE = 0,
> +	XGBE_MB_SUBCMD_RX_ADAP,
>  
>  	/* 10GbE SFP subcommands */
>  	XGBE_MB_SUBCMD_ACTIVE = 0,
> @@ -1316,6 +1317,10 @@ struct xgbe_prv_data {
>  
>  	bool debugfs_an_cdr_workaround;
>  	bool debugfs_an_cdr_track_early;
> +	bool en_rx_adap;

nit: there is a 1 byte hole here (on x86_64)

> +	int rx_adapt_retries;
> +	bool rx_adapt_done;
> +	bool mode_set;
>  };
>  
>  /* Function prototypes*/
> -- 
> 2.25.1
> 

Re: [PATCH net-next v2 2/2] amd-xgbe: add support for rx-adaptation

[Paolo Abeni]

On Wed, 2023-02-01 at 18:11 +0100, Simon Horman wrote:
> On Wed, Feb 01, 2023 at 11:19:32AM +0530, Raju Rangoju wrote:
> > The existing implementation for non-Autonegotiation 10G speed modes does
> > not enable RX adaptation in the Driver and FW. The RX Equalization
> > settings (AFE settings alone) are manually configured and the existing
> > link-up sequence in the driver does not perform rx adaptation process as
> > mentioned in the Synopsys databook. There's a customer request for 10G
> > backplane mode without Auto-negotiation and for the DAC cables of more
> > significant length that follow the non-Autonegotiation mode. These modes
> > require PHY to perform RX Adaptation.
> > 
> > The proposed logic adds the necessary changes to Yellow Carp devices to
> > ensure seamless RX Adaptation for 10G-SFI (LONG DAC) and 10G-KR without
> > AN (CL72 not present). The RX adaptation core algorithm is executed by
> > firmware, however, to achieve that a new mailbox sub-command is required
> > to be sent by the driver.
> > 
> > Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> > Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> > Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
> 
> ...
> 
> > diff --git a/drivers/net/ethernet/amd/xgbe/xgbe.h b/drivers/net/ethernet/amd/xgbe/xgbe.h
> > index 16e73df3e9b9..ad136ed493ed 100644
> > --- a/drivers/net/ethernet/amd/xgbe/xgbe.h
> > +++ b/drivers/net/ethernet/amd/xgbe/xgbe.h
> > @@ -625,6 +625,7 @@ enum xgbe_mb_cmd {
> >  
> >  enum xgbe_mb_subcmd {
> >  	XGBE_MB_SUBCMD_NONE = 0,
> > +	XGBE_MB_SUBCMD_RX_ADAP,
> >  
> >  	/* 10GbE SFP subcommands */
> >  	XGBE_MB_SUBCMD_ACTIVE = 0,
> > @@ -1316,6 +1317,10 @@ struct xgbe_prv_data {
> >  
> >  	bool debugfs_an_cdr_workaround;
> >  	bool debugfs_an_cdr_track_early;
> > +	bool en_rx_adap;
> 
> nit: there is a 1 byte hole here (on x86_64)

I think even in the current form is ok. The total size of the struct is
not going to change, due to alignment, and the fields will sit in the
same cacheline in both cases.

I guess the layout could be changed later if needed.

Thanks,

Paolo

Re: [PATCH net-next v2 2/2] amd-xgbe: add support for rx-adaptation

[Simon Horman]

On Thu, Feb 02, 2023 at 03:16:50PM +0100, Paolo Abeni wrote:
> On Wed, 2023-02-01 at 18:11 +0100, Simon Horman wrote:
> > On Wed, Feb 01, 2023 at 11:19:32AM +0530, Raju Rangoju wrote:
> > > The existing implementation for non-Autonegotiation 10G speed modes does
> > > not enable RX adaptation in the Driver and FW. The RX Equalization
> > > settings (AFE settings alone) are manually configured and the existing
> > > link-up sequence in the driver does not perform rx adaptation process as
> > > mentioned in the Synopsys databook. There's a customer request for 10G
> > > backplane mode without Auto-negotiation and for the DAC cables of more
> > > significant length that follow the non-Autonegotiation mode. These modes
> > > require PHY to perform RX Adaptation.
> > > 
> > > The proposed logic adds the necessary changes to Yellow Carp devices to
> > > ensure seamless RX Adaptation for 10G-SFI (LONG DAC) and 10G-KR without
> > > AN (CL72 not present). The RX adaptation core algorithm is executed by
> > > firmware, however, to achieve that a new mailbox sub-command is required
> > > to be sent by the driver.
> > > 
> > > Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> > > Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> > > Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
> > 
> > ...
> > 
> > > diff --git a/drivers/net/ethernet/amd/xgbe/xgbe.h b/drivers/net/ethernet/amd/xgbe/xgbe.h
> > > index 16e73df3e9b9..ad136ed493ed 100644
> > > --- a/drivers/net/ethernet/amd/xgbe/xgbe.h
> > > +++ b/drivers/net/ethernet/amd/xgbe/xgbe.h
> > > @@ -625,6 +625,7 @@ enum xgbe_mb_cmd {
> > >  
> > >  enum xgbe_mb_subcmd {
> > >  	XGBE_MB_SUBCMD_NONE = 0,
> > > +	XGBE_MB_SUBCMD_RX_ADAP,
> > >  
> > >  	/* 10GbE SFP subcommands */
> > >  	XGBE_MB_SUBCMD_ACTIVE = 0,
> > > @@ -1316,6 +1317,10 @@ struct xgbe_prv_data {
> > >  
> > >  	bool debugfs_an_cdr_workaround;
> > >  	bool debugfs_an_cdr_track_early;
> > > +	bool en_rx_adap;
> > 
> > nit: there is a 1 byte hole here (on x86_64)
> 
> I think even in the current form is ok. The total size of the struct is
> not going to change, due to alignment, and the fields will sit in the
> same cacheline in both cases.
> 
> I guess the layout could be changed later if needed.

Ok, I did think it was worth mentioning.
But I agree that it doesn't need to be changed at this time.

Re: [PATCH net-next v2 2/2] amd-xgbe: add support for rx-adaptation

[Raju Rangoju]

On 2/3/2023 1:39 PM, Simon Horman wrote:
> On Thu, Feb 02, 2023 at 03:16:50PM +0100, Paolo Abeni wrote:
>> On Wed, 2023-02-01 at 18:11 +0100, Simon Horman wrote:
>>> On Wed, Feb 01, 2023 at 11:19:32AM +0530, Raju Rangoju wrote:
>>>> The existing implementation for non-Autonegotiation 10G speed modes does
>>>> not enable RX adaptation in the Driver and FW. The RX Equalization
>>>> settings (AFE settings alone) are manually configured and the existing
>>>> link-up sequence in the driver does not perform rx adaptation process as
>>>> mentioned in the Synopsys databook. There's a customer request for 10G
>>>> backplane mode without Auto-negotiation and for the DAC cables of more
>>>> significant length that follow the non-Autonegotiation mode. These modes
>>>> require PHY to perform RX Adaptation.
>>>>
>>>> The proposed logic adds the necessary changes to Yellow Carp devices to
>>>> ensure seamless RX Adaptation for 10G-SFI (LONG DAC) and 10G-KR without
>>>> AN (CL72 not present). The RX adaptation core algorithm is executed by
>>>> firmware, however, to achieve that a new mailbox sub-command is required
>>>> to be sent by the driver.
>>>>
>>>> Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
>>>> Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
>>>> Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
>>>
>>> ...
>>>
>>>> diff --git a/drivers/net/ethernet/amd/xgbe/xgbe.h b/drivers/net/ethernet/amd/xgbe/xgbe.h
>>>> index 16e73df3e9b9..ad136ed493ed 100644
>>>> --- a/drivers/net/ethernet/amd/xgbe/xgbe.h
>>>> +++ b/drivers/net/ethernet/amd/xgbe/xgbe.h
>>>> @@ -625,6 +625,7 @@ enum xgbe_mb_cmd {
>>>>   
>>>>   enum xgbe_mb_subcmd {
>>>>   	XGBE_MB_SUBCMD_NONE = 0,
>>>> +	XGBE_MB_SUBCMD_RX_ADAP,
>>>>   
>>>>   	/* 10GbE SFP subcommands */
>>>>   	XGBE_MB_SUBCMD_ACTIVE = 0,
>>>> @@ -1316,6 +1317,10 @@ struct xgbe_prv_data {
>>>>   
>>>>   	bool debugfs_an_cdr_workaround;
>>>>   	bool debugfs_an_cdr_track_early;
>>>> +	bool en_rx_adap;
>>>
>>> nit: there is a 1 byte hole here (on x86_64)
>>
>> I think even in the current form is ok. The total size of the struct is
>> not going to change, due to alignment, and the fields will sit in the
>> same cacheline in both cases.
>>
>> I guess the layout could be changed later if needed.
> 
> Ok, I did think it was worth mentioning.
> But I agree that it doesn't need to be changed at this time.

Thanks Simon for taking time to review the patch. I'll try to handle it 
at a later point in time.

Re: [PATCH net-next v2 2/2] amd-xgbe: add support for rx-adaptation

[Simon Horman]

On Fri, Feb 03, 2023 at 01:49:01PM +0530, Raju Rangoju wrote:
> 
> 
> On 2/3/2023 1:39 PM, Simon Horman wrote:
> > On Thu, Feb 02, 2023 at 03:16:50PM +0100, Paolo Abeni wrote:
> > > On Wed, 2023-02-01 at 18:11 +0100, Simon Horman wrote:
> > > > On Wed, Feb 01, 2023 at 11:19:32AM +0530, Raju Rangoju wrote:
> > > > > The existing implementation for non-Autonegotiation 10G speed modes does
> > > > > not enable RX adaptation in the Driver and FW. The RX Equalization
> > > > > settings (AFE settings alone) are manually configured and the existing
> > > > > link-up sequence in the driver does not perform rx adaptation process as
> > > > > mentioned in the Synopsys databook. There's a customer request for 10G
> > > > > backplane mode without Auto-negotiation and for the DAC cables of more
> > > > > significant length that follow the non-Autonegotiation mode. These modes
> > > > > require PHY to perform RX Adaptation.
> > > > > 
> > > > > The proposed logic adds the necessary changes to Yellow Carp devices to
> > > > > ensure seamless RX Adaptation for 10G-SFI (LONG DAC) and 10G-KR without
> > > > > AN (CL72 not present). The RX adaptation core algorithm is executed by
> > > > > firmware, however, to achieve that a new mailbox sub-command is required
> > > > > to be sent by the driver.
> > > > > 
> > > > > Co-developed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> > > > > Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> > > > > Signed-off-by: Raju Rangoju <Raju.Rangoju@amd.com>
> > > > 
> > > > ...
> > > > 
> > > > > diff --git a/drivers/net/ethernet/amd/xgbe/xgbe.h b/drivers/net/ethernet/amd/xgbe/xgbe.h
> > > > > index 16e73df3e9b9..ad136ed493ed 100644
> > > > > --- a/drivers/net/ethernet/amd/xgbe/xgbe.h
> > > > > +++ b/drivers/net/ethernet/amd/xgbe/xgbe.h
> > > > > @@ -625,6 +625,7 @@ enum xgbe_mb_cmd {
> > > > >   enum xgbe_mb_subcmd {
> > > > >   	XGBE_MB_SUBCMD_NONE = 0,
> > > > > +	XGBE_MB_SUBCMD_RX_ADAP,
> > > > >   	/* 10GbE SFP subcommands */
> > > > >   	XGBE_MB_SUBCMD_ACTIVE = 0,
> > > > > @@ -1316,6 +1317,10 @@ struct xgbe_prv_data {
> > > > >   	bool debugfs_an_cdr_workaround;
> > > > >   	bool debugfs_an_cdr_track_early;
> > > > > +	bool en_rx_adap;
> > > > 
> > > > nit: there is a 1 byte hole here (on x86_64)
> > > 
> > > I think even in the current form is ok. The total size of the struct is
> > > not going to change, due to alignment, and the fields will sit in the
> > > same cacheline in both cases.
> > > 
> > > I guess the layout could be changed later if needed.
> > 
> > Ok, I did think it was worth mentioning.
> > But I agree that it doesn't need to be changed at this time.
> 
> Thanks Simon for taking time to review the patch. I'll try to handle it at a
> later point in time.

Thanks, much appreciated.

Re: [PATCH net-next v2 0/2] amd-xgbe: add support for 2.5GbE and rx-adaptation

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net-next.git (master)
by Paolo Abeni <pabeni@redhat.com>:

On Wed, 1 Feb 2023 11:19:30 +0530 you wrote:
> This patch series adds support for 2.GbE in 10GBaseT mode and
> rx-adaptation support for Yellow Carp devices.
> 
> 1) Support for 2.5GbE:
>    Add the necessary changes to the driver to fully recognize and enable
>    2.5GbE speed in 10GBaseT mode.
> 
> [...]

Here is the summary with links:
  - [net-next,v2,1/2] amd-xgbe: add 2.5GbE support to 10G BaseT mode
    https://git.kernel.org/netdev/net-next/c/3ee217c47b8b
  - [net-next,v2,2/2] amd-xgbe: add support for rx-adaptation
    https://git.kernel.org/netdev/net-next/c/4f3b20bfbb75

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html