Add bindings for the MediaTek XFI Ethernet SerDes T-PHY found in the
MediaTek MT7988 SoC which can operate at various interfaces modes:
via USXGMII PCS:
* USXGMII
* 10GBase-R
* 5GBase-R
via LynxI SGMII PCS:
* 2500Base-X
* 1000Base-X
* Cisco SGMII (MAC side)
Signed-off-by: Daniel Golle <[email protected]>
---
v2: unify filename and compatible as requested
.../phy/mediatek,mt7988-xfi-tphy.yaml | 77 +++++++++++++++++++
1 file changed, 77 insertions(+)
create mode 100644 Documentation/devicetree/bindings/phy/mediatek,mt7988-xfi-tphy.yaml
diff --git a/Documentation/devicetree/bindings/phy/mediatek,mt7988-xfi-tphy.yaml b/Documentation/devicetree/bindings/phy/mediatek,mt7988-xfi-tphy.yaml
new file mode 100644
index 0000000000000..e897118dcf7e6
--- /dev/null
+++ b/Documentation/devicetree/bindings/phy/mediatek,mt7988-xfi-tphy.yaml
@@ -0,0 +1,77 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/phy/mediatek,mt7988-xfi-tphy.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: MediaTek XFI T-PHY
+
+maintainers:
+ - Daniel Golle <[email protected]>
+
+description:
+ The MediaTek XFI SerDes T-PHY provides the physical SerDes lanes
+ used by the (10G/5G) USXGMII PCS and (1G/2.5G) LynxI PCS found in
+ MediaTek's 10G-capabale MT7988 SoC.
+
+properties:
+ compatible:
+ const: mediatek,mt7988-xfi-tphy
+
+ reg:
+ maxItems: 1
+
+ clocks:
+ items:
+ - description: XFI PHY clock
+ - description: XFI register clock
+
+ clock-names:
+ items:
+ - const: xfipll
+ - const: topxtal
+
+ resets:
+ items:
+ - description: PEXTP reset
+
+ mediatek,usxgmii-performance-errata:
+ $ref: /schemas/types.yaml#/definitions/flag
+ description:
+ One instance of the T-PHY on MT7988 suffers from a performance
+ problem in 10GBase-R mode which needs a work-around in the driver.
+ The work-around is enabled using this flag.
+
+ "#phy-cells":
+ const: 0
+
+required:
+ - compatible
+ - reg
+ - clocks
+ - clock-names
+ - resets
+ - "#phy-cells"
+
+additionalProperties: false
+
+examples:
+ - |
+ #include <dt-bindings/clock/mediatek,mt7988-clk.h>
+ soc {
+ #address-cells = <2>;
+ #size-cells = <2>;
+
+ phy@11f20000 {
+ compatible = "mediatek,mt7988-xfi-tphy";
+ reg = <0 0x11f20000 0 0x10000>;
+ clocks = <&xfi_pll CLK_XFIPLL_PLL_EN>,
+ <&topckgen CLK_TOP_XFI_PHY_0_XTAL_SEL>;
+ clock-names = "xfipll", "topxtal";
+ resets = <&watchdog 14>;
+ mediatek,usxgmii-performance-errata;
+ #phy-cells = <0>;
+ };
+ };
+
+...
--
2.43.0
Add driver for MediaTek's XFI T-PHY which can be found in the MT7988
SoC. The XFI T-PHY is a 10 Gigabit/s Ethernet SerDes PHY with muxes on
the internal side to be used with either USXGMII PCS or LynxI PCS,
depending on the selected PHY interface mode.
The PHY can operates only in PHY_MODE_ETHERNET, the submode is one of
PHY_INTERFACE_MODE_* corresponding to the supported modes:
* USXGMII \
* 10GBase-R }- USXGMII PCS - XGDM \
* 5GBase-R / \
}- Ethernet MAC
* 2500Base-X \ /
* 1000Base-X }- LynxI PCS - GDM /
* Cisco SGMII (MAC side) /
In order to work-around a performance issue present on the first of
two XFI T-PHYs present in MT7988, special tuning is applied which can be
selected by adding the 'mediatek,usxgmii-performance-errata' property to
the device tree node.
There is no documentation for most registers used for the
analog/tuning part, however, most of the registers have been partially
reverse-engineered from MediaTek's SDK implementation (an opaque
sequence of 32-bit register writes) and descriptions for all relevant
digital registers and bits such as resets and muxes have been supplied
by MediaTek.
Signed-off-by: Daniel Golle <[email protected]>
---
v2:
* use IO helpers from mtk-io.h instead of rolling my own
* use devm_clk_bulk_get()
* yse devm_platform_ioremap_resource()
* unify name and description everywhere
* invert bool is_xgmii into bool use_lynxi_pcs and add comments
describing the meaning of each of the stack variables
* not much we can do about remaining magic values unless MTK provides
definitions for them
MAINTAINERS | 1 +
drivers/phy/mediatek/Kconfig | 12 +
drivers/phy/mediatek/Makefile | 1 +
drivers/phy/mediatek/phy-mtk-xfi-tphy.c | 360 ++++++++++++++++++++++++
4 files changed, 374 insertions(+)
create mode 100644 drivers/phy/mediatek/phy-mtk-xfi-tphy.c
diff --git a/MAINTAINERS b/MAINTAINERS
index ff059e870a444..72270296de2bd 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -13729,6 +13729,7 @@ L: [email protected]
S: Maintained
F: drivers/net/phy/mediatek-ge-soc.c
F: drivers/net/phy/mediatek-ge.c
+F: drivers/phy/mediatek/phy-mtk-xfi-tphy.c
MEDIATEK I2C CONTROLLER DRIVER
M: Qii Wang <[email protected]>
diff --git a/drivers/phy/mediatek/Kconfig b/drivers/phy/mediatek/Kconfig
index 3125ecb5d119f..be2abf642f2f3 100644
--- a/drivers/phy/mediatek/Kconfig
+++ b/drivers/phy/mediatek/Kconfig
@@ -13,6 +13,18 @@ config PHY_MTK_PCIE
callback for PCIe GEN3 port, it supports software efuse
initialization.
+config PHY_MTK_XFI_TPHY
+ tristate "MediaTek 10GE SerDes XFI T-PHY driver"
+ depends on ARCH_MEDIATEK || COMPILE_TEST
+ depends on OF && OF_ADDRESS
+ depends on HAS_IOMEM
+ select GENERIC_PHY
+ help
+ Say 'Y' here to add support for MediaTek XFI T-PHY driver.
+ The driver provides access to the Ethernet SerDes T-PHY supporting
+ 1GE and 2.5GE modes via the LynxI PCS, and 5GE and 10GE modes
+ via the USXGMII PCS found in MediaTek SoCs with 10G Ethernet.
+
config PHY_MTK_TPHY
tristate "MediaTek T-PHY Driver"
depends on ARCH_MEDIATEK || COMPILE_TEST
diff --git a/drivers/phy/mediatek/Makefile b/drivers/phy/mediatek/Makefile
index c9a50395533eb..fa7217178e7f4 100644
--- a/drivers/phy/mediatek/Makefile
+++ b/drivers/phy/mediatek/Makefile
@@ -8,6 +8,7 @@ obj-$(CONFIG_PHY_MTK_PCIE) += phy-mtk-pcie.o
obj-$(CONFIG_PHY_MTK_TPHY) += phy-mtk-tphy.o
obj-$(CONFIG_PHY_MTK_UFS) += phy-mtk-ufs.o
obj-$(CONFIG_PHY_MTK_XSPHY) += phy-mtk-xsphy.o
+obj-$(CONFIG_PHY_MTK_XFI_TPHY) += phy-mtk-xfi-tphy.o
phy-mtk-hdmi-drv-y := phy-mtk-hdmi.o
phy-mtk-hdmi-drv-y += phy-mtk-hdmi-mt2701.o
diff --git a/drivers/phy/mediatek/phy-mtk-xfi-tphy.c b/drivers/phy/mediatek/phy-mtk-xfi-tphy.c
new file mode 100644
index 0000000000000..551d6cee33f94
--- /dev/null
+++ b/drivers/phy/mediatek/phy-mtk-xfi-tphy.c
@@ -0,0 +1,360 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/* MediaTek 10GE SerDes XFI T-PHY driver
+ *
+ * Copyright (c) 2024 Daniel Golle <[email protected]>
+ * Bc-bocun Chen <[email protected]>
+ * based on mtk_usxgmii.c and mtk_sgmii.c found in MediaTek's SDK (GPL-2.0)
+ * Copyright (c) 2022 MediaTek Inc.
+ * Author: Henry Yen <[email protected]>
+ */
+
+#include <linux/module.h>
+#include <linux/device.h>
+#include <linux/platform_device.h>
+#include <linux/of.h>
+#include <linux/io.h>
+#include <linux/clk.h>
+#include <linux/reset.h>
+#include <linux/phy.h>
+#include <linux/phy/phy.h>
+
+#include "phy-mtk-io.h"
+
+#define MTK_XFI_TPHY_NUM_CLOCKS 2
+
+#define REG_DIG_GLB_70 0x0070
+#define XTP_PCS_RX_EQ_IN_PROGRESS(x) FIELD_PREP(GENMASK(25, 24), (x))
+#define XTP_PCS_MODE_MASK GENMASK(17, 16)
+#define XTP_PCS_MODE(x) FIELD_PREP(GENMASK(17, 16), (x))
+#define XTP_PCS_RST_B BIT(15)
+#define XTP_FRC_PCS_RST_B BIT(14)
+#define XTP_PCS_PWD_SYNC_MASK GENMASK(13, 12)
+#define XTP_PCS_PWD_SYNC(x) FIELD_PREP(XTP_PCS_PWD_SYNC_MASK, (x))
+#define XTP_PCS_PWD_ASYNC_MASK GENMASK(11, 10)
+#define XTP_PCS_PWD_ASYNC(x) FIELD_PREP(XTP_PCS_PWD_ASYNC_MASK, (x))
+#define XTP_FRC_PCS_PWD_ASYNC BIT(8)
+#define XTP_PCS_UPDT BIT(4)
+#define XTP_PCS_IN_FR_RG BIT(0)
+
+#define REG_DIG_GLB_F4 0x00f4
+#define XFI_DPHY_PCS_SEL BIT(0)
+#define XFI_DPHY_PCS_SEL_SGMII FIELD_PREP(XFI_DPHY_PCS_SEL, 1)
+#define XFI_DPHY_PCS_SEL_USXGMII FIELD_PREP(XFI_DPHY_PCS_SEL, 0)
+#define XFI_DPHY_AD_SGDT_FRC_EN BIT(5)
+
+#define REG_DIG_LN_TRX_40 0x3040
+#define XTP_LN_FRC_TX_DATA_EN BIT(29)
+#define XTP_LN_TX_DATA_EN BIT(28)
+
+#define REG_DIG_LN_TRX_B0 0x30b0
+#define XTP_LN_FRC_TX_MACCK_EN BIT(5)
+#define XTP_LN_TX_MACCK_EN BIT(4)
+
+#define REG_ANA_GLB_D0 0x90d0
+#define XTP_GLB_USXGMII_SEL_MASK GENMASK(3, 1)
+#define XTP_GLB_USXGMII_SEL(x) FIELD_PREP(GENMASK(3, 1), (x))
+#define XTP_GLB_USXGMII_EN BIT(0)
+
+struct mtk_xfi_tphy {
+ void __iomem *base;
+ struct device *dev;
+ struct reset_control *reset;
+ struct clk_bulk_data clocks[MTK_XFI_TPHY_NUM_CLOCKS];
+ bool da_war;
+};
+
+static void mtk_xfi_tphy_setup(struct mtk_xfi_tphy *xfi_tphy,
+ phy_interface_t interface)
+{
+ /* Override 10GBase-R tuning value if work-around is selected */
+ bool da_war = (xfi_tphy->da_war && (interface == PHY_INTERFACE_MODE_10GBASER));
+ /* Bools to make setting up values for specific PHY speeds easier */
+ bool is_2p5g = (interface == PHY_INTERFACE_MODE_2500BASEX);
+ bool is_1g = (interface == PHY_INTERFACE_MODE_1000BASEX ||
+ interface == PHY_INTERFACE_MODE_SGMII);
+ bool is_10g = (interface == PHY_INTERFACE_MODE_10GBASER ||
+ interface == PHY_INTERFACE_MODE_USXGMII);
+ bool is_5g = (interface == PHY_INTERFACE_MODE_5GBASER);
+ /* Bool to configure input mux to either
+ * - USXGMII PCS (64b/66b coding) for 5G/10G
+ * - LynxI PCS (8b/10b coding) for 1G/2.5G
+ */
+ bool use_lynxi_pcs = (is_1g || is_2p5g);
+
+ dev_dbg(xfi_tphy->dev, "setting up for mode %s\n", phy_modes(interface));
+
+ /* Setup PLL setting */
+ mtk_phy_update_bits(xfi_tphy->base + 0x9024, 0x100000, is_10g ? 0x0 : 0x100000);
+ mtk_phy_update_bits(xfi_tphy->base + 0x2020, 0x202000, is_5g ? 0x202000 : 0x0);
+ mtk_phy_update_bits(xfi_tphy->base + 0x2030, 0x500, is_1g ? 0x0 : 0x500);
+ mtk_phy_update_bits(xfi_tphy->base + 0x2034, 0xa00, is_1g ? 0x0 : 0xa00);
+ mtk_phy_update_bits(xfi_tphy->base + 0x2040, 0x340000, is_1g ? 0x200000 : 0x140000);
+
+ /* Setup RXFE BW setting */
+ mtk_phy_update_bits(xfi_tphy->base + 0x50f0, 0xc10, is_1g ? 0x410 : is_5g ? 0x800 : 0x400);
+ mtk_phy_update_bits(xfi_tphy->base + 0x50e0, 0x4000, is_5g ? 0x0 : 0x4000);
+
+ /* Setup RX CDR setting */
+ mtk_phy_update_bits(xfi_tphy->base + 0x506c, 0x30000, is_5g ? 0x0 : 0x30000);
+ mtk_phy_update_bits(xfi_tphy->base + 0x5070, 0x670000, is_5g ? 0x620000 : 0x50000);
+ mtk_phy_update_bits(xfi_tphy->base + 0x5074, 0x180000, is_5g ? 0x180000 : 0x0);
+ mtk_phy_update_bits(xfi_tphy->base + 0x5078, 0xf000400, is_5g ? 0x8000000 :
+ 0x7000400);
+ mtk_phy_update_bits(xfi_tphy->base + 0x507c, 0x5000500, is_5g ? 0x4000400 :
+ 0x1000100);
+ mtk_phy_update_bits(xfi_tphy->base + 0x5080, 0x1410, is_1g ? 0x400 : is_5g ? 0x1010 : 0x0);
+ mtk_phy_update_bits(xfi_tphy->base + 0x5084, 0x30300, is_1g ? 0x30300 :
+ is_5g ? 0x30100 :
+ 0x100);
+ mtk_phy_update_bits(xfi_tphy->base + 0x5088, 0x60200, is_1g ? 0x20200 :
+ is_5g ? 0x40000 :
+ 0x20000);
+
+ /* Setting RXFE adaptation range setting */
+ mtk_phy_update_bits(xfi_tphy->base + 0x50e4, 0xc0000, is_5g ? 0x0 : 0xc0000);
+ mtk_phy_update_bits(xfi_tphy->base + 0x50e8, 0x40000, is_5g ? 0x0 : 0x40000);
+ mtk_phy_update_bits(xfi_tphy->base + 0x50ec, 0xa00, is_1g ? 0x200 : 0x800);
+ mtk_phy_update_bits(xfi_tphy->base + 0x50a8, 0xee0000, is_5g ? 0x800000 :
+ 0x6e0000);
+ mtk_phy_update_bits(xfi_tphy->base + 0x6004, 0x190000, is_5g ? 0x0 : 0x190000);
+
+ if (is_10g)
+ writel(0x01423342, xfi_tphy->base + 0x00f8);
+ else if (is_5g)
+ writel(0x00a132a1, xfi_tphy->base + 0x00f8);
+ else if (is_2p5g)
+ writel(0x009c329c, xfi_tphy->base + 0x00f8);
+ else
+ writel(0x00fa32fa, xfi_tphy->base + 0x00f8);
+
+ /* Force SGDT_OUT off and select PCS */
+ mtk_phy_update_bits(xfi_tphy->base + REG_DIG_GLB_F4,
+ XFI_DPHY_AD_SGDT_FRC_EN | XFI_DPHY_PCS_SEL,
+ XFI_DPHY_AD_SGDT_FRC_EN |
+ (use_lynxi_pcs ? XFI_DPHY_PCS_SEL_SGMII :
+ XFI_DPHY_PCS_SEL_USXGMII));
+
+ /* Force GLB_CKDET_OUT */
+ mtk_phy_set_bits(xfi_tphy->base + 0x0030, 0xc00);
+
+ /* Force AEQ on */
+ writel(XTP_PCS_RX_EQ_IN_PROGRESS(2) | XTP_PCS_PWD_SYNC(2) | XTP_PCS_PWD_ASYNC(2),
+ xfi_tphy->base + REG_DIG_GLB_70);
+
+ usleep_range(1, 5);
+
+ /* Setup TX DA default value */
+ mtk_phy_update_bits(xfi_tphy->base + 0x30b0, 0x30, 0x20);
+ writel(0x00008a01, xfi_tphy->base + 0x3028);
+ writel(0x0000a884, xfi_tphy->base + 0x302c);
+ writel(0x00083002, xfi_tphy->base + 0x3024);
+
+ /* Setup RG default value */
+ if (use_lynxi_pcs) {
+ writel(0x00011110, xfi_tphy->base + 0x3010);
+ writel(0x40704000, xfi_tphy->base + 0x3048);
+ } else {
+ writel(0x00022220, xfi_tphy->base + 0x3010);
+ writel(0x0f020a01, xfi_tphy->base + 0x5064);
+ writel(0x06100600, xfi_tphy->base + 0x50b4);
+ if (interface == PHY_INTERFACE_MODE_USXGMII)
+ writel(0x40704000, xfi_tphy->base + 0x3048);
+ else
+ writel(0x47684100, xfi_tphy->base + 0x3048);
+ }
+
+ if (is_1g)
+ writel(0x0000c000, xfi_tphy->base + 0x3064);
+
+ /* Setup RX EQ initial value */
+ mtk_phy_update_bits(xfi_tphy->base + 0x3050, 0xa8000000,
+ (interface != PHY_INTERFACE_MODE_10GBASER) ? 0xa8000000 : 0x0);
+ mtk_phy_update_bits(xfi_tphy->base + 0x3054, 0xaa,
+ (interface != PHY_INTERFACE_MODE_10GBASER) ? 0xaa : 0x0);
+
+ if (!use_lynxi_pcs)
+ writel(0x00000f00, xfi_tphy->base + 0x306c);
+ else if (is_2p5g)
+ writel(0x22000f00, xfi_tphy->base + 0x306c);
+ else
+ writel(0x20200f00, xfi_tphy->base + 0x306c);
+
+ mtk_phy_update_bits(xfi_tphy->base + 0xa008, 0x10000, da_war ? 0x10000 : 0x0);
+
+ mtk_phy_update_bits(xfi_tphy->base + 0xa060, 0x50000, use_lynxi_pcs ? 0x50000 : 0x40000);
+
+ /* Setup PHYA speed */
+ mtk_phy_update_bits(xfi_tphy->base + REG_ANA_GLB_D0,
+ XTP_GLB_USXGMII_SEL_MASK | XTP_GLB_USXGMII_EN,
+ is_10g ? XTP_GLB_USXGMII_SEL(0) :
+ is_5g ? XTP_GLB_USXGMII_SEL(1) :
+ is_2p5g ? XTP_GLB_USXGMII_SEL(2) :
+ XTP_GLB_USXGMII_SEL(3));
+ mtk_phy_set_bits(xfi_tphy->base + REG_ANA_GLB_D0, XTP_GLB_USXGMII_EN);
+
+ /* Release reset */
+ mtk_phy_set_bits(xfi_tphy->base + REG_DIG_GLB_70,
+ XTP_PCS_RST_B | XTP_FRC_PCS_RST_B);
+ usleep_range(150, 500);
+
+ /* Switch to P0 */
+ mtk_phy_update_bits(xfi_tphy->base + REG_DIG_GLB_70,
+ XTP_PCS_IN_FR_RG |
+ XTP_FRC_PCS_PWD_ASYNC |
+ XTP_PCS_PWD_ASYNC_MASK |
+ XTP_PCS_PWD_SYNC_MASK |
+ XTP_PCS_UPDT,
+ XTP_PCS_IN_FR_RG |
+ XTP_FRC_PCS_PWD_ASYNC |
+ XTP_PCS_UPDT);
+ usleep_range(1, 5);
+
+ mtk_phy_clear_bits(xfi_tphy->base + REG_DIG_GLB_70, XTP_PCS_UPDT);
+ usleep_range(15, 50);
+
+ if (use_lynxi_pcs) {
+ /* Switch to Gen2 */
+ mtk_phy_update_bits(xfi_tphy->base + REG_DIG_GLB_70,
+ XTP_PCS_MODE_MASK | XTP_PCS_UPDT,
+ XTP_PCS_MODE(1) | XTP_PCS_UPDT);
+ } else {
+ /* Switch to Gen3 */
+ mtk_phy_update_bits(xfi_tphy->base + REG_DIG_GLB_70,
+ XTP_PCS_MODE_MASK | XTP_PCS_UPDT,
+ XTP_PCS_MODE(2) | XTP_PCS_UPDT);
+ }
+ usleep_range(1, 5);
+
+ mtk_phy_clear_bits(xfi_tphy->base + REG_DIG_GLB_70, XTP_PCS_UPDT);
+
+ usleep_range(100, 500);
+
+ /* Enable MAC CK */
+ mtk_phy_set_bits(xfi_tphy->base + REG_DIG_LN_TRX_B0, XTP_LN_TX_MACCK_EN);
+ mtk_phy_clear_bits(xfi_tphy->base + REG_DIG_GLB_F4, XFI_DPHY_AD_SGDT_FRC_EN);
+
+ /* Enable TX data */
+ mtk_phy_set_bits(xfi_tphy->base + REG_DIG_LN_TRX_40,
+ XTP_LN_FRC_TX_DATA_EN | XTP_LN_TX_DATA_EN);
+ usleep_range(400, 1000);
+}
+
+static int mtk_xfi_tphy_set_mode(struct phy *phy, enum phy_mode mode, int
+ submode)
+{
+ struct mtk_xfi_tphy *xfi_tphy = phy_get_drvdata(phy);
+
+ if (mode != PHY_MODE_ETHERNET)
+ return -EINVAL;
+
+ switch (submode) {
+ case PHY_INTERFACE_MODE_1000BASEX:
+ case PHY_INTERFACE_MODE_2500BASEX:
+ case PHY_INTERFACE_MODE_SGMII:
+ case PHY_INTERFACE_MODE_5GBASER:
+ case PHY_INTERFACE_MODE_10GBASER:
+ case PHY_INTERFACE_MODE_USXGMII:
+ mtk_xfi_tphy_setup(xfi_tphy, submode);
+ return 0;
+ default:
+ return -EINVAL;
+ }
+}
+
+static int mtk_xfi_tphy_reset(struct phy *phy)
+{
+ struct mtk_xfi_tphy *xfi_tphy = phy_get_drvdata(phy);
+
+ reset_control_assert(xfi_tphy->reset);
+ usleep_range(100, 500);
+ reset_control_deassert(xfi_tphy->reset);
+ usleep_range(1, 10);
+
+ return 0;
+}
+
+static int mtk_xfi_tphy_power_on(struct phy *phy)
+{
+ struct mtk_xfi_tphy *xfi_tphy = phy_get_drvdata(phy);
+
+ return clk_bulk_prepare_enable(MTK_XFI_TPHY_NUM_CLOCKS, xfi_tphy->clocks);
+}
+
+static int mtk_xfi_tphy_power_off(struct phy *phy)
+{
+ struct mtk_xfi_tphy *xfi_tphy = phy_get_drvdata(phy);
+
+ clk_bulk_disable_unprepare(MTK_XFI_TPHY_NUM_CLOCKS, xfi_tphy->clocks);
+
+ return 0;
+}
+
+static const struct phy_ops mtk_xfi_tphy_ops = {
+ .power_on = mtk_xfi_tphy_power_on,
+ .power_off = mtk_xfi_tphy_power_off,
+ .set_mode = mtk_xfi_tphy_set_mode,
+ .reset = mtk_xfi_tphy_reset,
+ .owner = THIS_MODULE,
+};
+
+static int mtk_xfi_tphy_probe(struct platform_device *pdev)
+{
+ struct device_node *np = pdev->dev.of_node;
+ struct phy_provider *phy_provider;
+ struct mtk_xfi_tphy *xfi_tphy;
+ struct phy *phy;
+ int ret;
+
+ if (!np)
+ return -ENODEV;
+
+ xfi_tphy = devm_kzalloc(&pdev->dev, sizeof(*xfi_tphy), GFP_KERNEL);
+ if (!xfi_tphy)
+ return -ENOMEM;
+
+ xfi_tphy->base = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(xfi_tphy->base))
+ return PTR_ERR(xfi_tphy->base);
+
+ xfi_tphy->dev = &pdev->dev;
+ xfi_tphy->clocks[0].id = "topxtal";
+ xfi_tphy->clocks[1].id = "xfipll";
+ ret = devm_clk_bulk_get(&pdev->dev, MTK_XFI_TPHY_NUM_CLOCKS, xfi_tphy->clocks);
+ if (ret)
+ return ret;
+
+ xfi_tphy->reset = devm_reset_control_get_exclusive(&pdev->dev, NULL);
+ if (IS_ERR(xfi_tphy->reset))
+ return PTR_ERR(xfi_tphy->reset);
+
+ xfi_tphy->da_war = of_property_read_bool(np, "mediatek,usxgmii-performance-errata");
+
+ phy = devm_phy_create(&pdev->dev, NULL, &mtk_xfi_tphy_ops);
+ if (IS_ERR(phy))
+ return PTR_ERR(phy);
+
+ phy_set_drvdata(phy, xfi_tphy);
+ phy_provider = devm_of_phy_provider_register(&pdev->dev, of_phy_simple_xlate);
+
+ return PTR_ERR_OR_ZERO(phy_provider);
+}
+
+static const struct of_device_id mtk_xfi_tphy_match[] = {
+ { .compatible = "mediatek,mt7988-xfi-tphy", },
+ { /* sentinel */ }
+};
+MODULE_DEVICE_TABLE(of, mtk_xfi_tphy_match);
+
+static struct platform_driver mtk_xfi_tphy_driver = {
+ .probe = mtk_xfi_tphy_probe,
+ .driver = {
+ .name = "mtk-xfi-tphy",
+ .of_match_table = mtk_xfi_tphy_match,
+ },
+};
+module_platform_driver(mtk_xfi_tphy_driver);
+
+MODULE_DESCRIPTION("MediaTek 10GE SerDes XFI T-PHY driver");
+MODULE_AUTHOR("Daniel Golle <[email protected]>");
+MODULE_AUTHOR("Bc-bocun Chen <[email protected]>");
+MODULE_LICENSE("GPL");
--
2.43.0
Il 05/02/24 18:29, Daniel Golle ha scritto:
> Add driver for MediaTek's XFI T-PHY which can be found in the MT7988
> SoC. The XFI T-PHY is a 10 Gigabit/s Ethernet SerDes PHY with muxes on
> the internal side to be used with either USXGMII PCS or LynxI PCS,
> depending on the selected PHY interface mode.
>
> The PHY can operates only in PHY_MODE_ETHERNET, the submode is one of
> PHY_INTERFACE_MODE_* corresponding to the supported modes:
>
> * USXGMII \
> * 10GBase-R }- USXGMII PCS - XGDM \
> * 5GBase-R / \
> }- Ethernet MAC
> * 2500Base-X \ /
> * 1000Base-X }- LynxI PCS - GDM /
> * Cisco SGMII (MAC side) /
>
> In order to work-around a performance issue present on the first of
> two XFI T-PHYs present in MT7988, special tuning is applied which can be
> selected by adding the 'mediatek,usxgmii-performance-errata' property to
> the device tree node.
>
> There is no documentation for most registers used for the
> analog/tuning part, however, most of the registers have been partially
> reverse-engineered from MediaTek's SDK implementation (an opaque
> sequence of 32-bit register writes) and descriptions for all relevant
> digital registers and bits such as resets and muxes have been supplied
> by MediaTek.
>
> Signed-off-by: Daniel Golle <[email protected]>
Reviewed-by: AngeloGioacchino Del Regno <[email protected]>
Il 05/02/24 18:28, Daniel Golle ha scritto:
> Add bindings for the MediaTek XFI Ethernet SerDes T-PHY found in the
> MediaTek MT7988 SoC which can operate at various interfaces modes:
>
> via USXGMII PCS:
> * USXGMII
> * 10GBase-R
> * 5GBase-R
>
> via LynxI SGMII PCS:
> * 2500Base-X
> * 1000Base-X
> * Cisco SGMII (MAC side)
>
> Signed-off-by: Daniel Golle <[email protected]>
Reviewed-by: AngeloGioacchino Del Regno <[email protected]>
On 05/02/2024 18:28, Daniel Golle wrote:
> Add bindings for the MediaTek XFI Ethernet SerDes T-PHY found in the
> MediaTek MT7988 SoC which can operate at various interfaces modes:
>
> via USXGMII PCS:
> * USXGMII
> * 10GBase-R
> * 5GBase-R
>
> via LynxI SGMII PCS:
> * 2500Base-X
> * 1000Base-X
> * Cisco SGMII (MAC side)
>
> Signed-off-by: Daniel Golle <[email protected]>
> ---
> v2: unify filename and compatible as requested
Several comments, from me and Rob, were ignored. Please respond to them.
Best regards,
Krzysztof
Hi Krzysztof,
On Tue, Feb 06, 2024 at 11:53:55AM +0100, Krzysztof Kozlowski wrote:
> On 05/02/2024 18:28, Daniel Golle wrote:
> > Add bindings for the MediaTek XFI Ethernet SerDes T-PHY found in the
> > MediaTek MT7988 SoC which can operate at various interfaces modes:
> >
> > via USXGMII PCS:
> > * USXGMII
> > * 10GBase-R
> > * 5GBase-R
> >
> > via LynxI SGMII PCS:
> > * 2500Base-X
> > * 1000Base-X
> > * Cisco SGMII (MAC side)
> >
> > Signed-off-by: Daniel Golle <[email protected]>
> > ---
> > v2: unify filename and compatible as requested
>
> Several comments, from me and Rob, were ignored. Please respond to them.
I'm sorry if I have missed something. I just checked again on
patchwork, just in case I would have missed an email reply to this or
any of the preceding posts of this patch as part of the old series
going to netdev.
Comments you have made which I have addressed:
- removed $nodename
- use compatible as filename
And the only thing I found that I didn't either fix or reply to is this:
> Can you explain what is this issue and errata about (except performance)?
Not overwriting that (undocumented) value in that (undocumented)
register results in 10GBase-R having performance issues according to a
commit in MediaTek's SDK, see here:
https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/a500d94cd%5E%21/#F0
Maybe Bc or SkyLake of MediaTek (added to Cc) can explain this in more
detail?
What I did miss was Rob's comment at the very bottom of this reply:
> What is PEXTP?
I can again only answer by referencing to MediaTek's SDK sources:
https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/refs/heads/master/21.02/files/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_sgmii.c#96
Here this reset is called XFI_PEXTP0_GRST.
I personally find that name confusing (as this PHY has nothing to do with
_P_ci _EX_press) and have tried to get rid of it where it isn't either part
of official documentation or already merged drivers (like Sam's clock driver).
If there have been any other issues with this patch which I'm not aware
of, please point them out to me.
Thank you
Best regards
Daniel
Hi Daniel,
Hi Krzysztof,
On Tue, 2024-02-06 at 15:31 +0000, Daniel Golle wrote:
>
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
> Hi Krzysztof,
>
> On Tue, Feb 06, 2024 at 11:53:55AM +0100, Krzysztof Kozlowski wrote:
> > On 05/02/2024 18:28, Daniel Golle wrote:
> > > Add bindings for the MediaTek XFI Ethernet SerDes T-PHY found in
> the
> > > MediaTek MT7988 SoC which can operate at various interfaces
> modes:
> > >
> > > via USXGMII PCS:
> > > * USXGMII
> > > * 10GBase-R
> > > * 5GBase-R
> > >
> > > via LynxI SGMII PCS:
> > > * 2500Base-X
> > > * 1000Base-X
> > > * Cisco SGMII (MAC side)
> > >
> > > Signed-off-by: Daniel Golle <[email protected]>
> > > ---
> > > v2: unify filename and compatible as requested
> >
> > Several comments, from me and Rob, were ignored. Please respond to
> them.
>
> I'm sorry if I have missed something. I just checked again on
> patchwork, just in case I would have missed an email reply to this or
> any of the preceding posts of this patch as part of the old series
> going to netdev.
>
> Comments you have made which I have addressed:
> - removed $nodename
> - use compatible as filename
>
> And the only thing I found that I didn't either fix or reply to is
> this:
> > Can you explain what is this issue and errata about (except
> performance)?
>
> Not overwriting that (undocumented) value in that (undocumented)
> register results in 10GBase-R having performance issues according to
> a
> commit in MediaTek's SDK, see here:
>
>
https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/a500d94cd%5E%21/#F0
>
> Maybe Bc or SkyLake of MediaTek (added to Cc) can explain this in
> more
> detail?
Our QA team has identified a potential performance issue when
connecting XFI Port0 with an SFP optical fiber, which may violate the
SFP specification.
Therefore, we have adjusted the PHYA setting for this port.
>
> What I did miss was Rob's comment at the very bottom of this reply:
> > What is PEXTP?
>
> I can again only answer by referencing to MediaTek's SDK sources:
>
>
https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/refs/heads/master/21.02/files/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_sgmii.c#96
>
> Here this reset is called XFI_PEXTP0_GRST.
>
> I personally find that name confusing (as this PHY has nothing to do
> with
> _P_ci _EX_press) and have tried to get rid of it where it isn't
> either part
> of official documentation or already merged drivers (like Sam's clock
> driver).
>
> If there have been any other issues with this patch which I'm not
> aware
> of, please point them out to me.
>
>
> Thank you
>
>
> Best regards
>
>
> Daniel
Thanks
BC
On 06/02/2024 16:31, Daniel Golle wrote:
> Hi Krzysztof,
>
> On Tue, Feb 06, 2024 at 11:53:55AM +0100, Krzysztof Kozlowski wrote:
>> On 05/02/2024 18:28, Daniel Golle wrote:
>>> Add bindings for the MediaTek XFI Ethernet SerDes T-PHY found in the
>>> MediaTek MT7988 SoC which can operate at various interfaces modes:
>>>
>>> via USXGMII PCS:
>>> * USXGMII
>>> * 10GBase-R
>>> * 5GBase-R
>>>
>>> via LynxI SGMII PCS:
>>> * 2500Base-X
>>> * 1000Base-X
>>> * Cisco SGMII (MAC side)
>>>
>>> Signed-off-by: Daniel Golle <[email protected]>
>>> ---
>>> v2: unify filename and compatible as requested
>>
>> Several comments, from me and Rob, were ignored. Please respond to them.
>
> I'm sorry if I have missed something. I just checked again on
> patchwork, just in case I would have missed an email reply to this or
> any of the preceding posts of this patch as part of the old series
> going to netdev.
>
> Comments you have made which I have addressed:
> - removed $nodename
> - use compatible as filename
>
> And the only thing I found that I didn't either fix or reply to is this:
>> Can you explain what is this issue and errata about (except performance)?
>
> Not overwriting that (undocumented) value in that (undocumented)
> register results in 10GBase-R having performance issues according to a
> commit in MediaTek's SDK, see here:
>
> https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/a500d94cd%5E%21/#F0
>
> Maybe Bc or SkyLake of MediaTek (added to Cc) can explain this in more
> detail?
>
>
> What I did miss was Rob's comment at the very bottom of this reply:
>> What is PEXTP?
>
> I can again only answer by referencing to MediaTek's SDK sources:
>
> https://git01.mediatek.com/plugins/gitiles/openwrt/feeds/mtk-openwrt-feeds/+/refs/heads/master/21.02/files/target/linux/mediatek/files-5.4/drivers/net/ethernet/mediatek/mtk_sgmii.c#96
>
> Here this reset is called XFI_PEXTP0_GRST.
>
> I personally find that name confusing (as this PHY has nothing to do with
> _P_ci _EX_press) and have tried to get rid of it where it isn't either part
> of official documentation or already merged drivers (like Sam's clock driver).
>
> If there have been any other issues with this patch which I'm not aware
> of, please point them out to me.
These both cases should be explained in the binding somehow.
Best regards,
Krzysztof