[PATCH net-next v5 0/4] net: dsa: mxl862xx: SerDes ports

[PATCH net-next v5 0/4] net: dsa: mxl862xx: SerDes ports

[Daniel Golle]

Add support for the two SerDes PCS interfaces of the MxL862xx switch
ICs, which can both either be used to connect PHYs or SFP cages, or as
CPU port(s). 1000Base-X, 2500Base-X, 10GBase-R, 10GBase-KR, SGMII,
QSGMII and USXGMII (single 10G or quad 2.5G) are supported.

The firmware only added the API to directly control the PCS as of
version 1.0.84, so the PCS features are gated behind a version check.

As the driver is growing do some refactoring to break out the phylink
parts into mxl862xx-phylink.h.
---
Changes since v4:
 * use FIELD_GET/FIELD_PREP and macro definitions for the bitfields
   instead of endian-aware structs with bit-sized members
 * do not error out on old firmware, so driver at least probes and
   CPU port keeps working in firmware-configured mode. Issue a
   warning instead.

Changes since v3:
 * replace atomic_t serdes_refcount with a plain int guarded by a new
   serdes_lock mutex

Changes since v2:
 * get rid of endian-specific union handling firmware version
 * replace serdes_active bitmap with atomic_t serdes_refcount
 * defer mpcs->interface assignment until after firmware ack
 * handle firmware error codes in pcs_config
 * set st.usx_lane_mode in pcs_get_state
 * set lu.usx_subport and lu.usx_lane_mode in pcs_link_up
 * use phylink_mii_c22_pcs_encode_advertisement() in CL37 adv
 * rework commit message

Changes since v1:
 * drop custom ethtool stats (former patch 5/5)
 * add __{LE,BE}_BITFIELD layouts to ABI structs
 * per-sub-port QSGMII AN restart via usx_subport / usx_lane_mode
 * shared-SerDes refcount in pcs_disable via per-XPCS slot bitmap
 * let every sub-port call pcs_config
 * cache phy_interface_t instead of firmware type
 * skip pcs_link_up when inband-AN enabled
 * gate phylink_get_caps SerDes modes on same FW version as select_pcs
 * interpret xpcs_pcs_cfg.result as signed (s16)
 * drop dead MXL862XX_PCS_PORT macro
 * drop misleading "downshift detection" line from commit message

Daniel Golle (4):
  net: dsa: mxl862xx: store firmware version for feature gating
  net: dsa: mxl862xx: move phylink stubs to mxl862xx-phylink.c
  net: dsa: mxl862xx: move API macros to mxl862xx-host.h
  net: dsa: mxl862xx: add support for SerDes ports

 drivers/net/dsa/mxl862xx/Makefile           |   2 +-
 drivers/net/dsa/mxl862xx/mxl862xx-api.h     | 215 ++++++++++
 drivers/net/dsa/mxl862xx/mxl862xx-cmd.h     |   9 +
 drivers/net/dsa/mxl862xx/mxl862xx-host.h    |   8 +
 drivers/net/dsa/mxl862xx/mxl862xx-phylink.c | 433 ++++++++++++++++++++
 drivers/net/dsa/mxl862xx/mxl862xx-phylink.h |  21 +
 drivers/net/dsa/mxl862xx/mxl862xx.c         |  55 +--
 drivers/net/dsa/mxl862xx/mxl862xx.h         |  57 +++
 8 files changed, 754 insertions(+), 46 deletions(-)
 create mode 100644 drivers/net/dsa/mxl862xx/mxl862xx-phylink.c
 create mode 100644 drivers/net/dsa/mxl862xx/mxl862xx-phylink.h

-- 
2.54.0

[PATCH net-next v5 1/4] net: dsa: mxl862xx: store firmware version for feature gating

[Daniel Golle]

Query the firmware version at init (already done in wait_ready),
cache it in priv->fw_version, and provide MXL862XX_FW_VER_MIN()
for version-gated code paths throughout the driver.

MXL862XX_FW_VER() packs major/minor/revision into a u32 with
bitwise shifts so that versions compare with natural ordering,
independent of host endianness.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
---
v5: no changes
v4: no changes
v3: use bitwise shifts in macro instead of endian-specific union
v2: no changes

 drivers/net/dsa/mxl862xx/mxl862xx.c |  3 +++
 drivers/net/dsa/mxl862xx/mxl862xx.h | 23 +++++++++++++++++++++++
 2 files changed, 26 insertions(+)

diff --git a/drivers/net/dsa/mxl862xx/mxl862xx.c b/drivers/net/dsa/mxl862xx/mxl862xx.c
index b60482d93a85..2f22adedfbf6 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx.c
+++ b/drivers/net/dsa/mxl862xx/mxl862xx.c
@@ -257,6 +257,9 @@ static int mxl862xx_wait_ready(struct dsa_switch *ds)
 			 ver.iv_major, ver.iv_minor,
 			 le16_to_cpu(ver.iv_revision),
 			 le32_to_cpu(ver.iv_build_num));
+		priv->fw_version.major = ver.iv_major;
+		priv->fw_version.minor = ver.iv_minor;
+		priv->fw_version.revision = le16_to_cpu(ver.iv_revision);
 		return 0;
 
 not_ready_yet:
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx.h b/drivers/net/dsa/mxl862xx/mxl862xx.h
index 80053ab40e4c..e3db3711b245 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx.h
+++ b/drivers/net/dsa/mxl862xx/mxl862xx.h
@@ -3,6 +3,7 @@
 #ifndef __MXL862XX_H
 #define __MXL862XX_H
 
+#include <asm/byteorder.h>
 #include <linux/mdio.h>
 #include <linux/workqueue.h>
 #include <net/dsa.h>
@@ -241,6 +242,25 @@ struct mxl862xx_port {
 	spinlock_t stats_lock; /* protects stats accumulators */
 };
 
+/**
+ * struct mxl862xx_fw_version - firmware version for comparison and display
+ * @major: firmware major version
+ * @minor: firmware minor version
+ * @revision: firmware revision number
+ */
+struct mxl862xx_fw_version {
+	u8 major;
+	u8 minor;
+	u16 revision;
+};
+
+#define MXL862XX_FW_VER(maj, min, rev) \
+	(((u32)(maj) << 24) | ((u32)(min) << 16) | (rev))
+#define MXL862XX_FW_VER_MIN(priv, maj, min, rev) \
+	(MXL862XX_FW_VER((priv)->fw_version.major, (priv)->fw_version.minor, \
+			 (priv)->fw_version.revision) >= \
+	 MXL862XX_FW_VER(maj, min, rev))
+
 /* Bit indices for struct mxl862xx_priv::flags */
 #define MXL862XX_FLAG_CRC_ERR		0
 #define MXL862XX_FLAG_WORK_STOPPED	1
@@ -258,6 +278,8 @@ struct mxl862xx_port {
  * @drop_meter:         index of the single shared zero-rate firmware meter
  *                      used to unconditionally drop traffic (used to block
  *                      flooding)
+ * @fw_version:         cached firmware version, populated at probe and
+ *                      compared with MXL862XX_FW_VER_MIN()
  * @ports:              per-port state, indexed by switch port number
  * @bridges:            maps DSA bridge number to firmware bridge ID;
  *                      zero means no firmware bridge allocated for that
@@ -275,6 +297,7 @@ struct mxl862xx_priv {
 	struct work_struct crc_err_work;
 	unsigned long flags;
 	u16 drop_meter;
+	struct mxl862xx_fw_version fw_version;
 	struct mxl862xx_port ports[MXL862XX_MAX_PORTS];
 	u16 bridges[MXL862XX_MAX_BRIDGES + 1];
 	u16 evlan_ingress_size;
-- 
2.54.0

[PATCH net-next v5 2/4] net: dsa: mxl862xx: move phylink stubs to mxl862xx-phylink.c

[Daniel Golle]

Move the phylink MAC operations and get_caps callback from mxl862xx.c
into a dedicated mxl862xx-phylink.c file. This prepares for the SerDes
PCS implementation which adds substantial phylink/PCS code -- keeping
it in a separate file avoids function-position churn in the main
driver file.

No functional change.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
---
v5: no changes
v4: no changes
v3: no changes
v2: no changes

 drivers/net/dsa/mxl862xx/Makefile           |  2 +-
 drivers/net/dsa/mxl862xx/mxl862xx-phylink.c | 51 +++++++++++++++++++++
 drivers/net/dsa/mxl862xx/mxl862xx-phylink.h | 14 ++++++
 drivers/net/dsa/mxl862xx/mxl862xx.c         | 38 +--------------
 4 files changed, 67 insertions(+), 38 deletions(-)
 create mode 100644 drivers/net/dsa/mxl862xx/mxl862xx-phylink.c
 create mode 100644 drivers/net/dsa/mxl862xx/mxl862xx-phylink.h

diff --git a/drivers/net/dsa/mxl862xx/Makefile b/drivers/net/dsa/mxl862xx/Makefile
index d23dd3cd511d..a7be0e6669df 100644
--- a/drivers/net/dsa/mxl862xx/Makefile
+++ b/drivers/net/dsa/mxl862xx/Makefile
@@ -1,3 +1,3 @@
 # SPDX-License-Identifier: GPL-2.0
 obj-$(CONFIG_NET_DSA_MXL862) += mxl862xx_dsa.o
-mxl862xx_dsa-y := mxl862xx.o mxl862xx-host.o
+mxl862xx_dsa-y := mxl862xx.o mxl862xx-host.o mxl862xx-phylink.o
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx-phylink.c b/drivers/net/dsa/mxl862xx/mxl862xx-phylink.c
new file mode 100644
index 000000000000..f17c429d1f1d
--- /dev/null
+++ b/drivers/net/dsa/mxl862xx/mxl862xx-phylink.c
@@ -0,0 +1,51 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Phylink and PCS support for MaxLinear MxL862xx switch family
+ *
+ * Copyright (C) 2024 MaxLinear Inc.
+ * Copyright (C) 2025 John Crispin <john@phrozen.org>
+ * Copyright (C) 2025 Daniel Golle <daniel@makrotopia.org>
+ */
+
+#include <linux/phylink.h>
+#include <net/dsa.h>
+
+#include "mxl862xx.h"
+#include "mxl862xx-phylink.h"
+
+void mxl862xx_phylink_get_caps(struct dsa_switch *ds, int port,
+			       struct phylink_config *config)
+{
+	config->mac_capabilities = MAC_ASYM_PAUSE | MAC_SYM_PAUSE | MAC_10 |
+				   MAC_100 | MAC_1000 | MAC_2500FD;
+
+	__set_bit(PHY_INTERFACE_MODE_INTERNAL,
+		  config->supported_interfaces);
+}
+
+static void mxl862xx_phylink_mac_config(struct phylink_config *config,
+					unsigned int mode,
+					const struct phylink_link_state *state)
+{
+}
+
+static void mxl862xx_phylink_mac_link_down(struct phylink_config *config,
+					   unsigned int mode,
+					   phy_interface_t interface)
+{
+}
+
+static void mxl862xx_phylink_mac_link_up(struct phylink_config *config,
+					 struct phy_device *phydev,
+					 unsigned int mode,
+					 phy_interface_t interface,
+					 int speed, int duplex,
+					 bool tx_pause, bool rx_pause)
+{
+}
+
+const struct phylink_mac_ops mxl862xx_phylink_mac_ops = {
+	.mac_config = mxl862xx_phylink_mac_config,
+	.mac_link_down = mxl862xx_phylink_mac_link_down,
+	.mac_link_up = mxl862xx_phylink_mac_link_up,
+};
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx-phylink.h b/drivers/net/dsa/mxl862xx/mxl862xx-phylink.h
new file mode 100644
index 000000000000..c3d5215bdf60
--- /dev/null
+++ b/drivers/net/dsa/mxl862xx/mxl862xx-phylink.h
@@ -0,0 +1,14 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+
+#ifndef __MXL862XX_PHYLINK_H
+#define __MXL862XX_PHYLINK_H
+
+#include <linux/phylink.h>
+
+#include "mxl862xx.h"
+
+extern const struct phylink_mac_ops mxl862xx_phylink_mac_ops;
+void mxl862xx_phylink_get_caps(struct dsa_switch *ds, int port,
+			       struct phylink_config *config);
+
+#endif /* __MXL862XX_PHYLINK_H */
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx.c b/drivers/net/dsa/mxl862xx/mxl862xx.c
index 2f22adedfbf6..a193f3c07d35 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx.c
+++ b/drivers/net/dsa/mxl862xx/mxl862xx.c
@@ -22,6 +22,7 @@
 #include "mxl862xx-api.h"
 #include "mxl862xx-cmd.h"
 #include "mxl862xx-host.h"
+#include "mxl862xx-phylink.h"
 
 #define MXL862XX_API_WRITE(dev, cmd, data) \
 	mxl862xx_api_wrap(dev, cmd, &(data), sizeof((data)), false, false)
@@ -1424,16 +1425,6 @@ static void mxl862xx_port_teardown(struct dsa_switch *ds, int port)
 	priv->ports[port].setup_done = false;
 }
 
-static void mxl862xx_phylink_get_caps(struct dsa_switch *ds, int port,
-				      struct phylink_config *config)
-{
-	config->mac_capabilities = MAC_ASYM_PAUSE | MAC_SYM_PAUSE | MAC_10 |
-				   MAC_100 | MAC_1000 | MAC_2500FD;
-
-	__set_bit(PHY_INTERFACE_MODE_INTERNAL,
-		  config->supported_interfaces);
-}
-
 static int mxl862xx_get_fid(struct dsa_switch *ds, struct dsa_db db)
 {
 	struct mxl862xx_priv *priv = ds->priv;
@@ -2099,33 +2090,6 @@ static const struct dsa_switch_ops mxl862xx_switch_ops = {
 	.get_stats64 = mxl862xx_get_stats64,
 };
 
-static void mxl862xx_phylink_mac_config(struct phylink_config *config,
-					unsigned int mode,
-					const struct phylink_link_state *state)
-{
-}
-
-static void mxl862xx_phylink_mac_link_down(struct phylink_config *config,
-					   unsigned int mode,
-					   phy_interface_t interface)
-{
-}
-
-static void mxl862xx_phylink_mac_link_up(struct phylink_config *config,
-					 struct phy_device *phydev,
-					 unsigned int mode,
-					 phy_interface_t interface,
-					 int speed, int duplex,
-					 bool tx_pause, bool rx_pause)
-{
-}
-
-static const struct phylink_mac_ops mxl862xx_phylink_mac_ops = {
-	.mac_config = mxl862xx_phylink_mac_config,
-	.mac_link_down = mxl862xx_phylink_mac_link_down,
-	.mac_link_up = mxl862xx_phylink_mac_link_up,
-};
-
 static int mxl862xx_probe(struct mdio_device *mdiodev)
 {
 	struct device *dev = &mdiodev->dev;
-- 
2.54.0

[PATCH net-next v5 3/4] net: dsa: mxl862xx: move API macros to mxl862xx-host.h

[Daniel Golle]

Move the MXL862XX_API_WRITE, MXL862XX_API_READ and
MXL862XX_API_READ_QUIET convenience macros from mxl862xx.c to
mxl862xx-host.h next to the mxl862xx_api_wrap() prototype they wrap.
This makes them available to other compilation units that include
mxl862xx-host.h, which is needed once the SerDes PCS code in
mxl862xx-phylink.c also calls firmware commands.

No functional change.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
---
v5: no changes
v4: no changes
v3: no changes
v2: no changes

 drivers/net/dsa/mxl862xx/mxl862xx-host.h | 8 ++++++++
 drivers/net/dsa/mxl862xx/mxl862xx.c      | 7 -------
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/drivers/net/dsa/mxl862xx/mxl862xx-host.h b/drivers/net/dsa/mxl862xx/mxl862xx-host.h
index 84512a30bc18..66d6ae198aff 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx-host.h
+++ b/drivers/net/dsa/mxl862xx/mxl862xx-host.h
@@ -9,6 +9,14 @@ void mxl862xx_host_init(struct mxl862xx_priv *priv);
 void mxl862xx_host_shutdown(struct mxl862xx_priv *priv);
 int mxl862xx_api_wrap(struct mxl862xx_priv *priv, u16 cmd, void *data, u16 size,
 		      bool read, bool quiet);
+
+#define MXL862XX_API_WRITE(dev, cmd, data) \
+	mxl862xx_api_wrap(dev, cmd, &(data), sizeof((data)), false, false)
+#define MXL862XX_API_READ(dev, cmd, data) \
+	mxl862xx_api_wrap(dev, cmd, &(data), sizeof((data)), true, false)
+#define MXL862XX_API_READ_QUIET(dev, cmd, data) \
+	mxl862xx_api_wrap(dev, cmd, &(data), sizeof((data)), true, true)
+
 int mxl862xx_reset(struct mxl862xx_priv *priv);
 
 #endif /* __MXL862XX_HOST_H */
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx.c b/drivers/net/dsa/mxl862xx/mxl862xx.c
index a193f3c07d35..0b1a23364eb5 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx.c
+++ b/drivers/net/dsa/mxl862xx/mxl862xx.c
@@ -24,13 +24,6 @@
 #include "mxl862xx-host.h"
 #include "mxl862xx-phylink.h"
 
-#define MXL862XX_API_WRITE(dev, cmd, data) \
-	mxl862xx_api_wrap(dev, cmd, &(data), sizeof((data)), false, false)
-#define MXL862XX_API_READ(dev, cmd, data) \
-	mxl862xx_api_wrap(dev, cmd, &(data), sizeof((data)), true, false)
-#define MXL862XX_API_READ_QUIET(dev, cmd, data) \
-	mxl862xx_api_wrap(dev, cmd, &(data), sizeof((data)), true, true)
-
 /* Polling interval for RMON counter accumulation. At 2.5 Gbps with
  * minimum-size (64-byte) frames, a 32-bit packet counter wraps in ~880s.
  * 2s gives a comfortable margin.
-- 
2.54.0

[PATCH net-next v5 4/4] net: dsa: mxl862xx: add support for SerDes ports

[Daniel Golle]

The MxL862xx has two XPCS/SerDes interfaces (XPCS0 for ports 9-12,
XPCS1 for ports 13-16). Each can operate in various single-lane modes
(SGMII, 1000Base-X, 2500Base-X, 10GBase-R, 10GBase-KR, USXGMII) or as
QSGMII or 10G_QXGMII providing four sub-ports per interface.

Implement phylink PCS operations using the firmware's XPCS API:

  - pcs_config: configure negotiation mode and CL37/SGMII advertising.
  - pcs_get_state: read link state and the link-partner ability word
    from firmware and decode using phylink's standard CL37, SGMII, and
    USXGMII decoders.
  - pcs_an_restart: restart CL37 or CL73 auto-negotiation.
  - pcs_link_up: force speed/duplex for SGMII.
  - pcs_inband_caps: report per-mode in-band status capabilities.

Register a PCS instance for each SerDes interface and
QSGMII/10G_QXGMII sub-ports during setup. Advertise the supported
interface modes in phylink_get_caps based on port number.

Lacking support for expressing PHY-side role modes in Linux only the
MAC-side of SGMII, QSGMII, USXGMII and 10G_QXGMII are implemented for
now.

Signed-off-by: Daniel Golle <daniel@makrotopia.org>
---
v5:
 * use FIELD_GET/FIELD_PREP and macro definitions for the bitfields
   instead of endian-aware structs with bit-sized members
 * do not error out on old firmware, so driver at least probes and
   CPU port keeps working in firmware-configured mode. Issue a
   warning instead.

v4:
 * replace atomic_t serdes_refcount with a plain int guarded by a new
   serdes_lock mutex; pcs_disable now holds the lock across the count
   and the XPCS power-down so a sibling sub-port enable cannot race the
   transition to zero (the atomic only made the counter safe, not the
   decision-and-act)

v3:
 * replace serdes_active bitmap with atomic_t serdes_refcount
 * defer mpcs->interface assignment until after firmware ack
 * handle firmware error codes in pcs_config
 * set st.usx_lane_mode in pcs_get_state
 * set lu.usx_subport and lu.usx_lane_mode in pcs_link_up
 * use phylink_mii_c22_pcs_encode_advertisement() in CL37 adv
 * rework commit message

v2:
 * add __{LE,BE}_BITFIELD layouts to ABI structs
 * per-sub-port QSGMII AN restart via usx_subport / usx_lane_mode
 * shared-SerDes refcount in pcs_disable via per-XPCS slot bitmap
 * let every sub-port call pcs_config
 * cache phy_interface_t instead of firmware type
 * skip pcs_link_up when inband-AN enabled
 * gate phylink_get_caps SerDes modes on same FW version as select_pcs
 * interpret xpcs_pcs_cfg.result as signed (s16)
 * drop dead MXL862XX_PCS_PORT macro
 * drop misleading "downshift detection" line from commit message

 drivers/net/dsa/mxl862xx/mxl862xx-api.h     | 215 +++++++++++
 drivers/net/dsa/mxl862xx/mxl862xx-cmd.h     |   9 +
 drivers/net/dsa/mxl862xx/mxl862xx-phylink.c | 386 +++++++++++++++++++-
 drivers/net/dsa/mxl862xx/mxl862xx-phylink.h |   7 +
 drivers/net/dsa/mxl862xx/mxl862xx.c         |   7 +-
 drivers/net/dsa/mxl862xx/mxl862xx.h         |  34 ++
 6 files changed, 655 insertions(+), 3 deletions(-)

diff --git a/drivers/net/dsa/mxl862xx/mxl862xx-api.h b/drivers/net/dsa/mxl862xx/mxl862xx-api.h
index fb21ddc1bf1c..a180a5decffc 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx-api.h
+++ b/drivers/net/dsa/mxl862xx/mxl862xx-api.h
@@ -1366,4 +1366,219 @@ struct mxl862xx_rmon_port_cnt {
 	__le64 tx_good_bytes;
 } __packed;
 
+/* XPCS interface mode, MXL862XX_XPCS_*_INTERFACE field values */
+#define MXL862XX_XPCS_IF_SGMII		0
+#define MXL862XX_XPCS_IF_1000BASEX	1
+#define MXL862XX_XPCS_IF_2500BASEX	2
+#define MXL862XX_XPCS_IF_USXGMII	3	/* single or quad */
+#define MXL862XX_XPCS_IF_10GBASER	4
+#define MXL862XX_XPCS_IF_10GKR		5	/* 10GBASE-KR */
+#define MXL862XX_XPCS_IF_5GBASER	6
+#define MXL862XX_XPCS_IF_QSGMII		7
+
+/* PCS negotiation mode, MXL862XX_XPCS_CFG_NEG_MODE field values */
+#define MXL862XX_XPCS_NEG_NONE		0	/* no inband negotiation */
+#define MXL862XX_XPCS_NEG_INBAND_AN_OFF	1	/* inband, AN disabled */
+#define MXL862XX_XPCS_NEG_INBAND_AN_ON	2	/* inband, AN enabled */
+
+/*
+ * PCS protocol role, MXL862XX_XPCS_CFG_ROLE field value. Selects the role
+ * the XPCS plays in protocols with an asymmetric AN code word (Cisco SGMII
+ * / QSGMII / USXGMII), driving VR_MII_AN_CTRL.TX_CONFIG: MAC means the
+ * local end receives the partner's AN word, PHY means it sources one.
+ * Ignored for symmetric protocols (1000BASE-X, 2500BASE-X, 10GBASE-R/KR).
+ */
+#define MXL862XX_XPCS_ROLE_MAC		0	/* local end is MAC side */
+#define MXL862XX_XPCS_ROLE_PHY		1	/* local end is PHY side */
+
+/* USXGMII lane mode, MXL862XX_XPCS_*_USX_LANE_MODE field values */
+#define MXL862XX_XPCS_USX_SINGLE	0	/* single USXGMII lane */
+#define MXL862XX_XPCS_USX_QUAD		1	/* quad USXGMII, 4 ports/lane */
+
+/**
+ * union mxl862xx_xpcs_an_word - XPCS AN code word, tagged by interface mode
+ * @cl37: 16-bit base-page word exchanged over the CL37 hardware AN path
+ *        (SR_MII_AN_ADV on write, SR_MII_LP_BABL on read). Carries the
+ *        802.3 CL37 base page for 1000BASE-X/2500BASE-X and the Cisco
+ *        SGMII config word for SGMII/QSGMII.
+ * @usx: USXGMII 16-bit AN code word, MDIO_USXGMII_* layout
+ * @cl73: CL73 48-bit base page (10GBASE-KR), three 16-bit registers per
+ *        802.3 Annex 28C
+ * @cl73.adv1: CL73 SR_AN_ADV1 / SR_AN_LP_ABL1
+ * @cl73.adv2: CL73 SR_AN_ADV2 / SR_AN_LP_ABL2
+ * @cl73.adv3: CL73 SR_AN_ADV3 / SR_AN_LP_ABL3
+ *
+ * The host picks the right member based on the interface field of the
+ * surrounding struct (and, for the asymmetric protocols, on the role).
+ */
+union mxl862xx_xpcs_an_word {
+	__le16 cl37;
+	__le16 usx;
+	struct {
+		__le16 adv1;
+		__le16 adv2;
+		__le16 adv3;
+	} cl73;
+} __packed;
+
+/* PCS duplex mode, MXL862XX_XPCS_*_DUPLEX field values */
+#define MXL862XX_XPCS_DUPLEX_HALF	0
+#define MXL862XX_XPCS_DUPLEX_FULL	1
+
+/**
+ * enum mxl862xx_xpcs_loopback_mode - XPCS loopback mode
+ * @MXL862XX_XPCS_LB_DISABLE: disable all loopback
+ * @MXL862XX_XPCS_LB_PCS_SERIAL: PCS TX-to-RX serial loopback
+ * @MXL862XX_XPCS_LB_PCS_PARALLEL: PCS RX-to-TX parallel loopback
+ * @MXL862XX_XPCS_LB_PMA_SERIAL: PMA TX-to-RX serial loopback
+ * @MXL862XX_XPCS_LB_PMA_PARALLEL: PMA RX-to-TX parallel loopback
+ */
+enum mxl862xx_xpcs_loopback_mode {
+	MXL862XX_XPCS_LB_DISABLE = 0,
+	MXL862XX_XPCS_LB_PCS_SERIAL = 1,
+	MXL862XX_XPCS_LB_PCS_PARALLEL = 2,
+	MXL862XX_XPCS_LB_PMA_SERIAL = 3,
+	MXL862XX_XPCS_LB_PMA_PARALLEL = 4,
+};
+
+/* Fields of mxl862xx_xpcs_pcs_cfg.mode */
+#define MXL862XX_XPCS_CFG_PORT_ID	GENMASK(1, 0)
+#define MXL862XX_XPCS_CFG_INTERFACE	GENMASK(7, 2)
+#define MXL862XX_XPCS_CFG_NEG_MODE	GENMASK(9, 8)
+#define MXL862XX_XPCS_CFG_PERMIT_PAUSE	BIT(10)
+#define MXL862XX_XPCS_CFG_USX_LANE_MODE	GENMASK(12, 11)
+#define MXL862XX_XPCS_CFG_ROLE		BIT(13)
+#define MXL862XX_XPCS_CFG_USX_SUBPORT	GENMASK(15, 14)
+
+/**
+ * struct mxl862xx_xpcs_pcs_cfg - PCS configuration parameters
+ * @mode: Packed interface and negotiation parameters, see
+ *        MXL862XX_XPCS_CFG_*. port_id is the XPCS port index (0-3);
+ *        interface is the PCS interface mode (MXL862XX_XPCS_IF_*);
+ *        neg_mode is the negotiation mode (MXL862XX_XPCS_NEG_*);
+ *        permit_pause allows pause to MAC; usx_lane_mode is the USXGMII
+ *        lane mode (MXL862XX_XPCS_USX_*); role is the protocol role
+ *        (MXL862XX_XPCS_ROLE_*); usx_subport is the sub-port (0-3) within
+ *        the XPCS -- despite the name it also identifies the QSGMII
+ *        sub-port -- used by the firmware to set MAC pause per sub-port
+ *        and ignored for the XPCS-wide bringup, which is idempotent across
+ *        slots.
+ * @advertising: AN code word the local end transmits. The active union
+ *               member is selected by the interface field (and, for the
+ *               asymmetric protocols, by role). Ignored when the local end
+ *               does not transmit an AN word (role=MAC for SGMII/QSGMII/
+ *               USXGMII, 10GBASE-R, 5GBASE-R) or when neg_mode is not
+ *               INBAND_AN_ON. Pass all-zero to keep the firmware default
+ *               advertisement.
+ * @result: Firmware result. >0 means the host must follow with an AN
+ *          restart, 0 means no host follow-up is needed, <0 is an errno.
+ */
+struct mxl862xx_xpcs_pcs_cfg {
+	__le16 mode;
+	union mxl862xx_xpcs_an_word advertising;
+	__le16 result;
+} __packed;
+
+/* Fields of mxl862xx_xpcs_pcs_state.mode */
+#define MXL862XX_XPCS_ST_PORT_ID	GENMASK(1, 0)
+#define MXL862XX_XPCS_ST_INTERFACE	GENMASK(7, 2)
+#define MXL862XX_XPCS_ST_USX_LANE_MODE	GENMASK(9, 8)
+#define MXL862XX_XPCS_ST_USX_SUBPORT	GENMASK(11, 10)
+#define MXL862XX_XPCS_ST_LINK		BIT(12)
+#define MXL862XX_XPCS_ST_AN_COMPLETE	BIT(13)
+#define MXL862XX_XPCS_ST_DUPLEX		BIT(14)
+#define MXL862XX_XPCS_ST_PCS_FAULT	BIT(15)
+#define MXL862XX_XPCS_ST_PAUSE		GENMASK(17, 16)
+#define MXL862XX_XPCS_ST_LP_EEE_CAP	BIT(18)
+#define MXL862XX_XPCS_ST_LP_EEE_CS_CAP	BIT(19)
+
+/**
+ * struct mxl862xx_xpcs_pcs_state - PCS link state
+ * @mode: Packed input parameters and firmware status, see
+ *        MXL862XX_XPCS_ST_*. The host writes port_id (XPCS port index 0-3),
+ *        interface (MXL862XX_XPCS_IF_*), usx_lane_mode
+ *        (MXL862XX_XPCS_USX_*) and usx_subport (0-3); the firmware fills in
+ *        link, an_complete, duplex (MXL862XX_XPCS_DUPLEX_*), pcs_fault,
+ *        pause (bit 0 symmetric, bit 1 asymmetric), lp_eee_cap and
+ *        lp_eee_cs_cap.
+ * @speed: Resolved speed in Mbit/s (output)
+ * @lpa: Link partner ability word (output). Same union as
+ *       &union mxl862xx_xpcs_an_word; the host picks the member based on
+ *       the interface field.
+ */
+struct mxl862xx_xpcs_pcs_state {
+	__le32 mode;
+	__le16 speed; /* Mbit/s */
+	union mxl862xx_xpcs_an_word lpa;
+} __packed;
+
+/**
+ * struct mxl862xx_xpcs_pcs_disable - PCS disable parameters
+ * @port_id: XPCS port index
+ * @__pad: padding
+ * @result: Firmware result. 0 on success, <0 on error.
+ *
+ * Asserts IDDQ + PHY + XPCS resets to power down the SERDES when the
+ * port is admin-down or no module is plugged in. The next PCS config
+ * implicitly powers it back up and reprograms the desired interface.
+ */
+struct mxl862xx_xpcs_pcs_disable {
+	u8 port_id;
+	u8 __pad;
+	__le16 result;
+} __packed;
+
+/* Fields of mxl862xx_xpcs_an_restart.mode */
+#define MXL862XX_XPCS_ANR_PORT_ID	GENMASK(1, 0)
+#define MXL862XX_XPCS_ANR_INTERFACE	GENMASK(7, 2)
+#define MXL862XX_XPCS_ANR_USX_LANE_MODE	GENMASK(9, 8)
+#define MXL862XX_XPCS_ANR_USX_SUBPORT	GENMASK(11, 10)
+
+/**
+ * struct mxl862xx_xpcs_an_restart - AN restart parameters
+ * @mode: Packed input parameters, see MXL862XX_XPCS_ANR_*. port_id is the
+ *        XPCS port index (0-3); interface is the PCS interface mode
+ *        (MXL862XX_XPCS_IF_*); usx_lane_mode is the USX lane mode
+ *        (MXL862XX_XPCS_USX_*); usx_subport (0-3) selects the lane whose
+ *        AN is restarted for QSGMII and QUSXGMII and is ignored by
+ *        single-lane modes.
+ * @result: Firmware result. 0 on success, <0 on error.
+ *
+ * Restarts auto-negotiation on a single sub-port of the XPCS. The
+ * SERDES must already be configured.
+ */
+struct mxl862xx_xpcs_an_restart {
+	__le16 mode;
+	__le16 result;
+} __packed;
+
+/* Fields of mxl862xx_xpcs_pcs_link_up.mode */
+#define MXL862XX_XPCS_LU_PORT_ID	GENMASK(1, 0)
+#define MXL862XX_XPCS_LU_INTERFACE	GENMASK(7, 2)
+#define MXL862XX_XPCS_LU_DUPLEX		BIT(8)
+#define MXL862XX_XPCS_LU_USX_LANE_MODE	GENMASK(10, 9)
+#define MXL862XX_XPCS_LU_USX_SUBPORT	GENMASK(12, 11)
+
+/**
+ * struct mxl862xx_xpcs_pcs_link_up - PCS link-up parameters
+ * @mode: Packed input parameters, see MXL862XX_XPCS_LU_*. port_id is the
+ *        XPCS port index (0-3); interface is the PCS interface mode
+ *        (MXL862XX_XPCS_IF_*); duplex is the duplex mode
+ *        (MXL862XX_XPCS_DUPLEX_*); usx_lane_mode is the USX lane mode
+ *        (USXGMII only, ignored otherwise, MXL862XX_XPCS_USX_*);
+ *        usx_subport (0-3) selects the sub-port for QUSXGMII and QSGMII
+ *        (despite the name) and is ignored otherwise.
+ * @speed: Resolved speed in Mbit/s
+ * @result: Firmware result. 0 on success, <0 is errno.
+ *
+ * Called once per link-up event after the host has resolved the
+ * line-side speed/duplex (from the PHY's read_status, from a preceding
+ * PCS get-state, or from a fixed-link description).
+ */
+struct mxl862xx_xpcs_pcs_link_up {
+	__le16 mode;
+	__le16 speed; /* Mbit/s */
+	__le16 result;
+} __packed;
+
 #endif /* __MXL862XX_API_H */
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx-cmd.h b/drivers/net/dsa/mxl862xx/mxl862xx-cmd.h
index f1ea40aa7ea0..c87a955c13c4 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx-cmd.h
+++ b/drivers/net/dsa/mxl862xx/mxl862xx-cmd.h
@@ -24,6 +24,7 @@
 #define MXL862XX_SS_MAGIC		0x1600
 #define GPY_GPY2XX_MAGIC		0x1800
 #define SYS_MISC_MAGIC			0x1900
+#define MXL862XX_XPCS_MAGIC		0x1a00
 
 #define MXL862XX_COMMON_CFGGET		(MXL862XX_COMMON_MAGIC + 0x9)
 #define MXL862XX_COMMON_CFGSET		(MXL862XX_COMMON_MAGIC + 0xa)
@@ -71,6 +72,14 @@
 
 #define SYS_MISC_FW_VERSION		(SYS_MISC_MAGIC + 0x2)
 
+#define MXL862XX_XPCS_PCS_CONFIG	(MXL862XX_XPCS_MAGIC + 0x1)
+#define MXL862XX_XPCS_PCS_GET_STATE	(MXL862XX_XPCS_MAGIC + 0x2)
+#define MXL862XX_XPCS_PCS_DISABLE	(MXL862XX_XPCS_MAGIC + 0x4)
+#define MXL862XX_XPCS_AN_RESTART	(MXL862XX_XPCS_MAGIC + 0x5)
+#define MXL862XX_XPCS_PCS_LINK_UP	(MXL862XX_XPCS_MAGIC + 0x7)
+#define MXL862XX_XPCS_LOOPBACK		(MXL862XX_XPCS_MAGIC + 0x8)
+#define MXL862XX_XPCS_RESET		(MXL862XX_XPCS_MAGIC + 0x9)
+
 #define MMD_API_MAXIMUM_ID		0x7fff
 
 #endif /* __MXL862XX_CMD_H */
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx-phylink.c b/drivers/net/dsa/mxl862xx/mxl862xx-phylink.c
index f17c429d1f1d..3bcd72ecd95f 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx-phylink.c
+++ b/drivers/net/dsa/mxl862xx/mxl862xx-phylink.c
@@ -7,20 +7,401 @@
  * Copyright (C) 2025 Daniel Golle <daniel@makrotopia.org>
  */
 
+#include <linux/bitfield.h>
+#include <linux/mutex.h>
 #include <linux/phylink.h>
 #include <net/dsa.h>
 
 #include "mxl862xx.h"
+#include "mxl862xx-api.h"
+#include "mxl862xx-cmd.h"
+#include "mxl862xx-host.h"
 #include "mxl862xx-phylink.h"
 
 void mxl862xx_phylink_get_caps(struct dsa_switch *ds, int port,
 			       struct phylink_config *config)
 {
+	struct mxl862xx_priv *priv = ds->priv;
+
 	config->mac_capabilities = MAC_ASYM_PAUSE | MAC_SYM_PAUSE | MAC_10 |
 				   MAC_100 | MAC_1000 | MAC_2500FD;
 
-	__set_bit(PHY_INTERFACE_MODE_INTERNAL,
-		  config->supported_interfaces);
+	switch (port) {
+	case 1 ... 8:
+		__set_bit(PHY_INTERFACE_MODE_INTERNAL,
+			  config->supported_interfaces);
+		break;
+	case 9:
+	case 13:
+		__set_bit(PHY_INTERFACE_MODE_SGMII, config->supported_interfaces);
+		__set_bit(PHY_INTERFACE_MODE_1000BASEX, config->supported_interfaces);
+		__set_bit(PHY_INTERFACE_MODE_2500BASEX, config->supported_interfaces);
+		__set_bit(PHY_INTERFACE_MODE_10GBASER, config->supported_interfaces);
+		__set_bit(PHY_INTERFACE_MODE_10GKR, config->supported_interfaces);
+		__set_bit(PHY_INTERFACE_MODE_USXGMII, config->supported_interfaces);
+		fallthrough;
+	case 10 ... 12:
+	case 14 ... 16:
+		if (!MXL862XX_FW_VER_MIN(priv, 1, 0, 84))
+			break;
+		__set_bit(PHY_INTERFACE_MODE_QSGMII, config->supported_interfaces);
+		__set_bit(PHY_INTERFACE_MODE_10G_QXGMII, config->supported_interfaces);
+
+		break;
+	default:
+		break;
+	}
+
+	if (port == 9 || port == 13)
+		config->mac_capabilities |= MAC_10000FD | MAC_5000FD;
+}
+
+static struct mxl862xx_pcs *pcs_to_mxl862xx_pcs(struct phylink_pcs *pcs)
+{
+	return container_of(pcs, struct mxl862xx_pcs, pcs);
+}
+
+static int mxl862xx_xpcs_if_mode(phy_interface_t interface)
+{
+	switch (interface) {
+	case PHY_INTERFACE_MODE_SGMII:
+		return MXL862XX_XPCS_IF_SGMII;
+	case PHY_INTERFACE_MODE_QSGMII:
+		return MXL862XX_XPCS_IF_QSGMII;
+	case PHY_INTERFACE_MODE_1000BASEX:
+		return MXL862XX_XPCS_IF_1000BASEX;
+	case PHY_INTERFACE_MODE_2500BASEX:
+		return MXL862XX_XPCS_IF_2500BASEX;
+	case PHY_INTERFACE_MODE_USXGMII:
+	case PHY_INTERFACE_MODE_10G_QXGMII:
+		return MXL862XX_XPCS_IF_USXGMII;
+	case PHY_INTERFACE_MODE_10GBASER:
+		return MXL862XX_XPCS_IF_10GBASER;
+	case PHY_INTERFACE_MODE_10GKR:
+		return MXL862XX_XPCS_IF_10GKR;
+	default:
+		return -EINVAL;
+	}
+}
+
+static int mxl862xx_xpcs_neg_mode(unsigned int neg_mode)
+{
+	if (!(neg_mode & PHYLINK_PCS_NEG_INBAND))
+		return MXL862XX_XPCS_NEG_NONE;
+	if (neg_mode & PHYLINK_PCS_NEG_ENABLED)
+		return MXL862XX_XPCS_NEG_INBAND_AN_ON;
+	return MXL862XX_XPCS_NEG_INBAND_AN_OFF;
+}
+
+static int mxl862xx_pcs_enable(struct phylink_pcs *pcs)
+{
+	struct mxl862xx_pcs *mpcs = pcs_to_mxl862xx_pcs(pcs);
+
+	/* Bringup is done idempotently by pcs_config; just account this
+	 * sub-port so pcs_disable powers the shared XPCS down only after
+	 * the last sub-port has been released.
+	 */
+	mutex_lock(&mpcs->priv->serdes_lock);
+	mpcs->priv->serdes_refcount[mpcs->serdes_id]++;
+	mutex_unlock(&mpcs->priv->serdes_lock);
+
+	return 0;
+}
+
+static void mxl862xx_pcs_disable(struct phylink_pcs *pcs)
+{
+	struct mxl862xx_pcs *mpcs = pcs_to_mxl862xx_pcs(pcs);
+	struct mxl862xx_xpcs_pcs_disable dis = {};
+	struct mxl862xx_priv *priv = mpcs->priv;
+
+	dis.port_id = mpcs->serdes_id;
+
+	/* The SerDes is shared across QSGMII/QUSXGMII sub-ports; only
+	 * power it down once the last active sub-port goes away. Hold
+	 * serdes_lock across the count and the power-down so a sibling
+	 * sub-port enable cannot race the transition to zero.
+	 */
+	mutex_lock(&priv->serdes_lock);
+	if (--priv->serdes_refcount[mpcs->serdes_id] == 0)
+		MXL862XX_API_WRITE(priv, MXL862XX_XPCS_PCS_DISABLE, dis);
+	mutex_unlock(&priv->serdes_lock);
+}
+
+/* The XPCS firmware reports failures in the result field using its own
+ * libc errno values; ENOTSUP (134) in particular has no kernel errno.
+ * Translate the codes the firmware can actually return.
+ */
+static int mxl862xx_xpcs_errno(int result)
+{
+	switch (result) {
+	case -5:	/* firmware -EIO */
+		return -EIO;
+	case -134:	/* firmware -ENOTSUP */
+		return -EOPNOTSUPP;
+	default:	/* firmware -EINVAL and anything unexpected */
+		return -EINVAL;
+	}
+}
+
+static int mxl862xx_pcs_config(struct phylink_pcs *pcs, unsigned int neg_mode,
+			       phy_interface_t interface,
+			       const unsigned long *advertising,
+			       bool permit_pause_to_mac)
+{
+	struct mxl862xx_pcs *mpcs = pcs_to_mxl862xx_pcs(pcs);
+	struct mxl862xx_priv *priv = mpcs->priv;
+	struct mxl862xx_xpcs_pcs_cfg cfg = {};
+	int if_mode, lane, ret, adv;
+
+	if_mode = mxl862xx_xpcs_if_mode(interface);
+	if (if_mode < 0) {
+		dev_err(priv->ds->dev, "unsupported interface: %s\n",
+			phy_modes(interface));
+		return if_mode;
+	}
+
+	/* The XPCS bringup is per-instance and idempotent in the
+	 * firmware: every QSGMII/QUSXGMII sub-port may call pcs_config
+	 * and the firmware will skip the bringup if the requested mode
+	 * matches the cached one, then update MAC pause for the
+	 * sub-port indicated by @usx_subport.
+	 */
+	lane = (interface == PHY_INTERFACE_MODE_10G_QXGMII) ?
+	       MXL862XX_XPCS_USX_QUAD : MXL862XX_XPCS_USX_SINGLE;
+
+	cfg.mode = cpu_to_le16(FIELD_PREP(MXL862XX_XPCS_CFG_PORT_ID,
+					  mpcs->serdes_id) |
+			       FIELD_PREP(MXL862XX_XPCS_CFG_USX_SUBPORT,
+					  mpcs->slot) |
+			       FIELD_PREP(MXL862XX_XPCS_CFG_USX_LANE_MODE, lane) |
+			       FIELD_PREP(MXL862XX_XPCS_CFG_INTERFACE, if_mode) |
+			       FIELD_PREP(MXL862XX_XPCS_CFG_NEG_MODE,
+					  mxl862xx_xpcs_neg_mode(neg_mode)) |
+			       FIELD_PREP(MXL862XX_XPCS_CFG_ROLE,
+					  MXL862XX_XPCS_ROLE_MAC) |
+			       FIELD_PREP(MXL862XX_XPCS_CFG_PERMIT_PAUSE,
+					  permit_pause_to_mac));
+
+	if (neg_mode & PHYLINK_PCS_NEG_INBAND) {
+		adv = phylink_mii_c22_pcs_encode_advertisement(interface,
+							       advertising);
+		if (adv >= 0)
+			cfg.advertising.cl37 = cpu_to_le16(adv);
+	}
+
+	ret = MXL862XX_API_READ(priv, MXL862XX_XPCS_PCS_CONFIG, cfg);
+	if (ret)
+		return ret;
+
+	ret = (s16)le16_to_cpu(cfg.result);
+	if (ret < 0)
+		return mxl862xx_xpcs_errno(ret);
+
+	mpcs->interface = interface;
+	return ret > 0 ? 1 : 0;
+}
+
+static void mxl862xx_pcs_get_state(struct phylink_pcs *pcs,
+				   unsigned int neg_mode,
+				   struct phylink_link_state *state)
+{
+	struct mxl862xx_pcs *mpcs = pcs_to_mxl862xx_pcs(pcs);
+	struct mxl862xx_priv *priv = mpcs->priv;
+	struct mxl862xx_xpcs_pcs_state st = {};
+	int if_mode, lane, ret;
+	u32 mode;
+	u16 bmsr;
+
+	if_mode = mxl862xx_xpcs_if_mode(state->interface);
+	if (if_mode < 0)
+		return;
+
+	lane = (state->interface == PHY_INTERFACE_MODE_10G_QXGMII) ?
+	       MXL862XX_XPCS_USX_QUAD : MXL862XX_XPCS_USX_SINGLE;
+
+	st.mode = cpu_to_le32(FIELD_PREP(MXL862XX_XPCS_ST_PORT_ID,
+					 mpcs->serdes_id) |
+			      FIELD_PREP(MXL862XX_XPCS_ST_INTERFACE, if_mode) |
+			      FIELD_PREP(MXL862XX_XPCS_ST_USX_SUBPORT,
+					 mpcs->slot) |
+			      FIELD_PREP(MXL862XX_XPCS_ST_USX_LANE_MODE, lane));
+
+	ret = MXL862XX_API_READ(priv, MXL862XX_XPCS_PCS_GET_STATE, st);
+	if (ret)
+		return;
+
+	mode = le32_to_cpu(st.mode);
+	state->link = FIELD_GET(MXL862XX_XPCS_ST_LINK, mode) &&
+		      !FIELD_GET(MXL862XX_XPCS_ST_PCS_FAULT, mode);
+	state->an_complete = FIELD_GET(MXL862XX_XPCS_ST_AN_COMPLETE, mode);
+
+	switch (state->interface) {
+	case PHY_INTERFACE_MODE_1000BASEX:
+	case PHY_INTERFACE_MODE_2500BASEX:
+	case PHY_INTERFACE_MODE_SGMII:
+	case PHY_INTERFACE_MODE_QSGMII:
+		bmsr = (state->link ? BMSR_LSTATUS : 0) |
+		       (state->an_complete ? BMSR_ANEGCOMPLETE : 0);
+		phylink_mii_c22_pcs_decode_state(state, neg_mode, bmsr,
+						 le16_to_cpu(st.lpa.cl37));
+		break;
+
+	case PHY_INTERFACE_MODE_USXGMII:
+	case PHY_INTERFACE_MODE_10G_QXGMII:
+		if (state->link)
+			phylink_decode_usxgmii_word(state,
+						    le16_to_cpu(st.lpa.usx));
+		break;
+
+	case PHY_INTERFACE_MODE_10GBASER:
+	case PHY_INTERFACE_MODE_10GKR:
+		if (state->link) {
+			state->speed = SPEED_10000;
+			state->duplex = DUPLEX_FULL;
+		}
+		break;
+
+	default:
+		state->link = false;
+		break;
+	}
+}
+
+static void mxl862xx_pcs_an_restart(struct phylink_pcs *pcs)
+{
+	struct mxl862xx_pcs *mpcs = pcs_to_mxl862xx_pcs(pcs);
+	struct mxl862xx_priv *priv = mpcs->priv;
+	struct mxl862xx_xpcs_an_restart an = {};
+	int if_mode, lane;
+
+	if_mode = mxl862xx_xpcs_if_mode(mpcs->interface);
+	if (if_mode < 0)
+		return;
+
+	lane = (mpcs->interface == PHY_INTERFACE_MODE_10G_QXGMII) ?
+	       MXL862XX_XPCS_USX_QUAD : MXL862XX_XPCS_USX_SINGLE;
+
+	an.mode = cpu_to_le16(FIELD_PREP(MXL862XX_XPCS_ANR_PORT_ID,
+					 mpcs->serdes_id) |
+			      FIELD_PREP(MXL862XX_XPCS_ANR_INTERFACE, if_mode) |
+			      FIELD_PREP(MXL862XX_XPCS_ANR_USX_SUBPORT,
+					 mpcs->slot) |
+			      FIELD_PREP(MXL862XX_XPCS_ANR_USX_LANE_MODE, lane));
+
+	MXL862XX_API_WRITE(priv, MXL862XX_XPCS_AN_RESTART, an);
+}
+
+static void mxl862xx_pcs_link_up(struct phylink_pcs *pcs, unsigned int neg_mode,
+				 phy_interface_t interface, int speed,
+				 int duplex)
+{
+	struct mxl862xx_pcs *mpcs = pcs_to_mxl862xx_pcs(pcs);
+	struct mxl862xx_xpcs_pcs_link_up lu = {};
+	struct mxl862xx_priv *priv = mpcs->priv;
+	int if_mode, lane, dup;
+
+	/* With inband-AN enabled (role=MAC), the XPCS auto-resolves
+	 * speed/duplex from the partner's AN word and the firmware
+	 * short-circuits link_up.  Skip the firmware round-trip, same
+	 * as pcs-mtk-lynxi.
+	 */
+	if (neg_mode == PHYLINK_PCS_NEG_INBAND_ENABLED)
+		return;
+
+	if_mode = mxl862xx_xpcs_if_mode(interface);
+	if (if_mode < 0)
+		return;
+
+	lane = (interface == PHY_INTERFACE_MODE_10G_QXGMII) ?
+	       MXL862XX_XPCS_USX_QUAD : MXL862XX_XPCS_USX_SINGLE;
+	dup = (duplex == DUPLEX_FULL) ? MXL862XX_XPCS_DUPLEX_FULL :
+					MXL862XX_XPCS_DUPLEX_HALF;
+
+	lu.mode = cpu_to_le16(FIELD_PREP(MXL862XX_XPCS_LU_PORT_ID,
+					 mpcs->serdes_id) |
+			      FIELD_PREP(MXL862XX_XPCS_LU_INTERFACE, if_mode) |
+			      FIELD_PREP(MXL862XX_XPCS_LU_USX_SUBPORT,
+					 mpcs->slot) |
+			      FIELD_PREP(MXL862XX_XPCS_LU_USX_LANE_MODE, lane) |
+			      FIELD_PREP(MXL862XX_XPCS_LU_DUPLEX, dup));
+	lu.speed = cpu_to_le16(speed);
+
+	MXL862XX_API_WRITE(priv, MXL862XX_XPCS_PCS_LINK_UP, lu);
+}
+
+static unsigned int mxl862xx_pcs_inband_caps(struct phylink_pcs *pcs,
+					     phy_interface_t interface)
+{
+	switch (interface) {
+	case PHY_INTERFACE_MODE_SGMII:
+	case PHY_INTERFACE_MODE_QSGMII:
+	case PHY_INTERFACE_MODE_1000BASEX:
+	case PHY_INTERFACE_MODE_2500BASEX:
+		return LINK_INBAND_DISABLE | LINK_INBAND_ENABLE;
+	case PHY_INTERFACE_MODE_USXGMII:
+	case PHY_INTERFACE_MODE_10G_QXGMII:
+	case PHY_INTERFACE_MODE_10GKR:
+		return LINK_INBAND_ENABLE;
+	case PHY_INTERFACE_MODE_10GBASER:
+		return LINK_INBAND_DISABLE;
+	default:
+		return 0;
+	}
+}
+
+static const struct phylink_pcs_ops mxl862xx_pcs_ops = {
+	.pcs_enable = mxl862xx_pcs_enable,
+	.pcs_disable = mxl862xx_pcs_disable,
+	.pcs_config = mxl862xx_pcs_config,
+	.pcs_get_state = mxl862xx_pcs_get_state,
+	.pcs_an_restart = mxl862xx_pcs_an_restart,
+	.pcs_link_up = mxl862xx_pcs_link_up,
+	.pcs_inband_caps = mxl862xx_pcs_inband_caps,
+};
+
+void mxl862xx_setup_pcs(struct mxl862xx_priv *priv, struct mxl862xx_pcs *pcs,
+			int port)
+{
+	pcs->priv = priv;
+	pcs->serdes_id = MXL862XX_SERDES_PORT_ID(port);
+	pcs->slot = MXL862XX_SERDES_SLOT(port);
+	pcs->interface = PHY_INTERFACE_MODE_NA;
+
+	pcs->pcs.ops = &mxl862xx_pcs_ops;
+	pcs->pcs.poll = true;
+
+	__set_bit(PHY_INTERFACE_MODE_QSGMII, pcs->pcs.supported_interfaces);
+	__set_bit(PHY_INTERFACE_MODE_10G_QXGMII, pcs->pcs.supported_interfaces);
+	if (pcs->slot != 0)
+		return;
+
+	__set_bit(PHY_INTERFACE_MODE_SGMII, pcs->pcs.supported_interfaces);
+	__set_bit(PHY_INTERFACE_MODE_1000BASEX, pcs->pcs.supported_interfaces);
+	__set_bit(PHY_INTERFACE_MODE_2500BASEX, pcs->pcs.supported_interfaces);
+	__set_bit(PHY_INTERFACE_MODE_10GBASER, pcs->pcs.supported_interfaces);
+	__set_bit(PHY_INTERFACE_MODE_10GKR, pcs->pcs.supported_interfaces);
+	__set_bit(PHY_INTERFACE_MODE_USXGMII, pcs->pcs.supported_interfaces);
+}
+
+static struct phylink_pcs *
+mxl862xx_phylink_mac_select_pcs(struct phylink_config *config,
+				phy_interface_t interface)
+{
+	struct dsa_port *dp = dsa_phylink_to_port(config);
+	struct mxl862xx_priv *priv = dp->ds->priv;
+	int port = dp->index;
+
+	switch (port) {
+	case 9 ... 16:
+		if (!MXL862XX_FW_VER_MIN(priv, 1, 0, 84)) {
+			dev_warn_once(dp->ds->dev,
+				      "SerDes PCS unsupported on old firmware.\n");
+			return NULL;
+		}
+		return &priv->serdes_ports[port - 9].pcs;
+	default:
+		return NULL;
+	}
 }
 
 static void mxl862xx_phylink_mac_config(struct phylink_config *config,
@@ -48,4 +429,5 @@ const struct phylink_mac_ops mxl862xx_phylink_mac_ops = {
 	.mac_config = mxl862xx_phylink_mac_config,
 	.mac_link_down = mxl862xx_phylink_mac_link_down,
 	.mac_link_up = mxl862xx_phylink_mac_link_up,
+	.mac_select_pcs = mxl862xx_phylink_mac_select_pcs,
 };
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx-phylink.h b/drivers/net/dsa/mxl862xx/mxl862xx-phylink.h
index c3d5215bdf60..03bb9caad9aa 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx-phylink.h
+++ b/drivers/net/dsa/mxl862xx/mxl862xx-phylink.h
@@ -7,8 +7,15 @@
 
 #include "mxl862xx.h"
 
+#define MXL862XX_SERDES_SLOT(port) \
+	(((port) - MXL862XX_FIRST_SERDES_PORT) % MXL862XX_SERDES_SLOTS)
+#define MXL862XX_SERDES_PORT_ID(port) \
+	(((port) - MXL862XX_FIRST_SERDES_PORT) / MXL862XX_SERDES_SLOTS)
+
 extern const struct phylink_mac_ops mxl862xx_phylink_mac_ops;
 void mxl862xx_phylink_get_caps(struct dsa_switch *ds, int port,
 			       struct phylink_config *config);
+void mxl862xx_setup_pcs(struct mxl862xx_priv *priv, struct mxl862xx_pcs *pcs,
+			int port);
 
 #endif /* __MXL862XX_PHYLINK_H */
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx.c b/drivers/net/dsa/mxl862xx/mxl862xx.c
index 0b1a23364eb5..45d237b3a40f 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx.c
+++ b/drivers/net/dsa/mxl862xx/mxl862xx.c
@@ -622,7 +622,7 @@ static int mxl862xx_setup(struct dsa_switch *ds)
 	int n_user_ports = 0, max_vlans;
 	int ingress_finals, vid_rules;
 	struct dsa_port *dp;
-	int ret;
+	int ret, i;
 
 	ret = mxl862xx_reset(priv);
 	if (ret)
@@ -632,6 +632,11 @@ static int mxl862xx_setup(struct dsa_switch *ds)
 	if (ret)
 		return ret;
 
+	mutex_init(&priv->serdes_lock);
+	for (i = 0; i < ARRAY_SIZE(priv->serdes_ports); i++)
+		mxl862xx_setup_pcs(priv, &priv->serdes_ports[i],
+				   i + MXL862XX_FIRST_SERDES_PORT);
+
 	/* Calculate Extended VLAN block sizes.
 	 * With VLAN Filter handling VID membership checks:
 	 *   Ingress: only final catchall rules (PVID insertion, 802.1Q
diff --git a/drivers/net/dsa/mxl862xx/mxl862xx.h b/drivers/net/dsa/mxl862xx/mxl862xx.h
index e3db3711b245..432a5f3f2e08 100644
--- a/drivers/net/dsa/mxl862xx/mxl862xx.h
+++ b/drivers/net/dsa/mxl862xx/mxl862xx.h
@@ -11,6 +11,9 @@
 struct mxl862xx_priv;
 
 #define MXL862XX_MAX_PORTS		17
+#define MXL862XX_FIRST_SERDES_PORT	9
+#define MXL862XX_SERDES_SLOTS		4
+
 #define MXL862XX_DEFAULT_BRIDGE		0
 #define MXL862XX_MAX_BRIDGES		48
 #define MXL862XX_MAX_BRIDGE_PORTS	128
@@ -242,6 +245,26 @@ struct mxl862xx_port {
 	spinlock_t stats_lock; /* protects stats accumulators */
 };
 
+/**
+ * struct mxl862xx_pcs - link SerDes interfaces to bridge ports
+ * @pcs:       &struct phylink_pcs instance
+ * @priv:      pointer to &struct mxl862xx_priv
+ * @serdes_id: SerDes instance index (0 or 1)
+ * @slot:      slot within the SerDes (0-3 for QSGMII/QUSXGMII, 0 otherwise)
+ * @interface: cached PHY interface, last value passed to pcs_config().
+ *             %PHY_INTERFACE_MODE_NA before the first successful
+ *             pcs_config().  Used by pcs_an_restart() to populate the
+ *             firmware command and by pcs_disable() to skip the
+ *             firmware power-down for shared (QSGMII/QUSXGMII) modes.
+ */
+struct mxl862xx_pcs {
+	struct phylink_pcs pcs;
+	struct mxl862xx_priv *priv;
+	int serdes_id;
+	int slot;
+	phy_interface_t interface;
+};
+
 /**
  * struct mxl862xx_fw_version - firmware version for comparison and display
  * @major: firmware major version
@@ -280,6 +303,14 @@ struct mxl862xx_fw_version {
  *                      flooding)
  * @fw_version:         cached firmware version, populated at probe and
  *                      compared with MXL862XX_FW_VER_MIN()
+ * @serdes_ports:       SerDes interfaces incl. sub-interfaces in case of
+ *                      10G_QXGMII or QSGMII
+ * @serdes_refcount:    per-XPCS count of sub-ports enabled by phylink;
+ *                      pcs_disable powers an XPCS down when the count
+ *                      reaches zero. Protected by @serdes_lock.
+ * @serdes_lock:        serializes the @serdes_refcount transitions with
+ *                      the XPCS power-down so a sibling sub-port enable
+ *                      cannot race a power-down to zero
  * @ports:              per-port state, indexed by switch port number
  * @bridges:            maps DSA bridge number to firmware bridge ID;
  *                      zero means no firmware bridge allocated for that
@@ -298,6 +329,9 @@ struct mxl862xx_priv {
 	unsigned long flags;
 	u16 drop_meter;
 	struct mxl862xx_fw_version fw_version;
+	struct mxl862xx_pcs serdes_ports[8];
+	int serdes_refcount[2];
+	struct mutex serdes_lock;
 	struct mxl862xx_port ports[MXL862XX_MAX_PORTS];
 	u16 bridges[MXL862XX_MAX_BRIDGES + 1];
 	u16 evlan_ingress_size;
-- 
2.54.0