Netdev List
 help / color / mirror / Atom feed
* [PATCH net-next v2 0/3] net: pse-pd: realtek-pse-mcu: establish power management configuration
@ 2026-11-16 11:52 Jonas Jelonek
  2026-11-16 11:52 ` [PATCH net-next v2 1/3] dt-bindings: net: pse-pd: realtek,pse-mcu: add VMAIN supply Jonas Jelonek
                   ` (2 more replies)
  0 siblings, 3 replies; 6+ messages in thread
From: Jonas Jelonek @ 2026-11-16 11:52 UTC (permalink / raw)
  To: Oleksij Rempel, Kory Maincent, Andrew Lunn, David S . Miller,
	Eric Dumazet, Jakub Kicinski, Paolo Abeni, netdev, Rob Herring,
	Krzysztof Kozlowski, Conor Dooley
  Cc: linux-kernel, devicetree, Sander Vanheule, Russell Senior,
	Jonas Jelonek

Realtek PSE MCUs may start with a power-management configuration inherited
from firmware defaults or previous host configuration. This series
establishes an explicit configuration during probe: dynamic power management
with port priorities and, when a VMAIN regulator is provided, an aggregate
limit matching the board's usable PoE allowance.

The budget and management mode are configured before ungating ports.
VMAIN is used only for budget accounting, with managed reservation cleanup;
the rail is available independently of the host driver.

Validation includes W=1 builds and checkpatch.pl. OpenCode assisted with
review, implementation refinements and drafting this cover letter.

---
Changes in v2:
- Move the binding first and clarify accounting-only use of VMAIN.
- Preserved Acked-by from Rob and Reviewed-by from Kory
- Set the mode directly in the helper and shorten names to "mgmt".
- Remove explicit VMAIN enable/disable; retain managed budget cleanup.
- Drop the unrelated global-disable teardown patch.

v1: https://lore.kernel.org/netdev/20261103220312.860629-1-jonas@jonasjelonek.de/

---
Jonas Jelonek (2):
  dt-bindings: net: pse-pd: realtek,pse-mcu: add VMAIN supply
  net: pse-pd: realtek-pse-mcu: configure global power budget

Russell Senior (1):
  net: pse-pd: realtek-pse-mcu: ensure MCU does dynamic power management

 .../net/pse-pd/realtek,pse-mcu-gen1.yaml      |   9 +
 drivers/net/pse-pd/realtek-pse-mcu-core.c     | 155 +++++++++++++++++-
 drivers/net/pse-pd/realtek-pse-mcu.h          |   3 +
 3 files changed, 164 insertions(+), 3 deletions(-)


base-commit: 7ccc5e05043b1bb84326b19a2b05938d8f7a2ae6
-- 
2.53.0


^ permalink raw reply	[flat|nested] 6+ messages in thread

* [PATCH net-next v2 1/3] dt-bindings: net: pse-pd: realtek,pse-mcu: add VMAIN supply
  2026-11-16 11:52 [PATCH net-next v2 0/3] net: pse-pd: realtek-pse-mcu: establish power management configuration Jonas Jelonek
@ 2026-11-16 11:52 ` Jonas Jelonek
  2026-11-16 11:52 ` [PATCH net-next v2 2/3] net: pse-pd: realtek-pse-mcu: ensure MCU does dynamic power management Jonas Jelonek
  2026-11-16 11:52 ` [PATCH net-next v2 3/3] net: pse-pd: realtek-pse-mcu: configure global power budget Jonas Jelonek
  2 siblings, 0 replies; 6+ messages in thread
From: Jonas Jelonek @ 2026-11-16 11:52 UTC (permalink / raw)
  To: Oleksij Rempel, Kory Maincent, Andrew Lunn, David S . Miller,
	Eric Dumazet, Jakub Kicinski, Paolo Abeni, netdev, Rob Herring,
	Krzysztof Kozlowski, Conor Dooley
  Cc: linux-kernel, devicetree, Sander Vanheule, Russell Senior,
	Jonas Jelonek

Describe the VMAIN supply at the MCU controller level. The rail feeds the
PoE subsystem as a whole rather than an individual PI. Its regulator power
budget defines the aggregate limit programmed into the MCU.

Require a finite regulator power budget dedicated to this MCU, representing
the PoE power allocation of the system. The rail is available independently
of the host driver, which uses the supply reference for budget accounting.

Signed-off-by: Jonas Jelonek <jonas@jonasjelonek.de>
Reviewed-by: Kory Maincent <kory.maincent@bootlin.com>
Acked-by: Rob Herring (Arm) <robh@kernel.org>
---
 .../bindings/net/pse-pd/realtek,pse-mcu-gen1.yaml        | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/Documentation/devicetree/bindings/net/pse-pd/realtek,pse-mcu-gen1.yaml b/Documentation/devicetree/bindings/net/pse-pd/realtek,pse-mcu-gen1.yaml
index 526f8525bb4d..6acb5fb641c1 100644
--- a/Documentation/devicetree/bindings/net/pse-pd/realtek,pse-mcu-gen1.yaml
+++ b/Documentation/devicetree/bindings/net/pse-pd/realtek,pse-mcu-gen1.yaml
@@ -92,6 +92,15 @@ properties:
       asserted.
     maxItems: 1
 
+  vmain-supply:
+    description:
+      Regulator representing the PoE power allocation of the 44-57V VMAIN
+      supply. This MCU must be its only power-budget consumer. Its unclaimed
+      power budget is used to configure the MCU global PoE power budget. The
+      regulator must define a finite regulator-power-budget-milliwatt value.
+      The supply is available independently of the host driver and is used
+      for power-budget accounting only.
+
 required:
   - compatible
 
-- 
2.53.0


^ permalink raw reply related	[flat|nested] 6+ messages in thread

* [PATCH net-next v2 2/3] net: pse-pd: realtek-pse-mcu: ensure MCU does dynamic power management
  2026-11-16 11:52 [PATCH net-next v2 0/3] net: pse-pd: realtek-pse-mcu: establish power management configuration Jonas Jelonek
  2026-11-16 11:52 ` [PATCH net-next v2 1/3] dt-bindings: net: pse-pd: realtek,pse-mcu: add VMAIN supply Jonas Jelonek
@ 2026-11-16 11:52 ` Jonas Jelonek
  2026-11-19  5:54   ` netdev-bot+sashiko
  2026-11-16 11:52 ` [PATCH net-next v2 3/3] net: pse-pd: realtek-pse-mcu: configure global power budget Jonas Jelonek
  2 siblings, 1 reply; 6+ messages in thread
From: Jonas Jelonek @ 2026-11-16 11:52 UTC (permalink / raw)
  To: Oleksij Rempel, Kory Maincent, Andrew Lunn, David S . Miller,
	Eric Dumazet, Jakub Kicinski, Paolo Abeni, netdev, Rob Herring,
	Krzysztof Kozlowski, Conor Dooley
  Cc: linux-kernel, devicetree, Sander Vanheule, Russell Senior,
	Jonas Jelonek

From: Russell Senior <russell@personaltelco.net>

The MCU may have any power management mode configuration by default
but the driver should ensure that the MCU performs dynamic power
management with priorities as advertised. On certain devices, the
default MCU configuration ships with another default, possibly
preventing any power delivery.

Select the mode before deasserting the hardware port gate, so a failure
to establish the management policy does not ungate the ports.

Signed-off-by: Russell Senior <russell@personaltelco.net>
[jonas@jonasjelonek.de: adjust commit message, macro naming and comment;
 shorten command and helper names to "mgmt"; change error handling; select
 the mode before ungating ports; set dynamic mode with priorities directly
 in the helper]
Signed-off-by: Jonas Jelonek <jonas@jonasjelonek.de>
---
 drivers/net/pse-pd/realtek-pse-mcu-core.c | 32 +++++++++++++++++++++++
 1 file changed, 32 insertions(+)

diff --git a/drivers/net/pse-pd/realtek-pse-mcu-core.c b/drivers/net/pse-pd/realtek-pse-mcu-core.c
index c70b09ca2012..d8b6e0503939 100644
--- a/drivers/net/pse-pd/realtek-pse-mcu-core.c
+++ b/drivers/net/pse-pd/realtek-pse-mcu-core.c
@@ -68,10 +68,14 @@
 /* Nominal PSE rail; 802.3at/bt operating range. */
 #define RTPSE_MCU_PSE_VOLTAGE_UV		54000000
 
+/* RTPSE_MCU_CMD_SET_POWER_MGMT_MODE values */
+#define RTPSE_MCU_POWER_MGMT_DYNAMIC_PRIO	0x02	/* dynamic, with port priority */
+
 enum rtpse_mcu_cmd {
 	RTPSE_MCU_CMD_SET_GLOBAL_STATE,
 	RTPSE_MCU_CMD_GET_SYSTEM_INFO,
 	RTPSE_MCU_CMD_GET_EXT_CONFIG,
+	RTPSE_MCU_CMD_SET_POWER_MGMT_MODE,
 
 	RTPSE_MCU_CMD_PORT_ENABLE,
 	RTPSE_MCU_CMD_PORT_SET_POWER_LIMIT_TYPE,
@@ -369,6 +373,26 @@ static int rtpse_mcu_set_global_state(struct rtpse_mcu_ctrl *pse, bool enable)
 	return (resp.payload[0] == 0x0) ? 0 : -EIO;
 }
 
+static int rtpse_mcu_set_power_mgmt_mode(struct rtpse_mcu_ctrl *pse)
+{
+	struct rtpse_mcu_msg req, resp;
+	const struct rtpse_mcu_opcode *opc;
+	int ret;
+
+	opc = &pse->dialect->opcode[RTPSE_MCU_CMD_SET_POWER_MGMT_MODE];
+	if (!opc->valid)
+		return -EOPNOTSUPP;
+
+	rtpse_mcu_msg_init(&req, opc->op);
+	req.payload[0] = RTPSE_MCU_POWER_MGMT_DYNAMIC_PRIO;
+
+	ret = rtpse_mcu_do_xfer(pse, &req, &resp);
+	if (ret)
+		return ret;
+
+	return (resp.payload[0] == 0x0) ? 0 : -EIO;
+}
+
 /* Port operations */
 
 static int rtpse_mcu_port_get_status(struct rtpse_mcu_ctrl *pse, unsigned int port,
@@ -826,6 +850,12 @@ int rtpse_mcu_register(struct rtpse_mcu_ctrl *pse)
 	if (ret)
 		return ret;
 
+	/* ensure the MCU does dynamic power management with priorities */
+	ret = rtpse_mcu_set_power_mgmt_mode(pse);
+	if (ret)
+		return dev_err_probe(pse->dev, ret,
+				     "failed to set power management mode\n");
+
 	/*
 	 * Some boards gate all ports through a hardware line; deassert it only
 	 * after the MCU is confirmed, so a discover failure never ungates the
@@ -944,6 +974,7 @@ static const struct rtpse_mcu_dialect rtpse_mcu_dialect_gen2 = {
 		[RTPSE_MCU_CMD_SET_GLOBAL_STATE]	= RTPSE_MCU_OP(0x00),
 		[RTPSE_MCU_CMD_GET_SYSTEM_INFO]		= RTPSE_MCU_OP(0x40),
 		[RTPSE_MCU_CMD_GET_EXT_CONFIG]		= RTPSE_MCU_OP(0x4a),
+		[RTPSE_MCU_CMD_SET_POWER_MGMT_MODE]	= RTPSE_MCU_OP(0x10),
 
 		[RTPSE_MCU_CMD_PORT_ENABLE]		= RTPSE_MCU_OP(0x01),
 		[RTPSE_MCU_CMD_PORT_SET_POWER_LIMIT_TYPE] = RTPSE_MCU_OP(0x12),
@@ -964,6 +995,7 @@ static const struct rtpse_mcu_dialect rtpse_mcu_dialect_gen1 = {
 	.opcode = {
 		[RTPSE_MCU_CMD_GET_SYSTEM_INFO]		= RTPSE_MCU_OP(0x20),
 		[RTPSE_MCU_CMD_GET_EXT_CONFIG]		= RTPSE_MCU_OP(0x2b),
+		[RTPSE_MCU_CMD_SET_POWER_MGMT_MODE]	= RTPSE_MCU_OP(0x17),
 
 		[RTPSE_MCU_CMD_PORT_ENABLE]		= RTPSE_MCU_OP(0x00),
 		[RTPSE_MCU_CMD_PORT_SET_POWER_LIMIT_TYPE] = RTPSE_MCU_OP(0x15),
-- 
2.53.0


^ permalink raw reply related	[flat|nested] 6+ messages in thread

* [PATCH net-next v2 3/3] net: pse-pd: realtek-pse-mcu: configure global power budget
  2026-11-16 11:52 [PATCH net-next v2 0/3] net: pse-pd: realtek-pse-mcu: establish power management configuration Jonas Jelonek
  2026-11-16 11:52 ` [PATCH net-next v2 1/3] dt-bindings: net: pse-pd: realtek,pse-mcu: add VMAIN supply Jonas Jelonek
  2026-11-16 11:52 ` [PATCH net-next v2 2/3] net: pse-pd: realtek-pse-mcu: ensure MCU does dynamic power management Jonas Jelonek
@ 2026-11-16 11:52 ` Jonas Jelonek
  2026-11-19  5:54   ` netdev-bot+sashiko
  2 siblings, 1 reply; 6+ messages in thread
From: Jonas Jelonek @ 2026-11-16 11:52 UTC (permalink / raw)
  To: Oleksij Rempel, Kory Maincent, Andrew Lunn, David S . Miller,
	Eric Dumazet, Jakub Kicinski, Paolo Abeni, netdev, Rob Herring,
	Krzysztof Kozlowski, Conor Dooley
  Cc: linux-kernel, devicetree, Sander Vanheule, Russell Senior,
	Jonas Jelonek

Use the optional VMAIN supply's power budget to set the MCU's aggregate
PoE limit. Require a finite budget of at least 100 mW and round it down to
the protocol's 100 mW units. Reserve that amount through the regulator core
and program the limit at the MCU's current budget index before selecting
dynamic management and ungating the ports.

VMAIN must already be available: the MCU and downstream PSE managers
assume an active rail and provide no dedicated control for it. Use the
regulator only for budget accounting, leaving the rail externally
controlled or marked with regulator-always-on.

The regulator budget represents the usable PoE allowance. Vendor firmware
adds a device-specific guard on top: the UniFi USW Pro Max 24 PoE advertises
400 W but programs 404 W with a 4 W guard. The guard's effect on power
management is not sufficiently documented to apply a fixed percentage
across devices. Set it to zero rather than add an arbitrary margin to the
regulator-derived budget.

Register managed cleanup before hardware initialization so probe failures
and driver removal release the regulator reservation. Retain the programmed
MCU limit in either case. Boards without vmain-supply keep their existing
firmware-configured budget.

Signed-off-by: Jonas Jelonek <jonas@jonasjelonek.de>
---
 drivers/net/pse-pd/realtek-pse-mcu-core.c | 123 +++++++++++++++++++++-
 drivers/net/pse-pd/realtek-pse-mcu.h      |   3 +
 2 files changed, 123 insertions(+), 3 deletions(-)

diff --git a/drivers/net/pse-pd/realtek-pse-mcu-core.c b/drivers/net/pse-pd/realtek-pse-mcu-core.c
index d8b6e0503939..ae58581a0dad 100644
--- a/drivers/net/pse-pd/realtek-pse-mcu-core.c
+++ b/drivers/net/pse-pd/realtek-pse-mcu-core.c
@@ -34,10 +34,12 @@
 #include <linux/delay.h>
 #include <linux/gpio/consumer.h>
 #include <linux/jiffies.h>
+#include <linux/math.h>
 #include <linux/minmax.h>
 #include <linux/module.h>
 #include <linux/property.h>
 #include <linux/pse-pd/pse.h>
+#include <linux/regulator/consumer.h>
 #include <linux/unaligned.h>
 
 #include "realtek-pse-mcu.h"
@@ -62,6 +64,8 @@
 #define RTPSE_MCU_MAX_PORTS			48
 #define RTPSE_MCU_PORT_MAX_PRIORITY		3
 
+#define RTPSE_MCU_GLOBAL_BUDGET_LSB_MW		100
+
 /* Bounded resends when the MCU replies NOT_READY (busy). */
 #define RTPSE_MCU_NOT_READY_RETRIES		3
 
@@ -76,6 +80,8 @@ enum rtpse_mcu_cmd {
 	RTPSE_MCU_CMD_GET_SYSTEM_INFO,
 	RTPSE_MCU_CMD_GET_EXT_CONFIG,
 	RTPSE_MCU_CMD_SET_POWER_MGMT_MODE,
+	RTPSE_MCU_CMD_GET_POWER_STATUS,
+	RTPSE_MCU_CMD_SET_GLOBAL_BUDGET,
 
 	RTPSE_MCU_CMD_PORT_ENABLE,
 	RTPSE_MCU_CMD_PORT_SET_POWER_LIMIT_TYPE,
@@ -393,6 +399,84 @@ static int rtpse_mcu_set_power_mgmt_mode(struct rtpse_mcu_ctrl *pse)
 	return (resp.payload[0] == 0x0) ? 0 : -EIO;
 }
 
+static int rtpse_mcu_get_global_budget_idx(struct rtpse_mcu_ctrl *pse, u8 *budget_idx)
+{
+	struct rtpse_mcu_msg req, resp;
+	const struct rtpse_mcu_opcode *opc;
+	int ret;
+
+	opc = &pse->dialect->opcode[RTPSE_MCU_CMD_GET_POWER_STATUS];
+	if (!opc->valid)
+		return -EOPNOTSUPP;
+
+	rtpse_mcu_msg_init(&req, opc->op);
+	ret = rtpse_mcu_do_xfer(pse, &req, &resp);
+	if (ret)
+		return ret;
+
+	if (resp.payload[4] > 0x7)
+		return -ERANGE;
+
+	*budget_idx = resp.payload[4];
+	return 0;
+}
+
+static int rtpse_mcu_set_global_budget(struct rtpse_mcu_ctrl *pse, u8 budget_idx,
+				       u16 budget_raw)
+{
+	const struct rtpse_mcu_opcode *opc;
+	struct rtpse_mcu_msg req, resp;
+	int ret;
+
+	opc = &pse->dialect->opcode[RTPSE_MCU_CMD_SET_GLOBAL_BUDGET];
+	if (!opc->valid)
+		return -EOPNOTSUPP;
+
+	rtpse_mcu_msg_init(&req, opc->op);
+	req.payload[0] = budget_idx;
+	put_unaligned_be16(budget_raw, &req.payload[1]);
+	put_unaligned_be16(0, &req.payload[3]);
+
+	ret = rtpse_mcu_do_xfer(pse, &req, &resp);
+	if (ret)
+		return ret;
+	if (resp.payload[0] != budget_idx || resp.payload[1] != 0)
+		return -EIO;
+
+	return 0;
+}
+
+static int rtpse_mcu_apply_global_budget(struct rtpse_mcu_ctrl *pse,
+					 struct regulator *vmain)
+{
+	int budget, ret;
+	u16 budget_raw;
+	u8 budget_idx;
+
+	budget = regulator_get_unclaimed_power_budget(vmain);
+	if (budget == INT_MAX)
+		return -ENODATA;
+	if (budget < 0 || budget > U16_MAX * RTPSE_MCU_GLOBAL_BUDGET_LSB_MW)
+		return -ERANGE;
+
+	budget = rounddown(budget, RTPSE_MCU_GLOBAL_BUDGET_LSB_MW);
+	if (!budget)
+		return -ERANGE;
+	budget_raw = budget / RTPSE_MCU_GLOBAL_BUDGET_LSB_MW;
+
+	ret = regulator_request_power_budget(vmain, budget);
+	if (ret)
+		return ret;
+	pse->requested_budget_mW = budget;
+
+	ret = rtpse_mcu_get_global_budget_idx(pse, &budget_idx);
+	if (ret)
+		return ret;
+
+	/* Retain the programmed MCU limit even if later probe steps fail. */
+	return rtpse_mcu_set_global_budget(pse, budget_idx, budget_raw);
+}
+
 /* Port operations */
 
 static int rtpse_mcu_port_get_status(struct rtpse_mcu_ctrl *pse, unsigned int port,
@@ -801,6 +885,14 @@ static int rtpse_mcu_discover(struct rtpse_mcu_ctrl *pse, struct rtpse_mcu_info
 	return 0;
 }
 
+static void rtpse_mcu_free_global_budget(void *data)
+{
+	struct rtpse_mcu_ctrl *pse = data;
+
+	if (pse->requested_budget_mW)
+		regulator_free_power_budget(pse->vmain, pse->requested_budget_mW);
+}
+
 static void rtpse_mcu_global_disable(void *data)
 {
 	struct rtpse_mcu_ctrl *pse = data;
@@ -813,6 +905,7 @@ int rtpse_mcu_register(struct rtpse_mcu_ctrl *pse)
 	const struct rtpse_mcu_match_data *match;
 	struct rtpse_mcu_info info;
 	struct gpio_desc *gpiod;
+	struct regulator *vmain;
 	int ret;
 
 	BUILD_BUG_ON(sizeof(struct rtpse_mcu_msg) != RTPSE_MCU_MSG_SIZE);
@@ -837,6 +930,20 @@ int rtpse_mcu_register(struct rtpse_mcu_ctrl *pse)
 		return dev_err_probe(pse->dev, -EINVAL,
 				     "dialect for chip is incomplete\n");
 
+	vmain = devm_regulator_get_optional(pse->dev, "vmain");
+	if (PTR_ERR(vmain) == -ENODEV)
+		vmain = NULL;
+	else if (IS_ERR(vmain))
+		return dev_err_probe(pse->dev, PTR_ERR(vmain),
+				     "failed to get vmain supply\n");
+
+	if (vmain) {
+		pse->vmain = vmain;
+		ret = devm_add_action_or_reset(pse->dev, rtpse_mcu_free_global_budget, pse);
+		if (ret)
+			return ret;
+	}
+
 	/*
 	 * Release the MCU from reset before the first transaction; the
 	 * boot-retry loop in discover() waits for it to answer.
@@ -850,6 +957,13 @@ int rtpse_mcu_register(struct rtpse_mcu_ctrl *pse)
 	if (ret)
 		return ret;
 
+	if (vmain) {
+		ret = rtpse_mcu_apply_global_budget(pse, vmain);
+		if (ret)
+			return dev_err_probe(pse->dev, ret,
+					     "failed to apply global power budget\n");
+	}
+
 	/* ensure the MCU does dynamic power management with priorities */
 	ret = rtpse_mcu_set_power_mgmt_mode(pse);
 	if (ret)
@@ -858,9 +972,8 @@ int rtpse_mcu_register(struct rtpse_mcu_ctrl *pse)
 
 	/*
 	 * Some boards gate all ports through a hardware line; deassert it only
-	 * after the MCU is confirmed, so a discover failure never ungates the
-	 * ports. It is then left to the MCU - not re-gated on unbind or a later
-	 * probe error - so a driver reload doesn't black out PoE.
+	 * after discovery, budget setup and management-mode selection succeed.
+	 * Do not reassert it on unbind or a later probe error.
 	 */
 	gpiod = devm_gpiod_get_optional(pse->dev, "disable-ports", GPIOD_OUT_LOW);
 	if (IS_ERR(gpiod))
@@ -975,6 +1088,8 @@ static const struct rtpse_mcu_dialect rtpse_mcu_dialect_gen2 = {
 		[RTPSE_MCU_CMD_GET_SYSTEM_INFO]		= RTPSE_MCU_OP(0x40),
 		[RTPSE_MCU_CMD_GET_EXT_CONFIG]		= RTPSE_MCU_OP(0x4a),
 		[RTPSE_MCU_CMD_SET_POWER_MGMT_MODE]	= RTPSE_MCU_OP(0x10),
+		[RTPSE_MCU_CMD_GET_POWER_STATUS]	= RTPSE_MCU_OP(0x41),
+		[RTPSE_MCU_CMD_SET_GLOBAL_BUDGET]	= RTPSE_MCU_OP(0x04),
 
 		[RTPSE_MCU_CMD_PORT_ENABLE]		= RTPSE_MCU_OP(0x01),
 		[RTPSE_MCU_CMD_PORT_SET_POWER_LIMIT_TYPE] = RTPSE_MCU_OP(0x12),
@@ -996,6 +1111,8 @@ static const struct rtpse_mcu_dialect rtpse_mcu_dialect_gen1 = {
 		[RTPSE_MCU_CMD_GET_SYSTEM_INFO]		= RTPSE_MCU_OP(0x20),
 		[RTPSE_MCU_CMD_GET_EXT_CONFIG]		= RTPSE_MCU_OP(0x2b),
 		[RTPSE_MCU_CMD_SET_POWER_MGMT_MODE]	= RTPSE_MCU_OP(0x17),
+		[RTPSE_MCU_CMD_GET_POWER_STATUS]	= RTPSE_MCU_OP(0x23),
+		[RTPSE_MCU_CMD_SET_GLOBAL_BUDGET]	= RTPSE_MCU_OP(0x18),
 
 		[RTPSE_MCU_CMD_PORT_ENABLE]		= RTPSE_MCU_OP(0x00),
 		[RTPSE_MCU_CMD_PORT_SET_POWER_LIMIT_TYPE] = RTPSE_MCU_OP(0x15),
diff --git a/drivers/net/pse-pd/realtek-pse-mcu.h b/drivers/net/pse-pd/realtek-pse-mcu.h
index 5e31dd0993dd..56899ec4d76f 100644
--- a/drivers/net/pse-pd/realtek-pse-mcu.h
+++ b/drivers/net/pse-pd/realtek-pse-mcu.h
@@ -62,6 +62,7 @@ static inline bool rtpse_mcu_resp_is_final(const struct rtpse_mcu_msg *req,
 struct rtpse_mcu_dialect;
 struct rtpse_mcu_chip_info;
 struct rtpse_mcu_ctrl;
+struct regulator;
 
 /* Per-compatible match data (the of_match .data). */
 struct rtpse_mcu_match_data {
@@ -82,6 +83,8 @@ struct rtpse_mcu_ctrl {
 	const struct rtpse_mcu_dialect *dialect;
 	const struct rtpse_mcu_chip_info *chip;
 	const struct rtpse_mcu_transport_ops *transport;
+	struct regulator *vmain;
+	unsigned int requested_budget_mW;
 	u8 seq;			/* rolling request seq_num, echoed by the MCU */
 };
 
-- 
2.53.0


^ permalink raw reply related	[flat|nested] 6+ messages in thread

* Re: [PATCH net-next v2 2/3] net: pse-pd: realtek-pse-mcu: ensure MCU does dynamic power management
  2026-11-16 11:52 ` [PATCH net-next v2 2/3] net: pse-pd: realtek-pse-mcu: ensure MCU does dynamic power management Jonas Jelonek
@ 2026-11-19  5:54   ` netdev-bot+sashiko
  0 siblings, 0 replies; 6+ messages in thread
From: netdev-bot+sashiko @ 2026-11-19  5:54 UTC (permalink / raw)
  To: jonas
  Cc: o.rempel, kory.maincent, andrew+netdev, davem, edumazet, kuba,
	pabeni, netdev, robh, krzk+dt, conor+dt, linux-kernel, devicetree,
	sander, russell

Thank you for your contribution! Sashiko AI review found 3 potential
issue(s) to consider.

Critical: 0 · High: 0 · Medium: 1 · Low: 2

If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).

- [Medium] rtpse_mcu_register() (drivers/net/pse-pd/realtek-pse-mcu-
  core.c:854-857) now fails probe on any error from the new…
- [Low] The commit has no Fixes: tag, but it fixes a user-visible defect
  in the original driver commit 44566e614d84 ("net: pse-pd: add Realtek…
- [Low] In this commit, the comment above the 'disable-ports' GPIO
  request in rtpse_mcu_register() (drivers/net/pse-pd/realtek-pse-mcu-…

--

net: pse-pd: realtek-pse-mcu: ensure MCU does dynamic power management

This adds a command that switches the Realtek PSE MCU to dynamic power
management with port priorities. It uses opcode 0x17 on gen1 and 0x10 on
gen2. rtpse_mcu_register() sends the command after discovery and before
the disable-ports gpio is released, and probe fails if the command fails.

> The MCU may have any power management mode configuration by default
> but the driver should ensure that the MCU performs dynamic power
> management with priorities as advertised. On certain devices, the
> default MCU configuration ships with another default, possibly
> preventing any power delivery.

[Severity: Low]
This describes a user-visible problem in the original driver. Should the
patch carry a Fixes: tag?

Since commit 44566e614d84 ("net: pse-pd: add Realtek PSE MCU core"),
rtpse_mcu_register() has set PSE_BUDGET_EVAL_STRAT_DYNAMIC and
pis_prio_max. It has never put the MCU into the matching mode. That
commit's message says "Power budgeting is left to the MCU firmware; the
driver advertises PSE_BUDGET_EVAL_STRAT_DYNAMIC accordingly".

Without the tag, the stable and backport tooling won't pick this up for
kernels that already carry the driver. Perhaps:

Fixes: 44566e614d84 ("net: pse-pd: add Realtek PSE MCU core")

If the patch is meant to stay in net-next with the VMAIN budget work,
could the commit message still name the commit it fixes?

> diff --git a/drivers/net/pse-pd/realtek-pse-mcu-core.c b/drivers/net/pse-pd/realtek-pse-mcu-core.c
> index c70b09ca20122..d8b6e05039397 100644
> --- a/drivers/net/pse-pd/realtek-pse-mcu-core.c
> +++ b/drivers/net/pse-pd/realtek-pse-mcu-core.c

[ ... ]

> @@ -826,6 +850,12 @@ int rtpse_mcu_register(struct rtpse_mcu_ctrl *pse)
>  	if (ret)
>  		return ret;
>  
> +	/* ensure the MCU does dynamic power management with priorities */
> +	ret = rtpse_mcu_set_power_mgmt_mode(pse);
> +	if (ret)
> +		return dev_err_probe(pse->dev, ret,
> +				     "failed to set power management mode\n");

[Severity: Medium]
Could this break probe on boards that worked before this patch?

Both dialects now define the opcode (0x17 for gen1, 0x10 for gen2). That
means rtpse_mcu_set_power_mgmt_mode() never takes its -EOPNOTSUPP path.
It turns any non-zero status byte from the MCU into -EIO:

	return (resp.payload[0] == 0x0) ? 0 : -EIO;

A NOT_READY reply only gets the three resends inside
rtpse_mcu_do_xfer(). It does not get the bounded boot-retry window that
rtpse_mcu_discover() uses.

Some firmware variants may reject the command, or fail it with -EAGAIN,
-EBADMSG or -EBADE. In that case rtpse_mcu_register() returns before the
disable-ports gpio is requested and before devm_pse_controller_register()
runs.

On boards with the disable-ports gate, that leaves no PoE at all. On
boards without the gate, the MCU keeps its old policy but the kernel
registers no PSE controller. Before this patch, those boards probed as
long as the firmware default policy was usable.

Is it known that every supported gen1 and gen2 firmware accepts this
command? The changelog mentions "change error handling", so a softer
variant seems to have been considered. If warning and continuing is not
safe, could the commit message explain why the failure needs to be
fatal?

> +
>  	/*
>  	 * Some boards gate all ports through a hardware line; deassert it only
>  	 * after the MCU is confirmed, so a discover failure never ungates the

[Severity: Low]
This isn't a bug, but this comment still names only discovery as the step
that has to succeed before the ports are ungated. The new
rtpse_mcu_set_power_mgmt_mode() call above is now a precondition too.

The next patch in the series, "net: pse-pd: realtek-pse-mcu: configure
global power budget", rewrites this comment to "deassert it only after
discovery, budget setup and management-mode selection succeed". That
fixes it by the end of the series.

-- 
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261116115224.1454885-1-jonas%40jonasjelonek.de

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH net-next v2 3/3] net: pse-pd: realtek-pse-mcu: configure global power budget
  2026-11-16 11:52 ` [PATCH net-next v2 3/3] net: pse-pd: realtek-pse-mcu: configure global power budget Jonas Jelonek
@ 2026-11-19  5:54   ` netdev-bot+sashiko
  0 siblings, 0 replies; 6+ messages in thread
From: netdev-bot+sashiko @ 2026-11-19  5:54 UTC (permalink / raw)
  To: jonas
  Cc: o.rempel, kory.maincent, andrew+netdev, davem, edumazet, kuba,
	pabeni, netdev, robh, krzk+dt, conor+dt, linux-kernel, devicetree,
	sander, russell

Thank you for your contribution! Sashiko AI review found more severe
issue(s) in other patches of this series. This mail is only an FYI, in case
you want to address the nit picks below when you respin.

Low priority / nit picks for your consideration:

- [Low] rtpse_mcu_register() (drivers/net/pse-pd/realtek-pse-mcu-
  core.c:933) gets the optional 'vmain' supply with…
- [Low] rtpse_mcu_get_global_budget_idx() (drivers/net/pse-pd/realtek-
  pse-mcu-core.c:402-422) reads the MCU's current budget/power-bank…

--

net: pse-pd: realtek-pse-mcu: configure global power budget

This patch takes the power budget of the optional vmain supply and uses
it to program the MCU's aggregate PoE limit, in 100 mW units, at the
MCU's current budget index. The budget is reserved through the regulator
core and released by a devm action. The driver uses the regulator only
for budget accounting and never enables it.

> diff --git a/drivers/net/pse-pd/realtek-pse-mcu-core.c b/drivers/net/pse-pd/realtek-pse-mcu-core.c
> index d8b6e05039397..ae58581a0dad8 100644
> --- a/drivers/net/pse-pd/realtek-pse-mcu-core.c
> +++ b/drivers/net/pse-pd/realtek-pse-mcu-core.c

[ ... ]

> @@ -393,6 +399,84 @@ static int rtpse_mcu_set_power_mgmt_mode(struct rtpse_mcu_ctrl *pse)
>  	return (resp.payload[0] == 0x0) ? 0 : -EIO;
>  }
>  
> +static int rtpse_mcu_get_global_budget_idx(struct rtpse_mcu_ctrl *pse, u8 *budget_idx)
> +{

[ ... ]

> +	if (resp.payload[4] > 0x7)
> +		return -ERANGE;
> +
> +	*budget_idx = resp.payload[4];
> +	return 0;
> +}
> +

[ ... ]

> +static int rtpse_mcu_apply_global_budget(struct rtpse_mcu_ctrl *pse,
> +					 struct regulator *vmain)
> +{

[ ... ]

> +	ret = regulator_request_power_budget(vmain, budget);
> +	if (ret)
> +		return ret;
> +	pse->requested_budget_mW = budget;
> +
> +	ret = rtpse_mcu_get_global_budget_idx(pse, &budget_idx);
> +	if (ret)
> +		return ret;
> +
> +	/* Retain the programmed MCU limit even if later probe steps fail. */
> +	return rtpse_mcu_set_global_budget(pse, budget_idx, budget_raw);
> +}

[Severity: Low]
rtpse_mcu_get_global_budget_idx() accepts payload[4] values from 0 to 7,
which suggests the MCU has up to eight budget banks. The limit is written
only to the bank that is active at probe time, and the index is never
read again. The other banks keep their firmware or EEPROM limits.

What happens if the MCU firmware switches its active bank after probe?
For example, PSU presence inputs could change on a board with redundant
or hot-plug PSUs. Would the MCU then enforce a firmware default limit
that has nothing to do with requested_budget_mW? The reservation in the
regulator core and the limit enforced in hardware would then disagree.

It isn't clear that any of the supported boards switch banks at runtime.
Programming every bank with the same static value would not obviously be
correct either. Should this limitation at least be documented?

[ ... ]

> @@ -837,6 +930,20 @@ int rtpse_mcu_register(struct rtpse_mcu_ctrl *pse)
>  		return dev_err_probe(pse->dev, -EINVAL,
>  				     "dialect for chip is incomplete\n");
>  
> +	vmain = devm_regulator_get_optional(pse->dev, "vmain");
> +	if (PTR_ERR(vmain) == -ENODEV)
> +		vmain = NULL;
> +	else if (IS_ERR(vmain))
> +		return dev_err_probe(pse->dev, PTR_ERR(vmain),
> +				     "failed to get vmain supply\n");
> +

[Severity: Low]
Here vmain is used only for budget accounting. Nothing takes an enable
reference on it. There is no regulator_enable(), and no
regulator_is_enabled() check runs before the MCU budget is programmed,
the ports are ungated and the PSE controller is registered.

The PI regulators that pse_core registers take their supply from each
PI's vpwr, not from vmain, so vmain's use_count stays at zero.

Suppose a board describes vmain-supply as a switchable regulator without
regulator-always-on, such as a regulator-fixed with an enable GPIO, or
one marked only regulator-boot-on. Would regulator_late_cleanup() then
switch the PoE rail off after boot?

drivers/regulator/core.c:regulator_late_cleanup() {
    ...
	if (rdev->use_count)
		goto unlock;
    ...
	if (have_full_constraints()) {
		...
		ret = _regulator_do_disable(rdev);
    ...
}

In that case the MCU would still believe the programmed budget is
available, and the driver would not report an error.

The commit message says:

    Use the regulator only for budget accounting, leaving the rail
    externally controlled or marked with regulator-always-on.

The binding also says the supply is used for power-budget accounting
only, so such a DT would go against the documented expectation.

Would it be worth enforcing that requirement instead of assuming it?
That could be a regulator_is_enabled() check at probe, or a
regulator_enable() call, which would be a no-op on a truly fixed rail.

[ ... ]

-- 
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261116115224.1454885-1-jonas%40jonasjelonek.de

^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2026-11-19  5:54 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-16 11:52 [PATCH net-next v2 0/3] net: pse-pd: realtek-pse-mcu: establish power management configuration Jonas Jelonek
2026-11-16 11:52 ` [PATCH net-next v2 1/3] dt-bindings: net: pse-pd: realtek,pse-mcu: add VMAIN supply Jonas Jelonek
2026-11-16 11:52 ` [PATCH net-next v2 2/3] net: pse-pd: realtek-pse-mcu: ensure MCU does dynamic power management Jonas Jelonek
2026-11-19  5:54   ` netdev-bot+sashiko
2026-11-16 11:52 ` [PATCH net-next v2 3/3] net: pse-pd: realtek-pse-mcu: configure global power budget Jonas Jelonek
2026-11-19  5:54   ` netdev-bot+sashiko

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox