* [PATCH net-next v5 1/4] net: phy: Map C45 probe read errors to -ENODEV during bus scan
2026-11-20 19:43 [PATCH net-next v5 0/4] net: mdio: Clause 45 support for mdio-bcm-unimac Florian Fainelli
@ 2026-11-20 19:43 ` Florian Fainelli
2026-11-20 19:43 ` [PATCH net-next v5 2/4] net: mdio: mdio-bcm-unimac: Define MDIO_OP_SHIFT Florian Fainelli
` (2 subsequent siblings)
3 siblings, 0 replies; 8+ messages in thread
From: Florian Fainelli @ 2026-11-20 19:43 UTC (permalink / raw)
To: netdev
Cc: Florian Fainelli, Nicolai Buchwitz, Doug Berger,
Broadcom internal kernel review list, Andrew Lunn,
Heiner Kallweit, Russell King, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, open list, Mlody12 mlody12
get_phy_c22_id() maps a bus read error of -EIO or -ENODEV to -ENODEV, so
that mdiobus_scan_bus_c22() carries on with the next address when nothing
answers. get_phy_c45_ids() only does so for phy_c45_probe_present(); a
failed read of the devices in package registers is turned into -EIO, which
mdiobus_scan_bus_c45() treats as fatal and which makes the whole
__mdiobus_register() fail.
MDIO controllers that report a read from an empty address as a bus error
(e.g. MDIO_READ_FAIL on the UniMAC MDIO block) and that provide read_c45()
with unmasked addresses therefore cannot be registered: the C45 scan fails
on the first empty address. mdio-bcm-unimac does not provide read_c45()
yet, this is enabled by "net: mdio: mdio-bcm-unimac: Add Clause 45
support" later in this series.
Let get_phy_c45_devs_in_pkg() return the bus error and apply the same
mapping as get_phy_c22_id() to the two places that look for a device at an
address: the loop over the MMDs and the MMD 0 fallback. In the loop, an MMD
whose read fails that way is handled like one that reads as all Fs, except
that the remaining MMDs are not probed: the scan goes straight to MMD 0,
and -ENODEV is only returned if that does not answer either, so PHYs that
only respond on MMD 0 are still found while an empty address costs a
bounded number of bus reads.
Errors from the vendor MMD presence check done while reading the identifiers
and from the reads of the per-MMD identifiers themselves, which only happen
once a device has answered, are not mapped, so a PHY that fails midway is not
silently skipped. The kernel-doc therefore
documents "a negative error code" rather than -EIO as the only alternative.
This is not limited to mdiobus_scan_bus_c45(): get_phy_c45_ids() is reached
through get_phy_device(bus, addr, true) and phy_get_c45_ids() by every C45
caller. For those, a failed devices in package read is now reported as
-ENODEV instead of -EIO, which is what C22 callers already get:
- fwnode_mdiobus_register_phy() for "ethernet-phy-ieee802.3-c45" nodes:
__of_mdiobus_parse_phys() logs that the device is missing and registers
the bus without it, instead of failing the registration.
- sfp_sm_probe_phy(): a module whose PHY does not answer ends up with "no
PHY detected" after the usual retries, instead of going to SFP_S_FAIL.
- gpy_probe() through phy_get_c45_ids(): the driver core treats -ENODEV as
a silent match rejection rather than a probe failure.
Fixes: 17b447539408 ("net: phy: c45 scanning: Don't consider -ENODEV fatal")
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
---
drivers/net/phy/phy_device.c | 50 ++++++++++++++++++++++++++++--------
1 file changed, 40 insertions(+), 10 deletions(-)
diff --git a/drivers/net/phy/phy_device.c b/drivers/net/phy/phy_device.c
index b39e945cebd7..2f17c9d4a2b9 100644
--- a/drivers/net/phy/phy_device.c
+++ b/drivers/net/phy/phy_device.c
@@ -780,7 +780,7 @@ static int phy_c45_probe_present(struct mii_bus *bus, int prtad, int devad)
* Description: reads devices in package registers of a MMD at @dev_addr
* from PHY at @addr on @bus.
*
- * Returns: 0 on success, -EIO on failure.
+ * Returns: 0 on success, or the negative error code from the bus read.
*/
static int get_phy_c45_devs_in_pkg(struct mii_bus *bus, int addr, int dev_addr,
u32 *devices_in_package)
@@ -789,17 +789,30 @@ static int get_phy_c45_devs_in_pkg(struct mii_bus *bus, int addr, int dev_addr,
phy_reg = mdiobus_c45_read(bus, addr, dev_addr, MDIO_DEVS2);
if (phy_reg < 0)
- return -EIO;
+ return phy_reg;
*devices_in_package = phy_reg << 16;
phy_reg = mdiobus_c45_read(bus, addr, dev_addr, MDIO_DEVS1);
if (phy_reg < 0)
- return -EIO;
+ return phy_reg;
*devices_in_package |= phy_reg;
return 0;
}
+/* get_phy_c45_probe_err - map a bus error seen while looking for a device.
+ * @err: negative error code returned by a bus read
+ *
+ * Like get_phy_c22_id(), report -ENODEV when the bus read failed with -EIO or
+ * -ENODEV, which is how a controller signals that nothing answered at this
+ * address, so that bus scanning carries on with the next address. Anything
+ * else is reported as -EIO.
+ */
+static int get_phy_c45_probe_err(int err)
+{
+ return (err == -EIO || err == -ENODEV) ? -ENODEV : -EIO;
+}
+
/**
* get_phy_c45_ids - reads the specified addr for its 802.3-c45 IDs.
* @bus: the target MII bus
@@ -810,8 +823,10 @@ static int get_phy_c45_devs_in_pkg(struct mii_bus *bus, int addr, int dev_addr,
* the PHY identifiers for each device. Return the "devices in package"
* and identifiers in @c45_ids.
*
- * Returns zero on success, %-EIO on bus access error, or %-ENODEV if
- * the "devices in package" is invalid or no device responds.
+ * Returns zero on success, %-ENODEV if the "devices in package" is invalid or
+ * if no MMD responds (including when the bus reports -EIO or -ENODEV for the
+ * reads that look for a device), or a negative error code from a failed bus
+ * access otherwise (usually %-EIO).
*/
static int get_phy_c45_ids(struct mii_bus *bus, int addr,
struct phy_c45_device_ids *c45_ids)
@@ -844,8 +859,20 @@ static int get_phy_c45_ids(struct mii_bus *bus, int addr,
continue;
}
phy_reg = get_phy_c45_devs_in_pkg(bus, addr, i, &devs_in_pkg);
- if (phy_reg < 0)
- return -EIO;
+ if (phy_reg < 0) {
+ ret = get_phy_c45_probe_err(phy_reg);
+ if (ret != -ENODEV)
+ return ret;
+
+ /* The MMD did not answer: treat it like a read of all
+ * Fs, which is what controllers that do not check the
+ * turnaround return, and go straight to the MMD 0
+ * probe below. Do not probe the remaining MMDs, each
+ * one costs two bus reads on a slow bus.
+ */
+ devs_in_pkg = 0xffffffff;
+ break;
+ }
}
if ((devs_in_pkg & 0x1fffffff) == 0x1fffffff) {
@@ -855,7 +882,8 @@ static int get_phy_c45_ids(struct mii_bus *bus, int addr,
*/
phy_reg = get_phy_c45_devs_in_pkg(bus, addr, 0, &devs_in_pkg);
if (phy_reg < 0)
- return -EIO;
+ /* returning -ENODEV doesn't stop bus scanning */
+ return get_phy_c45_probe_err(phy_reg);
/* no device there, let's get out of here */
if ((devs_in_pkg & 0x1fffffff) == 0x1fffffff)
@@ -1077,8 +1105,10 @@ EXPORT_SYMBOL(phy_device_remove);
* phy_get_c45_ids - Read 802.3-c45 IDs for phy device.
* @phydev: phy_device structure to read 802.3-c45 IDs
*
- * Returns zero on success, %-EIO on bus access error, or %-ENODEV if
- * the "devices in package" is invalid.
+ * Returns zero on success, %-ENODEV if the "devices in package" is invalid or
+ * no MMD responds (this includes a bus read failing with -EIO or -ENODEV while
+ * looking for a device), or a negative error code from a failed bus access
+ * otherwise (usually %-EIO).
*/
int phy_get_c45_ids(struct phy_device *phydev)
{
--
2.34.1
^ permalink raw reply [flat|nested] 8+ messages in thread* [PATCH net-next v5 2/4] net: mdio: mdio-bcm-unimac: Define MDIO_OP_SHIFT
2026-11-20 19:43 [PATCH net-next v5 0/4] net: mdio: Clause 45 support for mdio-bcm-unimac Florian Fainelli
2026-11-20 19:43 ` [PATCH net-next v5 1/4] net: phy: Map C45 probe read errors to -ENODEV during bus scan Florian Fainelli
@ 2026-11-20 19:43 ` Florian Fainelli
2026-11-20 19:43 ` [PATCH net-next v5 3/4] net: mdio: mdio-bcm-unimac: Poll for completion at 1us granularity Florian Fainelli
2026-11-20 19:43 ` [PATCH net-next v5 4/4] net: mdio: mdio-bcm-unimac: Add Clause 45 support Florian Fainelli
3 siblings, 0 replies; 8+ messages in thread
From: Florian Fainelli @ 2026-11-20 19:43 UTC (permalink / raw)
To: netdev
Cc: Florian Fainelli, Andrew Lunn, Nicolai Buchwitz, Doug Berger,
Broadcom internal kernel review list, Heiner Kallweit,
Russell King, David S. Miller, Eric Dumazet, Jakub Kicinski,
Paolo Abeni, open list, Mlody12 mlody12
The MDIO_CMD operation code field starts at bit 26. Define
MDIO_OP_SHIFT for it and use it to build the opcodes instead of open
coding the shift.
While at it, rename MDIO_RD and MDIO_WR to MDIO_RD_C22 and MDIO_WR_C22 to
make it clear that these are the Clause 22 opcodes, in preparation for
adding the Clause 45 ones.
No functional change.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
---
drivers/net/mdio/mdio-bcm-unimac.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/drivers/net/mdio/mdio-bcm-unimac.c b/drivers/net/mdio/mdio-bcm-unimac.c
index 80fb3b5b52ab..5f79f31220f2 100644
--- a/drivers/net/mdio/mdio-bcm-unimac.c
+++ b/drivers/net/mdio/mdio-bcm-unimac.c
@@ -21,8 +21,9 @@
#define MDIO_CMD 0x00
#define MDIO_START_BUSY (1 << 29)
#define MDIO_READ_FAIL (1 << 28)
-#define MDIO_RD (2 << 26)
-#define MDIO_WR (1 << 26)
+#define MDIO_OP_SHIFT 26
+#define MDIO_RD_C22 (2 << MDIO_OP_SHIFT)
+#define MDIO_WR_C22 (1 << MDIO_OP_SHIFT)
#define MDIO_PMD_SHIFT 21
#define MDIO_PMD_MASK 0x1F
#define MDIO_REG_SHIFT 16
@@ -99,7 +100,8 @@ static int unimac_mdio_read(struct mii_bus *bus, int phy_id, int reg)
return ret;
/* Prepare the read operation */
- cmd = MDIO_RD | (phy_id << MDIO_PMD_SHIFT) | (reg << MDIO_REG_SHIFT);
+ cmd = MDIO_RD_C22 | (phy_id << MDIO_PMD_SHIFT) |
+ (reg << MDIO_REG_SHIFT);
unimac_mdio_writel(priv, cmd, MDIO_CMD);
/* Start MDIO transaction */
@@ -139,7 +141,7 @@ static int unimac_mdio_write(struct mii_bus *bus, int phy_id,
return ret;
/* Prepare the write operation */
- cmd = MDIO_WR | (phy_id << MDIO_PMD_SHIFT) |
+ cmd = MDIO_WR_C22 | (phy_id << MDIO_PMD_SHIFT) |
(reg << MDIO_REG_SHIFT) | (0xffff & val);
unimac_mdio_writel(priv, cmd, MDIO_CMD);
--
2.34.1
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH net-next v5 3/4] net: mdio: mdio-bcm-unimac: Poll for completion at 1us granularity
2026-11-20 19:43 [PATCH net-next v5 0/4] net: mdio: Clause 45 support for mdio-bcm-unimac Florian Fainelli
2026-11-20 19:43 ` [PATCH net-next v5 1/4] net: phy: Map C45 probe read errors to -ENODEV during bus scan Florian Fainelli
2026-11-20 19:43 ` [PATCH net-next v5 2/4] net: mdio: mdio-bcm-unimac: Define MDIO_OP_SHIFT Florian Fainelli
@ 2026-11-20 19:43 ` Florian Fainelli
2026-11-20 19:43 ` [PATCH net-next v5 4/4] net: mdio: mdio-bcm-unimac: Add Clause 45 support Florian Fainelli
3 siblings, 0 replies; 8+ messages in thread
From: Florian Fainelli @ 2026-11-20 19:43 UTC (permalink / raw)
To: netdev
Cc: Florian Fainelli, Andrew Lunn, Nicolai Buchwitz, Carlos Soto,
Doug Berger, Broadcom internal kernel review list,
Heiner Kallweit, Russell King, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, open list, Mlody12 mlody12
The current polling code unconditionally waits 30us, which was sized for
a ~25us Clause 22 transaction, and then polls MDIO_CMD with a sleep of up
to 2ms between samples.
Adding Clause 45 support means that every access becomes two back to back
transactions (an ADDR cycle followed by a READ/WRITE cycle), so a fixed
delay sized for a single C22 transaction is no longer meaningful, and a
sleep of up to 2ms between samples can add significant latency to each
access when the transaction has not completed at the first sample.
Instead, poll the MDIO_START_BUSY bit every microsecond so that the
completion is noticed as soon as it happens, regardless of the type of
transaction. This turns the wait into a busy wait, which is bounded by a
timeout reduced from 100ms to 1ms: a 64 bit MDIO frame takes about 33us at
the lowest MDIO clock the driver can program from a 250MHz reference clock,
so even a two frame C45 access completes well within that.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
Co-developed-by: Carlos Soto <carlos.soto@broadcom.com>
Signed-off-by: Carlos Soto <carlos.soto@broadcom.com>
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
---
drivers/net/mdio/mdio-bcm-unimac.c | 24 ++++++++++++++++--------
1 file changed, 16 insertions(+), 8 deletions(-)
diff --git a/drivers/net/mdio/mdio-bcm-unimac.c b/drivers/net/mdio/mdio-bcm-unimac.c
index 5f79f31220f2..f8e5c1811d96 100644
--- a/drivers/net/mdio/mdio-bcm-unimac.c
+++ b/drivers/net/mdio/mdio-bcm-unimac.c
@@ -6,8 +6,8 @@
*/
#include <linux/clk.h>
-#include <linux/delay.h>
#include <linux/io.h>
+#include <linux/iopoll.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/of.h>
@@ -36,6 +36,13 @@
#define MDIO_CLK_DIV_MASK 0x3F
#define MDIO_SUPP_PREAMBLE (1 << 12)
+/* A 64 bit MDIO frame takes ~25us at 2.5MHz and ~33us at the lowest MDIO
+ * clock the driver can program from a 250MHz reference clock. The timeout
+ * applies to the wait for one frame, so a 1ms timeout leaves a wide margin.
+ */
+#define MDIO_POLL_DELAY_US 1
+#define MDIO_POLL_TIMEOUT_US 1000
+
struct unimac_mdio_priv {
struct mii_bus *mii_bus;
void __iomem *base;
@@ -79,14 +86,15 @@ static int unimac_mdio_poll(void *wait_func_data)
struct unimac_mdio_priv *priv = wait_func_data;
u32 val;
- /*
- * C22 transactions should take ~25 usec, will need to adjust
- * if C45 support is added.
+ /* A frame should take ~25 usec. A C22 transaction is a single frame,
+ * a C45 transaction is an address frame followed by a data frame and
+ * each of them is waited for separately.
*/
- udelay(30);
-
- return read_poll_timeout(unimac_mdio_readl, val, !(val & MDIO_START_BUSY),
- 2000, 100000, false, priv, MDIO_CMD);
+ return read_poll_timeout_atomic(unimac_mdio_readl, val,
+ !(val & MDIO_START_BUSY),
+ MDIO_POLL_DELAY_US,
+ MDIO_POLL_TIMEOUT_US, false, priv,
+ MDIO_CMD);
}
static int unimac_mdio_read(struct mii_bus *bus, int phy_id, int reg)
--
2.34.1
^ permalink raw reply [flat|nested] 8+ messages in thread* [PATCH net-next v5 4/4] net: mdio: mdio-bcm-unimac: Add Clause 45 support
2026-11-20 19:43 [PATCH net-next v5 0/4] net: mdio: Clause 45 support for mdio-bcm-unimac Florian Fainelli
` (2 preceding siblings ...)
2026-11-20 19:43 ` [PATCH net-next v5 3/4] net: mdio: mdio-bcm-unimac: Poll for completion at 1us granularity Florian Fainelli
@ 2026-11-20 19:43 ` Florian Fainelli
2026-11-20 20:23 ` Nicolai Buchwitz
2026-11-20 20:29 ` Nicolai Buchwitz
3 siblings, 2 replies; 8+ messages in thread
From: Florian Fainelli @ 2026-11-20 19:43 UTC (permalink / raw)
To: netdev
Cc: Florian Fainelli, Carlos Soto, Justin Chen, Doug Berger,
Nicolai Buchwitz, Broadcom internal kernel review list,
Andrew Lunn, Heiner Kallweit, Russell King, David S. Miller,
Eric Dumazet, Jakub Kicinski, Paolo Abeni, open list,
Mlody12 mlody12
Add Clause 45 read/write support to the UniMAC MDIO bus controller driver
by implementing the read_c45/write_c45 bus callbacks. Each access performs
the two-step C45 sequence (ADDR cycle followed by a READ/WRITE cycle) under
the bus lock and switches MDIO_CFG between C22 and C45 mode as needed;
MDIO_CFG is only written when the mode actually changes.
Since MDIO_CFG also holds the clock divider, unimac_mdio_resume() now takes
the bus lock around unimac_mdio_clk_set() so that it cannot race with the
read-modify-write done by the accessors. Resume may therefore wait for an
MDIO access that is in flight.
Some wait_func implementations (e.g.: bcmgenet_mii_wait()) return 0 even
when the controller is still busy. For C45 this would let the data cycle
or the next access's MDIO_CFG switch start while a frame is in flight, and
return the command bits instead of the read data. Check that the controller
is idle after every cycle, C22 ones included, and report a timeout instead.
Also skip the BMSR workaround read in unimac_mdio_reset() for children
compatible with "ethernet-phy-ieee802.3-c45", since such PHYs are not
expected to respond to C22 accesses.
Note that providing read_c45/write_c45 also makes the MDIO core scan for
C45 PHYs at registration time, on addresses that are not masked and do not
already have a C22 device. This is not the case when the bus is registered
from a Device Tree node, which masks all addresses.
The platform device that GENET instantiates when it is probed without a
Device Tree node (e.g.: ACPI) does not mask any address, so it is scanned
too. Its wait_func is bcmgenet_mii_wait(), which sleeps for up to a timeout
per frame since GENET's interrupt is not registered yet when the bus is
registered. That makes the scan slower at probe time on such systems, the
same way the C22 scan already is.
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
Co-developed-by: Carlos Soto <carlos.soto@broadcom.com>
Signed-off-by: Carlos Soto <carlos.soto@broadcom.com>
Co-developed-by: Justin Chen <justin.chen@broadcom.com>
Signed-off-by: Justin Chen <justin.chen@broadcom.com>
---
drivers/net/mdio/mdio-bcm-unimac.c | 136 ++++++++++++++++++++++++++---
1 file changed, 126 insertions(+), 10 deletions(-)
diff --git a/drivers/net/mdio/mdio-bcm-unimac.c b/drivers/net/mdio/mdio-bcm-unimac.c
index f8e5c1811d96..b01e99327b56 100644
--- a/drivers/net/mdio/mdio-bcm-unimac.c
+++ b/drivers/net/mdio/mdio-bcm-unimac.c
@@ -24,10 +24,16 @@
#define MDIO_OP_SHIFT 26
#define MDIO_RD_C22 (2 << MDIO_OP_SHIFT)
#define MDIO_WR_C22 (1 << MDIO_OP_SHIFT)
+#define MDIO_ADDR_C45 (0 << MDIO_OP_SHIFT)
+#define MDIO_WR_C45 (1 << MDIO_OP_SHIFT)
+#define MDIO_RD_C45 (3 << MDIO_OP_SHIFT)
#define MDIO_PMD_SHIFT 21
#define MDIO_PMD_MASK 0x1F
#define MDIO_REG_SHIFT 16
#define MDIO_REG_MASK 0x1F
+#define MDIO_DEVAD_SHIFT 16
+#define MDIO_DEVAD_MASK GENMASK(20, 16)
+#define MDIO_C45_REG_MASK GENMASK(15, 0)
#define MDIO_CFG 0x04
#define MDIO_C22 (1 << 0)
@@ -97,7 +103,60 @@ static int unimac_mdio_poll(void *wait_func_data)
MDIO_CMD);
}
-static int unimac_mdio_read(struct mii_bus *bus, int phy_id, int reg)
+/* Wait for a cycle to complete. Some wait_func implementations (e.g.:
+ * bcmgenet_mii_wait()) return 0 even when they timed out with the controller
+ * still busy, in which case MDIO_CMD[15:0] holds the command rather than the
+ * data, and MDIO_CFG or MDIO_CMD must not be touched before the frame is
+ * over. Make sure the controller is idle after every cycle.
+ */
+static int unimac_mdio_wait(struct unimac_mdio_priv *priv)
+{
+ int ret;
+
+ ret = priv->wait_func(priv->wait_func_data);
+ if (ret)
+ return ret;
+
+ return unimac_mdio_poll(priv);
+}
+
+static u32 unimac_mdio_c45_devad(int devad)
+{
+ return ((u32)devad << MDIO_DEVAD_SHIFT) & MDIO_DEVAD_MASK;
+}
+
+static int unimac_mdio_addr_c45(struct unimac_mdio_priv *priv,
+ int phy_id, int devad, int reg)
+{
+ u32 cmd;
+
+ /* Prepare the addr c45 operation */
+ cmd = MDIO_ADDR_C45 | (phy_id << MDIO_PMD_SHIFT) |
+ unimac_mdio_c45_devad(devad) | (reg & MDIO_C45_REG_MASK);
+ unimac_mdio_writel(priv, cmd, MDIO_CMD);
+
+ /* Start MDIO transaction */
+ unimac_mdio_start(priv);
+
+ return unimac_mdio_wait(priv);
+}
+
+static void unimac_mdio_set_clause(struct unimac_mdio_priv *priv, bool is_c45)
+{
+ u32 cfg = unimac_mdio_readl(priv, MDIO_CFG);
+ u32 new_cfg = cfg;
+
+ if (is_c45)
+ new_cfg &= ~MDIO_C22;
+ else
+ new_cfg |= MDIO_C22;
+
+ if (new_cfg != cfg)
+ unimac_mdio_writel(priv, new_cfg, MDIO_CFG);
+}
+
+static int unimac_mdio_do_read(struct mii_bus *bus, int phy_id, int devad,
+ int reg, bool is_c45)
{
struct unimac_mdio_priv *priv = bus->priv;
int ret;
@@ -108,14 +167,25 @@ static int unimac_mdio_read(struct mii_bus *bus, int phy_id, int reg)
return ret;
/* Prepare the read operation */
- cmd = MDIO_RD_C22 | (phy_id << MDIO_PMD_SHIFT) |
- (reg << MDIO_REG_SHIFT);
+ unimac_mdio_set_clause(priv, is_c45);
+ if (is_c45) {
+ ret = unimac_mdio_addr_c45(priv, phy_id, devad, reg);
+ if (ret)
+ goto out;
+
+ cmd = MDIO_RD_C45 | (phy_id << MDIO_PMD_SHIFT) |
+ unimac_mdio_c45_devad(devad) |
+ (reg & MDIO_C45_REG_MASK);
+ } else {
+ cmd = MDIO_RD_C22 | (phy_id << MDIO_PMD_SHIFT) |
+ (reg << MDIO_REG_SHIFT);
+ }
unimac_mdio_writel(priv, cmd, MDIO_CMD);
/* Start MDIO transaction */
unimac_mdio_start(priv);
- ret = priv->wait_func(priv->wait_func_data);
+ ret = unimac_mdio_wait(priv);
if (ret)
goto out;
@@ -137,8 +207,8 @@ static int unimac_mdio_read(struct mii_bus *bus, int phy_id, int reg)
return ret;
}
-static int unimac_mdio_write(struct mii_bus *bus, int phy_id,
- int reg, u16 val)
+static int unimac_mdio_do_write(struct mii_bus *bus, int phy_id, int devad,
+ int reg, u16 val, bool is_c45)
{
struct unimac_mdio_priv *priv = bus->priv;
u32 cmd;
@@ -149,18 +219,50 @@ static int unimac_mdio_write(struct mii_bus *bus, int phy_id,
return ret;
/* Prepare the write operation */
- cmd = MDIO_WR_C22 | (phy_id << MDIO_PMD_SHIFT) |
- (reg << MDIO_REG_SHIFT) | (0xffff & val);
+ unimac_mdio_set_clause(priv, is_c45);
+ if (is_c45) {
+ ret = unimac_mdio_addr_c45(priv, phy_id, devad, reg);
+ if (ret)
+ goto out;
+
+ cmd = MDIO_WR_C45 | (phy_id << MDIO_PMD_SHIFT) |
+ unimac_mdio_c45_devad(devad) | (0xffff & val);
+ } else {
+ cmd = MDIO_WR_C22 | (phy_id << MDIO_PMD_SHIFT) |
+ (reg << MDIO_REG_SHIFT) | (0xffff & val);
+ }
unimac_mdio_writel(priv, cmd, MDIO_CMD);
unimac_mdio_start(priv);
- ret = priv->wait_func(priv->wait_func_data);
+ ret = unimac_mdio_wait(priv);
+out:
clk_disable_unprepare(priv->clk);
return ret;
}
+static int unimac_mdio_read(struct mii_bus *bus, int phy_id, int reg)
+{
+ return unimac_mdio_do_read(bus, phy_id, 0, reg, false);
+}
+
+static int unimac_mdio_write(struct mii_bus *bus, int phy_id, int reg, u16 val)
+{
+ return unimac_mdio_do_write(bus, phy_id, 0, reg, val, false);
+}
+
+static int unimac_mdio_read_c45(struct mii_bus *bus, int addr, int dev, int reg)
+{
+ return unimac_mdio_do_read(bus, addr, dev, reg, true);
+}
+
+static int unimac_mdio_write_c45(struct mii_bus *bus, int addr, int dev,
+ int reg, u16 val)
+{
+ return unimac_mdio_do_write(bus, addr, dev, reg, val, true);
+}
+
/* Workaround for integrated BCM7xxx Gigabit PHYs which have a problem with
* their internal MDIO management controller making them fail to successfully
* be read from or written to for the first transaction. We insert a dummy
@@ -191,6 +293,10 @@ static int unimac_mdio_reset(struct mii_bus *bus)
if (addr < 0)
continue;
+ if (of_device_is_compatible(child,
+ "ethernet-phy-ieee802.3-c45"))
+ continue;
+
read_mask |= 1 << addr;
}
}
@@ -302,6 +408,8 @@ static int unimac_mdio_probe(struct platform_device *pdev)
bus->parent = &pdev->dev;
bus->read = unimac_mdio_read;
bus->write = unimac_mdio_write;
+ bus->read_c45 = unimac_mdio_read_c45;
+ bus->write_c45 = unimac_mdio_write_c45;
bus->reset = unimac_mdio_reset;
snprintf(bus->id, MII_BUS_ID_SIZE, "%s-%d", pdev->name, pdev->id);
@@ -337,8 +445,16 @@ static void unimac_mdio_remove(struct platform_device *pdev)
static int __maybe_unused unimac_mdio_resume(struct device *d)
{
struct unimac_mdio_priv *priv = dev_get_drvdata(d);
+ int ret;
- return unimac_mdio_clk_set(priv);
+ /* MDIO_CFG is also modified by the MDIO accessors to switch between
+ * Clause 22 and Clause 45, which run under the bus lock.
+ */
+ mutex_lock(&priv->mii_bus->mdio_lock);
+ ret = unimac_mdio_clk_set(priv);
+ mutex_unlock(&priv->mii_bus->mdio_lock);
+
+ return ret;
}
static SIMPLE_DEV_PM_OPS(unimac_mdio_pm_ops,
--
2.34.1
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH net-next v5 4/4] net: mdio: mdio-bcm-unimac: Add Clause 45 support
2026-11-20 19:43 ` [PATCH net-next v5 4/4] net: mdio: mdio-bcm-unimac: Add Clause 45 support Florian Fainelli
@ 2026-11-20 20:23 ` Nicolai Buchwitz
2026-11-20 20:29 ` Nicolai Buchwitz
1 sibling, 0 replies; 8+ messages in thread
From: Nicolai Buchwitz @ 2026-11-20 20:23 UTC (permalink / raw)
To: Florian Fainelli
Cc: netdev, Carlos Soto, Justin Chen, Doug Berger,
Broadcom internal kernel review list, Andrew Lunn,
Heiner Kallweit, Russell King, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, linux-kernel, Mlody12 mlody12
On 20.11.2026 20:43, Florian Fainelli wrote:
> Add Clause 45 read/write support to the UniMAC MDIO bus controller
> driver
> by implementing the read_c45/write_c45 bus callbacks. Each access
> performs
> the two-step C45 sequence (ADDR cycle followed by a READ/WRITE cycle)
> under
> the bus lock and switches MDIO_CFG between C22 and C45 mode as needed;
> MDIO_CFG is only written when the mode actually changes.
>
> Since MDIO_CFG also holds the clock divider, unimac_mdio_resume() now
> takes
> the bus lock around unimac_mdio_clk_set() so that it cannot race with
> the
> read-modify-write done by the accessors. Resume may therefore wait for
> an
> MDIO access that is in flight.
>
> Some wait_func implementations (e.g.: bcmgenet_mii_wait()) return 0
> even
> when the controller is still busy. For C45 this would let the data
> cycle
> or the next access's MDIO_CFG switch start while a frame is in flight,
> and
> return the command bits instead of the read data. Check that the
> controller
> is idle after every cycle, C22 ones included, and report a timeout
> instead.
>
> Also skip the BMSR workaround read in unimac_mdio_reset() for children
> compatible with "ethernet-phy-ieee802.3-c45", since such PHYs are not
> expected to respond to C22 accesses.
>
> Note that providing read_c45/write_c45 also makes the MDIO core scan
> for
> C45 PHYs at registration time, on addresses that are not masked and do
> not
> already have a C22 device. This is not the case when the bus is
> registered
> from a Device Tree node, which masks all addresses.
>
> The platform device that GENET instantiates when it is probed without a
> Device Tree node (e.g.: ACPI) does not mask any address, so it is
> scanned
> too. Its wait_func is bcmgenet_mii_wait(), which sleeps for up to a
> timeout
> per frame since GENET's interrupt is not registered yet when the bus is
> registered. That makes the scan slower at probe time on such systems,
> the
> same way the C22 scan already is.
>
> Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
> Co-developed-by: Carlos Soto <carlos.soto@broadcom.com>
> Signed-off-by: Carlos Soto <carlos.soto@broadcom.com>
> Co-developed-by: Justin Chen <justin.chen@broadcom.com>
> Signed-off-by: Justin Chen <justin.chen@broadcom.com>
> ---
> drivers/net/mdio/mdio-bcm-unimac.c | 136 ++++++++++++++++++++++++++---
> 1 file changed, 126 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/net/mdio/mdio-bcm-unimac.c
> b/drivers/net/mdio/mdio-bcm-unimac.c
> index f8e5c1811d96..b01e99327b56 100644
> --- a/drivers/net/mdio/mdio-bcm-unimac.c
> +++ b/drivers/net/mdio/mdio-bcm-unimac.c
> @@ -24,10 +24,16 @@
> #define MDIO_OP_SHIFT 26
> #define MDIO_RD_C22 (2 << MDIO_OP_SHIFT)
> #define MDIO_WR_C22 (1 << MDIO_OP_SHIFT)
> +#define MDIO_ADDR_C45 (0 << MDIO_OP_SHIFT)
> +#define MDIO_WR_C45 (1 << MDIO_OP_SHIFT)
> +#define MDIO_RD_C45 (3 << MDIO_OP_SHIFT)
> #define MDIO_PMD_SHIFT 21
> #define MDIO_PMD_MASK 0x1F
> #define MDIO_REG_SHIFT 16
> #define MDIO_REG_MASK 0x1F
> +#define MDIO_DEVAD_SHIFT 16
> +#define MDIO_DEVAD_MASK GENMASK(20, 16)
> +#define MDIO_C45_REG_MASK GENMASK(15, 0)
>
> #define MDIO_CFG 0x04
> #define MDIO_C22 (1 << 0)
> @@ -97,7 +103,60 @@ static int unimac_mdio_poll(void *wait_func_data)
> MDIO_CMD);
> }
>
> -static int unimac_mdio_read(struct mii_bus *bus, int phy_id, int reg)
> +/* Wait for a cycle to complete. Some wait_func implementations (e.g.:
> + * bcmgenet_mii_wait()) return 0 even when they timed out with the
> controller
> + * still busy, in which case MDIO_CMD[15:0] holds the command rather
> than the
> + * data, and MDIO_CFG or MDIO_CMD must not be touched before the frame
> is
> + * over. Make sure the controller is idle after every cycle.
> + */
> +static int unimac_mdio_wait(struct unimac_mdio_priv *priv)
> +{
> + int ret;
> +
> + ret = priv->wait_func(priv->wait_func_data);
> + if (ret)
> + return ret;
> +
> + return unimac_mdio_poll(priv);
> +}
> +
> +static u32 unimac_mdio_c45_devad(int devad)
> +{
> + return ((u32)devad << MDIO_DEVAD_SHIFT) & MDIO_DEVAD_MASK;
> +}
> +
> +static int unimac_mdio_addr_c45(struct unimac_mdio_priv *priv,
> + int phy_id, int devad, int reg)
> +{
> + u32 cmd;
> +
> + /* Prepare the addr c45 operation */
> + cmd = MDIO_ADDR_C45 | (phy_id << MDIO_PMD_SHIFT) |
> + unimac_mdio_c45_devad(devad) | (reg & MDIO_C45_REG_MASK);
> + unimac_mdio_writel(priv, cmd, MDIO_CMD);
> +
> + /* Start MDIO transaction */
> + unimac_mdio_start(priv);
> +
> + return unimac_mdio_wait(priv);
> +}
> +
> +static void unimac_mdio_set_clause(struct unimac_mdio_priv *priv, bool
> is_c45)
> +{
> + u32 cfg = unimac_mdio_readl(priv, MDIO_CFG);
> + u32 new_cfg = cfg;
> +
> + if (is_c45)
> + new_cfg &= ~MDIO_C22;
> + else
> + new_cfg |= MDIO_C22;
> +
> + if (new_cfg != cfg)
> + unimac_mdio_writel(priv, new_cfg, MDIO_CFG);
> +}
> +
> +static int unimac_mdio_do_read(struct mii_bus *bus, int phy_id, int
> devad,
> + int reg, bool is_c45)
> {
> struct unimac_mdio_priv *priv = bus->priv;
> int ret;
> @@ -108,14 +167,25 @@ static int unimac_mdio_read(struct mii_bus *bus,
> int phy_id, int reg)
> return ret;
>
> /* Prepare the read operation */
> - cmd = MDIO_RD_C22 | (phy_id << MDIO_PMD_SHIFT) |
> - (reg << MDIO_REG_SHIFT);
> + unimac_mdio_set_clause(priv, is_c45);
> + if (is_c45) {
> + ret = unimac_mdio_addr_c45(priv, phy_id, devad, reg);
> + if (ret)
> + goto out;
> +
> + cmd = MDIO_RD_C45 | (phy_id << MDIO_PMD_SHIFT) |
> + unimac_mdio_c45_devad(devad) |
> + (reg & MDIO_C45_REG_MASK);
> + } else {
> + cmd = MDIO_RD_C22 | (phy_id << MDIO_PMD_SHIFT) |
> + (reg << MDIO_REG_SHIFT);
> + }
> unimac_mdio_writel(priv, cmd, MDIO_CMD);
>
> /* Start MDIO transaction */
> unimac_mdio_start(priv);
>
> - ret = priv->wait_func(priv->wait_func_data);
> + ret = unimac_mdio_wait(priv);
> if (ret)
> goto out;
>
> @@ -137,8 +207,8 @@ static int unimac_mdio_read(struct mii_bus *bus,
> int phy_id, int reg)
> return ret;
> }
>
> -static int unimac_mdio_write(struct mii_bus *bus, int phy_id,
> - int reg, u16 val)
> +static int unimac_mdio_do_write(struct mii_bus *bus, int phy_id, int
> devad,
> + int reg, u16 val, bool is_c45)
> {
> struct unimac_mdio_priv *priv = bus->priv;
> u32 cmd;
> @@ -149,18 +219,50 @@ static int unimac_mdio_write(struct mii_bus *bus,
> int phy_id,
> return ret;
>
> /* Prepare the write operation */
> - cmd = MDIO_WR_C22 | (phy_id << MDIO_PMD_SHIFT) |
> - (reg << MDIO_REG_SHIFT) | (0xffff & val);
> + unimac_mdio_set_clause(priv, is_c45);
> + if (is_c45) {
> + ret = unimac_mdio_addr_c45(priv, phy_id, devad, reg);
> + if (ret)
> + goto out;
> +
> + cmd = MDIO_WR_C45 | (phy_id << MDIO_PMD_SHIFT) |
> + unimac_mdio_c45_devad(devad) | (0xffff & val);
> + } else {
> + cmd = MDIO_WR_C22 | (phy_id << MDIO_PMD_SHIFT) |
> + (reg << MDIO_REG_SHIFT) | (0xffff & val);
> + }
> unimac_mdio_writel(priv, cmd, MDIO_CMD);
>
> unimac_mdio_start(priv);
>
> - ret = priv->wait_func(priv->wait_func_data);
> + ret = unimac_mdio_wait(priv);
> +out:
> clk_disable_unprepare(priv->clk);
>
> return ret;
> }
>
> +static int unimac_mdio_read(struct mii_bus *bus, int phy_id, int reg)
> +{
> + return unimac_mdio_do_read(bus, phy_id, 0, reg, false);
> +}
> +
> +static int unimac_mdio_write(struct mii_bus *bus, int phy_id, int reg,
> u16 val)
> +{
> + return unimac_mdio_do_write(bus, phy_id, 0, reg, val, false);
> +}
> +
> +static int unimac_mdio_read_c45(struct mii_bus *bus, int addr, int
> dev, int reg)
> +{
> + return unimac_mdio_do_read(bus, addr, dev, reg, true);
> +}
> +
> +static int unimac_mdio_write_c45(struct mii_bus *bus, int addr, int
> dev,
> + int reg, u16 val)
> +{
> + return unimac_mdio_do_write(bus, addr, dev, reg, val, true);
> +}
> +
> /* Workaround for integrated BCM7xxx Gigabit PHYs which have a problem
> with
> * their internal MDIO management controller making them fail to
> successfully
> * be read from or written to for the first transaction. We insert a
> dummy
> @@ -191,6 +293,10 @@ static int unimac_mdio_reset(struct mii_bus *bus)
> if (addr < 0)
> continue;
>
> + if (of_device_is_compatible(child,
> + "ethernet-phy-ieee802.3-c45"))
> + continue;
> +
> read_mask |= 1 << addr;
> }
> }
> @@ -302,6 +408,8 @@ static int unimac_mdio_probe(struct platform_device
> *pdev)
> bus->parent = &pdev->dev;
> bus->read = unimac_mdio_read;
> bus->write = unimac_mdio_write;
> + bus->read_c45 = unimac_mdio_read_c45;
> + bus->write_c45 = unimac_mdio_write_c45;
> bus->reset = unimac_mdio_reset;
> snprintf(bus->id, MII_BUS_ID_SIZE, "%s-%d", pdev->name, pdev->id);
>
> @@ -337,8 +445,16 @@ static void unimac_mdio_remove(struct
> platform_device *pdev)
> static int __maybe_unused unimac_mdio_resume(struct device *d)
> {
> struct unimac_mdio_priv *priv = dev_get_drvdata(d);
> + int ret;
>
> - return unimac_mdio_clk_set(priv);
> + /* MDIO_CFG is also modified by the MDIO accessors to switch between
> + * Clause 22 and Clause 45, which run under the bus lock.
> + */
> + mutex_lock(&priv->mii_bus->mdio_lock);
> + ret = unimac_mdio_clk_set(priv);
> + mutex_unlock(&priv->mii_bus->mdio_lock);
> +
> + return ret;
> }
>
> static SIMPLE_DEV_PM_OPS(unimac_mdio_pm_ops,
Reviewed-by: Nicolai Buchwitz <nb@tipi-net.de>
^ permalink raw reply [flat|nested] 8+ messages in thread* Re: [PATCH net-next v5 4/4] net: mdio: mdio-bcm-unimac: Add Clause 45 support
2026-11-20 19:43 ` [PATCH net-next v5 4/4] net: mdio: mdio-bcm-unimac: Add Clause 45 support Florian Fainelli
2026-11-20 20:23 ` Nicolai Buchwitz
@ 2026-11-20 20:29 ` Nicolai Buchwitz
2026-11-20 22:01 ` Florian Fainelli
1 sibling, 1 reply; 8+ messages in thread
From: Nicolai Buchwitz @ 2026-11-20 20:29 UTC (permalink / raw)
To: Florian Fainelli
Cc: netdev, Carlos Soto, Justin Chen, Doug Berger,
Broadcom internal kernel review list, Andrew Lunn,
Heiner Kallweit, Russell King, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, linux-kernel, Mlody12 mlody12
On 20.11.2026 20:43, Florian Fainelli wrote:
> Add Clause 45 read/write support to the UniMAC MDIO bus controller
> driver
> by implementing the read_c45/write_c45 bus callbacks. Each access
> performs
> the two-step C45 sequence (ADDR cycle followed by a READ/WRITE cycle)
> under
> the bus lock and switches MDIO_CFG between C22 and C45 mode as needed;
> MDIO_CFG is only written when the mode actually changes.
>
> Since MDIO_CFG also holds the clock divider, unimac_mdio_resume() now
> takes
> the bus lock around unimac_mdio_clk_set() so that it cannot race with
> the
> read-modify-write done by the accessors. Resume may therefore wait for
> an
> MDIO access that is in flight.
>
> Some wait_func implementations (e.g.: bcmgenet_mii_wait()) return 0
> even
> when the controller is still busy. For C45 this would let the data
> cycle
> or the next access's MDIO_CFG switch start while a frame is in flight,
> and
> return the command bits instead of the read data. Check that the
> controller
> is idle after every cycle, C22 ones included, and report a timeout
> instead.
>
> Also skip the BMSR workaround read in unimac_mdio_reset() for children
> compatible with "ethernet-phy-ieee802.3-c45", since such PHYs are not
> expected to respond to C22 accesses.
>
> Note that providing read_c45/write_c45 also makes the MDIO core scan
> for
> C45 PHYs at registration time, on addresses that are not masked and do
> not
> already have a C22 device. This is not the case when the bus is
> registered
> from a Device Tree node, which masks all addresses.
>
> The platform device that GENET instantiates when it is probed without a
> Device Tree node (e.g.: ACPI) does not mask any address, so it is
> scanned
> too. Its wait_func is bcmgenet_mii_wait(), which sleeps for up to a
> timeout
> per frame since GENET's interrupt is not registered yet when the bus is
> registered. That makes the scan slower at probe time on such systems,
> the
> same way the C22 scan already is.
>
> Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
> Co-developed-by: Carlos Soto <carlos.soto@broadcom.com>
> Signed-off-by: Carlos Soto <carlos.soto@broadcom.com>
> Co-developed-by: Justin Chen <justin.chen@broadcom.com>
> Signed-off-by: Justin Chen <justin.chen@broadcom.com>
Forgot to mention: Your tag needs to go last (v3 had it right).
You can keep my R-b as the changes are OK.
Thanks,
Nicolai
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH net-next v5 4/4] net: mdio: mdio-bcm-unimac: Add Clause 45 support
2026-11-20 20:29 ` Nicolai Buchwitz
@ 2026-11-20 22:01 ` Florian Fainelli
0 siblings, 0 replies; 8+ messages in thread
From: Florian Fainelli @ 2026-11-20 22:01 UTC (permalink / raw)
To: Nicolai Buchwitz
Cc: netdev, Carlos Soto, Justin Chen, Doug Berger,
Broadcom internal kernel review list, Andrew Lunn,
Heiner Kallweit, Russell King, David S. Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, linux-kernel, Mlody12 mlody12
On 11/20/26 12:29, Nicolai Buchwitz wrote:
> On 20.11.2026 20:43, Florian Fainelli wrote:
>> Add Clause 45 read/write support to the UniMAC MDIO bus controller driver
>> by implementing the read_c45/write_c45 bus callbacks. Each access
>> performs
>> the two-step C45 sequence (ADDR cycle followed by a READ/WRITE cycle)
>> under
>> the bus lock and switches MDIO_CFG between C22 and C45 mode as needed;
>> MDIO_CFG is only written when the mode actually changes.
>>
>> Since MDIO_CFG also holds the clock divider, unimac_mdio_resume() now
>> takes
>> the bus lock around unimac_mdio_clk_set() so that it cannot race with the
>> read-modify-write done by the accessors. Resume may therefore wait for an
>> MDIO access that is in flight.
>>
>> Some wait_func implementations (e.g.: bcmgenet_mii_wait()) return 0 even
>> when the controller is still busy. For C45 this would let the data cycle
>> or the next access's MDIO_CFG switch start while a frame is in flight,
>> and
>> return the command bits instead of the read data. Check that the
>> controller
>> is idle after every cycle, C22 ones included, and report a timeout
>> instead.
>>
>> Also skip the BMSR workaround read in unimac_mdio_reset() for children
>> compatible with "ethernet-phy-ieee802.3-c45", since such PHYs are not
>> expected to respond to C22 accesses.
>>
>> Note that providing read_c45/write_c45 also makes the MDIO core scan for
>> C45 PHYs at registration time, on addresses that are not masked and do
>> not
>> already have a C22 device. This is not the case when the bus is
>> registered
>> from a Device Tree node, which masks all addresses.
>>
>> The platform device that GENET instantiates when it is probed without a
>> Device Tree node (e.g.: ACPI) does not mask any address, so it is scanned
>> too. Its wait_func is bcmgenet_mii_wait(), which sleeps for up to a
>> timeout
>> per frame since GENET's interrupt is not registered yet when the bus is
>> registered. That makes the scan slower at probe time on such systems, the
>> same way the C22 scan already is.
>>
>> Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
>> Co-developed-by: Carlos Soto <carlos.soto@broadcom.com>
>> Signed-off-by: Carlos Soto <carlos.soto@broadcom.com>
>> Co-developed-by: Justin Chen <justin.chen@broadcom.com>
>> Signed-off-by: Justin Chen <justin.chen@broadcom.com>
>
> Forgot to mention: Your tag needs to go last (v3 had it right).
> You can keep my R-b as the changes are OK.
Indeed I did, thanks!
pw-bot: cr
--
Florian
^ permalink raw reply [flat|nested] 8+ messages in thread