[PATCH net] net: phy: aquantia: wait for the suspend/resume operations to finish

[PATCH net] net: phy: aquantia: wait for the suspend/resume operations to finish

[Ioana Ciornei]

The Aquantia datasheet notes that after issuing a Processor-Intensive
MDIO operation, like changing the low-power state of the device, the
driver should wait for the operation to finish before issuing a new MDIO
command.

The new aqr107_wait_processor_intensive_op() function is added which can
be used after these kind of MDIO operations. At the moment, we are only
adding it at the end of the suspend/resume calls.

The issue was identified on a board featuring the AQR113C PHY, on
which commands like 'ip link (..) up / down' issued without any delays
between them would render the link on the PHY to remain down.
The issue was easy to reproduce with a one-liner:
 $ ip link set dev ethX down; ip link set dev ethX up; \
 ip link set dev ethX down; ip link set dev ethX up;

Fixes: ac9e81c230eb ("net: phy: aquantia: add suspend / resume callbacks for AQR107 family")
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
---
 drivers/net/phy/aquantia_main.c | 56 ++++++++++++++++++++++++++++++---
 1 file changed, 52 insertions(+), 4 deletions(-)

diff --git a/drivers/net/phy/aquantia_main.c b/drivers/net/phy/aquantia_main.c
index 8b7a46db30e0..8537ce58a597 100644
--- a/drivers/net/phy/aquantia_main.c
+++ b/drivers/net/phy/aquantia_main.c
@@ -91,6 +91,9 @@
 #define VEND1_GLOBAL_FW_ID_MAJOR		GENMASK(15, 8)
 #define VEND1_GLOBAL_FW_ID_MINOR		GENMASK(7, 0)
 
+#define VEND1_GLOBAL_GEN_STAT2			0xc831
+#define VEND1_GLOBAL_GEN_STAT2_OP_IN_PROG	BIT(15)
+
 #define VEND1_GLOBAL_RSVD_STAT1			0xc885
 #define VEND1_GLOBAL_RSVD_STAT1_FW_BUILD_ID	GENMASK(7, 4)
 #define VEND1_GLOBAL_RSVD_STAT1_PROV_ID		GENMASK(3, 0)
@@ -125,6 +128,12 @@
 #define VEND1_GLOBAL_INT_VEND_MASK_GLOBAL2	BIT(1)
 #define VEND1_GLOBAL_INT_VEND_MASK_GLOBAL3	BIT(0)
 
+/* Sleep and timeout for checking if the Processor-Intensive
+ * MDIO operation is finished
+ */
+#define AQR107_OP_IN_PROG_SLEEP		100
+#define AQR107_OP_IN_PROG_TIMEOUT	50000
+
 struct aqr107_hw_stat {
 	const char *name;
 	int reg;
@@ -597,16 +606,55 @@ static void aqr107_link_change_notify(struct phy_device *phydev)
 		phydev_info(phydev, "Aquantia 1000Base-T2 mode active\n");
 }
 
+static int aqr107_read_vend1_gen_stat2(struct phy_device *phydev)
+{
+	return phy_read_mmd(phydev, MDIO_MMD_VEND1, VEND1_GLOBAL_GEN_STAT2);
+}
+
+static int aqr107_wait_processor_intensive_op(struct phy_device *phydev)
+{
+	int val, err;
+
+	/* The datasheet notes to wait at least 1ms after issuing a
+	 * processor intensive operation before checking.
+	 * We cannot use the 'sleep_before_read' parameter of read_poll_timeout
+	 * because that just determines the maximum time slept, not the minimum.
+	 */
+	usleep_range(1000, 5000);
+
+	err = readx_poll_timeout(aqr107_read_vend1_gen_stat2, phydev,
+				 val, !(val & VEND1_GLOBAL_GEN_STAT2_OP_IN_PROG),
+				 AQR107_OP_IN_PROG_SLEEP, AQR107_OP_IN_PROG_TIMEOUT);
+	if (err) {
+		phydev_err(phydev, "timeout: processor-intensive MDIO operation\n");
+		return err;
+	}
+
+	return 0;
+}
+
 static int aqr107_suspend(struct phy_device *phydev)
 {
-	return phy_set_bits_mmd(phydev, MDIO_MMD_VEND1, MDIO_CTRL1,
-				MDIO_CTRL1_LPOWER);
+	int err;
+
+	err = phy_set_bits_mmd(phydev, MDIO_MMD_VEND1, MDIO_CTRL1,
+			       MDIO_CTRL1_LPOWER);
+	if (err)
+		return err;
+
+	return aqr107_wait_processor_intensive_op(phydev);
 }
 
 static int aqr107_resume(struct phy_device *phydev)
 {
-	return phy_clear_bits_mmd(phydev, MDIO_MMD_VEND1, MDIO_CTRL1,
-				  MDIO_CTRL1_LPOWER);
+	int err;
+
+	err = phy_clear_bits_mmd(phydev, MDIO_MMD_VEND1, MDIO_CTRL1,
+				 MDIO_CTRL1_LPOWER);
+	if (err)
+		return err;
+
+	return aqr107_wait_processor_intensive_op(phydev);
 }
 
 static int aqr107_probe(struct phy_device *phydev)
-- 
2.33.1

Re: [PATCH net] net: phy: aquantia: wait for the suspend/resume operations to finish

[Ioana Ciornei]

On Mon, Sep 05, 2022 at 04:58:33PM +0300, Ioana Ciornei wrote:
> The Aquantia datasheet notes that after issuing a Processor-Intensive
> MDIO operation, like changing the low-power state of the device, the
> driver should wait for the operation to finish before issuing a new MDIO
> command.
> 
> The new aqr107_wait_processor_intensive_op() function is added which can
> be used after these kind of MDIO operations. At the moment, we are only
> adding it at the end of the suspend/resume calls.
> 
> The issue was identified on a board featuring the AQR113C PHY, on
> which commands like 'ip link (..) up / down' issued without any delays
> between them would render the link on the PHY to remain down.
> The issue was easy to reproduce with a one-liner:
>  $ ip link set dev ethX down; ip link set dev ethX up; \
>  ip link set dev ethX down; ip link set dev ethX up;
> 
> Fixes: ac9e81c230eb ("net: phy: aquantia: add suspend / resume callbacks for AQR107 family")
> Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>

I just stumbled upon the phy_read_mmd_poll_timeout() API which takes
care of polling on an MDIO register without having to declare a separate
read function.

I'll send a v2 on this with the change later today or tomorrow.

Ioana