[PATCH linux dev-5.10 07/14] pmbus: (max31785) Retry enabling fans after writing MFR_FAN_CONFIG

[Eddie James]

From: Andrew Jeffery <andrew at aj.id.au>

It has been observed across large fleets of systems that a small subset
of those systems occasionally loose control of some number of fans
across a BMC reboot (their hwmon fan attributes are missing from sysfs).

>From extensive testing and tracing it was discovered that writes
enabling a fan in FAN_CONFIG_1_2 failed to stick on the system under
test with a frequency of about 1 in 1000 re-binds of the driver.

The MAX31785 datasheet recommends in the documentation for
MFR_FAN_CONFIG that the asssociated fan(s) be disabled before updating
the register. The sequence in question implements this suggestion, and
the observed loss-of-fans symptom occurs when the write to re-enable the
fan in FAN_CONFIG_1_2 fails to stick.

The trace data suggests a one-shot retry is enough to successfully
update FAN_CONFIG_1_2. With the workaround, no loss of fans was observed
in over 20,000 consecutive rebinds of the driver.

OpenBMC-Staging-Count: 1
Signed-off-by: Andrew Jeffery <andrew at aj.id.au>
Signed-off-by: Joel Stanley <joel at jms.id.au>
---
 drivers/hwmon/pmbus/max31785.c | 23 +++++++++++++++++++----
 1 file changed, 19 insertions(+), 4 deletions(-)

diff --git a/drivers/hwmon/pmbus/max31785.c b/drivers/hwmon/pmbus/max31785.c
index 7518fff356f9..b37da2ec1ce4 100644
--- a/drivers/hwmon/pmbus/max31785.c
+++ b/drivers/hwmon/pmbus/max31785.c
@@ -398,6 +398,7 @@ static int max31785_of_fan_config(struct i2c_client *client,
 	u32 page;
 	u32 uval;
 	int ret;
+	int i;
 
 	if (!of_device_is_compatible(child, "pmbus-fan"))
 		return 0;
@@ -574,10 +575,24 @@ static int max31785_of_fan_config(struct i2c_client *client,
 	if (ret < 0)
 		return ret;
 
-	ret = max31785_i2c_smbus_write_byte_data(client, PMBUS_FAN_CONFIG_12,
-						 pb_cfg);
-	if (ret < 0)
-		return ret;
+	for (i = 0; i < 2; i++) {
+		ret = max31785_i2c_smbus_write_byte_data(client,
+							 PMBUS_FAN_CONFIG_12,
+							 pb_cfg);
+		if (ret < 0)
+			continue;
+
+		ret = max31785_i2c_smbus_read_byte_data(client,
+							PMBUS_FAN_CONFIG_12);
+		if (ret < 0)
+			continue;
+
+		if (ret == pb_cfg)
+			break;
+	}
+
+	if (i == 2)
+		return -EIO;
 
 	/*
 	 * Fans are on pages 0 - 5. If the page property of a fan node is
-- 
2.27.0