[PATCH] hwmon: (pmbus/bpa-rs600): Handle Vin readings >= 256V

[PATCH] hwmon: (pmbus/bpa-rs600): Handle Vin readings >= 256V

[Chris Packham]

The BPA-RS600 doesn't follow the PMBus spec for linear data.
Specifically it treats the mantissa as an unsigned 11-bit value instead
of a two's complement 11-bit value. At this point it's unclear whether
this only affects Vin or if Pin/Pout1 are affected as well. Erring on
the side of caution only Vin is dealt with here.

Fixes: 15b2703e5e02 ("hwmon: (pmbus) Add driver for BluTek BPA-RS600")
Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
---
 drivers/hwmon/pmbus/bpa-rs600.c | 29 +++++++++++++++++++++++++++++
 1 file changed, 29 insertions(+)

diff --git a/drivers/hwmon/pmbus/bpa-rs600.c b/drivers/hwmon/pmbus/bpa-rs600.c
index f6558ee9dec3..2be69fedfa36 100644
--- a/drivers/hwmon/pmbus/bpa-rs600.c
+++ b/drivers/hwmon/pmbus/bpa-rs600.c
@@ -46,6 +46,32 @@ static int bpa_rs600_read_byte_data(struct i2c_client *client, int page, int reg
 	return ret;
 }
 
+/*
+ * The BPA-RS600 violates the PMBus spec. Specifically it treats the
+ * mantissa as unsigned. Deal with this here to allow the PMBus core
+ * to work with correctly encoded data.
+ */
+static int bpa_rs600_read_vin(struct i2c_client *client)
+{
+	int ret, exponent, mantissa;
+
+	ret = pmbus_read_word_data(client, 0, 0xff, PMBUS_READ_VIN);
+	if (ret < 0)
+		return ret;
+
+	if (ret & BIT(10)) {
+		exponent = ret >> 11;
+		mantissa = ret & 0x7ff;
+
+		exponent++;
+		mantissa >>= 1;
+
+		ret = (exponent << 11) | mantissa;
+	}
+
+	return ret;
+}
+
 static int bpa_rs600_read_word_data(struct i2c_client *client, int page, int phase, int reg)
 {
 	int ret;
@@ -85,6 +111,9 @@ static int bpa_rs600_read_word_data(struct i2c_client *client, int page, int pha
 		/* These commands return data but it is invalid/un-documented */
 		ret = -ENXIO;
 		break;
+	case PMBUS_READ_VIN:
+		ret = bpa_rs600_read_vin(client);
+		break;
 	default:
 		if (reg >= PMBUS_VIRT_BASE)
 			ret = -ENXIO;
-- 
2.31.1

Re: [PATCH] hwmon: (pmbus/bpa-rs600): Handle Vin readings >= 256V

[Guenter Roeck]

On Wed, Jun 16, 2021 at 03:42:18PM +1200, Chris Packham wrote:
> The BPA-RS600 doesn't follow the PMBus spec for linear data.
> Specifically it treats the mantissa as an unsigned 11-bit value instead
> of a two's complement 11-bit value. At this point it's unclear whether
> this only affects Vin or if Pin/Pout1 are affected as well. Erring on
> the side of caution only Vin is dealt with here.
> 
> Fixes: 15b2703e5e02 ("hwmon: (pmbus) Add driver for BluTek BPA-RS600")
> Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>

Applied.

Thanks,
Guenter