Re: [PATCH 2/2] hwmon: pmbus: Add support for Sony APS-379

From: Chris Packham

Date: Tue Mar 31 2026 - 20:27:05 EST



On 01/04/2026 12:56, Guenter Roeck wrote:
> Hi Chris,
>
> On 3/31/26 16:19, Chris Packham wrote:
>> Add pmbus support for Sony APS-379 power supplies. There are a few PMBUS
>> commands that return data that is undocumented/invalid so these need to
>> be rejected with -ENXIO. The READ_VOUT command returns data in linear11
>> format instead of linear16 so we need to workaround this.
>>
>> Signed-off-by: Chris Packham <chris.packham@xxxxxxxxxxxxxxxxxxx>
>
> Feedback inline.
>
> Thanks,
> Guenter
>
>> ---
>>   drivers/hwmon/pmbus/Kconfig   |   6 ++
>>   drivers/hwmon/pmbus/Makefile  |   1 +
>>   drivers/hwmon/pmbus/aps-379.c | 196 ++++++++++++++++++++++++++++++++++
>>   3 files changed, 203 insertions(+)
>>   create mode 100644 drivers/hwmon/pmbus/aps-379.c
>>
>> diff --git a/drivers/hwmon/pmbus/Kconfig b/drivers/hwmon/pmbus/Kconfig
>> index fc1273abe357..29076921e330 100644
>> --- a/drivers/hwmon/pmbus/Kconfig
>> +++ b/drivers/hwmon/pmbus/Kconfig
>> @@ -77,6 +77,12 @@ config SENSORS_ADP1050_REGULATOR
>>         µModule regulators that can provide microprocessor power from
>> 54V
>>         power distribution architecture.
>>   +config SENSORS_APS_379
>> +    tristate "Sony APS-379 Power Supplies"
>> +    help
>> +      If you say yes here you get hardware monitoring support for Sony
>> +      APS-379 Power Supplies.
>> +
>>   config SENSORS_BEL_PFE
>>       tristate "Bel PFE Compatible Power Supplies"
>>       help
>> diff --git a/drivers/hwmon/pmbus/Makefile b/drivers/hwmon/pmbus/Makefile
>> index d6c86924f887..94f36c7069ec 100644
>> --- a/drivers/hwmon/pmbus/Makefile
>> +++ b/drivers/hwmon/pmbus/Makefile
>> @@ -9,6 +9,7 @@ obj-$(CONFIG_SENSORS_ACBEL_FSG032) += acbel-fsg032.o
>>   obj-$(CONFIG_SENSORS_ADM1266)    += adm1266.o
>>   obj-$(CONFIG_SENSORS_ADM1275)    += adm1275.o
>>   obj-$(CONFIG_SENSORS_ADP1050)    += adp1050.o
>> +obj-$(CONFIG_SENSORS_APS_379)    += aps-379.o
>>   obj-$(CONFIG_SENSORS_BEL_PFE)    += bel-pfe.o
>>   obj-$(CONFIG_SENSORS_BPA_RS600)    += bpa-rs600.o
>>   obj-$(CONFIG_SENSORS_DELTA_AHE50DC_FAN) += delta-ahe50dc-fan.o
>> diff --git a/drivers/hwmon/pmbus/aps-379.c
>> b/drivers/hwmon/pmbus/aps-379.c
>> new file mode 100644
>> index 000000000000..e4c4c2d12bc9
>> --- /dev/null
>> +++ b/drivers/hwmon/pmbus/aps-379.c
>
> Driver documentation is missing.
>
> This power supply does not seem to be documented anywhere, so this is
> actually quite
> important.

I'll see what I can add. I do have a list of the actual supported PMBus
commands so that would be helpful to share.

>  Having said this, the behavior seems quite similar to BluTek
> BPA-RS600. Are those
> power supplies from the same OEM ?

The BPA-RS600 is a smaller form factor. BluTek do make PSUs in the same
from factor (there's a DC input one we use that I might get around to
upstreaming one day), but they are genuinely different suppliers and the
definitely don't interop electrically. Part of the reason for us using
the APS-379 is manufactured by a Japanese company which lets us sell it
into markets that have country of origin restrictions (that's as much
politics as I care to discuss). They all have different
quirks/deviations from the PMBus spec.

I did of course refer to the bpa-rs600 driver as I was familiar with it.
I pulled the linear11 vout workaround from mp5990 and adapted it for my
needs.

>
>> @@ -0,0 +1,196 @@
>> +// SPDX-License-Identifier: GPL-2.0-or-later
>> +/*
>> + * Hardware monitoring driver for Sony APS-379 Power Supplies
>> + *
>> + * Copyright 2026 Allied Telesis Labs
>> + */
>> +
>> +#include <linux/i2c.h>
>> +#include <linux/init.h>
>> +#include <linux/kernel.h>
>> +#include <linux/module.h>
>> +#include <linux/pmbus.h>
>> +#include "pmbus.h"
>> +
>> +struct aps_379_data {
>> +    struct pmbus_driver_info info;
>> +    u8 vout_linear_exponent;
>> +};
>> +
>> +#define to_aps_379_data(x) container_of(x, struct aps_379_data, info)
>> +
>> +static const struct i2c_device_id aps_379_id[] = {
>> +    { "aps-379", 0 },
>> +    {},
>> +};
>> +
>> +static int aps_379_read_byte_data(struct i2c_client *client, int
>> page, int reg)
>> +{
>> +    const struct pmbus_driver_info *info =
>> pmbus_get_driver_info(client);
>> +    struct aps_379_data *data = to_aps_379_data(info);
>> +    int ret;
>> +
>> +    if (page > 0)
>> +        return -ENXIO;
>
> Unnecessary since there is only one page.
>
> Yes, I know, other drivers do it, but it is really pointless.
>
Ack.
>> +
>> +    switch (reg) {
>> +    case PMBUS_VOUT_MODE:
>> +        /*
>> +         * The VOUT format used by the chip is linear11,
>> +         * not linear16. Report that VOUT is in linear mode
>> +         * and return exponent value extracted while probing
>> +         * the chip.
>> +         */
>> +        ret = data->vout_linear_exponent;
>> +        break;
>> +    default:
>> +        ret = -ENODATA;
>> +        break;
>> +    }
>> +
>> +    return ret;
>> +}
>> +
>> +/*
>> + * The APS-379 uses linear11 format instead of linear16. We've
>> reported the exponent
>> + * via the PMBUS_VOUT_MODE so we just return the mantissa here.
>> + */
>> +static int aps_379_read_vout(struct i2c_client *client)
>> +{
>> +    int ret;
>> +    s32 mantissa;
>> +
>> +    ret = pmbus_read_word_data(client, 0, 0xff, PMBUS_READ_VOUT);
>> +    if (ret < 0)
>> +        return ret;
>> +
>> +    mantissa = ((s16)((ret & 0x7ff) << 5)) >> 5;
>
> sign_extend32() ?
I'll look into it.
>
> Also, is the exponent known to be static ? If not it may be necessary
> to adjust it. If yes, I'd suggest to add a comment above.

I can't be 100% sure but the datasheet says the "scaling factor" for
READ_VOUT is 2^-4 (before going on to explain how to do the the linear11
decoding). I've only ever read -4 as the exponent so I think it's OK to
assume a fixed exponent.

>
>> +    ret = mantissa;
>
> That assignment is really unnecessary.
Ack.
>
>> +
>> +    return ret;
>> +}
>> +
>> +static int aps_379_read_word_data(struct i2c_client *client, int
>> page, int phase, int reg)
>> +{
>> +    int ret;
>> +
>> +    if (page > 0)
>> +        return -ENXIO;
>
> Unnecessary.
>
Ack.
>> +
>> +    switch (reg) {
>> +    case PMBUS_VOUT_UV_WARN_LIMIT:
>> +    case PMBUS_VOUT_OV_WARN_LIMIT:
>> +    case PMBUS_VOUT_UV_FAULT_LIMIT:
>> +    case PMBUS_VOUT_OV_FAULT_LIMIT:
>> +    case PMBUS_PIN_OP_WARN_LIMIT:
>> +    case PMBUS_POUT_OP_WARN_LIMIT:
>> +    case PMBUS_MFR_IIN_MAX:
>> +    case PMBUS_MFR_PIN_MAX:
>> +    case PMBUS_MFR_VOUT_MIN:
>> +    case PMBUS_MFR_VOUT_MAX:
>> +    case PMBUS_MFR_IOUT_MAX:
>> +    case PMBUS_MFR_POUT_MAX:
>> +    case PMBUS_MFR_MAX_TEMP_1:
>> +        /* These commands return data but it is
>> invalid/un-documented */
>> +        ret = -ENXIO;
>> +        break;
>
> I'd suggest to return directly in this function. There is no real value
> to assign the return value to ret just to return it.
>
Ack.
>> +    case PMBUS_READ_VOUT:
>> +        ret = aps_379_read_vout(client);
>> +        break;
>> +    default:
>> +        if (reg >= PMBUS_VIRT_BASE)
>> +            ret = -ENXIO;
>> +        else
>> +            ret = -ENODATA;
>> +        break;
>> +    }
>> +
>> +    return ret;
>> +
>> +}
>> +
>> +static struct pmbus_driver_info aps_379_info = {
>> +    .pages = 1,
>> +    .format[PSC_VOLTAGE_OUT] = linear,
>> +    .format[PSC_CURRENT_OUT] = linear,
>> +    .format[PSC_POWER] = linear,
>> +    .format[PSC_TEMPERATURE] = linear,
>> +    .format[PSC_FAN] = linear,
>> +    .func[0] = PMBUS_HAVE_VOUT |
>> +        PMBUS_HAVE_IOUT |
>> +        PMBUS_HAVE_PIN | PMBUS_HAVE_POUT |
>> +        PMBUS_HAVE_TEMP |
>> +        PMBUS_HAVE_FAN12,
>> +    .read_byte_data = aps_379_read_byte_data,
>> +    .read_word_data = aps_379_read_word_data,
>> +};
>> +
>> +static int aps_379_probe(struct i2c_client *client)
>> +{
>> +    struct device *dev = &client->dev;
>> +    const struct i2c_device_id *mid;
>> +    struct pmbus_driver_info *info;
>> +    struct aps_379_data *data;
>> +    u8 buf[I2C_SMBUS_BLOCK_MAX + 1];
>> +    int ret;
>> +
>> +    data = devm_kzalloc(&client->dev, sizeof(*data), GFP_KERNEL);
>> +    if (!data)
>> +        return -ENOMEM;
>> +
>> +    memcpy(&data->info, &aps_379_info, sizeof(*info));
>> +    info = &data->info;
>> +
>> +    if (!i2c_check_functionality(client->adapter,
>> +                     I2C_FUNC_SMBUS_READ_BYTE_DATA
>> +                     | I2C_FUNC_SMBUS_READ_WORD_DATA
>> +                     | I2C_FUNC_SMBUS_READ_BLOCK_DATA))
>> +        return -ENODEV;
>> +
>> +    ret = i2c_smbus_read_block_data(client, PMBUS_MFR_MODEL, buf);
>> +    if (ret < 0) {
>> +        dev_err(dev, "Failed to read Manufacturer Model\n");
>> +        return ret;
>> +    }
>> +
>> +    for (mid = aps_379_id; mid->name[0]; mid++) {
>> +        if (!strncasecmp(buf, mid->name, strlen(mid->name)))
>> +            break;
>> +    }
>
> That seems to be excessive. There is only one power supply.
> If more are added in the future, a looop can be added. Right
> now a simple comparison should do.

Sure. One of the things I borrowed from the BPA driver.

>
> Thanks,
> Guenter
>
>> +    if (!mid->name[0]) {
>> +        buf[ret] = '\0';
>> +        dev_err(dev, "Unsupported Manufacturer Model '%s'\n", buf);
>> +        return -ENODEV;
>> +    }
>> +
>> +    ret = i2c_smbus_read_word_data(client, PMBUS_READ_VOUT);
>> +    if (ret < 0) {
>> +        dev_err(dev, "Can't get vout exponent.\n");
>> +        return ret;
>> +    }
>> +    data->vout_linear_exponent = (u8)((ret >> 11) & 0x1f);
>> +
>> +    return pmbus_do_probe(client, info);
>> +}
>> +
>> +static const struct of_device_id __maybe_unused aps_379_of_match[] = {
>> +    { .compatible = "sony,aps-379" },
>> +    {},
>> +};
>> +MODULE_DEVICE_TABLE(of, aps_379_of_match);
>> +
>> +static struct i2c_driver aps_379_driver = {
>> +    .driver = {
>> +        .name = "aps-379",
>> +        .of_match_table = of_match_ptr(aps_379_of_match),
>> +    },
>> +    .probe = aps_379_probe,
>> +    .id_table = aps_379_id,
>> +};
>> +
>> +module_i2c_driver(aps_379_driver);
>> +
>> +MODULE_AUTHOR("Chris Packham");
>> +MODULE_DESCRIPTION("PMBus driver for Sony APS-379");
>> +MODULE_LICENSE("GPL");
>> +MODULE_IMPORT_NS("PMBUS");
>