Re: [PATCH 1/3] hwmon: (pmbus) Add driver for Delta DPS-920AB PSU
From: Guenter Roeck
Date: Thu May 27 2021 - 10:50:09 EST
On 5/27/21 7:17 AM, Robert Marko wrote:
On Thu, May 27, 2021 at 4:03 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
On 5/27/21 3:59 AM, Robert Marko wrote:
On Thu, May 27, 2021 at 12:22 PM Robert Marko <robert.marko@xxxxxxxxxx> wrote:
On Mon, May 24, 2021 at 3:47 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
On 5/24/21 5:21 AM, Robert Marko wrote:
On Fri, May 21, 2021 at 4:46 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
On 5/21/21 4:56 AM, Robert Marko wrote:
On Fri, May 21, 2021 at 12:56 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
On 5/21/21 1:36 AM, Robert Marko wrote:
On Wed, May 19, 2021 at 3:19 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
On 5/19/21 5:38 AM, Robert Marko wrote:
On Fri, Apr 30, 2021 at 3:48 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
On Fri, Apr 30, 2021 at 03:27:33PM +0200, Robert Marko wrote:
This adds support for the Delta DPS-920AB PSU.
Only missing feature is fan control which the PSU supports.
Signed-off-by: Robert Marko <robert.marko@xxxxxxxxxx>
---
Documentation/hwmon/dps920ab.rst | 80 ++++++++++++++++++++++++++++++++
Documentation/hwmon/index.rst | 1 +
drivers/hwmon/pmbus/Kconfig | 9 ++++
drivers/hwmon/pmbus/Makefile | 1 +
drivers/hwmon/pmbus/dps920ab.c | 63 +++++++++++++++++++++++++
5 files changed, 154 insertions(+)
create mode 100644 Documentation/hwmon/dps920ab.rst
create mode 100644 drivers/hwmon/pmbus/dps920ab.c
diff --git a/Documentation/hwmon/dps920ab.rst b/Documentation/hwmon/dps920ab.rst
new file mode 100644
index 000000000000..df0aef530c7e
--- /dev/null
+++ b/Documentation/hwmon/dps920ab.rst
@@ -0,0 +1,80 @@
+.. SPDX-License-Identifier: GPL-2.0-or-later
+
+Kernel driver dps920ab
+========================
+
+Supported chips:
+
+ * Delta DPS920AB
+
+ Prefix: 'dps920ab'
+
+ Addresses scanned: -
+
+Authors:
+ Robert Marko <robert.marko@xxxxxxxxxx>
+
+
+Description
+-----------
+
+This driver implements support for Delta DPS920AB 920W 54V DC single output
+power supply with PMBus support.
+
+The driver is a client driver to the core PMBus driver.
+Please see Documentation/hwmon/pmbus.rst for details on PMBus client drivers.
+
+
+Usage Notes
+-----------
+
+This driver does not auto-detect devices. You will have to instantiate the
+devices explicitly. Please see Documentation/i2c/instantiating-devices.rst for
+details.
+
+
+Sysfs entries
+-------------
+
+======================= ======================================================
+curr1_label "iin"
+curr1_input Measured input current
+curr1_crit Critical maximum current
+curr1_crit_alarm Current critical high alarm
+
+curr2_label "iout1"
+curr2_input Measured output current
+curr2_crit Critical maximum current
+curr2_crit_alarm Current critical high alarm
+
+in1_label "vin"
+in1_input Measured input voltage
+in1_lcrit Critical minimum input voltage
+in1_lcrit_alarm Input voltage critical low alarm
+in1_crit Critical maximum input voltage
+in1_crit_alarm Input voltage critical high alarm
+
+in2_label "vout1"
+in2_input Measured output voltage
+in2_lcrit Critical minimum output voltage
+in2_lcrit_alarm Output voltage critical low alarm
+in2_crit Critical maximum output voltage
+in2_crit_alarm Output voltage critical high alarm
+
+power1_label "pin"
+power1_input Measured input power
+power1_alarm Input power high alarm
+
+power2_label "pout1"
+power2_input Measured output power
+
+temp[1-2]_input Measured temperature
+temp[1-2]_crit Critical high temperature
+temp[1-2]_crit_alarm Chip temperature critical high alarm
+temp[1-2]_max Maximum temperature
+temp[1-2]_max_alarm Chip temperature high alarm
+
+fan1_alarm Fan 1 warning.
+fan1_fault Fan 1 fault.
+fan1_input Fan 1 speed in RPM.
+======================= ======================================================
diff --git a/Documentation/hwmon/index.rst b/Documentation/hwmon/index.rst
index 8d5a2df1ecb6..b24436f22052 100644
--- a/Documentation/hwmon/index.rst
+++ b/Documentation/hwmon/index.rst
@@ -54,6 +54,7 @@ Hardware Monitoring Kernel Drivers
dell-smm-hwmon
dme1737
drivetemp
+ dps920ab
ds1621
ds620
emc1403
diff --git a/drivers/hwmon/pmbus/Kconfig b/drivers/hwmon/pmbus/Kconfig
index 32d2fc850621..865ade0aa205 100644
--- a/drivers/hwmon/pmbus/Kconfig
+++ b/drivers/hwmon/pmbus/Kconfig
@@ -66,6 +66,15 @@ config SENSORS_IBM_CFFPS
This driver can also be built as a module. If so, the module will
be called ibm-cffps.
+config SENSORS_DPS920AB
+ tristate "Delta DPS920AB Power Supply"
+ help
+ If you say yes here you get hardware monitoring support for Delta
+ DPS920AB Power Supplies.
+
+ This driver can also be built as a module. If so, the module will
+ be called dps920ab.
+
config SENSORS_INSPUR_IPSPS
tristate "INSPUR Power System Power Supply"
help
diff --git a/drivers/hwmon/pmbus/Makefile b/drivers/hwmon/pmbus/Makefile
index 6a4ba0fdc1db..f59ba0123d68 100644
--- a/drivers/hwmon/pmbus/Makefile
+++ b/drivers/hwmon/pmbus/Makefile
@@ -9,6 +9,7 @@ obj-$(CONFIG_SENSORS_ADM1266) += adm1266.o
obj-$(CONFIG_SENSORS_ADM1275) += adm1275.o
obj-$(CONFIG_SENSORS_BEL_PFE) += bel-pfe.o
obj-$(CONFIG_SENSORS_IBM_CFFPS) += ibm-cffps.o
+obj-$(CONFIG_SENSORS_DPS920AB) += dps920ab.o
obj-$(CONFIG_SENSORS_INSPUR_IPSPS) += inspur-ipsps.o
obj-$(CONFIG_SENSORS_IR35221) += ir35221.o
obj-$(CONFIG_SENSORS_IR38064) += ir38064.o
diff --git a/drivers/hwmon/pmbus/dps920ab.c b/drivers/hwmon/pmbus/dps920ab.c
new file mode 100644
index 000000000000..d579ed9f879c
--- /dev/null
+++ b/drivers/hwmon/pmbus/dps920ab.c
@@ -0,0 +1,63 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Driver for Delta DPS920AB PSU
+ *
+ * Copyright (C) 2021 Delta Networks, Inc.
+ * Copyright (C) 2021 Sartura Ltd.
+ */
+
+#include <linux/i2c.h>
+#include <linux/module.h>
+#include <linux/of_device.h>
+#include "pmbus.h"
+
+static struct pmbus_driver_info dps920ab_info = {
+ .pages = 1,
+
+ .format[PSC_VOLTAGE_IN] = linear,
+ .format[PSC_VOLTAGE_OUT] = linear,
+ .format[PSC_CURRENT_IN] = linear,
+ .format[PSC_CURRENT_OUT] = linear,
+ .format[PSC_POWER] = linear,
+ .format[PSC_FAN] = linear,
+ .format[PSC_TEMPERATURE] = linear,
+
+ .func[0] = PMBUS_HAVE_VIN | PMBUS_HAVE_IIN |
+ PMBUS_HAVE_VOUT | PMBUS_HAVE_STATUS_VOUT |
+ PMBUS_HAVE_IOUT | PMBUS_HAVE_STATUS_IOUT |
+ PMBUS_HAVE_TEMP | PMBUS_HAVE_TEMP2 |
+ PMBUS_HAVE_PIN | PMBUS_HAVE_POUT |
+ PMBUS_HAVE_FAN12 | PMBUS_HAVE_STATUS_FAN12 |
+ PMBUS_HAVE_STATUS_INPUT | PMBUS_HAVE_STATUS_TEMP,
+};
+
+static int dps920ab_probe(struct i2c_client *client)
+{
+ if (!i2c_check_functionality(client->adapter,
+ I2C_FUNC_SMBUS_BYTE_DATA |
+ I2C_FUNC_SMBUS_WORD_DATA))
+ return -ENODEV;
This check is done in pmbus_do_probe(), and repeating it here does not add
any value.
Ok, makes sense.
That makes me wonder: Is this driver needed in the first place, or could
it be added to drivers/hwmon/pmbus/pmbus.c ?
It could be added as a generic driver, but that creates properties in
the sysfs that this thing
does not support like 3 voltage readings, 3 fans, and so on.
Can you be more specific ? What additional unsupported attributes
are added, and why ? Are you saying the PSU reports registers as
existing which do not really exist ? If so, which registers are those ?
Sure, when core does probing and autodiscovery, then the following
sysfs attributes are created:
curr1_crit fan3_fault in3_min_alarm temp1_max
curr1_crit_alarm fan3_input in3_rated_max temp1_max_alarm
curr1_input fan3_target in3_rated_min temp1_min
curr1_label in1_crit name temp1_min_alarm
curr1_max in1_crit_alarm of_node temp1_rated_max
curr1_max_alarm in1_input power temp2_crit
curr1_rated_max in1_label power1_alarm temp2_crit_alarm
curr2_crit in1_lcrit power1_input temp2_input
curr2_crit_alarm in1_lcrit_alarm power1_label temp2_lcrit
curr2_input in1_max power1_max temp2_lcrit_alarm
curr2_label in1_max_alarm power1_rated_max temp2_max
curr2_lcrit in1_min power2_cap temp2_max_alarm
curr2_lcrit_alarm in1_min_alarm power2_cap_alarm temp2_min
curr2_max in1_rated_max power2_crit temp2_min_alarm
curr2_max_alarm in1_rated_min power2_crit_alarm temp2_rated_max
curr2_rated_max in2_input power2_input temp3_crit
device in2_label power2_label temp3_crit_alarm
fan1_alarm in3_crit power2_max temp3_input
fan1_fault in3_crit_alarm power2_max_alarm temp3_lcrit
fan1_input in3_input power2_rated_max temp3_lcrit_alarm
fan1_target in3_label subsystem temp3_max
fan2_alarm in3_lcrit temp1_crit temp3_max_alarm
fan2_fault in3_lcrit_alarm temp1_crit_alarm temp3_min
fan2_input in3_max temp1_input temp3_min_alarm
fan2_target in3_max_alarm temp1_lcrit temp3_rated_max
fan3_alarm in3_min temp1_lcrit_alarm uevent
The following return -1, or -500 so they are not supported.
* fan2
* in2
Weirdly, with the external driver both fan2 and in2 are enabled and work fine,
but when auto probing they are fan3 and in3.
temp3 actually seems to return a valid temperature despite it not being used in
the vendor driver that features were picked from.
Can you run "grep . *" in the hwmon directory so I can see actual values ?
Sure:
curr1_crit:-500
curr1_crit_alarm:0
curr1_input:195
curr1_label:iin
curr1_max:-500
curr1_max_alarm:0
curr1_rated_max:-500
curr2_crit:-500
curr2_crit_alarm:0
curr2_input:320
curr2_label:iout1
curr2_lcrit:-500
curr2_lcrit_alarm:0
curr2_max:18625
curr2_max_alarm:0
curr2_rated_max:16875
grep: device: Is a directory
fan1_alarm:0
fan1_fault:0
fan1_input:10000
fan1_target:0
fan2_alarm:1
fan2_fault:1
fan2_input:-1
fan2_target:-1
fan3_alarm:1
fan3_fault:1
fan3_input:-1
fan3_target:-1
in1_crit:-500
in1_crit_alarm:0
in1_input:245250
in1_label:vin
in1_lcrit:-500
in1_lcrit_alarm:0
in1_max:-500
in1_max_alarm:0
in1_min:-500
in1_min_alarm:0
in1_rated_max:-500
in1_rated_min:-500
in2_input:-500
in2_label:vcap
in3_crit:255996
in3_crit_alarm:0
in3_input:54511
in3_label:vout1
in3_lcrit:255996
in3_lcrit_alarm:0
in3_max:255996
in3_max_alarm:0
in3_min:255996
in3_min_alarm:0
in3_rated_max:56136
in3_rated_min:52863
name:dps920ab
grep: of_node: Is a directory
grep: power: Is a directory
power1_alarm:0
power1_input:33250000
power1_label:pin
power1_max:-500000
power1_rated_max:-500000
power2_cap:-500000
power2_cap_alarm:0
power2_crit:-500000
power2_crit_alarm:0
power2_input:17750000
power2_label:pout1
power2_max:-500000
power2_max_alarm:0
power2_rated_max:920000000
grep: subsystem: Is a directory
temp1_crit:-500
temp1_crit_alarm:0
temp1_input:23000
temp1_lcrit:-500
temp1_lcrit_alarm:0
temp1_max:-500
temp1_max_alarm:0
temp1_min:-500
temp1_min_alarm:0
temp1_rated_max:-500
temp2_crit:-500
temp2_crit_alarm:0
temp2_input:26000
temp2_lcrit:-500
temp2_lcrit_alarm:0
temp2_max:-500
temp2_max_alarm:0
temp2_min:-500
temp2_min_alarm:0
temp2_rated_max:-500
temp3_crit:-500
temp3_crit_alarm:0
temp3_input:30000
temp3_lcrit:-500
temp3_lcrit_alarm:0
temp3_max:-500
temp3_max_alarm:0
temp3_min:-500
temp3_min_alarm:0
temp3_rated_max:-500
uevent:OF_NAME=psu
uevent:OF_FULLNAME=/ap806/config-space@f0000000/i2c@511000/psu@5a
uevent:OF_COMPATIBLE_0=delta,dps920ab
uevent:OF_COMPATIBLE_N=1
Ok, good enough. It looks like the PSU reports values for pretty much everything,
including registers which don't exist. With that in mind, please check
the attributes generated by your driver - I suspect that some of the limit
attributes are not really supported (maybe none of them is supported).
Yeah, I also think that none of those limits are actually supported.
Does the core expose a way to not register those?
Have the read_word function return -ENXIO for unsupported registers
(and add a comment to the code explaining why you do that).
See drivers/hwmon/pmbus/fsp-3y.c for an example.
Thanks, will implement that.
In the meantime I finally got the PMBus command list that
the PSU supports.
It implements PMBus 1.2, but it also claims that MFR_ID,
MFR_MODEL and MFR_REVISION are supported.
Block read returns what looks like a correct length, but the buffer
is empty.
[ 195.246464] dps920ab 0-005a: PMBUS_MFR_ID length: 5
[ 195.251495] dps920ab 0-005a: PMBUS_MFR_ID:
[ 195.300789] dps920ab 0-005b: PMBUS_MFR_ID length: 5
[ 195.305762] dps920ab 0-005b: PMBUS_MFR_ID:
[ 272.795222] dps920ab 0-005a: PMBUS_MFR_ID length: 5
[ 272.800190] dps920ab 0-005a: PMBUS_MFR_ID:
[ 272.804862] dps920ab 0-005a: PMBUS_MFR_MODEL length: 11
[ 272.810172] dps920ab 0-005a: PMBUS_MFR_MODEL:
[ 272.815098] dps920ab 0-005a: PMBUS_MFR_REVISION length: 2
[ 272.820600] dps920ab 0-005a: PMBUS_MFR_REVISION:
[ 272.874060] dps920ab 0-005b: PMBUS_MFR_ID length: 5
[ 272.879102] dps920ab 0-005b: PMBUS_MFR_ID:
[ 272.883833] dps920ab 0-005b: PMBUS_MFR_MODEL length: 11
[ 272.889151] dps920ab 0-005b: PMBUS_MFR_MODEL:
[ 272.894097] dps920ab 0-005b: PMBUS_MFR_REVISION length: 2
[ 272.899548] dps920ab 0-005b: PMBUS_MFR_REVISION:
However, manually reading using i2cget in word mode returns
something that looks ok.
i2cget -y 0 0x5a 0x9a w
0x440b
This would be D is ASCII, it looks like any kind of block read fails actually.
Ok, so after using i2c_smbus_read_i2c_block_data() I can actually read
the model kind of OK.
There actually are 10 bytes of valid data, no idea why
i2c_smbus_read_block_data() does not work.
[ 721.194706] dps920ab 0-005a: MFR MODEL:
DPS-920AB
[ 721.245502] dps920ab 0-005b: MFR MODEL:
DPS-920AB
It behaves weirdly when printing by creating a newline before and
stuff like that.
I suspect it prints the first byte, which should be the length.
I also suspect that the i2c controller driver has a bug and doesn't
support SMBus block data correctly. Did you try i2cdump with i and s
options ?
That could be the case, I just wanted to see whether the controller or
PSU are at fault.
I tried using s with i2cdump and it will error out on the controller
not supporting block reads.
The controller most likely has a bug in the driver.
I tried applying the block support for mv64xx as well:
https://patchwork.ozlabs.org/project/linux-i2c/patch/20200118115820.9080-1-fuga@xxxxxxxxxxxxxx/
That patch would be needed, but it looks buggy to me. This chunk:
+ drv_data->effective_length = data+1;
+ drv_data->bytes_left = data+1;
+ drv_data->msg->len = data+1;
should be:
+ drv_data->effective_length = data+1;
+ drv_data->bytes_left = data;
+ drv_data->msg->len = data+1;
It should also make sure that 'data' is not larger than I2C_SMBUS_BLOCK_MAX,
and bail out if it isn't.
Yeah, I did not check the contents, I just saw 2 reviews and tested it
since it can't hurt.
That patch doesn't work at all. Make the above change, and also change
the type of effective_length from u32 to int, and try again.
Guenter
Do you see a difference with the patch applied ?
No, unfortunately not, i2cdump in s mode returns all zeroes.
I can see that it gets the correct length but returns all zeroes.
I did some development on the driver, added whitelisting for
word-based registers.
Found that this thing actually has the third temperature sensor after
all and a lot of warnings
and stuff are supported.
I need to review the byte-based registers and whitelist those if needed.
Tried adding PWM/DC control for the fan as it supports it, but it was
a bit confusing so I
gave up on that for now.
Regards,
Robert
Guenter