2022-04-27 09:32:06

by Zev Weiss

[permalink] [raw]
Subject: [PATCH v3 0/6] hwmon: (nct6775) Convert to regmap, add i2c support

Hello,

This is v3 of my effort to add i2c support to the nct6775 hwmon
driver.

Changes since v2 [0]:
- Fixed wrong parenthesization in nct6775_write_value()
- Moved DT binding to trivial-devices.yml instead of a dedicated file
[Guenter]
- Renamed drivers and Kconfig symbols to keep existing platform
driver as "nct6775" (SENSORS_NCT6775) and the core module as
"nct6775-core" (SENSORS_NCT6775_CORE) [Guenter]
- Fixed SENSORS_NCT6775_CORE Kconfig to select REGMAP, removed
erroneous REGMAP_I2C selection from SENSORS_NCT6775_I2C [Guenter]

Changes since v1 [1]:
- Added preparatory patch converting driver to regmap API [Guenter]
- Replaced ENOSPC with ENOBUFS and removed WARN_ON() in
nct6775_add_attr_group() [Guenter]
- Added dedicated symbol namespace [Guenter]
- Removed nct6775_write_temp() and nct6775_update_device() symbol
exports [Guenter]
- Reordered patches to put dt-bindings patch first [Krzysztof]

[0] https://lore.kernel.org/linux-hwmon/[email protected]/
[1] https://lore.kernel.org/linux-hwmon/[email protected]/

A slightly edited version of the previous cover letter follows:


This patch series augments the existing nct6775 driver with support
for the hardware's i2c interface; along the way it converts the driver
to use the regmap API, and splits the LPC-specific platform driver
into a separate module from the interface-independent core.

Thus far the nct6775 driver has only supported the LPC interface,
which is the main interface by which the Super-I/O chip is typically
connected to the host (x86) processor.

However, these chips also provide an i2c interface, which can provide
a way for a BMC to also monitor sensor readings from them. On some
systems (such as the ASRock Rack ROMED8HM3 and X570-D4U) this may be
the only way for the BMC to monitor host CPU temperatures (e.g. to
indirectly access a TSI interface); this functionality is thus an
important component of enabling OpenBMC to support such systems.

In such an arrangement the Super-I/O chip is simultaneously controlled
by two independent processors (the host and the BMC) which typically
do not coordinate their accesses with each other. In order to avoid
conflicts between the two, the i2c driver avoids all writes to the
device, since the BMC's needs with the hardware are merely that it be
able to retrieve sensor readings. This allows the host processor to
remain ultimately in control of the chip and unaware of the BMC's use
of it at all.

The sole exception to the "no writes" rule for the i2c driver is for
the bank-select register -- while I haven't been able to find any
explicit statement in the Nuvoton datasheets guaranteeing this,
testing via manual register accesses (as detailed in [2]) has
indicated that, as one might hope, the i2c interface has its own
bank-select register independent of the one used by the LPC interface.

In terms of code structure, the approach taken in this series is to
first convert the driver's register accesses to the regmap API, and
then split the LPC-specific parts of it out into a separate module
(retaining the current "nct6775" name), leaving the
interface-independent parts in a generic driver (called nct6775-core).
The nct6775-i2c driver is then added as an additional consumer of the
nct6775-core module's functionality (essentially just providing its
own set of regmap read/write callback functions).

The first patch adds the chips supported by the nct6775 driver to
Documentation/device-tree-bindings/trivial-devices.yml. The second
patch contains the change to convert all register accesses to use a
regmap. The third and fourth patches make some relatively small
infrastructural changes to the driver. The core/platform driver split
is in the fifth patch, and the final patch adds the i2c driver itself.

The nct6775 and nct6775-i2c drivers have both been tested on the
NCT6779D in an ASRock ROMED8HM3 system and the NCT6798 [3] in an
ASRock X570-D4U (the latter thanks to Renze, CCed); both seem to work
as expected on both systems. I don't have access to any asuswmi
hardware, so testing of the nct6775-platform driver on that to ensure
it doesn't break there would be appreciated (Oleksandr, perhaps?).


Thanks,
Zev

[2] https://lore.kernel.org/linux-hwmon/[email protected]/
[3] Though it's physically labeled (mislabeled?) as an NCT6796, for
what that's worth.

Zev Weiss (6):
dt-bindings: trivial-devices: Add Nuvoton Super I/O chips
hwmon: (nct6775) Convert register access to regmap API
hwmon: (nct6775) Rearrange attr-group initialization
hwmon: (nct6775) Add read-only mode
hwmon: (nct6775) Split core and platform driver
hwmon: (nct6775) Add i2c driver

.../devicetree/bindings/trivial-devices.yaml | 14 +
MAINTAINERS | 10 +-
drivers/hwmon/Kconfig | 30 +-
drivers/hwmon/Makefile | 2 +
drivers/hwmon/{nct6775.c => nct6775-core.c} | 2310 +++-----
drivers/hwmon/nct6775-core.h | 252 +
drivers/hwmon/nct6775-i2c.c | 179 +
drivers/hwmon/nct6775.c | 4652 ++---------------
8 files changed, 1410 insertions(+), 6039 deletions(-)
copy drivers/hwmon/{nct6775.c => nct6775-core.c} (69%)
create mode 100644 drivers/hwmon/nct6775-core.h
create mode 100644 drivers/hwmon/nct6775-i2c.c

--
2.36.0


2022-04-27 10:57:18

by Zev Weiss

[permalink] [raw]
Subject: [PATCH v3 6/6] hwmon: (nct6775) Add i2c driver

This driver provides an i2c I/O mechanism for the core nct6775 driver,
as might be used by a BMC. Because the Super I/O chip is shared with
the host CPU in such a scenario (and the host should ultimately be in
control of it), the i2c driver is strictly read-only to avoid
interfering with any usage by the host (aside from the bank-select
register, which seems to be replicated for the i2c interface).

Signed-off-by: Zev Weiss <[email protected]>
Tested-by: Renze Nicolai <[email protected]>
---
MAINTAINERS | 6 ++
drivers/hwmon/Kconfig | 16 ++++
drivers/hwmon/Makefile | 1 +
drivers/hwmon/nct6775-i2c.c | 179 ++++++++++++++++++++++++++++++++++++
4 files changed, 202 insertions(+)
create mode 100644 drivers/hwmon/nct6775-i2c.c

diff --git a/MAINTAINERS b/MAINTAINERS
index ca1d930d3d88..9819c5ade556 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -13538,6 +13538,12 @@ F: drivers/hwmon/nct6775-core.c
F: drivers/hwmon/nct6775-core.h
F: drivers/hwmon/nct6775.c

+NCT6775 HARDWARE MONITOR DRIVER - I2C DRIVER
+M: Zev Weiss <[email protected]>
+L: [email protected]
+S: Maintained
+F: drivers/hwmon/nct6775-i2c.c
+
NETDEVSIM
M: Jakub Kicinski <[email protected]>
S: Maintained
diff --git a/drivers/hwmon/Kconfig b/drivers/hwmon/Kconfig
index 85c22bba439b..a68ec731608d 100644
--- a/drivers/hwmon/Kconfig
+++ b/drivers/hwmon/Kconfig
@@ -1489,6 +1489,22 @@ config SENSORS_NCT6775
This driver can also be built as a module. If so, the module
will be called nct6775.

+config SENSORS_NCT6775_I2C
+ tristate "I2C driver for Nuvoton NCT6775F and compatibles"
+ depends on I2C
+ select SENSORS_NCT6775_CORE
+ help
+ If you say yes here you get support for the hardware monitoring
+ functionality of the Nuvoton NCT6106D, NCT6775F, NCT6776F, NCT6779D,
+ NCT6791D, NCT6792D, NCT6793D, NCT6795D, NCT6796D, and compatible
+ Super-I/O chips via their I2C interface.
+
+ If you're not building a kernel for a BMC, this is probably
+ not the driver you want (see CONFIG_SENSORS_NCT6775).
+
+ This driver can also be built as a module. If so, the module
+ will be called nct6775-i2c.
+
config SENSORS_NCT7802
tristate "Nuvoton NCT7802Y"
depends on I2C
diff --git a/drivers/hwmon/Makefile b/drivers/hwmon/Makefile
index 86a3573be7e8..1ad5777a5d03 100644
--- a/drivers/hwmon/Makefile
+++ b/drivers/hwmon/Makefile
@@ -156,6 +156,7 @@ obj-$(CONFIG_SENSORS_MR75203) += mr75203.o
obj-$(CONFIG_SENSORS_NCT6683) += nct6683.o
obj-$(CONFIG_SENSORS_NCT6775_CORE) += nct6775-core.o
obj-$(CONFIG_SENSORS_NCT6775) += nct6775.o
+obj-$(CONFIG_SENSORS_NCT6775_I2C) += nct6775-i2c.o
obj-$(CONFIG_SENSORS_NCT7802) += nct7802.o
obj-$(CONFIG_SENSORS_NCT7904) += nct7904.o
obj-$(CONFIG_SENSORS_NPCM7XX) += npcm750-pwm-fan.o
diff --git a/drivers/hwmon/nct6775-i2c.c b/drivers/hwmon/nct6775-i2c.c
new file mode 100644
index 000000000000..059d650a1920
--- /dev/null
+++ b/drivers/hwmon/nct6775-i2c.c
@@ -0,0 +1,179 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * nct6775-i2c - I2C driver for the hardware monitoring functionality of
+ * Nuvoton NCT677x Super-I/O chips
+ *
+ * Copyright (C) 2022 Zev Weiss <[email protected]>
+ *
+ * This driver interacts with the chip via it's "back door" i2c interface, as
+ * is often exposed to a BMC. Because the host may still be operating the
+ * chip via the ("front door") LPC interface, this driver cannot assume that
+ * it actually has full control of the chip, and in particular must avoid
+ * making any changes that could confuse the host's LPC usage of it. It thus
+ * operates in a strictly read-only fashion, with the only exception being the
+ * bank-select register (which seems, thankfully, to be replicated for the i2c
+ * interface so it doesn't affect the LPC interface).
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/i2c.h>
+#include <linux/hwmon.h>
+#include <linux/hwmon-sysfs.h>
+#include <linux/err.h>
+#include <linux/of_device.h>
+#include <linux/regmap.h>
+#include "nct6775-core.h"
+
+static int nct6775_i2c_read(void *ctx, unsigned int reg, unsigned int *val)
+{
+ int ret;
+ u32 tmp;
+ u8 bank = reg >> 8;
+ struct nct6775_data *data = ctx;
+ struct i2c_client *client = data->driver_data;
+
+ if (bank != data->bank) {
+ ret = i2c_smbus_write_byte_data(client, NCT6775_REG_BANK, bank);
+ if (ret)
+ return ret;
+ data->bank = bank;
+ }
+
+ ret = i2c_smbus_read_byte_data(client, reg & 0xff);
+ if (ret < 0)
+ return ret;
+ tmp = ret;
+
+ if (nct6775_reg_is_word_sized(data, reg)) {
+ ret = i2c_smbus_read_byte_data(client, (reg & 0xff) + 1);
+ if (ret < 0)
+ return ret;
+ tmp = (tmp << 8) | ret;
+ }
+
+ *val = tmp;
+ return 0;
+}
+
+/*
+ * The write operation is a dummy so as not to disturb anything being done
+ * with the chip via LPC.
+ */
+static int nct6775_i2c_write(void *ctx, unsigned int reg, unsigned int value)
+{
+ struct nct6775_data *data = ctx;
+ struct i2c_client *client = data->driver_data;
+
+ dev_dbg(&client->dev, "skipping attempted write: %02x -> %03x\n", value, reg);
+
+ /*
+ * This is a lie, but writing anything but the bank-select register is
+ * something this driver shouldn't be doing.
+ */
+ return 0;
+}
+
+static const struct of_device_id __maybe_unused nct6775_i2c_of_match[] = {
+ { .compatible = "nuvoton,nct6106", .data = (void *)nct6106, },
+ { .compatible = "nuvoton,nct6116", .data = (void *)nct6116, },
+ { .compatible = "nuvoton,nct6775", .data = (void *)nct6775, },
+ { .compatible = "nuvoton,nct6776", .data = (void *)nct6776, },
+ { .compatible = "nuvoton,nct6779", .data = (void *)nct6779, },
+ { .compatible = "nuvoton,nct6791", .data = (void *)nct6791, },
+ { .compatible = "nuvoton,nct6792", .data = (void *)nct6792, },
+ { .compatible = "nuvoton,nct6793", .data = (void *)nct6793, },
+ { .compatible = "nuvoton,nct6795", .data = (void *)nct6795, },
+ { .compatible = "nuvoton,nct6796", .data = (void *)nct6796, },
+ { .compatible = "nuvoton,nct6797", .data = (void *)nct6797, },
+ { .compatible = "nuvoton,nct6798", .data = (void *)nct6798, },
+ { },
+};
+MODULE_DEVICE_TABLE(of, nct6775_i2c_of_match);
+
+static const struct i2c_device_id nct6775_i2c_id[] = {
+ { "nct6106", nct6106 },
+ { "nct6116", nct6116 },
+ { "nct6775", nct6775 },
+ { "nct6776", nct6776 },
+ { "nct6779", nct6779 },
+ { "nct6791", nct6791 },
+ { "nct6792", nct6792 },
+ { "nct6793", nct6793 },
+ { "nct6795", nct6795 },
+ { "nct6796", nct6796 },
+ { "nct6797", nct6797 },
+ { "nct6798", nct6798 },
+ { }
+};
+MODULE_DEVICE_TABLE(i2c, nct6775_i2c_id);
+
+static int nct6775_i2c_probe_init(struct nct6775_data *data)
+{
+ /*
+ * The i2c interface doesn't provide access to the control registers
+ * needed to determine the presence of other fans, but fans 1 and 2
+ * are (in principle) always there.
+ *
+ * In practice this is perhaps a little silly, because the system
+ * using this driver is mostly likely a BMC, and hence probably has
+ * totally separate fan tachs & pwms of its own that are actually
+ * controlling/monitoring the fans -- these are thus unlikely to be
+ * doing anything actually useful.
+ */
+ data->has_fan = 0x03;
+ data->has_fan_min = 0x03;
+ data->has_pwm = 0x03;
+ return 0;
+}
+
+static const struct regmap_config nct6775_i2c_regmap_config = {
+ .reg_bits = 16,
+ .val_bits = 16,
+ .reg_read = nct6775_i2c_read,
+ .reg_write = nct6775_i2c_write,
+};
+
+static int nct6775_i2c_probe(struct i2c_client *client)
+{
+ struct nct6775_data *data;
+ const struct of_device_id *of_id;
+ const struct i2c_device_id *i2c_id;
+ struct device *dev = &client->dev;
+
+ of_id = of_match_device(nct6775_i2c_of_match, dev);
+ i2c_id = i2c_match_id(nct6775_i2c_id, client);
+
+ if (of_id && (unsigned long)of_id->data != i2c_id->driver_data)
+ dev_notice(dev, "Device mismatch: %s in device tree, %s detected\n",
+ of_id->name, i2c_id->name);
+
+ data = devm_kzalloc(&client->dev, sizeof(*data), GFP_KERNEL);
+ if (!data)
+ return -ENOMEM;
+
+ data->kind = i2c_id->driver_data;
+
+ data->read_only = true;
+ data->driver_data = client;
+ data->driver_init = nct6775_i2c_probe_init;
+
+ return nct6775_probe(dev, data, &nct6775_i2c_regmap_config);
+}
+
+static struct i2c_driver nct6775_i2c_driver = {
+ .class = I2C_CLASS_HWMON,
+ .driver = {
+ .name = "nct6775-i2c",
+ .of_match_table = of_match_ptr(nct6775_i2c_of_match),
+ },
+ .probe_new = nct6775_i2c_probe,
+ .id_table = nct6775_i2c_id,
+};
+
+module_i2c_driver(nct6775_i2c_driver);
+
+MODULE_AUTHOR("Zev Weiss <[email protected]>");
+MODULE_DESCRIPTION("I2C driver for NCT6775F and compatible chips");
+MODULE_LICENSE("GPL");
+MODULE_IMPORT_NS(HWMON_NCT6775);
--
2.36.0