Hi Alex,

On Thu, Jun 08, 2017 at 08:44:35PM +0300, Alex A. Mihaylov wrote:

08.06.17 15:48, Sebastian Reichel wrote:

quoted

On Fri, Jun 02, 2017 at 10:06:29AM +0300, Alex A. Mihaylov wrote:

quoted

diff --git a/drivers/power/supply/max1721x_battery.c b/drivers/power/supply/max1721x_battery.c
new file mode 100644
index 0000000000..aa0effec3d
--- /dev/null
+++ b/drivers/power/supply/max1721x_battery.c

@@ -0,0 +1,357 @@
+/*
+ * 1-wire client/driver for the Maxim Semicondactor
+ * MAX17211/MAX17215 Standalone Fuel Gauge IC
+ *
+ * Copyright (C) 2017 OAO Radioavionica
+ * Author: Alex A. Mihaylov <minimumlaw@rambler.ru>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/param.h>

param?

Ok, sorry. This really not need. I remove this in next revision.

quoted

quoted

+#include <linux/pm.h>
+#include <linux/regmap.h>
+#include <linux/platform_device.h>
+#include <linux/power_supply.h>
+#include <linux/idr.h>
+
+#include "../../w1/w1.h"

This will conflict with public w1 interface patch
https://www.spinics.net/lists/kernel/msg2524566.html

This patch should be on top of that patch.

Ok. No problem. I can fix this here. I can fix this in regmap-w1. Just tell
me which of the patches will be applied first. If the one to which you
refer, I will resend the patches immediately after it appears at least in
-rc.

quoted

quoted

+#include "../../w1/slaves/w1_max1721x.h"

Let's merge those defines into the driver. They
are not used anywhere else.

Theory, Maxim integrated have MAX17201/MAX17205 with I2C interface. This may
required for feature i2c driver.

This would end up in the same driver with only probe function being
different.

quoted

quoted

+
+/* Model Gauge M5 Register Memory Map access */
+static const struct regmap_range max1721x_regs_allow[] = {
+	/* M5 Model Gauge Algorithm area */
+	regmap_reg_range(0x00, 0x23),
+	regmap_reg_range(0x27, 0x2F),
+	regmap_reg_range(0x32, 0x32),
+	regmap_reg_range(0x35, 0x36),
+	regmap_reg_range(0x38, 0x3A),
+	regmap_reg_range(0x3D, 0x3F),
+	regmap_reg_range(0x42, 0x42),
+	regmap_reg_range(0x45, 0x46),
+	regmap_reg_range(0x4A, 0x4A),
+	regmap_reg_range(0x4D, 0x4D),
+	regmap_reg_range(0xB0, 0xB0),
+	regmap_reg_range(0xB4, 0xB4),
+	regmap_reg_range(0xB8, 0xBE),
+	regmap_reg_range(0xD1, 0xDA),
+	regmap_reg_range(0xDC, 0xDF),
+	/* Factory settins area */
+	regmap_reg_range(0x180, 0x1DF),
+	{ }
+};
+
+static const struct regmap_range max1721x_regs_deny[] = {
+	regmap_reg_range(0x24, 0x26),
+	regmap_reg_range(0x30, 0x31),
+	regmap_reg_range(0x33, 0x34),
+	regmap_reg_range(0x37, 0x37),
+	regmap_reg_range(0x3B, 0x3C),
+	regmap_reg_range(0x40, 0x41),
+	regmap_reg_range(0x43, 0x44),
+	regmap_reg_range(0x47, 0x49),
+	regmap_reg_range(0x4B, 0x4C),
+	regmap_reg_range(0x4E, 0xAF),
+	regmap_reg_range(0xB1, 0xB3),
+	regmap_reg_range(0xB5, 0xB7),
+	regmap_reg_range(0xBF, 0xD0),
+	regmap_reg_range(0xDB, 0xDB),
+	regmap_reg_range(0xE0, 0x17F),
+	{ }
+};
+
+static const struct regmap_access_table max1721x_regs = {
+	.yes_ranges	= max1721x_regs_allow,
+	.n_yes_ranges	= ARRAY_SIZE(max1721x_regs_allow),
+	.no_ranges	= max1721x_regs_deny,
+	.n_no_ranges	= ARRAY_SIZE(max1721x_regs_deny),
+};

It should be enough to specify the yes_range. Unspecified
values will be no implicitly.

I can remove this. I just desribe all registers hole described in datasheet.
I hope this reduce memory in regmap infrastructure.

That's what I suspected.

quoted

quoted

+/* W1 regmap config */
+static const struct regmap_config max1721x_regmap_w1_config = {
+	.reg_bits = 16,
+	.val_bits = 16,
+	.volatile_table = &max1721x_regs,
+	.max_register = MAX1721X_MAX_REG_NR,
+};

Are the non-volatile registers missing? Then you probably also
want to specify .rd_table with the same access table, so that
dumping registers via debugfs work correctly. Did you try to
cat /sys/kernel/debug/regmap/<your-device>/registers?

Ok, I try this. Non-volatile registers present (Rsense, manufaturer, device
name, serial number). I not read this register until probe step, so I not
put them into nonvolatile regmap table. But I can do this. May be it's more
correctly, than desribe registers hole.

Register hole table should be used for the rd_table. You can skip
configuration of the volatile_table, if you do not enable caching
via max1721x_regmap_w1_config.cache_type. Enabling the caching is
only sensible, if you do not mark all registers as volatile ;)

quoted

quoted

+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Alex A. Mihaylov [off-list ref]");
+MODULE_DESCRIPTION("Maxim MAX17211/MAX17215 Fuel Gauage IC driver");
+MODULE_ALIAS("platform:max1721x-battery");

Otherwise looks good.

BTW. I try send RFC with alternative realisation of this driver into
linux-pm:
[1] http://marc.info/?l=linux-pm&m=149422406914440
[2] http://marc.info/?l=linux-pm&m=149422407014444

Ah I skipped this one, since there was a newer revision.
Yes, this is what I had in mind when I talked about merging
the w1 driver into the power-supply driver.

This code maped to thermal zones, not used platform device and put
max172xx-battery.h into include/linux/power. All know troubel in [1].

So here are answers to your questions

1. Endian for W1-regmap
Code writen for use nativ endian for host (master) device. But waiting for
little-endian slave divice. I don't know W1 slave device with big-endian
addrees or data format. May be regmap_bus or regmap_config will be checked
for host/device endian requirements?

regmap_bus.reg_format_endian_default provides a default endianess for
the bus, which can be overwritten via regmap_config.val_format_endian.
Since Mark queued your patch everything seems to be alright (I did not
review it).

2. W1 Family/device infrastructure
All present power_supply class drivers in vanilla kernel use w1-slave devices.
All of them create platform_device with name "chip-battery.X.auto", and 
power_supply class driver use this platform device. Why used this way? I
write code, allocated power_supply at W1 slave (family) device register.
This work as expected.

Historical reasons. The patch [0], which I already mentioned is the
first step to move bq27xxx w1 driver into the power-supply subsystem.

3. Device names
All W1 device have unical 64 bit ID (8-bit family, 48-bit serial number,
8 bit CRC). W1 infrastructure show 56 bits (family-serial_nimber) as w1
slave device name. I use this (unical) device name as power_supply battery
name. This work. But in /sys/class/power_supply is placed machine readabe 
subdir "26-HexDigString", instead of human-readable "chip-battery.X.auto".
But in /sys/class/power_supply/26-HexDigString/type file still content
human-readable type "Battery".
Thermal zone working, but in /sys/class/thermal/thermal_zoneX/type also
placed "26-HexDigString", instead of "chip-battery.X.auto".
I can rename "26-HexDigString" to "battery-26-HexDigString" or 
"26-HexDigString-battery", but this exceeded 20 chars thermal zone device
name. So, thermal zone will have to be disabled.

I think max1721x-<serial> would be better. If the name is too long, just
disable the thermal zone for now. We have more w1/psy drivers, that have problems
due to the thermal zone device name length limitation.

[0] https://www.spinics.net/lists/kernel/msg2524566.html

-- Sebastian