Re: [PATCH 1/3] hwmon: max31827: Add PEC support

[Guenter Roeck]

On 12/19/23 17:49, Matyas, Daniel wrote:
> > -----Original Message-----
> > From: Guenter Roeck <groeck7@xxxxxxxxx> On Behalf Of Guenter Roeck
> > Sent: Monday, December 18, 2023 9:01 PM
> > To: Matyas, Daniel <Daniel.Matyas@xxxxxxxxxx>
> > Cc: Jean Delvare <jdelvare@xxxxxxxx>; Rob Herring
> > <robh+dt@xxxxxxxxxx>; Krzysztof Kozlowski
> > <krzysztof.kozlowski+dt@xxxxxxxxxx>; Conor Dooley
> > <conor+dt@xxxxxxxxxx>; Jonathan Corbet <corbet@xxxxxxx>; linux-
> > hwmon@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; linux-
> > kernel@xxxxxxxxxxxxxxx; linux-doc@xxxxxxxxxxxxxxx
> > Subject: Re: [PATCH 1/3] hwmon: max31827: Add PEC support
> >
> > [External]
> >
> > On 12/18/23 09:59, Matyas, Daniel wrote:
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------------------------------------------------------------------
> > > ----------
> > > *Von:* Guenter Roeck <groeck7@xxxxxxxxx> im Auftrag von Guenter
> > Roeck
> > > <linux@xxxxxxxxxxxx>
> > > *Gesendet:* Montag, Dezember 18, 2023 6:26:57 nachm.
> > > *An:* Matyas, Daniel <Daniel.Matyas@xxxxxxxxxx>
> > > *Cc:* Jean Delvare <jdelvare@xxxxxxxx>; Rob Herring
> > > <robh+dt@xxxxxxxxxx>; Krzysztof Kozlowski
> > > <krzysztof.kozlowski+dt@xxxxxxxxxx>; Conor Dooley
> > > <conor+dt@xxxxxxxxxx>; Jonathan Corbet <corbet@xxxxxxx>;
> > > linux-hwmon@xxxxxxxxxxxxxxx <linux-hwmon@xxxxxxxxxxxxxxx>;
> > > devicetree@xxxxxxxxxxxxxxx <devicetree@xxxxxxxxxxxxxxx>;
> > > linux-kernel@xxxxxxxxxxxxxxx <linux-kernel@xxxxxxxxxxxxxxx>;
> > > linux-doc@xxxxxxxxxxxxxxx <linux-doc@xxxxxxxxxxxxxxx>
> > > *Betreff:* Re: [PATCH 1/3] hwmon: max31827: Add PEC support
> > >
> > > [External]
> > >
> > > On 12/18/23 06:55, Matyas, Daniel wrote:
> > > [ ... ]
> > > > > On top of that, it is not clear why regmap can't be used in the first
> > place.
> > > > > It seems that the major change is that one needs to read the
> > > > > configuration register after a write to see if there was a PEC
> > > > > error. It is not immediately obvious why that additional read (if
> > > > > indeed necessary) would require regmap support to be dropped.
> > > >
> > > > I tried out writing and and reading with regmap, but it is not working
> > properly. Even if I modify the client flag, I still receive only 2 bytes of data
> > (a word). I should be receiving 2+1 bytes = data + CRC-8.
> > > > With i2c_smbus reads and writes, when I set the flag, I receive the 2+1
> > bytes, as expected.
> > > >
> > >
> > > The SMBus code in drivers/i2c/i2c-core-smbus.c is supposed to check if
> > > the received PEC is correct for SMBus transfers. Are you saying that
> > > this doesn't work, or that regmap doesn't use SMBus functions to
> > > communicate with the chip ?
> > >
> > > Thanks,
> > > Guenter
> > >
> > >
> > > I am 70% sure, that the regmap does not use SMBus functions.
> >
> > It should.
> >
> > $ git grep smbus drivers/base/regmap/regmap-i2c.c
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_smbus_byte_reg_read(void *context, unsigned int reg,
> > drivers/base/regmap/regmap-i2c.c:       ret =
> > i2c_smbus_read_byte_data(i2c, reg);
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_smbus_byte_reg_write(void *context, unsigned int reg,
> > drivers/base/regmap/regmap-i2c.c:       return
> > i2c_smbus_write_byte_data(i2c, reg, val);
> > drivers/base/regmap/regmap-i2c.c:static const struct regmap_bus
> > regmap_smbus_byte = {
> > drivers/base/regmap/regmap-i2c.c:       .reg_write =
> > regmap_smbus_byte_reg_write,
> > drivers/base/regmap/regmap-i2c.c:       .reg_read =
> > regmap_smbus_byte_reg_read,
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_smbus_word_reg_read(void *context, unsigned int reg,
> > drivers/base/regmap/regmap-i2c.c:       ret =
> > i2c_smbus_read_word_data(i2c, reg);
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_smbus_word_reg_write(void *context, unsigned int reg,
> > drivers/base/regmap/regmap-i2c.c:       return
> > i2c_smbus_write_word_data(i2c, reg, val);
> > drivers/base/regmap/regmap-i2c.c:static const struct regmap_bus
> > regmap_smbus_word = {
> > drivers/base/regmap/regmap-i2c.c:       .reg_write =
> > regmap_smbus_word_reg_write,
> > drivers/base/regmap/regmap-i2c.c:       .reg_read =
> > regmap_smbus_word_reg_read,
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_smbus_word_read_swapped(void *context, unsigned int reg,
> > drivers/base/regmap/regmap-i2c.c:       ret =
> > i2c_smbus_read_word_swapped(i2c, reg);
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_smbus_word_write_swapped(void *context, unsigned int reg,
> > drivers/base/regmap/regmap-i2c.c:       return
> > i2c_smbus_write_word_swapped(i2c, reg, val);
> > drivers/base/regmap/regmap-i2c.c:static const struct regmap_bus
> > regmap_smbus_word_swapped = {
> > drivers/base/regmap/regmap-i2c.c:       .reg_write =
> > regmap_smbus_word_write_swapped,
> > drivers/base/regmap/regmap-i2c.c:       .reg_read =
> > regmap_smbus_word_read_swapped,
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_i2c_smbus_i2c_write(void *context, const void *data,
> > drivers/base/regmap/regmap-i2c.c:       return
> > i2c_smbus_write_i2c_block_data(i2c, ((u8 *)data)[0], count,
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_i2c_smbus_i2c_read(void *context, const void *reg,
> > drivers/base/regmap/regmap-i2c.c:       ret =
> > i2c_smbus_read_i2c_block_data(i2c, ((u8 *)reg)[0], val_size, val);
> > drivers/base/regmap/regmap-i2c.c:static const struct regmap_bus
> > regmap_i2c_smbus_i2c_block = {
> > drivers/base/regmap/regmap-i2c.c:       .write =
> > regmap_i2c_smbus_i2c_write,
> > drivers/base/regmap/regmap-i2c.c:       .read =
> > regmap_i2c_smbus_i2c_read,
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_i2c_smbus_i2c_write_reg16(void *context, const void *data,
> > drivers/base/regmap/regmap-i2c.c:       return
> > i2c_smbus_write_i2c_block_data(i2c, ((u8 *)data)[0], count,
> > drivers/base/regmap/regmap-i2c.c:static int
> > regmap_i2c_smbus_i2c_read_reg16(void *context, const void *reg,
> > drivers/base/regmap/regmap-i2c.c:       ret =
> > i2c_smbus_write_byte_data(i2c, ((u16 *)reg)[0] & 0xff,
> > drivers/base/regmap/regmap-i2c.c:               ret =
> > i2c_smbus_read_byte(i2c);
> > drivers/base/regmap/regmap-i2c.c:static const struct regmap_bus
> > regmap_i2c_smbus_i2c_block_reg16 = {
> > drivers/base/regmap/regmap-i2c.c:       .write =
> > regmap_i2c_smbus_i2c_write_reg16,
> > drivers/base/regmap/regmap-i2c.c:       .read =
> > regmap_i2c_smbus_i2c_read_reg16,
> > drivers/base/regmap/regmap-i2c.c:               bus =
> > &regmap_i2c_smbus_i2c_block;
> > drivers/base/regmap/regmap-i2c.c:               bus =
> > &regmap_i2c_smbus_i2c_block_reg16;
> > drivers/base/regmap/regmap-i2c.c:                       bus =
> > &regmap_smbus_word;
> > drivers/base/regmap/regmap-i2c.c:                       bus =
> > &regmap_smbus_word_swapped;
> > drivers/base/regmap/regmap-i2c.c:               bus = &regmap_smbus_byte;
> >
> > If that doesn't work for some reason, I'd rather figure out why instead of
> > starting to drop regmap support.
> >
> > Guenter
>
> I tried to figure it out and this is what I came up with. The code snippet below is from drivers/base/regmap/regmap-i2c.c:
>
> static const struct regmap_bus *regmap_get_i2c_bus(struct i2c_client *i2c,
> 					const struct regmap_config *config)
> {
> 	const struct i2c_adapter_quirks *quirks;
> 	const struct regmap_bus *bus = NULL;
> 	struct regmap_bus *ret_bus;
> 	u16 max_read = 0, max_write = 0;
>
> 	if (i2c_check_functionality(i2c->adapter, I2C_FUNC_I2C))
> 		bus = &regmap_i2c;
> 	else if (config->val_bits == 8 && config->reg_bits == 8 &&
> 		 i2c_check_functionality(i2c->adapter,
> 					 I2C_FUNC_SMBUS_I2C_BLOCK))
> 		bus = &regmap_i2c_smbus_i2c_block;
> 	else if (config->val_bits == 8 && config->reg_bits == 16 &&
> 		i2c_check_functionality(i2c->adapter,
> 					I2C_FUNC_SMBUS_I2C_BLOCK))
> 		bus = &regmap_i2c_smbus_i2c_block_reg16;
> 	else if (config->val_bits == 16 && config->reg_bits == 8 &&
> 		 i2c_check_functionality(i2c->adapter,
> 					 I2C_FUNC_SMBUS_WORD_DATA))
> 		switch (regmap_get_val_endian(&i2c->dev, NULL, config)) {
> 		case REGMAP_ENDIAN_LITTLE:
> 			bus = &regmap_smbus_word;
> 			break;
> 		case REGMAP_ENDIAN_BIG:
> 			bus = &regmap_smbus_word_swapped;
> 			break;
> 		default:		/* everything else is not supported */
> 			break;
> 		}
>
> This is executed when regmap is initialized. My adapter has the I2C_FUNC_I2C functionality (I use a raspberry pi 4), so it seems to me like regmap_i2c is loaded as the bus. This uses i2c_transfer internally to read and write.
>
> For PEC I need regmap_smbus_word. This uses i2c_smbus_xfer internally. Unlike i2c_transfer, i2c_smbus_xfer can be used to send and receive PEC byte.
>
> What should I do?

It seems to me that regmap should not use functions not supporting PEC if
I2C_CLIENT_PEC is enabled on an i2c device. Of course, that is tricky if
not impossible to implement because the flag can be set at runtime but
the bus function assignment in regmap is static.

The only alternative I can think of is to define driver specific regmap
access functions to re-implement the regmap_smbus_word access functions.
That is less than perfect but better than to drop regmap access entirely.

Copying Mark Brown for additional input.

Thanks,
Guenter