Without this Linux driver reports that the currently set fan speed is 0.

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

According to PMBUS v1.3.1 part II specification the used prefix 0xff for extended
commands is reserved for future use:
"The PMBUS_COMMAND_EXT is reserved for future use to extend the PMBus command set
beyond the available 256 command codes."

As an alternative the standard points the MFR_SPECIFIC_COMMAND_EXT to be available
with 0xFE as the first byte of a command.
The same standard also points that standard commands from 0xc4 can be defined
by a manufacturer. To be compliant with the standard and to simplify the design
I have decided to move all used extended command to the standard range.
Be aware that this brings some non-backward compatible changes.

Changes include:
- main FW
- bootloader
- test_master
- documentation

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

These coefficients has to be fixed, there are just values to export values other than 0

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

To be called at the read of FAN_COMMAND_n. For some reason it fixes
random restarts after the write of FAN_COMMAND_n.

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

STATUS_BYTE is read only, STATUS_CML is R/W, can be cleared on write
with selected bits

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

also:
- merge cmd_data_lengths and ext_cmd_data_lengths
- sort commands (main_fw)
- align columns in cmds_cmds arrays

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

The size of a string is passed as a first byte

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

Signed-off-by: Adam Wujek <dev_public@wujek.eu>

They were

When the BE_COMPLIANT_USE_LINEAR16 switch is set to False, causing
the (non-standard-compliant) LINEAR11 to be used for the output
voltage instead of LINEAR16, the mode field of the VOUT_MODE register
shall return an invalid value (3'b111) to indicate non-compliance.

Especially functions that *set* TMR'ed variables *must* use it,
otherwise they only set one of the three copies, which makes the data
invalid.

When a bus error happened, the I2C peripheral's error bit remained
set. This caused it to respond erroneously in future transactions.

Now, when a bus error occurs, we're clearing the error bit and setting
the error flags in the STATUS_BYTE and STATUS_CML PMBus registers.
After the user reads those registers (using commands 0x78 and 0x7E),
their error flags are cleared.